Signal and Information Processing, Networking and Computers. Proceedings of the 9th International Conference on Signal and Information Processing, Networking and Computers (ICSINC) 9789811947742, 9789811947759

Citation preview

Lecture Notes in Electrical Engineering 895

Songlin Sun · Tao Hong · Peng Yu · Jiaqi Zou Editors

Signal and Information Processing, Networking and Computers Proceedings of the 9th International Conference on Signal and Information Processing, Networking and Computers (ICSINC)

Lecture Notes in Electrical Engineering Volume 895

Series Editors Leopoldo Angrisani, Department of Electrical and Information Technologies Engineering, University of Napoli Federico II, Naples, Italy Marco Arteaga, Departament de Control y Robótica, Universidad Nacional Autónoma de México, Coyoacán, Mexico Bijaya Ketan Panigrahi, Electrical Engineering, Indian Institute of Technology Delhi, New Delhi, Delhi, India Samarjit Chakraborty, Fakultät für Elektrotechnik und Informationstechnik, TU München, Munich, Germany Jiming Chen, Zhejiang University, Hangzhou, Zhejiang, China Shanben Chen, Materials Science and Engineering, Shanghai Jiao Tong University, Shanghai, China Tan Kay Chen, Department of Electrical and Computer Engineering, National University of Singapore, Singapore, Singapore Rüdiger Dillmann, Humanoids and Intelligent Systems Laboratory, Karlsruhe Institute for Technology, Karlsruhe, Germany Haibin Duan, Beijing University of Aeronautics and Astronautics, Beijing, China Gianluigi Ferrari, Università di Parma, Parma, Italy Manuel Ferre, Centre for Automation and Robotics CAR (UPM-CSIC), Universidad Politécnica de Madrid, Madrid, Spain Sandra Hirche, Department of Electrical Engineering and Information Science, Technische Universität München, Munich, Germany Faryar Jabbari, Department of Mechanical and Aerospace Engineering, University of California, Irvine, CA, USA Limin Jia, State Key Laboratory of Rail Traffic Control and Safety, Beijing Jiaotong University, Beijing, China Janusz Kacprzyk, Systems Research Institute, Polish Academy of Sciences, Warsaw, Poland Alaa Khamis, German University in Egypt El Tagamoa El Khames, New Cairo City, Egypt Torsten Kroeger, Stanford University, Stanford, CA, USA Yong Li, Hunan University, Changsha, Hunan, China Qilian Liang, Department of Electrical Engineering, University of Texas at Arlington, Arlington, TX, USA Ferran Martín, Departament d’Enginyeria Electrònica, Universitat Autònoma de Barcelona, Bellaterra, Barcelona, Spain Tan Cher Ming, College of Engineering, Nanyang Technological University, Singapore, Singapore Wolfgang Minker, Institute of Information Technology, University of Ulm, Ulm, Germany Pradeep Misra, Department of Electrical Engineering, Wright State University, Dayton, OH, USA Sebastian Möller, Quality and Usability Laboratory, TU Berlin, Berlin, Germany Subhas Mukhopadhyay, School of Engineering & Advanced Technology, Massey University, Palmerston North, Manawatu-Wanganui, New Zealand Cun-Zheng Ning, Electrical Engineering, Arizona State University, Tempe, AZ, USA Toyoaki Nishida, Graduate School of Informatics, Kyoto University, Kyoto, Japan Luca Oneto, Department of Informatics, Bioengineering., Robotics, University of Genova, Genova, Genova, Italy Federica Pascucci, Dipartimento di Ingegneria, Università degli Studi “Roma Tre”, Rome, Italy Yong Qin, State Key Laboratory of Rail Traffic Control and Safety, Beijing Jiaotong University, Beijing, China Gan Woon Seng, School of Electrical & Electronic Engineering, Nanyang Technological University, Singapore, Singapore Joachim Speidel, Institute of Telecommunications, Universität Stuttgart, Stuttgart, Germany Germano Veiga, Campus da FEUP, INESC Porto, Porto, Portugal Haitao Wu, Academy of Opto-electronics, Chinese Academy of Sciences, Beijing, China Walter Zamboni, DIEM - Università degli studi di Salerno, Fisciano, Salerno, Italy Junjie James Zhang, Charlotte, NC, USA

The book series Lecture Notes in Electrical Engineering (LNEE) publishes the latest developments in Electrical Engineering—quickly, informally and in high quality. While original research reported in proceedings and monographs has traditionally formed the core of LNEE, we also encourage authors to submit books devoted to supporting student education and professional training in the various fields and applications areas of electrical engineering. The series cover classical and emerging topics concerning:

• • • • • • • • • • • •

Communication Engineering, Information Theory and Networks Electronics Engineering and Microelectronics Signal, Image and Speech Processing Wireless and Mobile Communication Circuits and Systems Energy Systems, Power Electronics and Electrical Machines Electro-optical Engineering Instrumentation Engineering Avionics Engineering Control Systems Internet-of-Things and Cybersecurity Biomedical Devices, MEMS and NEMS

For general information about this book series, comments or suggestions, please contact leontina. [email protected]. To submit a proposal or request further information, please contact the Publishing Editor in your country: China Jasmine Dou, Editor ([email protected]) India, Japan, Rest of Asia Swati Meherishi, Editorial Director ([email protected]) Southeast Asia, Australia, New Zealand Ramesh Nath Premnath, Editor ([email protected]) USA, Canada Michael Luby, Senior Editor ([email protected]) All other Countries Leontina Di Cecco, Senior Editor ([email protected]) ** This series is indexed by EI Compendex and Scopus databases. **

More information about this series at https://link.springer.com/bookseries/7818

Songlin Sun Tao Hong Peng Yu Jiaqi Zou •

•

•

Editors

Signal and Information Processing, Networking and Computers Proceedings of the 9th International Conference on Signal and Information Processing, Networking and Computers (ICSINC)

123

Editors Songlin Sun Beijing University of Posts and Telecommunications Beijing, China Peng Yu Beijing University of Posts and Telecommunications Beijing, China

Tao Hong Beihang University Beijing, China Jiaqi Zou Beijing University of Posts and Telecommunications Beijing, China

ISSN 1876-1100 ISSN 1876-1119 (electronic) Lecture Notes in Electrical Engineering ISBN 978-981-19-4774-2 ISBN 978-981-19-4775-9 (eBook) https://doi.org/10.1007/978-981-19-4775-9 © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 This work is subject to copyright. All rights are solely and exclusively licensed by the Publisher, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed. The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. The publisher, the authors, and the editors are safe to assume that the advice and information in this book are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the editors give a warranty, expressed or implied, with respect to the material contained herein or for any errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. This Springer imprint is published by the registered company Springer Nature Singapore Pte Ltd. The registered company address is: 152 Beach Road, #21-01/04 Gateway East, Singapore 189721, Singapore

Preface

It is our great honor to welcome you to the 9th International Conference on Signal and Information Processing, Network and Computers (ICSINC 2021 Winter). The ICSINC 2021 Winter Committee has been monitoring the evolving COVID-19 pandemic. We have decided to delay the edition of this conference to December. ICSINC 2021 Winter provides a forum for researchers, engineers and industry experts to discuss recent development, new ideas and breakthrough in signal and information processing schemes, computer theory, space technologies, big data and so on. ICSINC 2021 Winter received 330 papers submitted by authors, and 163 papers were accepted and included in the final conference proceedings. On behalf of the ICSINC 2021 Winter Committee, we would like to express our sincere appreciation to the TPC members and reviewers for their tremendous efforts. Especially, we appreciate the sponsor, Springer, for the generous support and advice. Finally, we would also like to thank all the authors for their excellent work and cooperation. Xuesong Qiu XueTian Zhu Bo Rong ICSINC 2021 Winter General Co-chairs

v

Committee Members

International Steering Committee Songlin Sun Michel Kadoch Ju Liu Takeo Fujii Jiaxun Zhang Yi Qian

Beijing University of Posts and Telecommunications, China École de technologie supérieure, University of Quebec, Canada Shandong University, China The University of Electro-Communications, Japan China Academy of Space Technology, China University of Nebraska–Lincoln, USA

General Co-chairs Xuesong Qiu XueTian Zhu Bo Rong

Beijing University of Posts and Telecommunications, China China Unicom Research Institute, China Communications Research Centre Canada

Technical Program Committee Chairs Shuai Han Peng Yu Weiliang Xie

Harbin Institute of Technology, China Beijing University of Posts and Telecommunications, China China Telecom Research Institute, China

vii

viii

Committee Members

Publicity Chairs Tao Hong Lei Feng

Sponsor Springer

Beihang University, China Beijing University of Posts and Telecommunications, China

Contents

Wireless Communication Research on Key Technical Solutions for 5G Co-construction and Sharing Network . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Guiqing Liu, Pingping Lin, and Weiliang Xie

3

Virtual Network Service Failure Recovery Algorithm Based on Routing Survivability in IPv6 Network . . . . . . . . . . . . . . . . . . . . . . . . . Dan Luo, Kaibo Zhou, Zhonghua Liang, and Lei Chen

12

Efficient Physical-Layer Authentication with a Lightweight C&S Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Kefeng Pan and Xiaoying Qiu

19

Recent Advances of Rock Engineering and Communication Technologies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Junjie Wang, Michel Kadoch, and Zhaohua Li

25

Joint TDOA and FDOA Estimation Based on Keystone Transform and Chirp-Z Transform . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Xiaoyang Chen, Lede Qiu, Shuai Li, Ming Li, and Yihao Song

31

Industrial Wisdom Based on 5G Customized Network . . . . . . . . . . . . . . Dan Liu, Qiuhong Zheng, Peng Ding, and Yun Shen

37

Implementation of DOA Estimation Algorithm Based on FPGA . . . . . . Hengyuan Zhou, Xiaojun Jing, Bingyang Li, Zesheng Zhou, and Bogan Li

46

Research on Dynamic Spectrum Allocation of Space-Air-Ground Integration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Lei Liu and Qiang Liu

53

Research on Intelligent Access of Space-Air-Ground Integrated Network . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Lei Liu and Qiang Liu

66

ix

x

Contents

Spectrum Sensing Based on Federated Learning with Value Evaluation Mechanism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Zheng Liu, Junsheng Mu, Fangpei Zhang, Xiaojun Jing, and Bohan Li

80

Application of Artificial Intelligence for Space-Air-Ground-Sea Integrated Network . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Shaowei Zhang, Lei Liu, and Mohamed Cheriet

88

Machine Learning Based 5G RAN Slicing for Channel Evaluation in Mobile State . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103 Fangpei Zhang, Xiaojun Jing, Junsheng Mu, Jia Zhu, and Bohan Li Use Case Analysis and Architecture Design for 5G Emergency Communications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108 Hongbiao Jiao, Jijiang Hou, and Chengli Mei A Resource Allocation Method for Power Backhaul Network Based on Flexible Ethernet . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120 Dayang Wang, Song Jiang, Yong Dai, Wei Li, Huichen Xu, Lin Cong, Ying Wang, and Peng Yu Cooperative Routing Algorithm for Space-Based Information Network Based on Traffic Forecast . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129 Zhongxiang Jia, Ying Wang, and Hongyang Liu Exploration on the Practice Teaching of Environmental Design Network Based on Mobile Internet Technology . . . . . . . . . . . . . . . . . . . 139 Wei Meng and Hui Liu Modern Information Technology Develops Intelligent Elderly Care Service Industry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145 Xiu Sun Construction of Piano Live Broadcasting Platform Based on Wireless Network Communication Technology . . . . . . . . . . . . . . . . . . . . . . . . . . . 150 Heda Zhang Value Education System of College Students Based on Mobile Internet Technology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158 Shizhe Zhang The Application of Computer Virtual Reality Technology in the Athletic Training of Colleges and Universities . . . . . . . . . . . . . . . . . . . . 167 Juan Yin The Application of Intelligent Mobile Internet Methods in the Development of Smart Physical Education . . . . . . . . . . . . . . . . . . . . . . . 174 Yongcai Zheng

Contents

xi

Survey on Wireless Power Transfer in Future Mobile Communication Network . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182 Heng Wang, Xu Xia, Wen Qi, and Yanxia Xing Application of Modern Information Technology in Promoting the Reform of Art Teaching . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 191 Huiying Chen Application of Digital Media Technology in Modern Art Design . . . . . . 196 Huiying Chen Testing Method of Shipborne Radar in Virtual Verification System . . . 201 Fangning Tian Coexistence Analysis Between HIBS System and IMT System Below 3GHz Band . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 209 Li Wang, Cheng Wang, Weidong Wang, and Zhiyan Fan Photovoltaic Power Prediction Based on Wavelet Analysis . . . . . . . . . . 216 Lianhe Li, Jihan Cao, Tao Hong, Mingshu Lu, Weiting Zhao, and Linquan Fang Energy-Efficient Networking for Emergency Communications with Air Base Stations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 223 Zifan Li, Bozhong Li, Hongxi Zhou, Yuanlong Peng, Fang Chen, Jingyue Tian, and Peng Yu Space-Air-Ground Integrated Network Driven by 6G Technology . . . . . 232 Pingping Lin, Lei Liu, Xi Meng, and Michel Kadoch Federated Learning Based 6G NTN Dynamic Spectrum Access . . . . . . . 243 Pingping Lin, Xi Meng, Lei Liu, and Michel Kadoch Green Communication Architecture Based on Cloud Radio Access Network for Demand Response Resources of Virtual Power Plant . . . . . 251 Zhengyuan Liu, Chuan Liu, Xinyue Zhao, Fei Zhou, and Shidong Liu Real-Time Bandwidth Prediction and Allocation Method for Smart Grid Communication Network Services . . . . . . . . . . . . . . . . . . . . . . . . . 259 Bozhong Li, Fang Chen, Zifan Li, Xingyu Zhao, Zhuojun Jin, and Zhengyuan Liu Attack Portrait and Replay Based on Multi-spatial Data in Grid System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 267 Zhiyuan Pan, Lisong Shao, Xinxin Song, and Hui Guo Radar Signal Classification Based on Bispectrum Feature and Convolutional Neural Network . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 275 Haoyun Liu, Zesheng Zhou, Bingyang Li, Jia Zhu, Xiaojun Jing, and Bohan Li

xii

Contents

Blockchain and Edge Computing Blockchain-Based Power Internet of Things Data Access Control Mechanism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 285 Xinyan Wang, Zheng Jia, Qi Wang, Dong Li, Jing Zhang, Xin Chen, and Han Yan Consortium Blockchain Based Anonymous and Trusted Authentication Mechanism for IoT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 292 Tianhong Su, Wenjing Li, Di Liu, Shaoyong Guo, and Linna Ruan Domain Name Management Architecture Based on Blockchain . . . . . . . 303 Zhenjiang Ma, Feng Qi, and Wenjing Li A Novel Layered GSP Incentive Mechanism for Federated Learning Combined with Blockchain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 310 Jiangfeng Sun, Guangwei Zhan, Jiaxi Liu, and Yu Feng Vehicle Searching in Underground Parking Lots Based on Blockchain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 319 Haoming Zhang, Xiaojun Jing, Quan Zhou, Junsheng Mu, and Bohan Li Blockchain and Knowledge Graph Fusion Network Architecture Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 325 Mengqi Han, Xiaojun Jing, Jia Zhu, and Junsheng Mu Blockchain Performance Optimization Mechanism Based on Caching Strategy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 331 Xinyan Wang, Jing Zhang, Jizhao Lu, Beibei Zhu, Xiao Feng, and Han Yan Multi Energy Coordinated Dispatching of Virtual Power Plant Based on Blockchain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 338 Zuhao Wang, Bing Zhou, Xusheng Yang, Na Li, Bing Wang, Shaojie Yang, Yu Chen, and Ao Xiong Research on Distributed Energy Trading Mode and Mechanism Based on Blockchain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 345 Chang Liu, Xuwei Xia, Yong Yan, Junwei Ma, and Feng Qi BlockChain-Based Power Communication Network Cross-Domain Service Function Chain Orchestration Algorithm . . . . . . . . . . . . . . . . . . 351 Xinyan Wang, Zheng Jia, Wencui Li, Beibei Zhu, Feifei Zhang, Ying Zhu, and Peng Lin Research on Intelligent Intrusion Detection Method of Power Information Network Under Cloud Computing . . . . . . . . . . . . . . . . . . . 359 Chenyi Xia, Feixiang Ao, Jianping Xu, and Baoming Yao

Contents

xiii

Research on Cloud-Edge Collaboration Architecture for Intelligent Acquisition of Digital City Information Based on 5G Customized Network . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 365 Peng Ding, Qiuhong Zheng, Yun Shen, Dan Liu, and Shuntian Feng Identification Hierarchical Cooperative Caching Strategy Based on Edge Computing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 377 Yutong Wen, Wei Bai, Xin Xu, Yang Lu, and Shaoyong Guo Intelligent Park Organism Based on 5G Edge-Cloud Collaboration . . . . 386 Yuying Xue, Peng Ding, Yun Shen, Huibin Duan, Xuezhi Zhang, and Yaqi Song Portable Citrus Detection System Combining UAV and Edge Equipment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 398 Heqing Huang and Michel Kadoch Q-learning Based Computation Offloading Algorithm in Mobile Edge Computing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 404 Cheng Zhong, Shaoyong Guo, Pengcheng Lu, and Sujie Shao Data Analysis I Research on Smart City Platform Based on 3D Video Fusion . . . . . . . . 413 Lai Wei Research on Intelligent Finance in the Era of Big Data . . . . . . . . . . . . . 425 Xuemei Wu, Quan Zhou, Ronghui Zhang, and Bohan Li Data Mining and Reasoning of Radar Radiation Sources Based on Knowledge Graph . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 431 Lingxiao Li, Junsheng Mu, Fangpei Zhang, Xiaojun Jing, and Bohan Li Exploration of English Learning in Cloud Classroom APP Based on Information Technology Platform . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 439 Liping Zhang and Kaitlyn Huseyin Data Analysis of University Innovation Fusion Based on Big Data Technology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 448 Yan Zhang and John Madurai Investigation and Analysis of Online Teaching Documents Based on Data Mining Technology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 454 Yuan Liu and Baoquan Men Evaluation Index System of Student Achievement Based on Big Data Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 461 Jiexiu Ming and Riley Asghar

xiv

Contents

“VR + VCD” Information Technology to Realize the Teaching System Innovation Exploration and Algorithm Design . . . . . . . . . . . . . . . . . . . . 469 Jing Xie Data Asset Model Construction Based on Naive Bayes Algorithm Technology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 478 Lei Wang and Güzin Mayzus Cultural Creative Products Based on Information Processing Technology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 486 Xiaofei Zhou and Can Chen Research on Online Teaching Model Mining Based on Network Database . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 495 Baoquan Men and Yuan Liu Construction of Hotel Management Software Model Based on Network Information Technology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 502 Zhixue Qin Terminal Model of Japanese Listening Resource System Based on Digital Audio Technology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 509 Tingting Wu AR Technology of English Stereo Teaching Material Based on Computer Graphics Technology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 516 Jingjing Wang The Integration of Computer Network Technology and Innovation and Entrepreneurship Vocational Education Under the Background of “Internet +” . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 523 Wenting Yu Discussion on Online Learning Software Platform Based on Network Communication Technology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 531 Maozheng Fu, Zhenrong Luo, and Laiyan Yun Popular Vocal Music Recommendation System Based on Particle Swarm Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 539 Xiafei Fan Application of Intelligent Seismic Fracture Identification Technology to Permeability Prediction in Wubaochang Area, Eastern Sichuan . . . . 550 Yanguo Qiao, Zhigang Liu, Wei Luo, Chenrui Li, and Peter Fokker A Corpus-Based Evaluation System of Thesis Graduation Resources Under Intelligent Information Technology . . . . . . . . . . . . . . . . . . . . . . . 558 Yilong Yang, Qing He, Yadan Li, and Shinian Wu

Contents

xv

Computer Management Information System in University Management Mode Based on Information Technology . . . . . . . . . . . . . . 564 Wenjuan Miao Intelligent Technology in the All-Media Era Promotes the Spread of Chinese Culture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 573 Junmei Luo Problems and Improvement Strategies of Higher Mathematics Course Based on Data Mining Resources . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 580 Di Li Analysis on Characteristics of Customer Satisfaction to Gymnasiums Based on the Holographic Projection Technology . . . . . . . . . . . . . . . . . 589 Yan Yan Construction of Sports Network Information Platform Based on Literature Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 596 Yingyue Hu The Application of Intelligent Data Mining Model Technology in the Study of Physical Training Video System . . . . . . . . . . . . . . . . . . . . . . . . 605 Jinyuan Zhu and Jijun Chu Innovation of Internet Sports Model Based on Literature Analysis Technology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 612 Xinying Liu Interactive Innovation of Student Management Information Based on Internet Analysis Technology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 618 Liyuan Wang Data Analysis of the Influence of Internet Multimedia Communication Technology on the Quality of College Students . . . . . . . . . . . . . . . . . . . 625 Zheng Fu Discussion on Russian Online Blended Learning Mode Based on Internet Technology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 633 Bingqing Li Artificial Intelligence VR Technology Cultivates College Students’ Entrepreneurship Ability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 640 Wenbin Wu, Bixi Wang, and Isabella Sabet AWE Feedback on the Effectiveness of the Automatic Scoring System for English Writing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 647 Xiangyu Zhao and Madeline Kidston Research on Demand-Side Reforms and Measures Based on Big Data Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 655 Qinggui Huang

xvi

Contents

Informatization Teaching Reform of Accounting Courses Under Modern Information Technology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 663 Xiaoyu Yan, Hongyan Li, and Olivia Pratap The Application of Digital Technology in the Teaching of Oracle Bone Inscriptions in Middle Schools from the Perspective of Computer . . . . . 669 Yongming Chen Exploration of Financial App Software Simulation Practice Learning in the Internet Era . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 677 Xuemei Shen Application of AutoCAD in the Drawing of Archaeological Objects——Post-production of Complete Line Drawings of Standard Objects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 686 Qiwang Zhao and Qianyun Lyu The Application of Intelligent Equipment System Based on Information Technology in College Practice Teaching . . . . . . . . . . . . . . 695 Yanan Yin Development of Football Computer Experimental System Based on Literature Mining and Analysis Technology . . . . . . . . . . . . . . . . . . . . . 702 Yaoduo Xu Analyze the Application of Photovoltaic Coupling in Smart Rural Housing Based on the Data Survey Results . . . . . . . . . . . . . . . . . . . . . . 709 Xiangzhe Liu Analysis of Intelligent English Learning System Based on Intelligent Data Collection Technology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 716 Jing Liang Analysis on the Form of Electronic Music Network Learning Based on Network Information Technology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 723 Yixin Cui Data Analysis II Innovation of Teaching Management Model Based on Internet Technology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 733 Libin Duan Design of Engineering Cost Database Software for Nuclear Power Plants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 741 Yang Zhao, Yutong Cai, Erying Yi, and Xue Liu Analysis of the Reform of Vocal Music Teaching by Using Network Platform in the New Media Era . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 747 Yang Chen

Contents

xvii

A Design Based on Big Data Processing Frame for Data Mid-platform in Time of IoT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 755 Xiao Wang, Yingfu Sai, Lai Man, Xingze Zhang, Yan Zhao, and Shawn Wilson Innovative Design of University Dance Course System Based on Big Data Analysis Technology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 763 Man Lou The Integrated Development of Children’s Drama Education Based on Internet We-Media Technology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 770 Yun Du and Willie Schaik The Innovative Model of Teaching Combining Artificial Intelligence and Educational Big Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 779 Miaohong Wen Design of Campus Employment Information Service System Based on Service Design Concept . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 787 Yangning Zheng Investigation and Analysis of Library Information Service Based on Information Technology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 796 Haipeng Zhao Data Analysis of Village Cultural Knowledge Map Based on VR Virtual Technology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 805 Yanliang Chen Research on Sports Teaching App Based on Internet Statistical Data Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 813 Guochao Yu and Qin Fei Based on the Intelligent Statistical Software Stata15.0 to Study the Impact of Executive Compensation Incentives and Managerial Capabilities on Business Performance in Artificial Intelligence Companies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 822 Tian Huang, Yuxuan Yao, and Fred Sadegh Research on the Transformation Carrier of Scientific Research Achievements in Colleges and Universities Based on Computer Technology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 831 Zhibin Li, Peng Zhang, and Yadan Xu Analysis on the Application of Big Data Technology in the Curriculum System of Aesthetic Education in Universities . . . . . . . . . . . . . . . . . . . . 838 Tianyu Jiang and Ning Yang

xviii

Contents

Using Digital Technology to Analyze the Degree of Polymerization of Tooth Preparation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 845 Na Xie, Xiaoting Ji, Danyang Wang, Zixia Li, and Brent Tahack Using Computer Data Analysis Technology to Analyze the Credit Decision-Making Problem of Small and Micro Enterprises with Grey Correlation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 853 Zhihui Zhang, Shuanglin Shu, Yanang Li, Guanchen Liu, and Luming Yin International Political Economy Analysis of Free Trade Area Construction Under the Background of Big Data . . . . . . . . . . . . . . . . . . 858 Zehao Wu Informatization Reform of Market Research Courses in Undergraduate Colleges Based on Informatization Teaching Platform . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 866 Shenghui Jin and Hongyan Li Interactive Training of School Enterprise Cooperation of Hotel Management Major in Higher Vocational Colleges Under the Background of Information Age . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 873 Chunhua Wu Intelligent Innovation Management Measures of Rural Agricultural Economy from the Perspective of Information Technology . . . . . . . . . . 881 Jin Ma Construction and Research of 4E Performance Evaluation Model of Public Rental Housing Based on Decision-Making Support System . . . . 890 Yongxian Ming, Yuqi Dong, Leshi Guo, Shitao Lin, Haile Andargie, and Yu Zhou Using Big Data Analysis Technology to Analyze the Impact of Household Leverage Ratio on House Price Bubble . . . . . . . . . . . . . . . . . 900 Xuhesheng Chen Application and Development of Computer Information Network Technology in Management . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 910 Siyu Ning Exploration of Biochemistry Teaching Mode Based on Big Data Technology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 917 Feng Zhou, Lei Xiong, Huiming Hu, Yuexing Ma, and Yang Liu The Application and Practice of Data Mining Technology in the Integrated Teaching of Psychology and Computer . . . . . . . . . . . . . . . . . 923 Weinan Dong The Application of Multimedia Technology in Interactive Teaching Approach in Senior Middle School English Teaching . . . . . . . . . . . . . . . 933 Xuemei Chen, Dong Wang, Xia Wang, and Tingyong Ma

Contents

xix

Internet Shopping Willingness Based on Internet Information Technology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 939 Ziyue Hong and Xiangdong Yang The Application of Computer Technology in the Teaching of Human Resources Management in Colleges . . . . . . . . . . . . . . . . . . . . . . . . . . . . 946 Hongbo Zhu Development of Hainan Cruise Tourism Industry Based on Big Data Tourism Demand Forecast . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 954 Xiaomei Yang and Liping Chen Innovative Research on Multimedia English Teaching Curriculum Design Under the Background of Computer Big Data . . . . . . . . . . . . . . 962 Min Su Learning Effect of College Students Under SPOC Mode Based on Data Analysis Technology of SPSS Software . . . . . . . . . . . . . . . . . . . . . . . . . 970 Zhihai Hu and Lili Zhu Discussion on the Relationship Between Innovation and Enterprise Development Based on Big Data Analysis Technology . . . . . . . . . . . . . . 978 Lvqing Zhang and Adrian Dehghani Research on Big Data Collection and Application of Building Material Price Based on Focused Web Crawler . . . . . . . . . . . . . . . . . . . . . . . . . . 984 Shao-lin Zhang, Mei Rong, Rui-lin Dong, Jun Huang, and Peng-cheng Zhi Carrier Network Fault Diagnosis Algorithm Based on Network Characteristics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 990 Zhan Shi, Yutu Liang, Ying Zeng, and Linna Ruan Deterministic Technology and Its Application in the Mobile Network . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 996 Jinyan Li, Wen Qi, Heng Wang, and Yanxia Xing Joint Resource and Power Allocation Algorithm for Multi Cell Slicing Service in Power Grid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1006 Guoyi Zhang, Xudong Wang, Yu Zhou, Kunyi Xie, Fanqin Zhou, Wenwei Tao, and Yang Cao Research on Integrated Engineering Economic Platform Under the Background of Digital Nuclear Power . . . . . . . . . . . . . . . . . . . . . . . . . . 1013 Le Wang, Si-si Zhao, and Hai-yang Cao Deep Learning and Machine Learning Improving Wireless Devices Identification Using Deep Learning Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1021 Kefeng Pan and Xiaoying Qiu

xx

Contents

A Fairness-Aware UAV Trajectory Design with Reinforcement Learning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1026 Bowei Xu, Liang Peng, Xiaoxiang Wang, Lixin Jiang, Yunxia Zhang, and Peng Zhang ISAR Image Target Location Algorithm Based on Template Matching . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1032 Lingkang Kong, Jingcheng Zhao, Xu Chao, Tao Hong, and Michel Kadoch Facial Expression Recognition via Re-parameterized Feature Fusion Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1038 Aoran Chen, Hai Huang, Yang Yang, Bohan Li, and Xiaojun Jing AffectRAF: A Dataset Designed Based on Facial Expression Recognition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1044 Sijie Wei, Xiaojun Jing, Aoran Chen, Qianqian Chen, Junsheng Mu, and Bohan Li Optical Music Recognition Based Deep Neural Networks . . . . . . . . . . . 1051 Yaqi Song, Yun Shen, Peng Ding, Xuezhi Zhang, Xiaohou Shi, and Yuying Xue Research on UAV Communication Based on Artificial Intelligence . . . . 1060 Shaowei Zhang, Lei Liu, and Qiang Liu Automatic Modulation Classification Based on Knowledge Graph . . . . . 1073 Jiawei Pan, Junsheng Mu, and Xiaojun Jing Clockwork Convolutional Recurrent Neural Network for Traffic Prediction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1079 Yu Yan, Yang Yang, Zhipeng Gao, Bowen Gao, and Rui Lv Constructing a Modular Curriculum System of College Education Based on BP Neural Network Algorithm . . . . . . . . . . . . . . . . . . . . . . . . 1087 Wenbo Wang, Chunmiao Li, Mingzhu Zheng, Daiping Cai, and Emily Squires A Time-Delay-Sensitive Traffic Data Forwarding Scheme Based on Deep Reinforcement Learning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1095 Wenjing Li, Wei Deng, Ningchi Zhang, Yanru Wang, Liuwang Wang, and Wenjie Ma Research on Computer Aided Translation Transfer Based on Intelligent Speech Recognition Technology . . . . . . . . . . . . . . . . . . . . . . . 1101 Xueshuo Ma Computer CT Imaging Technology in the Detection and Analysis of Coronary Artery Disease . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1110 Aiju Guo

Contents

xxi

Construction of Psychological Evaluation System for Coal Mine Rescue Workers Based on BP Neural Network Algorithm . . . . . . . . . . . 1118 Zhichen Zhang Research on the Application of Artificial Intelligence Technology in the Cultivation of Oral English Communicative Competence . . . . . . . . . 1126 Jingtai Li and Simin Tang Intelligent Teaching Mode of Chinese Language and Literature Based on Artificial Intelligence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1132 Dan Wang Design Form of Landscape Sculpture Based on the Aid of Computer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1141 Yisheng Zhang Online Intelligent Classroom Teaching Innovation of Pharmacy Based on Artificial Intelligence Technology . . . . . . . . . . . . . . . . . . . . . . . . . . . 1147 Lin Wang, Weiwei Bian, Yiting Fu, Yanxiang Li, Jingjing Ming, and Chunzhen Zhao Artificial Intelligence-Based SDA Technology Improves the Deasphalting Effect and Mechanism of Inferior Solvents . . . . . . . . . . . . 1155 Bo Tian, Chaohe Yang, Jie Liu, and Hao Zhang Diversified Characteristics and Performance of Network Visual Communication Design Based on Internet Technology . . . . . . . . . . . . . . 1165 Hongbo Sun and Andrew Mattek Reform of Modern Art Education System Based on Neural Network Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1173 Xin Sun and Ning Yang Analysis on the Application of Evolutionary Computing Technology in Innovative Art Teaching System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1181 Tianyu Jiang and Xin Sun Application of Artificial Intelligence System in the Design of Education Management System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1189 Qin Yin, Lijiao Xu, and Zhentao Dai Ethical Reflection on the Deep Embedding of Artificial Intelligence in University Teaching . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1197 Jie Zhang and Dennis Andwari Artificial Intelligence-Based Electric Energy Meter Operating Error Monitoring Data Fitting System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1205 Zhengang Shi, Chaofei Wu, Wenjie Fu, Peng Tao, Linhao Zhang, and Bo Gao

xxii

Contents

Intelligent Monitoring System of Aquaculture Water Environment Based on Internet of Things . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1212 Yang Yang China’s Inter-provincial Green Economy Efficiency and Forecast Based on SBM-DES and Deep Neural Network . . . . . . . . . . . . . . . . . . . 1222 Chao Yang, Feng He, Anitha Moosa, and Rong Liu Track and Field Competition Track and Field Monitoring System Based on TEB Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1231 Wenchao Li Strategies and Paths for Training Network and New Media Professionals in the Era of Big Data Technology . . . . . . . . . . . . . . . . . . 1240 Rong Fan Soybean Futures Forecasting Based on EMD and Transformer . . . . . . . 1245 Jie Zhang, Liulin Zhen, and Dongsheng Zhai Cable Price Forecast Based on Neural Network Models . . . . . . . . . . . . 1254 Yi Fu, Qi-fan Zhang, and Chi-yu Zhang An Effective End-To-End Resource Elastic Allocation Mechanism for SDN Based on Deep Learning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1261 Run Ma, Xiaobo Li, Shuang Wu, Guoli Feng, Shengjie Wang, Xinnan Ha, and Peng Lin Power Load Forecasting Algorithm Based on Regression Support Vector Machine . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1268 Xi Chen, Jiao Peng, Yue He, Bo Zhang, Dan Jiang, and Peng Lin Author Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1275

Wireless Communication

Research on Key Technical Solutions for 5G Co-construction and Sharing Network Guiqing Liu1 , Pingping Lin2(B) , and Weiliang Xie2 1 China Telecom Corporation, Beijing, China 2 China Telecom Research Institute, Beijing, China

[email protected]

Abstract. Since 5G networks are mostly deployed in high frequency bands, in order to achieve the same coverage as 4G, the number of base stations will be greatly increased. Coupled with factors such as the high price of 5G base stations, high power consumption, and difficulty in site selection, it is very meaningful to explore the co-construction and sharing network of multiple operators, which can reduce network construction costs and enhance network benefits. The article focused on several key technical solutions for 5G co-construction and sharing networks, including network architecture, NSA sharing technology solutions, and SA sharing evolution solutions. There were two main 5G shared network solutions, access network sharing and roaming in different networks. The sharing network of China Telecom and China Unicom adopted the access network sharing scheme, which included the initial NSA network sharing and target SA network sharing. The technical solutions of NSA network included two implementation modes of single anchor and double anchor, as well as the voice fallback solution. The article also introduced the evolution of NSA sharing network to SA network and the voice of SA sharing network. Keywords: 5G · Co-construction and sharing · Access network sharing · NSA · SA · Voice

1 Introduction In the 5G era, the State Council put forward the strategic requirements of 5G leadership, and through innovation, openness, and sharing to create a national 5G network to meet the needs of people in different regions and at different stages of development. However, 5G network construction is facing challenges in many aspects. First, the frequency band for deploying 5G has been improved compared to 4G. Although 5G has introduced Multiple Input Multiple Output (Massive MIMO) technology [1], it is limited by the shortcomings of uplink coverage and the coverage distance of a single station has shrunk. The number of 5G base stations has multiplied and the introduction of Massive MIMO technology has increased the number of AAU (Active Antenna Unit) channels. co-construction and sharing between operators may be the only way to reduce costs and increase efficiency [2, 3]. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 S. Sun et al. (Eds.): ICSINC 2021, LNEE 895, pp. 3–11, 2022. https://doi.org/10.1007/978-981-19-4775-9_1

4

G. Liu et al.

5G networking strategies include SA (Stand Alone) and NSA (Non-Stand Alone). The scheme needs to fully consider evolution capabilities and future competitiveness. The principle of protecting the 4G user experience on the existing network is avoiding or reducing the impact on existing 4G users as much as possible [3–5]. In view of the different 5G networking modes of NSA and SA, the co-construction and sharing technology also has differences in technical solutions [6, 7].

2 Analysis of 5G Shared Network Solutions In 2019, China Telecom and China Unicom will jointly promote the cooperation and coconstruction of 5G networks across the country [8]. China Telecom and China Unicom have been actively exploring ways to co-construction and sharing 5G networks. Coconstruction and sharing involves many aspects, such as the sharing of site infrastructure and the sharing of technical solutions. The sharing of site infrastructure is a relatively common way. Towers, computer rooms, power supplies, shelters can all be shared, but the specific network elements of each operator, such as BBU, AAU, RRU and other equipment, are also operated independently. It can effectively reduce the cost of basic physical equipment. At the same time, each operator’s radio equipment is independent, operation and maintenance are simple, and no specific network equipment is involved, but the power supply and transmission are the superposition of multiple sets of operator equipment. There are two main network sharing solutions: the access network sharing solution and roaming in different networks solution. 2.1 Access Network Sharing The network topology of the solution is shown in Fig. 1.

Operator A Core Network

Operator B Core Network

Bearer Network

Shared Base Station

Operator A UE

Operator B UE

Fig. 1. Access network sharing topology.

3GPP recommends Multi-Operator Core Network (MOCN) and Gateway Core Network (GWCN) two modes of sharing network architecture [9, 10]. Access network sharing can be divided into independent carrier and sharing carrier according to different sharing carrier configurations. The independent carrier scheme is to configure

Research on Key Technical Solutions

5

two carriers and broadcast their respective PLMNs on different carriers. The cells are independent, and schedule independent frequency resources. There is no competition for services and no need to consider resource allocation strategies. It is suitable for areas with high traffic volume and some business innovation scenarios. The sharing carrier scheme is to configure one or two carriers to realize the sharing of frequency resources. Two PLMNs are broadcast in the shared cell, and the same cell-level characteristic parameters are used. The specific parameters need to be negotiated and configured by both operators. At the same time, it is necessary to negotiate the allocation of air interface resources and adopt the same QoS (Quality of Service) strategy. For base stations that share a single carrier, because two operators use the same frequency, one of them will introduce inter-frequency networking over the shared boundaries. The carrier bandwidth can be configured, which is suitable for areas with low or high traffic. 2.2 Roaming in Different Networks The roaming in different networks solution is shown in Fig. 2. The roaming in different networks sharing solution can provide 5G sharing user access by broadcasting the sharing operator’s PLMN through the sharing base station. Users are still showing services provided by the home operator, and spending accounts, billing and policy control are still implemented in their respective core networks. In the NSA phase, the 4G core network is connected and the new network number is broadcast for the other operator’s 5G users to access. In the SA phase, the 5G core network is connected and the new network number is broadcast for the other operator’s 5G users to access. Compared with international roaming, there is 2/3/4G network coverage in the roaming area, and it is necessary to distinguish between 5G and non-5G users.

Operator A Core Network

Operator B Core Network

Bearer Network

Shared Base Station Operator A UE

Operator B UE

Fig. 2. Roaming in different networks topology

The main difference between roaming in different networks and the access network sharing scheme is whether the sharing base station is connected to the core network of the contractor or the core network of both operators. In the roaming solution, the sharing base station is only connected to the core network of the contractor, and its network configuration and resource scheduling are controlled by the contractor. The

6

G. Liu et al.

support for the subsequent evolution of network slicing and MEC and other emerging services of SA is poor. Both schemes are technically feasible, but the roaming in different networks solution has a certain impact on the service experience, which is embodied in the following aspects: (1) The delay of some services is longer than that of users of this network; (2) The boundary of the roaming area may be stuck or dropped; (3) It is impossible to support the ownership of roaming users of the sharing operator.

3 NSA Network Sharing Technology Solution NSA network is a networking method that builds 5G network on the basis of 4G network and integrates 4G network with 5G network. A variety of network architectures are defined in the protocol, and option 3× is prefered [11, 12]. This standardization is the earliest to complete, which is conducive to rapid commercial use. The network changes are small, the network construction speed is fast, and the investment is relatively small. There is no requirement for 5G coverage, and dual connections are supported for offloading, and the user experience is good. Of course, there are certain shortcomings. 5G base stations and existing 4G base stations must work together. The flexibility is low because 4G and 5G come from the same manufacturer. It is unable to support the relevant new functions and new services introduced by the 5G core network. NSA networking is much more complicated than SA networking. Because the architecture uses 4G network, the massive IOT access and low latency characteristics of 5G networks cannot be used. 3.1 Single Anchor Implementation of NSA Sharing In the single anchor implementation of NSA network sharing, a single-anchor sharing carrier scheme or a single-anchor independent carrier scheme can be configured. The schematic diagram of the single anchor implementation is shown in Fig. 3. In this solution, both the 5G base station and the 4G anchor base station need to be shared, and the 5G base station and the 4G anchor base station must be equipment from the same manufacturer. In the areas of different 4G manufacturers of both operators, the sharing

Operator A EPC

Operator B EPC

Bearer Network

4G eNB

Operator A UE

X2 5G gNB

Both 4G and 5G need to be shared and made from the same manufacturer.

Operator B UE

Fig. 3. Single anchor implementation of NSA sharing network

Research on Key Technical Solutions

7

of 5G networks can be realized by transforming existing 4G base stations or building new 4G anchor base stations. 5G users share 5G and 4G at the same time, and 4G users do not share it. The built 1.8 GHz 4G base station can be transformed. The contractor’s original 1.8 GHz 4G base station is equipped with anchors and newly built 5G base stations. The shared 4G anchor station needs to broadcast the old network number of the contractor and the new network number of the sharing operator. Configure the other operator’s new network number as the equivalent PLMN for the 5G UE, and always display the LOGO of your own network. The 4G base station of the sharing party needs to be configured to support the identification of 5G NSA UE, and to configure the frequency priority for it to point to the anchor frequency band configured by the contractor. The shared 4G users use the home 4G network and do not use the 4G anchor shared network. The comparison between the reconstruction scheme and the new scheme is shown in Table 1. Table 1. Reconstruction & new-build. Reconstruction (1.8G + 3.5G) New-build (2.1G + 3.5G) Implementation

1.8G base station reconstruction

Build 2.1G anchor station and 5G base station at the same time

Expansion requirements

Need to be expanded according to the number of inbound users

2.1G bearer NSA network, no need to expand

User experience

4G user experience declines

The original 4G user experience remains unaffected

Interoperability

4G/5G user interoperability is 4G/5G interoperability complex and difficult to relationship is simple and easy to optimize optimize

Evolution

Some old 4G stations do not support upgrades, and the evolution is difficult

2.1G adopts new equipment and can be DSS mode to smoothly evolve to 5G

3.2 Double Anchors Implementation of NSA Sharing Double anchors of NSA sharing is realized by sharing only 5G base stations, not 4G base stations. The 5G base stations are respectively connected to their 4G anchor base stations, and the 4G base stations of both operators are used as anchor, and the same 5G base station is anchored at the same time, as shown in Fig. 4. Since there may be problems in the interoperability of the X2 interface between different manufacturers, the 4G base station and the 5G base station must be the same manufacturer, and the 4G base stations of the two operators must also be the same manufacturer. In the area of the same manufacturer of 4G base stations of both operators, the co-construction and sharing of 5G networks can be quickly realized under the condition of small transformation. In the 5G coverage area, under the guidance of the anchor base station, 5G NR connections

8

G. Liu et al.

are added as needed. When the UE is moving, based on the 4G situation of the local network, the handover between anchors is performed. First, the 5G NR connection is disconnected, the 4G connection is retained, and the 4G handover is performed. After the handover is completed, if there is 5G coverage, the 5G NR connection is activated again.

Operator A EPC

Operator B EPC

Bearer Network

Operator A 4G eNB

X2

X2

Operator B 4G eNB

5G shared gNB Operator A UE

Operator B UE

Fig. 4. Double anchors implementation of NSA sharing

The double anchors implementation requires that the 4G base stations and 5G base stations of the two operators are of the same manufacturer. It is only applicable to areas where the 4G base stations of the two operators are the same manufacturer or areas where 4G anchor base stations are newly built. Currently, the proportion of the same manufacturer in the country is not Very high, the application range is limited, but the network configuration and transformation are relatively simple. The single anchor mode only requires the same manufacturer configuration between the 4G anchor base station and the newly-built 5G base station. The 4G network can be of different manufacturers, so it is suitable for most 4G networks of different manufacturers, but the network configuration is relatively complicated. Both the single anchor solution and the double anchors solution are technically mature and available, and the operator can choose which solution according to the actual situation of the 4G base station in the existing network. 3.3 Voice Solution of NSA Sharing In the NSA networking mode, 4G and 5G adopt a dual-connection mode, and the voice service mainly uses VoLTE provided by 4G. Under the single anchor configuration network, there are two voice bearing solutions for 5G users. The difference is whether to return to the home network for voice services. One solution is that the 4G anchor station directly bears the VoLTE service, and the voice service of the 5G UE is directly carried by the contractor’s 4G network. Its advantages are low service initiation delay and support for concurrent services of VoLTE and 5G services. The disadvantages are that the anchor carrier has a larger service volume, poor user experience, and inter-network settlement is required. The second solution is that VoLTE services return to the sharer’ 4G local network and return to the 4G/5G network of the contractor after the service

Research on Key Technical Solutions

9

is released. The advantage is that the VoLTE service is independent, no settlement is required and the quality of the service is guaranteed by the network. The disadvantage is that the service initiation delay increases and the 5G data service cannot be concurrently used during VoLTE. It is necessary to increase the 4G network configuration of the contractor, and it needs to support the function of configuring the neighbor cell list based on PLMN and setting the frequency priority based on PLMN and business. When the contractor’s 4G anchor station receives a VoLTE voice service request from the sharer’s 5G UE, the contractor’s 4G anchor station triggers the sharer’s 5G UE to return to its 4G for VoLTE service, and the sharer’s VoLTE user needs to perform an additional inter-frequency handover. So the VoLTE access delay will be increased. Under the double anchors configuration network, since the 4G anchor stations are owned by their respective operators and are independent of each other, VoLTE services can also be carried on their respective 4G networks. The voice solution is relatively simple and does not involve inter-network settlement and other issues.

4 SA Network Sharing Technology Solution With the construction and development of 5G networks, in order to meet the requirements of the three major business scenarios of eMBB, URLLC and mMTC, and to realize the application of technologies such as network slicing and MEC, the network needs to evolve from NSA sharing to SA sharing. Because most 5G UE sold in the early stage of 5G only support NSA, the compatibility of NSA UE needs to be considered. At the same time, it is also necessary to consider the issue of foreign NSA users roaming to China. Therefore, in the process of evolving from the NSA network to the SA network, the existence of the NSA/SA dual protocol stack sharing network will continue for a long time. The NSA/SA dual-mode sharing network structure is a transitional solution. The shared 5G NR base station needs to access the four core networks of two operators. The shared network architecture is very complex, and its network architecture is shown in Fig. 5.

Operator A EPC

Operator A 5GC

Operator B EPC

Operator B 5GC

X2

4G shared eNB

5G shared gNB

Fig. 5. NSA/SA dual protocol stack sharing network architecture

With the development of the industry chain, the massive promotion of SA UE, and the later elimination and updating of NSA UE, SA network sharing will inevitably become the target network architecture. In SA network, the access network sharing architecture

10

G. Liu et al.

is the same as that of NSA. Both the contractor and the sharer need to build a new 5G core network, and the user, billing, and policy control are still implemented in their respective core networks. At the same time, the 5G base station needs to configure different parameters according to the PLMN to realize interoperability with the 4G base stations of both operators. The voice service of the SA sharing network is also much simpler. Initially, VoNR is not yet mature, and the voice service needs to fall back to the 4G network for VoLTE through EPS Fallback [13]. When both 5G and 4G base stations are shared, VoLTE services fall back to the contractor’s 4G network. When the terminal initiates a voice call, the gNB triggers a handover when the IMS voice channel is established on the NR. At this time, the gNB initiates a redirection or Inter-RAT handover request to the 5GC, and then falls back to the LTE network, and VoLTE provides voice services. After the VoNR technology matures, voice can be directly carried on the 5G network through VoNR, and both voice and data services are carried on the 5G network, so that the voice quality and delay can be improved to a certain extent. Figure 6 shows a schematic diagram of SA network voice bearer.

VoNR

NR

5GC

EPS Fallback

N26

EPC

LTE VoLTE

Fig. 6. SA voice bearer diagram

5 Conclusion China Telecom and China Unicom’s 5G co-construction and sharing has created a precedent for domestic operators to co-construct and share. It is of great significance to the development of the country, the operators themselves, and the healthy development of 5G. The 5G co-construction and sharing technical solution framework is very feasible. This article introduced sharing network solutions, including access network sharing solutions and roaming in different networks solutions. In the early stage of 5G network construction, due to the immature industry chain, only NSA deployment was possible. There are two implementation solutions for NSA deployment, single anchor and double anchors. The double anchors solution is only applicable to devices from the same manufacturer of two operators, so the deployment conditions are limited. The NSA stage sharing technical solution is complicated, the design transformation workload is large, and the network management and optimization are difficult. Under the condition that SA has commercial capabilities, it will evolve and upgrade to SA co-construction and sharing as soon as possible to improve network quality and enhance network competitiveness. As a basic service, it is necessary to give priority to ensuring the voice service

Research on Key Technical Solutions

11

experience. Under the NSA sharing network, the voice needs to fall back to the 4G network for VoLTE, and it can go back to the 4G anchor base station or return to the home 4G network. Each has its advantages and disadvantages. After evolving to the SA sharing network, the current voice scheme is to fall back to the respective 4G network through the EPS Fallback method and VoLTE provides voice services. VoNR is the target voice bearing method of SA sharing network. The network sharing of China Telecom and China Unicom has brought operators a new model of network construction and operation, which is of far-reaching significance to the development of 5G.

References 1. Larsson, E.G., Edfors, O., Tufvesson, F., et al.: Massive MIMO for next generation wireless systems. IEEE Commun. Mag. 52(2), 186–195 (2014) 2. Lin, P.: Analysis on power configuration in 5G co-construction and sharing network. In: IEEE/CIC International Conference on Communications in China (ICCC Workshops), pp. 349–352 (2021). https://doi.org/10.1109/ICCCWorkshops52231.2021.9538894 3. Lin, P., Hou, J., Xu, X.: Research on 5G SA Mobility Management. In: 2021 International Wireless Communications and Mobile Computing (IWCMC), pp. 503–507 (2021). https:// doi.org/10.1109/IWCMC51323.2021.9498977 4. Zhu, C., Wang, Q., Li, X.: Mobile Communications After 5G 2020. People’s Posts and Telecommunications Press, Beijing (2016) 5. Ying, L., Wu, J., Peng, J.: Practice of 5G NSA co-construction and sharing scheme. Post Telecommun. Des. Technol. 12, 22–27 (2020) 6. Liu, X., Sun, X., Du, Z.: 5G Wireless System Design and International Standards. People’s Posts and Telecommunications Press, Beijing (2019) 7. Wang, D., Xu, G., Wei, D., et al.: 5G Wireless Network Technology and Planning Design. Renmin Post and Telecommunications Publishing House, Beijing (2019) 8. Liu, Q., Xiang, C., Duan, J., et al.: Scheme of co-construction and sharing alignment between China Unicom and China telecom in 5G NSA stage. Electron. Prod. World 27(08), 77–81 (2020) 9. 3GPP TS 23.251 V16.0.0, Network Sharing; Architecture and functional description (2020) 10. 3GPP TS 23.501 V16.2.0; Technical specification group services and system aspects; system architecture for the 5G system (5GS) (2019) 11. 3GPP TR 38.801 V14.0.0; Technical Specification Group Radio Access Network; Study on new radio access technology: radio access architecture and interfaces (2017) 12. Dahlman, E., Parkvall, S., Skold, J.: 5G NR: The Next Generation Wireless Access Technology. China Machine Press, Beijing (2019) 13. Ma, J., Yang, Z., Zhu, X.: Discussion on 4G solutions for 5G voice fallback. Mobile Commun. 43(4), 37–42 (2019)

Virtual Network Service Failure Recovery Algorithm Based on Routing Survivability in IPv6 Network Dan Luo(B) , Kaibo Zhou, Zhonghua Liang, and Lei Chen China Academy of Information and Communications Technology, Beijing 100191, China [email protected]

Abstract. How to recover failed network services when recovery resources are limited is a common problem encountered in ipv6 networks. In order to solve this problem, this article makes innovations from the two dimensions of restoring important nodes, restoring cut-set links, and puts forward a service failure recovery algorithm. When restoring nodes, restore the node with the largest loss first, and then restore the central node, which helps restore more links. When restoring links, priority is given to restoring unallocated links in the same cut to improve link survivability. Through experiments and comparisons, the algorithm of this article has a better virtual network service failure recovery rate under different network scales and different amounts of recovery resources. Keywords: IPv6 network · Failed network · Routing survivability · Survivability · Failure recovery

1 Introduction In the IPv6 scenario, network services have increasingly stringent requirements for network service quality. In order to meet the requirements of network services, network operators adopt network virtualization technology to improve the utilization and availability of network resources [1, 2]. The factor related to the quality of service of the network is the probability of network failure. Therefore, measures to improve network service quality can be carried out in two dimensions: reducing the probability of failure and taking measures to quickly recover from failures. In terms of reducing the probability of failure, the main way is how to improve the quality of network equipment and reduce the misconfiguration of network equipment. In terms of taking measures to quickly recover from failures, the main way is to take the most efficient measures to recover the failed resources after the failure occurs. After the failure of network resources, if the network services affected by the failure are quickly restored, many scholars and experts have conducted research and have achieved more research results. Literature [3] designs a failure recovery algorithm based on the characteristics of virtualization dynamic migration and on-demand expansion, which improves the recovery ability of failure resources. Literature [4] analyzes the © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 S. Sun et al. (Eds.): ICSINC 2021, LNEE 895, pp. 12–18, 2022. https://doi.org/10.1007/978-981-19-4775-9_2

Virtual Network Service Failure Recovery Algorithm

13

location of network resources and their importance in network reliability, and designs backup reliability algorithm, which improves the recovery ability of failure networks. To improve the recovery rate of failure resources, literature [5] uses linear programming algorithm. For the problem of low efficiency of resource in a large-scale network, literature [6] analyzed network attributes and proposed a regional resource recovery strategy. For the prioritize the restoration of key network devices, literature [7] prioritizes the restoration of failure network resources with higher evaluation values in the network topology. Literature [8] uses network connectivity characteristics to analyze, identify key network resources and perform resource recovery. For a specific network environment, literature [9] analyzed the failure characteristics in the SDN network environment, and designed a recovery algorithm for failure resources, which improved the recovery ability of customer link resources. This article makes innovations from the two dimensions of restoring important nodes and restoring cut-set links, designs a routing survivability failure recovery algorithm. When restoring nodes, restore the node with the largest loss first, and then restore the central node, which helps restore more links. When restoring links, priority is given to restoring unallocated links in the same cut to improve link survivability.

2 Network Environment The network environment is the basis and conditions for network operation. A detailed understanding of network characteristics is a prerequisite and condition for completing network failure recovery. In terms of network environment, network elements related to failure recovery include network nodes and network links. The functions of network nodes mainly include providing computing resources and processing services according to business needs. The function of the network link mainly includes the provision of network connectivity according to the connectivity needs of the business, and the transmission of business data according to the bandwidth needs of the business. It can be known through the analysis of network elements and their functions. The attributes contained in network nodes are mainly computing resource attributes. The attributes included in the network link are mainly bandwidth resource attributes. Use G S = (N S , E S ) to represent the underlying network. Use CPU (nSi ) to represent the computing resources of the underlying node. Use bw(ejS ) to represent the bandwidth resource of the underlying link. Use G V = (N V , E V ) to represent a virtual network. Use CPU (nVi ) to represent the computing resources of the virtual node. Use bw(ejV ) to represent the bandwidth resource of the virtual link. In order to associate the underlying node with the underlying link, use N V → N S to represent the resource allocation from the virtual node to the underlying link. Use E V → E S to represent the resource allocation from the virtual link to the underlying link.

3 Analysis of Resource Characteristics 3.1 Node Importance In order to judge the importance of the node, this article first analyzes the relevant attributes of the node. First of all, from the description of the network environment,

14

D. Luo et al.

the number of network resources is an important factor in evaluating the importance of network resources. When the network resources are sufficient, sufficient resources can be provided for the business to meet the business demand for resources. When the network resources are lacking, not only will the service be unavailable, but the service will be prone to failure due to the high utilization rate, thereby reducing the service quality of the service. Use formula (1) to calculate the number of used resources of the node. It can be seen from formula (1) that the used resources are the sum of the computing resources allocated to the virtual nodes. Use formula (2) to calculate the number of used resources of the link. It can be seen from formula (2) that the used resources are the sum of the bandwidth resources allocated to the virtual link.  cpu(nVj ) (1) NUsed (nSi ) = V S nj ∈Ni

LUsed (nSi ) =



ejV ∈EiS

bw(ejV )

(2)

Use formula (3) to calculate the distance from the current bottom node to other nodes in the bottom network. It can be seen from formula (3) that the fewer the number of hops from each bottom node to other nodes, the closer the distance between the bottom node and other nodes. According to the definition of formula (3), if we know the distance between the current node and the underlying node that has allocated resources, we can better allocate resources for the new virtual node, thereby significantly reducing the resource overhead of link bandwidth. Core(nSi ) =  LS(nSi ) =



nSj ∈E S

nSsi ∈SMAP

1 hops(nSi , nSj )

(3)

hops(nSi , nSsi )

(4)

After analyzing and calculating the attributes of the node resources, the various attributes are summarized by formula (5) to calculate the importance of each underlying node. It can be seen from formula (5) that it is necessary to calculate the attributes of the four bottom nodes in order to better analyze the attributes of the bottom nodes. IMPOT (nSi ) =

NUsed (nSi ) + LUsed (nSi ) LS(nSi )

• Core(nSi )

(5)

3.2 Node Recovery Value The node loss value is related to the probability of node failure. The higher the probability of a node failure, the more likely the current node will fail. The probability of node nSi failure is defined as the node failure risk factor Ri , which is used to describe the probability of node nSi failure (formula 6). pj is the normalized value of the hops among the faulty substrate node and other faulty underlying nodes. The larger the value, the faultier nodes around the faulty node. If the current node is restored, the probability of service recovery

Virtual Network Service Failure Recovery Algorithm

15

is small. Because more nodes around the node have failed. At this time, if only the current node is restored, other nodes still fail. phi represents the recovery probability of node nSi , and its value is obtained from historical experience and is inversely proportional to the difficulty of recovery of node nSi . If it is less difficult to recover the node nSi , the value is larger. Ri =

phi pi

(6)

4 Algorithm The cut sets of graphs include point cut sets and edge cut sets. This article studies the survivability of routing, so the relevant knowledge of edge cut sets is used, hereinafter referred to as cut sets. Here, the edge cut set refers to a subset of the set formed by the links of the graph. If all the links in this subset are deleted, the graph becomes a disconnected graph at this time. Therefore, for a virtual network topology, if the underlying link does not belong to the underlying link in the same cut set, the current virtual network topology has routing survivability. The detailed process of the algorithm is shown in Table 1. Table 1. Algorithm Input: Network, Failure underlying nodes, Failure underlying links, Available node recovery resources, Available link recovery resources Output: The set of recovered faulty underlying nodes, Set of recovered faulty underlying links 1. According to the node characteristics, restore the failed node 1.1 calculate the node importance using formula (5); 1.2 calculate the recovery value using formula (6); 1.3 According to the node recovery value, sort the fault nodes in descending order; 1.4 Recover the failed underlying nodes from the collective in turn until the available node resources are used up 2. According to the characteristics of the underlying link, restore the faulty underlying link 2.1 Find faulty virtual networks that do not have faulty underlying nodes to form the set; 2.2 Recover the failed link of each virtual network (1) Take out the unavailable virtual link; (2) Check whether there is a cut set containing this virtual link, and a virtual link contained in the cut set has been mapped; (3) If there is such a cut set, then the underlying link of the virtual link that has been mapped in the cut set will be marked. If such a cut set does not exist and needs to be restored, go to step (5) (4) Determine whether the current link is mapped to the underlying link. If not, the virtual eged has routing survivability and does not need to be restored. If there is, it indicates that the virtual link does not have routing survivability and needs to be restored (5) The algorithm with the shortest hop distance is used to find the optimal underlying link resource for the virtual link. If there is a faulty link in the shortest path, the available link recovery resources are used to restore until the resources are used up; (6) Determine whether there is still a virtual link that needs to be restored. If there is, return to step (2)

16

D. Luo et al.

Step 1 is to restore the failed underlying node according to the characteristics of the underlying node, including the following four sub processes. S , calculate the three indicators of the allocated resource, node (1) For node in Nfault centrality, and node concentration, and solve the importance of the failure node. S , calculate the node recovery value of each failed underlying node. (2) For node in Nfault (3) According to the node recovery value, sort the failed nodes in the failed underlying S S to obtain the sorted set Nfault−order . nodes set Nfault S (4) Recovery of failed underlying nodes: sequentially recover from the set of Nfault−order until the available node resources N are used up.

Step 2 is to restore the failed link. (1) Dig out the faulty virtual network, it does not have the faulty underlying node to v = {G1v , G2v , ..., Gxv }. The virtual link with failed underlying node form a set GNode does not need to be restored, because this service is unavailable when the virtual node fails. (2) Based on routing survivability, the faulty link in each virtual network is restored.

5 Performance In the process of analyzing the algorithm VNSFRAoRS in this paper, the GT-ITM [10] tool is used to generate the network environment. Because network failure recovery requires a bearer network and service network. Therefore, the network environment generated by the GT-ITM [10] tool has two types: bearer network and service network. Because the size of the business volume is related to the size of the bearer network and the business network. Therefore, the network environment of different scales is simulated in the network environment. Because the number of network failures and the resources required to recover from failures are also related to the scale of the network, when verifying the performance of the algorithm, the algorithm was also analyzed from different scales of network environments. Because this article solves the problem of failure recovery, there needs to be a network element failure. When simulating failures, select links to simulate failures from all underlying links with a probability of [0.02, 0.05]. The contrast algorithm is the heuristic algorithm VNSFRAoHRM. The algorithm uses heuristic recovery strategy to maximize the number of businesses for maximizing the number of recovery services. In order to analyze the recovery ability of the faulty network, the experiment analyzed the number of network nodes and the fault recovery rate. It can be seen from Fig. 1 that the larger the number of network nodes, the weaker the resilience of the algorithm. The algorithm VNSFRAoRS optimizes the utilization value of recovery resources by considering the recovery value of the underlying nodes and the routing survivability, thereby increasing the recovery rate of network failures (see Fig. 1).

Virtual Network Service Failure Recovery Algorithm

17

Fig. 1. The relationship between the number of network nodes and the failure recovery rate

In order to analyze the recovery capability of a faulty network, the experiment analyzes the relationship between the number of recovery resources and the failure recovery rate. It can be seen from Fig. 2 that the larger the number of recovery resources, the stronger the recovery capability of the algorithm. The algorithm VNSFRAoRS recovers more important network resources, thereby enhancing the use value of the restoration resources.

Fig. 2. The relationship between the number of recovery resources and the failure recovery rate

6 Conclusion IPv6 overcomes the defect of insufficient IPv4 network addresses, provides the possibility for the access of massive network resources, and improves the service quality, and the creation speed of services. However, when recovery resources are limited, existing studies still need to optimization the recovery ability of failure. This article designs a virtual equipment service failure recovery mechanism about routing survivability. It can comprehensively evaluate the network characteristics of the recovery node and the

18

D. Luo et al.

recovery link, so as to optimize ability of the faulty. However, virtual link recovery is only based on the routing survivability, and the relationship between the underlying links corresponding to the virtual link is not analyzed. In the next step, the relationship between the underlying links corresponding to the virtual links that need to be restored will be further explored, and the underlying links that need to be restored will be further increased. Acknowledgment. This work is supported by State Grid Science and Technology Supported Project (5700-201999496A-0-0-00).

References 1. Peng, M., Li, Y., Jiang, J., et al.: Heterogeneous cloud radio access networks: a new perspective for enhancing spectral and energy efficiencies. IEEE Wirel. Commun. 21(6), 126–135 (2014) 2. Tang, J., Tay, W.P., Quek, T.Q.S.: Cross-layer resource allocation in cloud radio access network. In: 2014 IEEE Global Conference on Signal and Information Processing (GlobalSIP), pp. 158– 162. IEEE (2014) 3. Hawilo, H., Shami, A., Mirahmadi, M., et al.: NFV: state of the art, challenges, and implementation in next generation mobile networks (vEPC). IEEE Netw. 28(6), 18–26 (2014) 4. Wang, X.: Network recovery and augmentation under geographically correlated region failures. In: 2011 IEEE Global Telecommunications Conference-GLOBECOM 2011, pp. 1–5. IEEE (2011) 5. Mijumbi, R., Serrat, J., Gorricho, J.L., et al.: Design and evaluation of algorithms for mapping and scheduling of virtual network functions. In: Proceedings of the 2015 1st IEEE Conference on Network Softwarization (NetSoft), pp. 1–9. IEEE (2015) 6. Yu, H., Yang, C.: Partial network recovery to maximize traffic demand. IEEE Commun. Lett. 15(12), 1388–1390 (2011) 7. Bao, N., Yuan, Y., Liu, Z., et al.: Network service recovery scheme of optical data center based on link lifetime. J. Commun. 39(8), 125–132 (2018) 8. Pan, Z., Liu, Q., Wang, X.: Software defined network customer information link fault recovery simulation. Comput. Simul. 35(5), 241–244 (2018) 9. Zegura, E.W., Calvert, K.L., Bhattacharjee, S.: How to model an internetwork. In: Proceedings of IEEE INFOCOM’96. Conference on Computer Communications, vol. 2, pp. 594–602. IEEE (1996)

Efficient Physical-Layer Authentication with a Lightweight C&S Model Kefeng Pan1 and Xiaoying Qiu2(B) 1 National Engineering and Research Center of Software Engineering Peking University,

Beijing 100871, China 2 School of Information Management, Beijing Information Science and Technology University,

Beijing 100192, China [email protected]

Abstract. The authentication process of Internet of things (IoT) devices in wireless communication scenarios is complicated, and the existing algorithms have the problem of relying on manual selection for device fingerprint extraction. In this paper, we analyze the superiority of wireless channel information as device characteristic fingerprint, and propose a lightweight authentication model. We have also established the connection between the distribution function of the channel estimation matrix and the convolutional mapping, and then proposed a lightweight and fast authentication framework based on Neural Network learning. Based on various measurement data collected by the National Institute of Standard and Technology (NIST), the performance of the proposed C&S algorithm is analyzed by studying the detection accuracy and miss detection rate. Keywords: Impersonation attacks · Physical layer authentication · Hypothesis testing · Spoofing detection

1 Introduction Traditional authentication protocols rely on cryptography-based encryption mechanisms, allowing requesters to prove their identity to the verifier [1, 2]. Although cryptography-based methods are well integrated and implemented in existing applications, there may be certain limitations when using the above methods in new scenarios (such as smart cities, smart transportation, smart buildings, etc.) [3, 4]. The 5th generation wireless networks (5G) and above will connect to billions of devices [5–8]. The fundamental reason for the security defects of wireless systems is the open broadcasting of wireless signal propagation, intermittent communication of devices, heterogeneous network architecture, and a large number of miniaturized sensor devices. In IoT applications, resource-constrained devices need to be deployed. In this case, due to the limited computing and memory of the embedded device, the use of a passwordbased classic authentication protocol may cause trouble. Although encryption technologies have been extensively studied in [9–11], their computational overhead are very undesirable. Existing authentication frameworks need to generate, refresh and deliver © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 S. Sun et al. (Eds.): ICSINC 2021, LNEE 895, pp. 19–24, 2022. https://doi.org/10.1007/978-981-19-4775-9_3

20

K. Pan and X. Qiu

keys multiple times. In addition, the key transmission process may lead to potential key leakage [12–14]. The keyless authentication technology based on channel information can provide lightweight security by utilizing the properties of channel reciprocity [15–17]. Its main advantage are low complexity and no assumptions about the attacker’s computing power. In general, identity authentication is one of the most important tasks in secure communication. At the physical layer, authentication is to distinguish the source of a message based on the unique characteristics of the channel [18]. This paper proposes a new lightweight authentication scheme based on C&S analysis. By identifying the channel estimation sequence of each transmitter (i.e., IoT device), it provides fingerprint information that is difficult to imitate, thereby providing seamless protection for wireless communication. We explored channel reciprocity to obtain fingerprints. This paper is based on the multi-dimensional physical layer feature acquisition technology of C&S analysis, which includes three stages: channel estimation, feature extraction and classification authentication. The organization of this paper is as follows: Sect. 2 presents the system model. The C&S model-based authentication scheme is proposed in Sect. 3. Following up from that, the security performances are discussed in Sect. 4. Finally, Sect. 5 concludes the paper.

2 System Model and Problem Statement Consider a typical wireless communication scenario model, as shown in Fig. 1. Various IoT sensor devices communicate with base stations that may be attacked by spoofing. The base station acts as a legal receiver, receiving the signal sent by the device node and using the authentication approach to distinguish the source of the received message to defend against active attacks such as impersonation attacks and witch attacks.

Fig. 1. A typical system model for spoofing attacks.

Due to channel estimation errors, perfect channel reciprocity is usually impractical. The channel estimates between IoT devices can be expressed as  T H l [t1 ] = H l1 [t1 ], H l2 [t1 ], · · · H lN [t1 ] (1) 







where N represents the number of channel characteristic, L indicates the number of IoT nodes, T is the transpose of a vector, t1 denotes the instantaneous moment of sampling.

Efficient Physical-Layer Authentication

21

In order to study the impact of wireless channel-based security authentication technology on the detection rate, This subsection builds a binary hypothesis test model to heal with the verification problem. The mathematical expression is as follows: H0 : HT (t + 1) = HAl (t + 1)

(2)

H1 : HT (t + 1) = HE (t + 1)

(3)

From the point of view of distinguishing signal source, for the transmitter L of legitimate signal, we assume that the estimated value of channel information is H Al , and that of the illegal attacker is H E . Under this assumption, the authentication process is modeled as the threshold test method of hypothesis testing. H0 : |HAl (t) − HT (t + 1)| ≤ τ

(4)

H1 : |HAl (t) − HT (t + 1)| > τ

(5)

where H 0 indicates that there is no malicious attack, H 1 denotes that there is malicious attack, and Tau represents the authentication threshold. When the difference between two successive frames is less than Tau, it indicates that the channel information estimation matrix between the two adjacent frames is very similar, so it can be judged that the current signal comes from the legitimate transmitting equipment L.

3 Authentication Strategy Based on C&S Algorithm This paper proposes a lightweight authentication algorithm based on convolutional neural network-based SVM Algorithm (C&S), which mainly contains a neural network module for feature learning and high-dimensional mapping relationship construction, and then support vector machine is used to optimize the optimal nonlinear boundary. We first collect the estimated value of channel information in each frame and perform normalization processing. Since the normalized wireless channel sequence can be regarded as a binary sequence, the characteristic fingerprint is an important feature for accurately identifying malicious node. Therefore, the goal of the CNN module is to obtain highdimensional feature fingerprints from channel estimates through input data, and to take the feature mapping relationship into account for identity verification. 3.1 Model Training Stage The C&S algorithm model is trained by the feature finger-print extracted from the channel information estimate, that is, the high-dimensional feature training set extracted from the convolutional neural network is used to model and classify the SVM, and the parameters of the whole verification model are jointly optimized. Through iterative solution, the relevant parameters of the optimal hyperplane are obtained, so that the classification interval of the support vector machine is maximized, and the highest classification accuracy of the C&S algorithm is obtained. According to the obtained optimal partitioning hyperplane, the received unknown channel information estimated value is verified, thereby distinguishing the identities of different IoT devices, and realizing security authentication based on channel information.

22

K. Pan and X. Qiu

3.2 Model Detection Stage We divide the training set into two subsets: 80% for training and 20% for testing. In a binary classification setting, the classifier in the training set. Then the safety performance of the trained classifier is evaluated on the test set. While training the C&S algorithm, we performed 10 cross-validation and grid search to optimize the hyperparameters. In each case, choose the best performing model to evaluate the test set. In the identity verification scheme based on the C&S algorithm, the specific time sequence change process between consecutive signals is not considered, and the focus is to use the similarity between two consecutive frames to verify the identity of the IoT device. Analyzed from the perspective of machine learning, the system includes the data acquisition process of collecting the signal preamble, the feature data extraction process of the channel information estimation value, the data labeling process of adding sample labels to the channel timing feature, and it also includes the SVM algorithm. During the identity verification process, based on the C&S algorithm, two types of information are fed back for identity verification, one is the feature label of the unknown signal, which characterizes the legality of the unknown signal, and the other is the time sliding window input into the model at the next moment. Here, the proposed algorithm is called C&S algorithm.

4 Prototype and Performance Evaluation As shown in Fig. 2, the detection results based on SVM algorithm are shown. It can be seen from the results that with the increase of SNR, the detection accuracy of SVM continues to improve. When SNR > 40 dB, the detection result of the authentication system reaches 100%, but when the SNR is less than 10 dB, the accuracy of the authentication system is only about 70%. In addition, the results of detection under different training times are also compared. Blue is the test result of 10,000 times of training. Green is the test result training times, the higher the accuracy of model detection. For the model detection results with low SNR, the accuracy of model detection can reach more than 98% after more than 15000 rounds of training.

Fig. 2. Identification accuracy based on SVM algorithm.

Efficient Physical-Layer Authentication

23

Figure 3 shows the simulation results based on C&S algorithm. In Fig. 3, the xcoordinate is the SNR and the y coordinate is the accuracy of detection. The detection accuracy of the C&S scheme increases with the increase of SNR. However, for the SNR is less than 30 dB, the accuracy is only about 70%. Similarly, the detection performance of the C&S model-based method under different training times is also studied and analyzed in Fig. 3.

Fig. 3. Accuracy of identification and detection based on C&S algorithm.

5 Conclusion This paper studies device authentication in wireless industrial Internet of Things networks, and proposes a lightweight model based on C&S algorithm. The feature extraction mechanism based on neural network algorithm is studied to map a limited number of channel estimation data to a high dimensional space. The C&S algorithm combines the high-dimensional extraction algorithm and the optimal nonlinear boundary optimization module. Finally, the validity of the lightweight authentication scheme based on C&S algorithm is verified by using NIST data sets sampled in an industrial environment.

References 1. Xie, N., Chen, C., Ming, Z.: Security model of authentication at the physical layer and performance analysis over fading channels. IEEE Trans. Dependable Secure Comput. 18(1), 253–268 (2021) 2. Xiao, L., Lu, X., Xu, T., Zhuang, W., Dai, H.: Reinforcement learning-based physical-layer authentication for controller area networks. IEEE Trans. Inf. Forensics Secur. 16, 2535–2547 (2021) 3. Qadri, Y.A., Nauman, A., Zikria, Y.B., Vasilakos, A.V., Kim, S.W.: The future of healthcare Internet of Things: a survey of emerging technologies. IEEE Commun. Surv. Tutor. 22(2), 1121–1167 (2020). https://doi.org/10.1109/COMST.2020.2973314

24

K. Pan and X. Qiu

4. Wang, R., Liu, H., Wang, H., Yang, Q., Wu, D.: Distributed security architecture based on blockchain for connected health: architecture, challenges, and approaches. IEEE Wirel. Commun. 26(6), 30–36 (2019). https://doi.org/10.1109/MWC.001.1900108 5. Xu, T.: Waveform-defined security: a framework for secure communications. In: IEEE/IET 12th International Symposium on Communication Systems, Networks and Digital Signal Processing (CSNDSP2020), Porto, Portugal, July 2020 6. Wang, N., Wang, P., AlipourFanid, A., Jiao, L., Zeng, K.: Physical-layer security of 5G wireless networks for IoT: challenges and opportunities. IEEE Internet Things J. 6(5), 8169– 8181 (2019) 7. Gao, D., Sun, Q., Hu, B., Zhang, S.: A framework for agricultural pest and disease monitoring based on Internet-of-Things and unmanned aerial vehicles. Sensors 20, 1487 (2020). https:// doi.org/10.3390/s20051487 8. Qiu, X., et al.: Wireless user authentication based on KLT and Gaussian mixture model. In: 2019 IEEE Wireless Communications and Networking Conference (WCNC), Marrakesh, Morocco, pp. 1–5 (2019) 9. Weinand, A., Karrenbauer, M., Lianghai, J., Schotten, H.D.: Physical layer authentication for mission critical machine type communication using Gaussian mixture model based clustering. In: Proceedings of IEEE 85th Vehicular Technology Conference (VTC Spring), pp. 1–5, June 2017 10. Buczak, A.L., Guven, E.: A survey of data mining and machine learning methods for cyber security intrusion detection. IEEE Commun. Surv. Tutor. 18(2), 1153–1176 (2016) 11. Farnaaz, N., Jabbar, M.A.: Random forest modeling for network intrusion detection system. Procedia Comput. Sci. 89, 213–217 (2016) 12. Van Huynh, N., Nguyen, D.N., Hoang, D.T., Dutkiewicz, E.: “Jam Me If You Can”: defeating jammer with deep dueling neural network architecture and ambient backscattering augmented communications. IEEE J. Sel. Areas Commun. 37(11), 2603–2620 (2019) 13. Choudhary, S., Kesswani, N.: A survey: intrusion detection techniques for Internet of Things. Int. J. Inf. Secur. Priv. 13, 86–105 (2019) 14. Hamamreh, J.M., Furqan, H.M., Arslan, H.: Classifications and applications of physical layer security techniques for confidentiality: a comprehensive survey. IEEE Commun. Surv. Tutor. 21(2), 1773–1828 (2019) 15. Pei, C., Zhang, N., Shen, X.S., Mark, J.W.: Channel-based physical layer authentication. In: 2014 IEEE Global Communications Conference, Austin, TX, pp. 4114–4119 (2014) 16. Henningsen, S., Dietzel, S., Scheuermann, B.: Misbehavior detection in industrial wireless networks: challenges and directions. Mobile Netw. Appl. 23(5), 1330–1336 (2018). https:// doi.org/10.1007/s11036-018-1040-0 17. Mahmood, A., Aman, W., Iqbal, M.O., Rahman, M.M.U., Abbasi, Q.: Channel impulse response-based distributed physical layer authentication. In: Proceedings of IEEE 85th Vehicular Technology Conference (VTC Spring), pp. 1–5, June 2017 18. Qiu, X., Dai, J., Hayes, M.: A learning approach for physical layer authentication using adaptive neural network. IEEE Access 8, 26139–26149 (2020)

Recent Advances of Rock Engineering and Communication Technologies Junjie Wang1 , Michel Kadoch2 , and Zhaohua Li1,3(B) 1 College of Civil Engineering, Yango University, Fuzhou 350015, China

[email protected]

2 Synchromedia Laboratory, École de Technologie Supérieure, Université du Québec,

Montreal, QC, Canada 3 Yunnan Innovation Institute, BUAA, Kunming 650233, China

Abstract. This paper takes a review of the advances of rock engineering and communication technologies, and the 2017 ISRM international symposium with the theme ‘Rock Mechanics for Africa’, held in Cape Town, South Africa in October 2017, is also retrospected. The advances of rock engineering and communication technologies applied in the domain in Africa were focused, and 50 articles, of the 98 papers embodies in total, were from Africa. As the mining is a primary impetus of the economies in African countries, many articles deal with rock engineering research and case studies for mines. Some advanced communication techniques and devices are applied in mining engineering and expected results are obtained. This paper reviews the main contributions of this symposium, and concluded the recent applications of the communication technologies in rock engineering. Keywords: ISRM · Communication · Data acquisition · Monitoring · Micro-seismic activity

1 ISRM International Symposium AfriRock 2017 The ISRM International Symposium on Rock Mechanics, hosted by the South African National Groups, was held in Cape Town, South Africa, in October 2017. 245 experts from 25 countries attended the symposium. 98 papers were embodies, and the numbers per theme and continent were shown in Table 1. In accordance with the previous ISRM board meeting and council meeting, the winner of the Rocha Medal in 2018, Dr. Michael du Plessis, and the candidate of the Franklin Lecture in 2018, Prof. Hide Yasuhara, were selected, respectively; the International Society of Rock Mechanics was renamed as the International Society of Rock Mechanics and Rock Engineering, and the acronym remained as ISRM; Prof. Re¸sat Ulusay was elected as the next ISRM President; and Chinese and Japanese groups were rewarded as the ‘Best National Group’. The theme of this symposium was “Rock Mechanics for Africa”, and the main topics included support techniques, micro-seismic activities, surface mining slope stability, underground mining and tunneling projects, numerical modelling, experimental techniques, petroleum, water conservancy and other fields of theoretical and engineering issues. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 S. Sun et al. (Eds.): ICSINC 2021, LNEE 895, pp. 25–30, 2022. https://doi.org/10.1007/978-981-19-4775-9_4

26

J. Wang et al. Table 1. Paper numbers per theme and continent Theme Keynote lectures Support techniques Micro-seismic activities Surface mining slopes stability Underground mining Numerical modeling Investigation technology Geotechnical investigations Tunnels and others

Paper number 4 9 11 8 19 11 11 15 10

Continent Africa Asia Europe North America South America Oceanica

Paper number 50 18 16 8 2 4

1.1 Micro-seismic Activities This sub-symposium received 11 articles about the micro-seismic activity. In these studies, mining and tunnelling-induced micro-seismicity were deeply investigated in experimental, numerical and micro-seismicity monitoring approaches. Scholars from South Africa, China, Austria, Australia, Singapore, Japan and Canada attended the sub-symposium. To study the relationship between the heterogeneity of rock and rockburst propensity at grain scale, several artificial sample sets consisting of a very fine-grained fibreless Ultra-High Performance Concrete (UHPC) [1], and a constant volumetric fraction of different coarse rock grains as aggregate were produced and tested, by research group of Klammer [2]. The findings of the study show that the material’s heterogeneity at grain scale has a major impact on the failure behaviour. With increasing stiffness heterogeneity between the rock matrix and the admixed grains, the failure mode becomes more ductile, and a lower uniaxial compressive strength is reached. Therefore, the material is less prone to rockburst, as less elastic strain energy is developed prior to failure. The results agree with various rockburst parameters, which were additionally evaluated and compared. Deep excavations in rock masses may break the frictional equilibrium of nearby faults, and result in the seismicity. Based on the unloading-induced direct shear model [3], Wu [4] conducted experimental and numerical studies on a simulated granular fault, to investigate the mechanism of excavation-induced seismicity. A series of laboratory experiments was used to investigate the effects of initial shear stress, initial normal stress, and its unloading rate on the frictional instability of the fault. And a numerical simulation was carried out to interpret stress variation and particle evolution during the unloading process. According to the results, both normal and shear stresses decrease sharply, when the fault is approaching a critical stress state, which is due to a decrease in inter-particle force and particle contact breakage. The evolution of the state of the fault depends on the initial stress condition and excavation process. The initial shear stress and the rate of normal stress unloading are the most significant factors for the instability of fault, whereas the magnitude of the initial normal stress is not a critical factor in this respect. A greater initial normal stress and a lower initial shear stress provide a favourable environment for accumulation of higher strain energy in adjacent rock, leading to larger slip displacement. A larger normal stress unloading rate can also

Recent Advances of Rock Engineering and Communication Technologies

27

cause higher strain energy and a larger slip displacement. Therefore, in tunnel excavations near discontinuities, the control of excavation rates is an important measure for reducing the occurrence of excavation-induced micro-seisms. In addition, Durrheim [5] summarised the findings of Japanese-South African collaborative project, titled “Observational studies in South African mines to mitigate seismic risks”, which lasted 5 years. In this project, acoustic emissions, surface displacement monitoring, and ground stress monitoring were used to perform an in-depth study on mining-induced seismicity and rockbursts [6]. Lynch [7] analysed seismic source mechanisms and the position of micro-seisms in mines using reflected waves recorded in seismograms. Butler [8] reviewed the latest developments of three-dimensional arrayed micro-seismic monitoring systems installed in boreholes during the initial stages of mine construction; this system has been successfully applied in the Nickel Rim Deep (NRD) project. 1.2 Surface Mining Slope Stability In total, 8 research articles were received in this sub-conference. The contents covered slope design principles, case study of pit slope design, identification of potential failure mechanisms, measurement techniques of 3D displacement vector and so on. Scholars from South Africa, Zambia, China, and Italy participated in the discussion and communication. Open-pit mining slopes are often threatened by a host of geological hazards, including landslides and rockfalls. Rock masses are usually segmented by discontinuities such as joints and faults, and experience the weathering of varying degrees. The determination of orientation and parameters of rock mass is crucial, yet challenging issues for studies on the Stability of rock slopes. A total of 8 articles about the stability of open-pit mining slopes were collected for this symposium, which discuss in detail, laser probing techniques and rockfall analysis and prediction, the design of open-pit mining slopes, artificial neural networks and analyses on the stability of open-pit mining slopes, and qualitative analyses on the stability of rock slopes controlled by primary shear zones. Tsao et al., 2017 [9] has introduced a method for investigating potential rockfalls, by means of airborne LiDAR and two different types of ground-based LiDAR. LiDAR can be used to construct high-resolution point cloud topographic maps, which can be used to determine the orientation (strike, dip and inclination) of geological discontinuities in rock masses and reveal potential rockfalls. This supplementary information is useful for the design and construction of slope protection systems. In a case example, the position and mode of occurrence of potential rockfalls were analysed, using stereographic projections based on the occurrence of discontinuities. The Rockfall in the Rocscience suite (a program of rockfall analysis) was then used to analyse rockfall paths, and the probability for the occurrence of rockfalls in each slope of a mountainous terrain. The protective and support measurements such as wire meshes and rock bolts were then proposed, according to the conditions of each site. Regarding the design of open-pit mining slopes, Pillay [10] used the Mining Rock Mass Classification (MRMR) System, proposed by Laubscher in 1990 [11], to convert the rock strength parameters, obtained from laboratory experiments, into the strength

28

J. Wang et al.

parameters of rock masses, segmented by discontinuities, and then used numerical analyses to provide suggestions for slope design, in the Otjikoto open pit gold mine in Namibia. Epiga [12] used decision-analysis method to design the shape of working slopes and toes of the Mamatwan open pit mine in South Africa. Regarding the stability of open-pit mining slopes, Fakir [13] used artificial neural networks to study 18 key parameters for the slope stability issue; although discontinuity parameters were found to be the most important factors for slope instability, the importance of other parameters cannot be neglected (see Fig. 1). The stabilities of 15 real open-pit mining slopes were then analyzed, by calculating the Open-pit Mining Slope Stability Indexes (OMSSI), and the correct predictions in 13 of these 15 slopes were obtained. Although this novel empirical method cannot replace the conventional methods, it has nonetheless provided a new approach to analyze the stability of open-pit mining slopes.

Fig. 1. Parameter dominance within the rock engineering system [30]

In addition, Romer [14] used the conventional limit equilibrium method to perform a preliminary analysis on slope stabilities, and then incorporated numerical methods with higher levels of complexity to analyze a single or multiple instability-inducing factors, thus obtaining results with higher levels of credibility. Lefu [15] performed detailed preliminary geological surveys, rock mass classifications and discontinuity surveys, and then used numerical back analysis to investigate the instability of slopes, controlled by primary shear zones. Excellent results were obtained in the approach. 1.3 Data Acquisition for Numerical Modelling Numerical modelling is an important research tool for rock mechanics and rock engineering. The acquisition of reasonable parameters is of utmost importance for ensuring the effectiveness of numerical analyses. With the continued development of numerical simulation techniques, the accuracy and complexity requirements of numerical model parameters have also increased in tow. Remote sensing techniques such as radar scanning, digital photogrammetry and infrared thermography could be used to acquire various

Recent Advances of Rock Engineering and Communication Technologies

29

types of data from dangerous and inaccessible locations. Donati [16] noted that, while remote sensing techniques have become a mainstream method in rock engineering, the selection of parameter types and acquisition methods in a scientific manner remains necessary for reducing the model and parameter uncertainties in numerical analyses, and to increase the reliability and realism of numerical slope simulations. For example, when an existing landslide is being studied, it is extremely important to restore the terrain features that existed prior to the landslide. Airborne LiDAR, vertical aerial photography and relevant historical photographs are effective methods for acquiring the initial parameters for numerical model. The strength parameters of rocks and discontinuities may be obtained using conventional methods or laboratory experiments. For the rock masses with important discontinuity, the parameters of rock specimens must be reduced using various empirical methods, in order to convert rock specimen parameters into effective rock mass parameters; alternatively, rock mass parameters may be determined via numerical inversions. A simplified model may then be used in a preliminary analysis to validate these parameters. In the analysis of large-scale engineering problems, large-scale topographic maps and geo-morphological analyses that reflect on the geological and tectonic setting of the engineering project are necessary. A preliminary geo-morphological analysis can help to identify latent factors for instability and failure such as old landslide zones, underground mining, and erosion. A preliminary kinematic analysis (e.g., key block theory) could be used to identify the key blocks that determine the occurrence of instability.

2 Conclusions 8 themes were discussed in depth during this international symposium, including the support techniques, micro-seismic activities, surface mining slope stability, underground mining, numerical modeling, investigation technology, geotechnical investigations, tunnels, and the communication technologies applied in rock engineering were also highlighted.

References 1. Yu, R., Spiesz, P., Brouwers, H.J.H.: Development of an eco-friendly Ultra-HighPerformance-Concrete (UHPC) with efficient cement and mineral admixtures uses. Cem. Concr. Compos. 55(1), 383–394 (2015) 2. Klammer, A., Peintner, C., Lagger, M., et al.: Investigations of rockburst propensity of artificial samples containing different aggregates. In: Conference Proceedings of the ISRM International Symposium AfriRock 2017, vol. 1, pp. 189–201 (2017) 3. Wu, W., Zhao, Y., Li, X., Zhao, J.: An unload-induced direct-shear model for granular particle friction in rock discontinuities. Rev. Sci. Instrum. 85(9), 093902–093906 (2014) 4. Wu, W., Zhao, Z., Duan, K.: Unloading-induced instability of a simulated granular fault and implications for excavation-induced seismicity. Tunn. Undergr. Space Technol. 63, 154–161 (2017) 5. Durrheim, R.J., Ogasawara, H., Nakatani, M., et al.: Observational studies in South Africa mines to mitigate seismic risks: challenges and achievements. In: Conference Proceedings of the ISRM International Symposium AfriRock 2017, vol. 1, pp. 175–189 (2017)

30

J. Wang et al.

6. Moriya, H., Naoi, M., Nakatani, M., Aswegen, G.V., et al.: Delineation of large localized damage structures forming ahead of an active mining front by using advanced acoustic emission mapping techniques. Int. J. Rock Mech. Min. Sci. 79, 157–165 (2015) 7. Lynch, R.A., Jackson, S., Brijraj, S., Moolla, S.: Location and source mechanism of a microseismic event in a mine using a single seismogram. In: Conference Proceedings of the ISRM International Symposium AfriRock 2017, vol. 1, pp. 249–257 (2017) 8. Bulter, T., Simser, B.: Early access microseismic monitoring using sensors installed in long boreholes. In: Conference Proceedings of the ISRM International Symposium AfriRock 2017, vol. 1, pp. 225–237 (2017) 9. Tsao, M.-C., Wang, T.-T.: Identification of potential rock wedge controlled rockfalls using groundbased LiDAR Techniques: a case study in Eastern Taiwan. In: Conference Proceedings of the ISRM International Symposium AfriRock 2017, vol. 1, pp. 287–297 (2017) 10. Pillay, O., Armstrong, R., Terbrugge, P.J.: Otjikoto gold mine - a case study in pit slope design. In: Conference Proceedings of the ISRM International Symposium AfriRock 2017, vol. 1, pp. 317–331 (2017) 11. Lauchester, D.H.: A geomechanics classification system for the rating of rock mass in mine design. J. S. Afr. Inst. Min. Metall. 90(10), 257–273 (1990) 12. Epiga, S.N., Rupprecht, S.M.: Slope design considerations for shallow open pit mines: a case study at Mamatwan mine, Northern Cape Province, South Africa. In: Conference Proceedings of the ISRM International Symposium AfriRock 2017, vol. 1, pp. 277–287 (2017) 13. Fakir, M., Ferrentinou, M.: A holistic open pit mine slope stability index using artificial neural networks. In: Conference Proceedings of the ISRM International Symposium AfriRock 2017, vol. 1, pp. 331–351 (2017) 14. Romer, C., Ferrentinou, M.: The significance of identifying potential failure mechanisms from conceptual to design level for open pit rock slopes. In: Conference Proceedings of the ISRM International Symposium AfriRock 2017, vol. 1, pp. 351–365 (2017) 15. Lefu, N., Hingston, E.D.C., Lephatsoe, N.: Investigation of failures associated with a major shear zone in the Main Pit Cut 3 West, at Letseng Diamond Mine, Lesotho. In: Conference Proceedings of the ISRM International Symposium AfriRock 2017, vol. 1, pp. 365–377 (2017) 16. Donati, D., Stead, D., Ghirotti, M., et al.: A structural investigation of the Hope Slide, British Columbia, using terrestrial photogrammetry and rock mass characterization. Rendiconti online della Societa Geologica Italiana 24, 107–109 (2013)

Joint TDOA and FDOA Estimation Based on Keystone Transform and Chirp-Z Transform Xiaoyang Chen1,2 , Lede Qiu1,2(B) , Shuai Li1,2 , Ming Li1,2 , and Yihao Song1,2 1 Institute of Telecommunication and Navigation Satellites, China Academy of Space

Technology, Beijing 100094, China [email protected] 2 Innovation Center of Satellite Communication System, CNSA, Beijing 100094, China

Abstract. The traditional time difference of arrival (TDOA) and frequency difference of arrival (FDOA) estimation method suffers from the huge computational complexity and the estimate stepwise caused by the discrete parameter space. Moreover, the estimation performance deteriorates severely when the target moves rapidly. Therefore, a high accuracy method for joint TDOA and FDOA estimation based on the keystone transform and chirp-z transform is proposed. Firstly, the received signal data is divided into two-dimensional array of Fast Time (FT) and Slow Time (ST). And then the keystone transform is used to compensate the time delay migration due to the movement of target. Based on the compensated data, the coarse estimation of two parameters can be obtained by the utilization of the twodimensional fast Fourier transform. In the following, the two-dimensional chirp-z transform is applied to reduce the search step of the region around the coarse estimation, so that the fine estimation of two parameters can be obtained. Finally, the quadratic function fitting is used to get the accurate estimation. Numerical simulations demonstrate the superiority of this method. Keywords: TDOA · FDOA · Keystone transform · Chirp-z transform

1 Introduction In the field of location of non-cooperative emitters, TDOA and FDOA are two important parameters. Accurate parameter estimation can significantly improve the localization accuracy [1]. For joint estimation of TDOA and FDOA, the traditional cross ambiguity function (CAF) is a maximum likelihood method [2]. However, CAF method needs many computing resources, so it is impossible for real-time computation. In order to abate computational complexity, a new coherent processing approach was extended to the field of estimation of TDOA and FDOA [3–5], which has been widely used in the maneuvering target detection [6], and the synthetic aperture radar (SAR) [7, 8]. The method divides the received signal into two-dimensional array of FT and ST, the Fast Time denotes a small segment of signal and the Slow Time denotes the whole signal length. In each Fast Time segment, the signal is viewed to have a fixed TDOA and a fixed FDOA. Meanwhile, the TDOA and FDOA change along the Slow Time. The focus © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 S. Sun et al. (Eds.): ICSINC 2021, LNEE 895, pp. 31–36, 2022. https://doi.org/10.1007/978-981-19-4775-9_5

32

X. Chen et al.

of existing papers is to compensate the time delay migration (TM) and the Doppler shift migration (DM), such as the keystone transform (KT) [3, 6, 7], the second-order keystone transform (SKT) [4, 5], the Lv’s distribution (LVD) [5, 6, 8] and so on. However, the discrete parameter space of TDOA and FDOA leads the estimate stepwise, whereas the true value as continuous so that the quantitative error cannot be avoided. This paper proposes a high-accuracy joint TDOA and FDOA estimation method. The received signal is divided into two-dimensional array of FT and ST. The keystone transform is used to compensate the TM, and two-dimensional fast Fourier transform (2D-FFT) is utilized to get the coarse estimation. And based on the coarse estimation, the two-dimensional chirp-z transform (2D-CZT) is used to reduce the search step and get the fine estimation. Then, the quadratic function fitting is applied to get the accurate estimation.

2 Signal Model Assuming that the noises, the variation of the amplitude and phase are ignored, the two received signals s1 (t), s2 (t) are [3] s1 (t) = a(t) s2 (t) = a(t + D(t))ej2π fc D(t)

(1)

where fc denotes the carrier frequency. And D(t) denotes time delay. It should be noted that D(t) is time-varying due to the relative motion between receiving machine and emitter, which can be expanded as Taylor series D(t) = τ0 + vt + ϕ(t)

(2)

where ϕ(t) is the higher-order terms, v is the generalized velocity, τ0 is the time delay when t = 0. Then the signal can be divided into two-dimensional array of FT and ST [3–5]. The TDOA and FDOA are assumed to be constant in a Fast Time segment t, but change along the Slow Time axis tm . Let D(t) = D(tm ), the signal model can be expressed in two-dimensional form, which is denoted as s1 (t, tm ) = a(t) s2 (t, tm ) = a(t + τ0 + vtm )ej2π fc (τ0 +vtm )

(3)

3 The Proposed Method After the signal partition, there are three steps about the proposed method. In the first step, KT is used to correct the linear time delay migration. Then based on the compensated date, 2D-FFT is utilized to obtain the coarse estimation. The second step is to reduce the search step by 2D-CZT, so that the fine estimation can be obtained from the finer parameter space grids. The third step is to apply quadratic function fitting to get the accurate estimation.

Joint TDOA and FDOA Estimation

33

3.1 Coarse Estimation Define the mixing product as s(f , tm ) = FFTt∗ [s1 (t, tm )]  FFTt [s2 (t, tm )] = |A(f )|2 ej2π fc (τ0 +vtm ) ej2π f (τ0 +vtm ) (4) where A(f ) is the expression of a(t) in frequency domain, and f is the FT frequency. FFTt [·] represents the FFT operation in FT domain.  denotes the Hadamard product. ∗ is the complex conjugation. The expression of (4) in FT time domain is s(t, tm ) = raa (t + τ0 + vtm )ej2π fc (τ0 +vtm )

(5)

where raa (t) = E(a(x + t)a(x)) is the time domain expression of |A(f )|2 , and the autocorrelation function of a(t) in physical meaning. The focus position of raa (t) in FT time domain is the time-varying TDOA. And the phase term reflects a single-frequency signal along the ST axis, whose frequency is FDOA. Utilize the keystone transform to clear off the linking between f and tm . And the KT is defined as fc tn = (f + fc )tm

(6)

where tn is the time with respect to ST after the KT, so (4) can be changed as s(f , tn ) = |A(f )|2 ej2π(fc +f )τ0 ej2π fc vtn

(7)

Performing the generalized 2D-FFT on (7) along the FT axis and the ST axis, we have s(t, fn ) = raa (t + τ0 )δ(fn + fc v)

(8)

where fn is the frequency with respect to ST. It is obvious from (8) that τ0 and fc v can be obtained, which correspond to the initial TDOA and FDOA respectively. 3.2 Fine Estimation As shown in above, the calculation of TDOA and FDOA is carried out by FFT, so the “Picket Fence Effect” degrades the accuracy of the final result. Note that the resolution of FT time domain is τ = 1/fs , where fs is the sampling frequency. And the resolution of ST frequency domain is f = 1/T , where T is the total integral time. To improve the resolution, sampling rate and the length of signal should be increased, which will greatly increase the cost. So the 2D-CZT is applied to magnify the region of interest around the peak, which comes from the result of the coarse estimation. And the fine estimation can be gotten from the new peak position. The 2D-CZT of (7) is defined as t 2 /2

S(t, fn ) = czt[s(f , tn )] = W1

f 2 /2

W2n

fs /2 T   f =−fs /2 tn =0

−(t−f )2 /2

g(f , tn )W1

−(fn −tn )2 /2

W2

(9)

34

X. Chen et al.

where S(t, fn ) is the result of 2D-CZT of s(f , tn ) and the expression of s(t, fn ) with smaller search step. And t, f are time and frequency with respect to FT. tn , fn are −f

f 2 /2

t 2 /2

n time and frequency with respect to ST. g(f , tn ) = s(f , tn )A1 A−t W2n and 2 W1 the terms A1 , A2 ,W1 ,W2 are complex exponential scalars, whose expressions are W1 = exp[−j2π(T2 − T1 )/(PT  )], A1 = exp[j2π T1 /T  ], W2 = exp[−j2π(F2 − F1 )/(Qfs,m )], A2 = exp[j2π F1 /fs,m ], where T  denotes the length of a Fast Time segment. T2 ,T1 are two endpoints of zoom interval of TDOA. fs,m denotes the sampling frequency in Slow Time axis. F2 , F1 are two endpoints of zoom interval of FDOA. P, Q is the interval length with respect to TDOA and FDOA.

3.3 Quadratic Function Fitting The quadratic function is used for interpolation to get the accurate estimation. Based on the fine estimation, two slices can be gotten which along the two axes of the array respectively. Select the largest value and the adjacent two points in a slice [(xk−1 , yk−1 ), (xk , yk ), (xk+1 , yk+1 )], so that the real peak can be accurately located which is halfway between (xk , yk ) and the other point. The expression of the quadratic function is defined as a1 x2 + a2 x + a3 = y

(10)

where a1 , a2 , a3 are the coefficient of the quadratic function. So real peak position is xmax = −a2 /(2a1 ), which is the accurate estimation. Table 1. Computational complexity comparison The method

The computational complexity

The proposed method

O(11MN log2 N + 25MN )

KTM

O(5MN log2 N + 9MN )

CAF

o(2LNt Nf )

4 Computational Complexity Analysis This part discusses the computational load and compares method in this paper with the CAF [2] and the method based on keystone transform (KTM) [3]. Denote the length of FT and ST as N , M respectively. Denote the search length of TDOA and FDOA as Nt , Nf and the length of signal as L in CAF. The computational complexity comparison shows in the Table 1.

Joint TDOA and FDOA Estimation

35

Fig. 1. The simulation result: (a) panorama of the coarse estimation; (b) the detail of the spectral peak of the coarse estimation; (c) the result of 2D-CZT, the display region is the same as the (b)

Fig. 2. (a) RMSEs of TDOA versus SNR; (b) RMSEs of FDOA versus SNR

5 Numerical Simulations This part uses simulations to demonstrate the enforcement of method in this paper and compare it with conventional method. The signal modulation mode is the binary phase shift keying (BPSK). The bandwidth is Bs = 600 kHz and the carrier frequency is fc = 300 MHz.The sampling frequency is fs = 2 MHz, the total time is T = 1 s. And the length of ST is M = 1000, the length of FT is N = 2000. The initial true value is TDOA = 1.19799945 × 10−4 s and FDOA = 453.506919 Hz. The results of the coarse estimation and the fine estimation with a relatively high SNR are shown in Fig. 1. It is found that the coarse estimation can form a peak at (TDOA = 1.2 × 10−4 s, FDOA = 454 Hz), whereas the low resolution limits the accuracy. And the fine estimation can overcome the disadvantage, the detail of the peak can be completely observed by the 2D-CZT, which magnifies the resolution of the region around the peak by 100 times. And the new peak position after the 2D-CZT is (TDOA = 1.1980 × 10−4 s, FDOA = 453.51 Hz), so the quantization error is decreased. The accuracy of the CAF [2], the KTM [3] and the method in the paper are presented in Fig. 2. And 100 Monte Carlo trials are calculated for each SNR values. Note that the SNR r here is the equivalent input SNR of two received signals. And it can be expressed as r = 2r1 r2 /(r1 + r2 + 1), where ri denotes the input SNR of si (t), i = 1, 2. And the RMSE is the evaluation norm of the estimation accuracy. As shown in the Fig. 2, due to the existence of the time delay migration, the error of the CAF is the highest. Compared with CAF, KTM has better estimation performance due to compensation time delay migration. Based on the compensated data, the other two methods are inferior to the

36

X. Chen et al.

proposed method, this is because that the search step of the proposed method is smaller and the quadratic function fitting is applied. So the proposed method can efficaciously enhance the accuracy of TDOA and FDOA estimation.

6 Conclusion Based on KT and 2D-CZT, a high-accuracy estimation method for TDOA and FDOA has been investigated. The proposed method firstly divides the received signal into twodimensional array of FT and ST. And the KT is used to compensate the linear time delay migration so that the coarse estimation of two parameters can be obtained by 2DFFT. Sequentially, the 2D-CZT is used to reduce the search step based on the coarse estimation, and the fine estimation can be acquired. Then the accurate estimation can be found by the quadratic function fitting. The results of numerical experiments prove the superiority of the proposed method, and the accuracy can approach the Cramer-Rao lower bounds (CRLB).

References 1. Li, J., Guo, F., Yang, L., Jiang, W., Pang, H.: On the use of calibration sensors in source localization using TDOA and FDOA measurements. Digit. Signal Process. Rev. J. 27, 33–43 (2014) 2. Stein, S.: Algorithms for ambiguity function processing. IEEE Trans. Acoust. Speech Signal Process. 29, 588–599 (1981) 3. Xiao, X., Guo, F., Feng, D.: Low-complexity methods for joint delay and Doppler estimation of unknown wideband signals. IET Radar Sonar Navig. 12, 398–406 (2018) 4. Xiao, X., Guo, F.: Joint estimation of time delay, Doppler velocity and Doppler rate of unknown wideband signals. Circ. Syst. Signal Process. 38(1), 85–104 (2018). https://doi.org/10.1007/ s00034-018-0816-6 5. Liu, Z., Hu, D., Zhao, Y., Zhao, Y.: Computationally efficient TDOA, FDOA and differential Doppler rate estimation algorithm for passive emitter localization. Digit. Signal Process. Rev. J. 96, 102598 (2020) 6. Li, X., Cui, G., Yi, W., Kong, L.: Manoeuvring target detection based on keystone transform and Lv’s distribution. IET Radar Sonar Navig. 10, 1234–1242 (2016) 7. Huang, P., Liao, G., Yang, Z., Xia, X.G., Ma, J., Zhang, X.: An approach for refocusing of ground moving target without target motion parameter estimation. IEEE Trans. Geosci. Remote Sens. 55, 336–350 (2017) 8. Li, X., Kong, L., Cui, G., Yi, W., Yang, Y.: ISAR imaging of maneuvering target with complex motions based on ACCF-LVD. Digit. Signal Process. Rev. J. 46, 191–200 (2015)

Industrial Wisdom Based on 5G Customized Network Dan Liu(B) , Qiuhong Zheng, Peng Ding, and Yun Shen Department of Service and Application Innovation, China Telecom Corporation Research Institute, Beijing, China {liudan11,zhengqh,dingpeng6,shenyun6}@chinatelecom.cn

Abstract. With the rapid development of precision electronics manufacturing, electronic products are increasingly integrated. PCB (printed circuit board) integrated circuits, especially ultra-large-scale integrated circuits in the information industry have become more and more important. And in the field of precision instruments, the requirements of PCB products are even higher. The process of the PCB product is complex and prone to stub defects, slope defects, false solder, less tin and other surface defects. In practical industrial scenarios, deep-learning based AI industrial vision can solve the above defect detection problems in PCB production process, but AI industrial vision technology exist some problems such as high real-time requirement, business fragmentation, edge data heterogeneity, privacy security and data silos in the process of factory production line application. Based on 5G customized network, Sedna platform and industrial quality inspection AI model technology, this paper sinks AI capabilities to industrial production lines through MEC, builds industrial-grade machine vision solutions, and creates a single-industry multi-plant intelligent industrial life-form ecology. Keywords: 5G customized network · Sedna · Cloud edge collaboration · Industrial internet

1 Introduction Nowadays, 5G+industrial Internet industry is accelerating. The 5G+industrial Internet industrial ecology is growing. Basic telecommunication enterprises and industrial enterprises have accelerated their docking, with more than 1,500 5G+industrial Internet projects under construction. Network service providers and equipment vendors are actively participating in the implementation of 5G+industrial Internet applications and exploring new business cooperation models. In various market segments of the Industrial Internet, AI visual quality inspection has attracted more market attention due to its relatively emerging gap in the market. According to IDC data, the market size of China’s industrial AI quality inspection software and services (excluding hardware equipment) in 2019 has reached 110 million U.S. dollars, and the market is still in the fragmentation stage. In the future, the AI industrial quality inspection market will become more mature and will continue to grow. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 S. Sun et al. (Eds.): ICSINC 2021, LNEE 895, pp. 37–45, 2022. https://doi.org/10.1007/978-981-19-4775-9_6

38

D. Liu et al.

Deepen the expansion to more industries and scenarios. According to the forecast data of the Foresight Industry Research Institute, the global testing industry market will reach 220.3 billion euros in 2021, and it will increase year by year in the future. In various industries, due to the advantages of traditional machine vision applications such as good foundation, large customer volume, and relatively high input and output, PCB AOI/SPI inspection is one of the scenarios that has attracted the most manufacturers to participate. With the acceleration of 5G construction, the number of 5G+industrial Internet projects and application scenarios are expanding. The field of industrial product quality inspection has a huge market scale at home and abroad, and the AI industrial quality inspection market is maturing. However, the stock of industrial plants is huge, while the digital transformation of industrial enterprises is still in the early stage. In the actual industrial scenario, AI industrial vision based on deep learning can solve the problem of PCB production line quality inspection, but the following bottlenecks exist in the production applications of factories: (1) Real-time: The performance of production-line-level reasoning needs to reach millisecond-level processing speed, with stringent ultra-low real-time requirements, real-time processing and interconnection, and shipment after inspection; (2) Business fragmentation: At present, most cloud network solutions on the market are complex and decentralized service methods. In the face of multi-operator network services and multi-cloud computing service providers, it is inconvenient to form a complete and fast deployment plan; (3) Heterogeneous edge data: Industrial vision based on deep learning has been widely used in quality inspection scenarios for industrial production lines, but there are problems such as heterogeneous edge data distribution, few edge data samples, and cold start of models, resulting in poor application model effects; (4) Privacy and security, data islands: In a typical industrial vision system solution, there is a single plant data sample is insufficient, industrial production line data due to privacy and security is not factory, resulting in poor generalization ability of the application. And it cannot use the data of each plant to optimize the model upgrade. In order to realize the deep integration of AI applications and industrial scenes and solve the pain points of industrial vision applications in the field of industrial quality inspection, this paper proposes a solution of 5G custom network and edge cloud collaborative AI platform Sedna for PCB product quality inspection scenarios to create a smart industrial life form with cross-production line and cross-plant collaboration.

2 The Design Concept of the Industrial Wisdom Based on the existing industrial vision system, as shown in Fig. 1, by introducing China Telecom 5G network and edge-cloud collaborative AI platform Sedna, a cloud-edge-end collaborative intelligent industrial vision quality inspection system is built to realize the information linkage of multiple production lines and plants combined with Riseconda 5G intelligent gateway and Deloitte factory AOI/SPI inspection equipment based on the end + 5G network + edge + cloud collaboration [1]. Then a smart industrial life form is created and an industry ecology and promote resource integration is formed.

Industrial Wisdom Based on 5G Customized Network

39

Fig. 1. 5G + Sedna intelligent industrial vision solution framework

2.1 China Telecom’s 5G Customized Network Provides Cloud and Network Support for Smart Industrial Life Forms Figure 2 shows the three modes and deployment methods of China Telecom’s 5G customized network [2, 3]. The 5G customized network has the advantages of flexible customized services, reliable secure access, and comprehensive performance guarantees. A large network covers a single plant area and multiple plant areas, and realizes the linkage of information between a single plant area and multiple plant areas.

Fig. 2. China Telecom 5G customized network

• Based on China Telecom’s large network, customized end-to-end differentiated dedicated network channels for factory production lines, providing reliable and efficient network guarantees for production lines. • UPF equipment and MEC platform provide localized deployment of computing resources, complete delay-sensitive computing at the edge, improve enterprise data security and processing efficiency, and reduce end-to-end delay. 2.2 Sedna, an Edge-Cloud Collaborative AI Platform, Provides Platform Support for Smart Industrial Life Forms Sedna is the industry’s first edge-cloud collaborative AI framework open-sourced by Huawei Cloud, which realizes distributed collaboration of AI models based on KubeEdge, supports incremental learning, federated learning, collaborative reasoning, and lifelong learning capabilities of AI applications across edge clouds and edges. It improves model performance on the premise of ensuring low resource consumption and

40

D. Liu et al.

data privacy security on the edge side. And rapid deployment of AI applications and minimal operation and maintenance can be achieved. Based on Sedna, it can achieve: • Lifelong learning solves the problems of few edge data samples and cold start through continuous increment, and solves the problems of statistical distribution heterogeneity and catastrophic forgetting through multi-category models, making edge AI smarter and smarter. • Federated Learning uses multi-production line collaboration to solve the problem of fewer samples on a single production line, efficiently and rationally use various resources of the end, edge, and cloud to achieve an edge AI system with high performance, low cost, privacy and security. 2.3 Application Ability: PCB Board Solder Joint Quality Inspection PCB board solder joints have many types of defects and uneven data distribution, based on traditional inspection technology cannot meet the requirements of high quality. Relying on AI deep vision inspection technology, it can reduce the quality inspection leakage and false detection under complex texture and background interference. In the production environment, PCB production quality inspection is not isolated, but to link up with the upstream and downstream of the production line, to create a “industrial vision - industrial automation - industrial platform” PCB quality inspection solutions, from product loading, testing, sorting, discharging, to AI analysis and visualization of the results to achieve the automation of the whole process.

3 Deployment of the Industrial Wisdom Based on the characteristics of 5G network with large bandwidth and low latency, 5G wireless private network is used to replace the original wired network to provide wireless network solutions for machine vision, so that machine vision has a more flexible deployment mode. It can move according to the deployment site, and no need to adjust the network wiring. The vision training that consumes a lot of arithmetic power is put into the edge cloud processing, and the Riseconda 5G intelligent industrial gateway deployed on the machine side is used for data collection and real-time inference. With the lifelong learning and federal learning features of Huawei Sedna edge cloud collaboration platform, the self-renewal and self-optimization of the model is completed, solving the problem of insufficient data samples from a single plant area, realizing the comprehensive iterative evolution of AI capabilities, and promoting the continuous development of intelligent industrial life forms. 3.1 Lifelong Learning Realizes Closed-Loop Update of Application Model The current classical paradigm of edge-cloud collaborative machine learning is to run machine learning algorithms to build a model on a given dataset on the cloud, and the model is directly applied to multiple inference tasks on multiple edges. This learning paradigm is called closed learning [4] because it does not take into account knowledge

Industrial Wisdom Based on 5G Customized Network

41

learned from other scenarios and historical knowledge learned in the past, and thus still has many challenges in terms of cost, performance, and security: data silos, small samples, data heterogeneity, restricted resources, etc. Specifically, the edge of the cloud collaborative AI framework Sedna lifelong learning applied to the PCB industrial production line, each process part of the fault data there are small sample problems and data heterogeneity, for example, the link product quality visual inspection in the slope and warp problem data more, the production line false solder, less tin problem data less; different plants of the same production line due to differences in equipment data heterogeneity. For example, there are more data of slope and foot warping in the visual inspection of product quality, and less data of false solder and less tin in the production line; there are data heterogeneity problems in the same production line and different factories due to equipment differences. Facing the above problems, Sedna edge-cloud collaborative lifelong learning firstly completes continuous inference and training based on the cooperation of cloud-side computing power and edge-side data, and is able to become increasingly good at model training while inference is running. At the same time, the cloud-side knowledge base is used as the center to persist and maintain the cloud-side knowledge, realize the knowledge sharing across the cloud-side and handle the side-side tasks. The edge-side is able to discover and handle unknown tasks in the cloud-side knowledge base. Unknown tasks are new tasks discovered during operation or testing, such as tasks whose application scenarios or models are outside the current knowledge of the knowledge base. Based on Sedna’s lifelong learning capability, models can be adaptively optimized in the cloud or on the edge side, learning as they are used with increasingly accurate model results.

Fig. 3. Sedna lifelong learning

As shown in Fig. 3, Sedna’s lifelong learning task is divided into three stages [5]: training, evaluation and deployment, and maintaining a globally available knowledge base (KB) to serve each lifelong learning task. Sedna’s lifelong learning process is: (1) Start the training worker to perform multi-task migration learning based on the developer’s AI-based model and training data set to realize task knowledge induction, including: sample attributes, AI models, model hyper-parameters, etc. (2) After the training completes the update of the knowledge base, the evaluation worker for the evaluation data set is started, and based on the evaluation strategy defined

42

D. Liu et al.

by the deployer, it is judged that the task model meets the scheduling deployment. After Global Manager module captures the completion status of the evaluation task, it notifies Edge to initialize and start Inference Service for reasoning service. The application calls the model reasoning interface for reasoning, and judges the cloud on unknown tasks. (3) By docking with a third-party marking system and knowledge base-based migration learning, Local Controller monitors new data changes based on pre-configured rules and triggers training workers to perform incremental learning according to the configured strategy. After retraining is completed, it is re-issued to the edge side. In summary, based on the 5G + Sedna lifelong learning solution, the edge data or model parameters are transmitted back to the cloud side through the 5G network in a single plant, and the pain points such as small samples and edge data heterogeneity of industrial production line detection are solved through the lifelong learning technology, and the flexible deployment scheme and low latency transmission based on the 5G wireless network realize the seamless collaborative detection at the cloud side. 3.2 Federated Learning Breaks Multi-plant Data Silos For edge AI, data is naturally generated at the edge. In several different QA regions, it is usually difficult to train a good model on a single edge node because the raw image data is not willing to be shared to the cloud and the amount of data on a single production line is limited, but the factory needs them to work together to train a model with high detection accuracy to meet the production demand. Federated learning can break the data silos of multiple plants, make full use of the data resources of dispersed plants for model optimization, and improve the generalization ability of detection. As shown in Fig. 4, each plant separately downloads the latest detection model from the central cloud combined with the local data set for model retraining, and after the training parameters are transferred to the central cloud, the

Fig. 4. Sedna federated learning

Industrial Wisdom Based on 5G Customized Network

43

central cloud performs parameter aggregation to obtain a detection model with stronger generalization capability. The single plant nodes decide independently when to train. Through federated learning, a large amount of weight data is exchanged frequently in the process of training models, and the data transmission bandwidth between edge clouds is smaller and more unstable compared to the central cloud, so the efficiency of gradient transmission in the process of federated learning cloud edge collaboration is improved by 5G custom network. When deployed in practice, industrial cameras in production lines constantly push product inspection image data. Factories have high demand for raw data not to be out of edge and high privacy requirements. Federated learning allows for joint training and optimization of models across multiple production lines and plants, and Sedna can quickly deploy federated learning tasks by simply configuring the required training scripts, training data and aggregation algorithms.

4 Implementation Based on the above 5G custom network to build a smart industrial life body solution, a set of flexible manufacturing SMT production line based on 5G MEC cloud-side collaboration + industrial vision inspection has been built inside the factory; in factories in different regions, the China Telecom 5G custom network neighboring mode and Sedna federal learning features have been used to realize resource sharing in multiple factories. 4.1 Implementation of the Deployment of Internal Production Lines in the Factory This paper combines 5G MEC and AI industrial vision capabilities with the production line. As shown in Fig. 5, a set of flexible manufacturing SMT production lines based on 5G MEC cloud edge collaboration + industrial vision inspection are built in the factory [6]. The overall architecture is shown in the figure below:

Fig. 5. PCB production line quality inspection framework

Take the production process of SMT line in workshop 1 as an example, after the PCB board enters the line, it passes through the processes of solder paste printing, SMT placement soldering, wave reflow soldering, etc., and finally passes through the

44

D. Liu et al.

assembly, finished product testing and packaging trip finished products together with other parts. With Sedna Lifelong Learning, multiple communications are required for model iterations in the cloud and at the edge. In industrial edge computing scenarios, the communication between a large number of devices and the cloud has high bandwidth requirements, requiring a large number of nodes to be able to upload messages to the cloud simultaneously. The bandwidth limitation leads to a significant slowdown in training speed, making the convergence of the model slower. The 5G network is used to enhance the communication rate and bandwidth to solve the communication overhead problem in training and accelerate the convergence speed of the model. 4.2 Implementation of Multi-plant Deployment At present, there are many types of inspection types in the field of industrial quality inspection, and the number of samples of product defect types in a single plant is uneven. The adoption of multi-plant collaboration enables intelligent AI applications to cope with more types and complex inspection tasks, meet the special needs of more flexible manufacturing, and fully tap the data gold mine in the industrial field. Through multi-plant collaboration, it realizes comprehensive iterative evolution of AI capabilities, generates industry standard inspection models, creates intelligent industrial lifeforms, and promotes continuous evolution of industrial quality inspection applications. China Telecom’s self-researched lightweight 5G UPF supports sensitive local triage of production line data. Relying on Sedna, the edge cloud collaborative AI platform, data can be involved in model training without leaving the factory, safeguarding enterprise data security and realizing the ability of cross-cloud edge collaborative training for AI inspection.

Fig. 6. Implementation of the industrial wisdom

As shown in Fig. 6, using China Telecom’s 5G custom network than neighbor mode, shared MECs can be deployed in the same provincial plant to divert service data for local processing in the MEC, and the plant can also deploy independent MEC resource

Industrial Wisdom Based on 5G Customized Network

45

pools separately, and the SkyCloud MEPM node can perform overall scheduling and unified management of multiple geographically distributed edge MECs to break through the information and resource barriers between multiple plants and realize the resource scheduling, task allocation and collaborative management for multiple plants. Multiplant MECs collaborate to realize intelligent capability and application sharing of industrial life-form clusters, adaptive optimization training of capability and privacy protection of industrial data. Through the deployment of 5G custom network combined with the federal learning characteristics of Sedna platform to open up the information and resource barriers of multi-plant QA, solving the problem of data silos and promoting the integration of quality resources. The solution in this paper realizes unmanned, automated and intelligent defect detection for PCBs, effectively ensuring the quality of soldering and improving the production line’s accurate control of production quality. After verification, the quality of semifinished products after the furnace once through rate steadily increased, defective rate reduced by 1%, effective productivity increased by 15%, SMT personnel streamlined 68%, to achieve the whole process without personnel contact operations, the whole process information real-time traceability. In summary, the deployment of industrial AI visual quality inspection capability based on 5G custom network has the cornerstone to support the implementation in terms of network capability, computing capability, collaboration capability, security protection, etc. After application practice, the solution can effectively improve the efficiency of industrial quality inspection and realize the lean digitalization of the production and manufacturing process.

5 Conclusion In this paper, the ‘5G custom network + edge collaboration Sedna platform + AI quality inspection model’ solution allows factory quality inspection to enhance flexibility and efficiency through ‘end + edge + 5G network + cloud’ collaboration. It achieves data interoperability, improves the intelligence level and productivity of factory production, reduces maintenance and scale deployment costs, and has certain industrial and social benefits.

References 1. Baidu Cloud: 5G+AI Intelligent Industrial Vision Solution White Paper (2020) 2. China Telecom: China Telecom 5G Customized Network Product Manual (2020) 3. Jacobs, I.S., Bean, C.P.: Fine particles, thin films and exchange anisotropy. In: Rado, G.T., Suhl, H. (eds.) Magnetism, vol. III, pp. 271–350. Academic, New York (1963) 4. Liu, B.: Lifelong machine learning: a paradigm for continuous learning. Front. Comp. Sci. 11(3), 359–361 (2017) 5. Zheng, Z., Pu, J., Liu, L., et al.: Contextual anomaly detection in solder paste inspection with multi-task learning. ACM Trans. Intell. Syst. Technol. (TIST) 11(6), 1–17 (2020) 6. Yun, S., Peng, D., Yuying, X., Liangliang, L., Yong, Y.: Technical solution for flexible intelligent manufacturing based on 5G edge-cloud collaboration. Mobile Commun. 45(02), 18–23 (2021)

Implementation of DOA Estimation Algorithm Based on FPGA Hengyuan Zhou1 , Xiaojun Jing1(B) , Bingyang Li2 , Zesheng Zhou3 , and Bogan Li4 1 Beijing University of Posts and Telecommunications, Beijing, China

{zhouhengyuan,jxiaojun}@bupt.edu.cn

2 Jiangsu Automation Research Institute, Lianyungang, Jiangsu, China 3 Shandong Institute of Aerospace Electronics Technology, Yantai, Shandong, China 4 University of Southampton, Southampton, UK

[email protected]

Abstract. Direction of arrival (DOA) estimation is the prerequisite for beamforming, which largely determines the performance of smart antennas. However, DOA estimation algorithms usually need a large amount of computation while making enormous demands on real-time processing, which poses challenges to hardware implementation. Field Programmable Gate Array (FPGA) is widely used in the implementation of DOA estimation algorithms in recent years due to its advantages of high throughput rate, parallelizable computing, and design flexibility. But the traditional method of using hardware description language (HDL) to implement algorithms on FPGA has disadvantages such as high complexity and long timeto-market. Since the 21st century, high-level synthesis (HLS) tools have gradually developed, allowing designers to define their algorithms with a higher abstraction level so that effectively reduces workload and development time. Keywords: Direction of arrival · MUSIC · FPGA · HLS

1 Introduction The direction of arrival (DOA) estimation is of great significance and is used in many applications. DOA estimation can enhance the sensing ability of the communication system, such as sensing the direction of the vehicle relative to the roadside unit (RSU) in the vehicular communication [1] and sensing the location of the device in the communication system in which the unmanned aerial vehicle participates [2, 3]. But DOA estimation generally requires a large amount of computation. Researchers initially used the digital signal processor (DSP) chips to implement the algorithm. The estimated time is at the level of milliseconds [4, 5]. Recently, with the development of field programmable gate array (FPGA), the advantage of parallel computing has emerged. Special designs such as confidentiality can also be realized. More importantly, the delay jitter of algorithms on FPGA is weak, which is very suitable for implementing various communication algorithms. By designing flexible, fast, stable, and parallel processing algorithms, FPGAs have gradually replaced DSP chips in the field of DOA estimation, compressing DOA estimation time to the microsecond level [6]. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 S. Sun et al. (Eds.): ICSINC 2021, LNEE 895, pp. 46–52, 2022. https://doi.org/10.1007/978-981-19-4775-9_7

Implementation of DOA Estimation Algorithm Based on FPGA

47

The shortcomings of hardware description language (HDL) have gradually been exposed with the geometric growth of the circuit scale. First, these languages are verbose and error-prone, have rough syntax. What’s more, they may produce sub-optimal, faulty hardware, which is usually difficult to debug [7]. High-level synthesis (HLS) tools enable engineers to use software programs to specify hardware functions. Compared with HDL, the abstraction level raises from the register-transfer level (RTL) to the algorithm or behavior level. NEC’s research shows that it takes about 300,000 lines of HDL code to realize a million-gate FPGA design, and the use of modern HLS tools can easily increase the code density by 7 to 10 times, requiring only 30,000 to 40,000 lines of code [8]. HLS tools enable algorithm engineers to use high-level languages to develop FPGA algorithms without fully grasping relevant hardware knowledge. This paper verifies the feasibility of using C language to implement the MUSIC algorithm in Vivado HLS and explores the effect of parallel optimization instructions.

2 Design of HLS Project 2.1 Algorithm Implement MUSIC algorithm parameters implemented on FPGA: the number of antenna elements is 4, the number of signals is 2, the number of snapshots is 128, and the resolution is 1° so that there are 181 space spectrum points. The variables and main loops are shown in Table 1 and Table 2 respectively. The specific process is as follows (Fig. 1): Table 1. Variables used in the implementation process Name[Size]

Data

a_theta[4][181]

Direction vector

X[4][128]

Received signal

Rx[4][4]

The covariance matrix of X

RxU[4][4]

The feature vector of Rx

En[4][2]

Noise subspace

a[4][1]

Direction vector in a certain direction

aH_En[1][2]

Product of the conjugate transpose of the direction vector and the noise subspace

aH_En_EnH_a[1][1] Square of the aH_En’s norm P_theta[181]

Space spectrum

log_P_theta[181]

The logarithm of the space spectrum

48

H. Zhou et al. Table 2. The main loops in implementation Name

Function

calculate_P_theta

Calculate space spectrum

read_a_theta

Read direction vector in a certain direction

calculate_log_P_theta Calculate the logarithm of the space spectrum

multiply X and its conjugate transpose to get the covariance matrix Rx

get X’s feature vector Matrix RxU

assign_En: obtain the noise subspace En from RxU

read_a_theta: read a specific direction vector a from a_theta

calculate_P_theta: calculate the space spectrum

do matrix multiplication to obtain aH_En

do matrix multiplication to obtain aH_En_EnH_a calculate_log_P_theta: calculate logarithm of the space spectrum

search peaks to obtain DOA

Fig. 1. Algorithm implementation flowchart

1. Call the matrix multiplication function to multiply the received signal matrix (X) and its conjugate transpose to obtain the covariance matrix (Rx). 2. Perform eigenvalue decomposition on the covariance matrix (Rx). Get the eigenvector matrix (RxU) of the covariance matrix. 3. Take the last 2 columns of the eigenvector matrix to get the noise subspace matrix (En). 4. Calculate the space spectrum (calculate_P_theta). The inner loop obtains the direction vector array (a) through the loop named read_a_theta. Then the outer loop multiplies the conjugate transpose matrix of direction vector and En to calculate array aH_En. Multiply aH_En and its conjugate transpose to calculate aH_En_EnH_a, and store aH_En_EnH_a in the space spectrum array (P_theta). 5. Search peaks on the space spectrum array to obtain DOA. All matrix multiplications in the algorithm are implemented by the matrix multiplication function named matrix_multiply provided in Vivado HLS, which can easily realize various matrix transposition and parallel optimization. Eigenvalue decomposition is the most important and most complex part of the MUSIC algorithm. In this experiment, we call the svd_top function in the Vivado HLS algorithm library using the bilateral Jacobi

Implementation of DOA Estimation Algorithm Based on FPGA

49

method to realize the eigenvalue decomposition. The bilateral Jacobi method is more accurate than other singular value decomposition methods and is convenient for parallel calculation, so it is more suitable for implementation on FPGA. 2.2 Parallel Optimization The optimizations directives for loops in Vivado HLS mainly include pipelining, unrolling, merging, and dataflow. Pipelining adds registers before devices used in the loop so that a cycle can start after the previous one releases the necessary resources. Several adjacent cycles overlap in time, greatly improving resource utilization and reducing the delay. Unrolling allocates several times the hardware resources required for a single cycle to the entire loop so that multiple cycles can be performed on different hardware resources at the same time to reduce the time consumption. The loop to calculate space spectrum (calculate_P_theta) is unrolled by 2 times and pipelined. The inner loop reading direction vector (read_a_theta) is fully unrolled. The loop to calculate the logarithm of the space spectrum is pipelined. Partition is the most significant directive for arrays. Due to the access bottleneck of a single block of RAM, a single array is divided into several pieces and stored in different blocks of RAM or registers in a specific way, which can improve the throughput of the array. Partition methods include block partition, cyclic partition, and complete partition, etc. The direction vector array (a) used to calculate the space spectrum needs to be read and written frequently, therefore it is completely partitioned and stored in the registers to speed up the operation. The matrix multiplication operation of the direction vector (a) and the noise subspace matrix (En) can be decomposed into the multiplications of array a and each column of En. The operation can be accelerated by reading an entire column of the noise subspace matrix. Therefore, the noise subspace matrix (En) is divided so that the elements in the same row are stored in the same block of RAM.

3 Simulation and Analysis 3.1 Accuracy Compared with MATLAB The simulation uses a 4-element uniform linear array with a half wavelength element spacing. Its number of snapshots is 128. The SNR is 20 dB. Space spectrums generated by FPGA simulation and MATLAB simulation are shown in Fig. 2. They almost coincide. The difference mainly appears where the value is small. When FPGA performs floatingpoint numbers or some non-linear calculations, accuracy may be lost. But it does not affect the DOA estimate result.

50

H. Zhou et al.

Fig. 2. Space spectrums generated by FPGA simulation and MATLAB simulation

3.2 Estimation Speed The difference before and after optimization is mainly manifested in the number of resources and clock cycles required by each loop and function. Figure 3 shows the delays of the main loops before and after the optimization. The division of the noise subspace matrix and the unrolling in the matrix multiplication function cause Vivado HLS to automatically divide the noise subspace matrix completely and store it in the registers. Compared with the default solution, the single-cycle delay of the loop (calculate_P_theta) has been reduced from 41 clock cycles to 22 clock cycles. The initial interval (II) before and after optimization is 41 clock cycles and 4 clock cycles respectively, and the number of cycles drops from 181 to 90. Ideally, the initial interval should be 1 clock cycle when the inner loop is fully unrolled. But if it is fully unrolled, it will take up too many resources and not meet the resource constraints. Partial unrolling, retaining part of the serial operation in the matrix multiplication makes II reduce to 4, and the delay of the entire loop body is reduced from 7421 to 379. The loop to calculate the logarithm of the space spectrum (calculate_log_P_theta) involves non-linear floating-point operations such as division and logarithm. If optimization is not performed, the delay will be huge. After optimization, the iteration latency remains unchanged. The II is reduced from 28 to 1 clock cycle, and the number of cycles is reduced from 181 to 90. Therefore, the overall delay of the loop body is reduced from 5068 to 117, and the delay is reduced by more than 97%. As shown in Fig. 4, in comparison to the result before optimization, although the usage of program block memory, look-up tables, registers, and DSP blocks all increased, the overall latency dropped from 29023 to 16501, a drop of more than 40%.

Implementation of DOA Estimation Algorithm Based on FPGA

51

Fig. 3. Delay of the main loops before and after optimization

Fig. 4. Resources utilization and entire latency before and after optimization

4 Conclusion In the process of implementing the MUSIC algorithm, the advantages of HLS tools can be summarized in the following four points. First, it uses a high-level language instead of HDL, which has a higher level of abstraction and a smaller workload. Second, HLS tools are similar to an algorithm development environment, rather than a hardware development environment. Thirdly, the HLS tool uses independent optimization instructions to control FPGA synthesis, which allows developers to get different synthesis results without modifying the source code. Finally, there are abundant algorithm library resources in HLS tools. They can be easily called and integrated into various designs. In summary, using HLS tools to write FPGA prototypes of DOA algorithms such as MUSIC is very attractive. It can be verified relatively quickly and obtain considerable performance, which is an effective method.

References 1. Mu, J., Gong, Y., Zhang, F., Cui, Y., Zheng, F., Jing, X.: Integrated sensing and communicationenabled predictive beamforming with deep learning in vehicular networks. IEEE Commun. Lett. 25, 3301–3304 (2021) 2. Gao, N., Li, X., Jin, S., Matthaiou, M.: 3-D deployment of UAV swarm for massive MIMO communications. IEEE J. Sel. Areas Commun. 39(10), 3022–3034 (2021) 3. Gao, N., Jin, S., Li, X., Matthaiou, M.: Aerial RIS-assisted high altitude platform communications. IEEE Wirel. Commun. Lett. 10(10), 2096–2100 (2021) 4. Huang, Q.: The research of DOA estimation algorithm based on DSP. University of Electronic Science and Technology of China (2010)

52

H. Zhou et al.

5. Zhou, Z.Y.: The research and implementation of multi-source super-resolution DOA estimation algorithm based on DSP system. Huazhong University of Science and Technology (2012) 6. Liu, T.: Implementation of classical DOA algorithm based on FPGA. Harbin Institute of Technology (2016) 7. Zwagerman, M.D.: High-level synthesis, a use case comparison with hardware description language. Grand Valley State University (2015) 8. Wakabayashi, K.: C-based behavioral synthesis and verification. ASP-DAC (2004)

Research on Dynamic Spectrum Allocation of Space-Air-Ground Integration Lei Liu1(B) and Qiang Liu2 1 Beijing Research Institute, China Telecom Corporation Limited, Beijing, China

[email protected] 2 University of Quebec, Montreal, Canada

Abstract. The SAG integrated network can achieve global three-dimensional coverage and mobile ultra-wide-area broadband access capabilities. The integration of space-based satellite networks and ground-based cellular networks can provide a better user experience, which has become the core development direction of 6G networks. First of all, the organizational form and spectrum requirements of SAG integration have been sorted out and analyzed. Then, because of the problem of co-frequency interference caused by the sharing of frequency spectrum by each communication subsystem in the integration of SAG, the spectrum sharing of UAV formation and satellite system is considered, and energy-efficient resource allocation technology is studied. The relevant conclusions can provide a reference for improving the spectrum resource allocation of the SAG integrated network. Keywords: Space-Air-Ground integration · Spectrum allocation · 6G · UAV

1 Introduction With the advancement of information communication and aerospace technology, the global information development field has been fully expanded to human production, life, and scientific research, including land, sea, sky, and space. The Space-Air-Ground (SAG) integrated network is composed of a space-based network which composed of space satellite interconnected and a ground-based network [1]. The integrated development of space-based networks and ground-based networks has increased the wide-area coverage of the network. This has obvious advantages for realizing communication and information services in remote areas and has become an important development area for wide-area communication guarantees and information applications [2]. With the commercialization of 5G technology, mobile communications are expanding from satisfying the communication needs of humans to being able to provide wireless connections for the Internet of Vehicles (IoV), Internet of Things (IoT), and the Industrial Internet [3]. New service targets are often beyond the scope of conventional activities, and the current network is still dominated by the coverage of urban areas and some suburbs [4]. Remote areas still lack large broadband and high-quality communication services, which will lead the future development direction of 6G. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 S. Sun et al. (Eds.): ICSINC 2021, LNEE 895, pp. 53–65, 2022. https://doi.org/10.1007/978-981-19-4775-9_8

54

L. Liu and Q. Liu

The SAG integration is a possible form of 6G in the future. It can use the advantages of space-based and ground-based systems to comprehensively use satellites, unmanned aerial vehicles (UAV), and ground facilities to cover the global area in a three-dimensional and efficient manner to meet the ubiquitous needs of future communication networks [5]. Satellite coverage is large, but the communication rate is limited, and the communication delay is relatively large. The ground network can directly use the new 5G technology, but the deployment of network facilities in remote areas is limited. In recent years, UAV-assisted communication and High Altitude Platform Station (HAPS) have become important ways to compensate for insufficient ground network coverage and excessive satellite network communication delays [6, 7]. However, the integration of the original communication sub-networks in different spaces has the problem of how to schedule and use spectrum resources [8]. The SAG integration design should fully consider the advantages and disadvantages of different subsystems, and integrate all subsystems with open and flexible overall network architecture. The 6G technology based on SAG integration makes the unified design of the access network and the core network possible. In addition, to achieve SAG integrated networking, the first task is to solve the problem of spectrum usage. Under the condition of a limited frequency spectrum, sharing the frequency spectrum flexibly is the only way. For a simple network of integrated networks of individual drones, Hua M et al. consider multi-point collaboration models of drones and ground base stations, by optimizing transmit power, enhances the service performance of ground users, and reduces the same frequency interference to the satellite system [9]. Li X et al. considers the offshore scene, under the interference constraints of the drone, optimizes the drone flight trajectory and resource allocation strategy, and realize the accompanying coverage of the drone to the target vessel [10]. In order to avoid the same frequency interference of integrated network, KONG H et al. attempts to use free space optical communication to establish a satellite and drone aerial link, while the transport link of the drone and the ground user communicates with the ground user [11]. These studies have laid an important foundation for integrated network design. However, a single drone is often limited, so a plurality of drone forming proceeds are a feasible way to improve network mobile coverage. Zhang S et al. considers two programs for forwarding satellite data to realize rapid reconstruction of communication networks in the disaster [12]. Liu C et al. consider using multiple drones into aerial multi-cells, and adjusts the energy efficiency of the drone formation communication by optimizing the transmission power of the drone [13]. Due to the wider space distribution, spectrum sharing of drone formations in integrated networks will lead to more complex same-frequency interference, which is not an effective problem in existing research. In this paper, the organizational form of the SAG integrated communication network is discussed first. Then the HAPS spectrum requirements in the SAG integrated network are analyzed. The theoretical derivation and basis of the last SAG integrated spectrum dynamic allocation are finally given.

Research on Dynamic Spectrum Allocation

55

2 SAG Integrated Communication Network The SAG integrated network relies on the ground-based network, expands on the spacebased network, and adopts a unified structure, technical system, and standard specifications. The SAG integration is interconnected by a space-based network, ground-based Internet, and mobile communication network, as shown in Fig. 1.

Fig. 1. SAG integrated network structure

A space-based network is composed of space segment and ground segment. The space-based constellation network can include Geostationary Earth Orbit (GEO) satellites node, Medium Earth Orbit (MEO) satellites node, and Low Earth Orbit (LEO) satellites node. It can also contain only one type of satellite node in GEO, MEO, or LEO. The gateway (GW) node network of the ground segment is formed by interconnecting satellite ground GW stations related to the space-based constellation network. On the one hand, it interconnects and exchanges information with the space-based constellation network; on the other hand, it interconnects and exchanges information with the terrestrial Internet and mobile communication network. The ground cellular mobile network in the ground-based network is the land-based public mobile communication system with the widest coverage. A fixed base station is set up in each cell to provide users with access and information forwarding services. The ground base station is generally connected to the core network by wire. The core network is mainly responsible for user subscription management, Internet access, and other services, mobility management, and session management. There are many solutions for the integration of space-based networks and groundbased cellular mobile networks. And these different integration architectures will likely

56

L. Liu and Q. Liu

coexist for a long time in the evolution process, and will eventually achieve deep integration. The simplest way of integration is that the satellite network serves as a backhaul for ground base stations and core networks or as a backup for ground wired backhaul. In addition, satellites can access the 6G core network through non-3GPP access methods, and share the core network with the ground mobile network. The satellite can also be connected to the 6G core network as a special 6G base station through the 3GPP access method, which is a deep integration method of the satellite network and the ground network.

3 SAG Integration Spectrum Requirements HAPS is the connecting layer in the SAG integrated network, which is 20 km to 50 km above the ground. It provides broadband access services for ground-based equipment in the ground coverage area, provides connections for network access points to connect to the backbone network, and provides emergency communication services for temporary deployment. HAPS communication services include special application services and connectivity services. Special application services are mainly emergency communications. Connectivity services refer to the provision of broadband connections between nodes in areas where the communication infrastructure is assumed to be difficult, including connections between ground user equipment (GUE) and gateways and user equipment. Due to the obvious differences in the use scenarios of connectivity services defined by countries, the results of spectrum use demand analysis are also clearly different. The specific numerical ranges are shown in Table 1. Table 1. Frequency requirements of HAPS communication system (MHz) Service type

Forward link

Reverse link

GW → HAPS

HAPS → GUE

GUE → HAPS

HAPS → GW

Special application

110

15

15

110

Connectivity

247–2727

164–938

24–240

35–480

The number of ground users covered by each HAPS is. c = π × a2 × b,

(1)

where a is the radius of the coverage area in kilometers, and b is the number of users per square kilometer, that is, the user density. Suppose the penetration rate is d , then the number of connected users per HAPS is e = c × d.

(2)

The amount of forwarding link data is h = f × g,

(3)

Research on Dynamic Spectrum Allocation

57

where f represents the amount of data used by each user per month in GB, and g is the forward link ratio. When the data rate ratio i and the utilization rate j during busy and idle hours are known, the capacity demand per user (in kbps) can be expressed as k=

h × 8 × i × 100 × 106 . 30 × 24 × 3600 × j

(4)

Therefore, we can measure the forward link capacity of each HAPS platform as l = e × k,

(5)

and the spectrum demand ratio of each HAPS platform as n=

l×m , 100

(6)

where m is the forward/reverse link ratio. The spectrum efficiency of the forward and reverse links of the HAPS system is shown in Table 2. Table 2. Spectrum efficiency of forward and reverse links Link

Forward link

Reverse link

GW → HAPS HAPS → GUE GUE → HAPS HAPS → GW System spectrum efficiency 5.5 (bit/s/Hz)

2.4 (low)

2.4 (low)

3.3 (average)

3.3 (average)

5.5 (high)

5.5 (high)

5.5

The position between the ground GW and the HAPS is relatively fixed, and the communication link uses a higher-gain directional antenna, so the highest value of 5.5 bit/s/Hz is adopted. There are three recommended values (low, average, high) for the link between GUE and HAPS to estimate the system bandwidth requirements under different channel conditions. The frequency resources that can match the requirements of the satellite mobile communication capabilities of the SAG integrated network are mainly concentrated in the L-band. In the L-band, most of the frequency bands from 1518 to 1559 MHz (uplink) and 1610 to 1675 MHz (downlink) are allocated for mobile satellite services. In the frequency range of 1518 to 1675 MHz, frequency resources are also allocated for fixed, mobile, satellite meteorological services, and mobile satellite services must share frequency resources with these also majorly allocated radio services within the framework of the radio regulations. Table 3 analyzes the proportion of the main frequency interval groups in the current declared data and the declaration stage of the corresponding data. If notification data (N) has been declared for this frequency range, count the countries with the most frequency

58

L. Liu and Q. Liu Table 3. Proportion of L-band interval declaration and its data declaration stage

Frequency groups

GSO network Percentage

Country/region

Declaration stage

Percentage

NGSO network Country/region

Declaration stage

1518–1525

6%

United Kingdom

N (1 group)

14%

France

C

1525–1559

44%

United Kingdom

N (65 groups) 35%

France/Russia

N (1 group)

1610–1626.5

3%

China

N (8 groups)

United States /France

N (2 groups)

5%

1626.5–1660.5

35%

United Kingdom

N (35 groups) 26%

France

C

1668–1675

11%

United Kingdom /United Arab Emirates

N (1 group)

France

C

17%

groups. If the frequency range has only coordinated data (C) declaration, then count the countries with the earliest receipt time of the frequency group. From the point of view of the coordination status of each frequency group, the Geostationary Stationary Orbit (GSO) satellite network has N declarations in all frequency ranges, and the UK has the largest amount of data in all four frequency groups. For the Non-Geostationary Stationary Orbit (NGSO) satellite network, France occupies a dominant position in coordination. France has a C declaration with the highest coordination status among the frequency groups that have no N declaration. In terms of the proportion of frequency declarations, the most concentrated declaration frequency groups are 1525 to 1559 MHz and 1626.5 to 1660.5 MHz, which account for almost one-third of the number of declarations in satellite networks.

4 Dynamic Spectrum Allocation As a key technique for cognitive radio, dynamic spectrum allocation technology can greatly improve the utilization efficiency of spectrum resources, improve the unevenness of current spectrum resources in development and utilization, so the industry has caused extensive attention and in-depth study. At present, research based on nonsmart technology-based dynamic spectrum allocation algorithms can be divided into the following three directions: based on chart, game theory and transaction theory. The dynamic spectrum allocation algorithm based on the chart discussion is the vertex coloring problem in the map discussion, and each cognitive radio user and its available channel are used as the vertices in the figure. When the user cannot share the same channel, it is connected The spectral assignment process is abstracted to a coloring process of this vertex called the interference map. The vertex coloring of the interference map is a np-hard problem. It is difficult to obtain the best solution. PENG et al. proposed the heuristic algorithm for seeking secondary solution, the algorithm needs to set different application environments in advance to set up the different node settings Level, a high priority node priority to allocate spectrum, when the channel is more complex, and the convergence speed is slow [14]. Liao Chulin, etc., proposes a method of decomposing complex interference

Research on Dynamic Spectrum Allocation

59

maps as simple graphs, which will transform the sequential dyeing of nodes into simple charts in parallel, improve the time overhead brought by sequential dyeing [15]. Wang et al. Proposes a list of coloring algorithms, and after each round of random allocation channels, the channel is deleted in the list, enhances the convergence speed [16]. Liu Peng et al. proposed a dynamic spectrum allocation algorithm based on quantum genetic and picture coloring method, combining small habitatics and quantum genetic algorithm, the algorithm can be solved in partial optimal problems, by dynamically adjusting rotary door and improving chromosome threshold Increase the overall convergence speed [17]. He Jianqiang et al. proposed an improvement method based on color sensitive map, with maximizing bandwidth as a target function, and take the most fair guidelines in secondary distribution, superior to a single color sensitive map coloring algorithm and maximum fair guidelines algorithm [18]. The dynamic spectrum allocation algorithm for obtaining maximum spectrum utilization efficiency in the multi-cognitive radio user competition spectrum has achieved good results. Neel et al. I analyze the application prospects in cognitive radio systems in cognitive radios, and proposed dynamic spectrum allocation under complete potential models [19, 20]. The dynamic spectrum allocation will eventually converge to Nash Equares, and then analyze the use of repetition Game, Short Academic Game, S-Model Game, Submarine Cognitive Radio Model Convergence. Teng Zhijun et al. proposed a distributed algorithm based on the potential, and the convergence is verified by simulation [21]. Xu et al. proposed a game theory of improved pricing functions to dynamically spectral allocation models, and verified under static game and dynamic games [22]. In addition to the method based on chart and game theory, the dynamic spectrum allocation algorithms based on spectrum market theory and auction mechanism have also developed a lot of results. The dynamic spectrum allocation method based on the auction theory will vase the active cognitive user as auction bidder, regard the idle spectrum cognitive user as auction selection, the base station as auction and distribution process. Chen et al. Proposed a spectrum auction algorithm based on the simplified. Vickrey-Clark-Grov-ES (VGG) model, which proposes a new price method based on the first pricing closed auction according to the number of cumulative participation and successful access, which reduces the spectrum Communication interrupt during switching and improves the fairness of spectral assignment. Zhou et al. proposed a trusted dual spectrum auction model to solve the incomplessability problem in spectrum repetition and double auction. Wang et al. takes the maximization of spectral utilization as a target function, introducing approximate integrity concepts, taking into account spectrum utilization and integrity, and maximizes spectral auctioneer profit [16]. Based on the auction theory, although it can converge to maximize spectral utilization efficiency under defined primary user conditions, lack of flexibility. Although the above algorithm can solve the spectral utilization of dynamic spectrum allocation and the constraints and optimization problems between user communication efficiency and network communication performance, there is a problem with flexibility, slow convergence and unable to meet the requirements under distributed conditions. This centralized distribution method is relatively high for communication conditions between the control center and the user and the accuracy of the spectrum perception, and the difficulty is difficult in actual.

60

L. Liu and Q. Liu

With the rapid development of machine learning research in the study and other machine learning research in recent years, the intelligent dynamic spectrum allocation method based on the machine learning algorithm has gradually attracted more and more researchers.

5 Dynamic Spectrum Allocation Method Based on Multi-intelligent Body Strength Based on traditional algorithms such as chart coloring, game theory and trading theory requires the distribution of spectrum resources using a central control entity. The common problems of these methods are mainly to take up a large number of resources and the user’s communication between the spectrum allocation control centers, and these algorithms must be re-allocated when the environment changes, so the time overhead is relatively large, and the practical application is not reached. Real-time requirements for dynamic spectrum allocation. The utilization of multi-agent reinforcement learning methods can solve such problems, and the intelligent body can be distributed according to the training income based on the experience of the channel environment, and converge to optimal. When the external environment changes, each user (agent) can respond quickly according to the welltrained strategy, and quickly converge. This intelligent dynamic spectrum allocation method has a huge advantage over the real-time algorithm for the adaptability of the dynamic environment and the real-time performance of the spectrum. 5.1 Dynamic Spectrum Allocation Model Analysis Based on DEC-POMDP Research on dynamic spectrum allocation model is the basis for studying dynamic allocation algorithms. It is also an important aspect of cognitive radio theory research. Three layered access models, wherein the dedicated model is divided into spectral property model and dynamic proprietary model; the layered access model is divided into the spectrum underlay access model and the opportunistic spectrum access model. Guo Bingjie proposed to add a confirmation character status word based on the time slots of each user in the DEC-POMDP dynamic spectrum allocation model, and increase the observation channel status as 4: idle, busy, success, failure. In addition, in the observing information species, it is also added to the current observation time slot observation channel as a busy number of times and the total observation of the channel, and the number of channels of each channel is characterized by joining the statistic. However, the model does not consider the quality of service quality of the user service in the design of the reward function, only the basis of the success of the access channel as a reward, not only in the actual dynamic spectrum allocation, not only should the user can access the spectrum, It is also necessary to consider the effects of interference caused by access to the same channel (especially for the master user) to QoS, and weigh the trade-off under this constraint.

Research on Dynamic Spectrum Allocation

61

5.2 Dynamic Spectrum Allocation Method Based on DEC-POMDP Model At present, DEC-POMDP model and MARL dynamic spectrum allocation algorithm are divided into: a method based on independent Q-learning, based on cooperative Q-learning methods, based on joint Q-learning methods, and execution method of concentrated training distribution based on multi-agent actor-critic. Dynamic Spectrum Allocation Method Based on IQL. Based on independent Qlearning, each agent performs status value estimation and strategy based on the information of independent observations, and converges to stable points through a large number of training. TENG et al. proposed a dynamic spectrum allocation based on IQLbased bidding mechanism. Secondary users learn optimal bidding strategies through IQL algorithms [21], the primary user generates acceptable price vectors to ensure their own interests, the algorithm is effectively improved according to the secondary user policy. Bidding efficiency; WU, etc. According to the mutual interference between the user due to spectrum access behavior, WU, the user’s learning method K-Means and IQL algorithm are combined with the IQL algorithm, and the user is clustered. After reducing the number of intelligent, policies are performed with a variable learning rate IQL method. The dynamic spectrum allocation method based on the IQL algorithm ignores the nature of the non-Marcov chain having a variation of the external environment for a single user, and its state transfer model is not smooth, and the user cooperation is not considered in the optimization of the value function. The balanced strategy is constrained, so the number of users applies is small, the convergence speed is slow, and it is often not necessarily converged to the optimal strategy, and the secondary strategy is often obtained. Dynamic Spectrum Allocation Method Based on CQL. Based on cooperation Qlearning methods not only considering the current state itself, but also the factors of other user actions, but also considers the strategic trend of other intelligent body, making the Q function of separate users can converge faster. To the stable point (or Nash equilibrium point).The CQL algorithm needs to obtain all other intelligent operations and Q functions and the Q function of the independent Q function and the combined status information of the environment. Under distributed decision conditions, the overall status is actually not easy; this complete information interaction is actually Communication networks will cause a lot of communication overhead, which is difficult to implement. Dynamic Spectrum Allocation Method Based on JQL. JQL-based approach is a way to concentrate on training concentration. This method treats all users’ actions as unified actions in the global environment, so the partial observational Markov decisionmaking problem will be simplified. For the general Markov decision-making problem, you can directly apply a single intelligent body strength to learn. Wang et al. As a centralized training focused, and verified the convergence of the algorithm in the experimental environment, compared with the optimal short-term algorithm under the whittle index heuristic algorithm and channel positive correlation, indicating that DQN can converge the result of the optimum algorithm [16]. However, this JQL algorithm first needs to make centralized decisions. In each state, it must ensure that the center’s full control

62

L. Liu and Q. Liu

of the user, there is a shortcoming of communication overhead; secondly, the algorithm requires full perceive information about the environment, due to multipath, Shadow fading and path loss, this complete perception of the environment is difficult to do in practice; in addition to the number of users increase, its evaluation and decision-making action space dimension expires an exponential growth, and it is easy to cause value to function. It is difficult to train. Therefore, it is suitable for solving fewer problems with fewer users, and is not suitable for solving a dynamic spectrum allocation problem that users with a large number of ultra-intensive networks (UDNs).

6 SAG Integration Spectrum Allocation The SAG integrated network of UAV formations and satellites sharing spectrum is shown in Fig. 2. Among them, ground-based base stations cover areas within 100 km, and satellites cover all areas, but mainly remote users. UAV formations are deployed in an on-demand manner to fill the blind spots of broadband coverage in ground-based networks. UAV formations often move randomly in a large area, and satellite systems cover a large area. This makes it impossible for satellites or UAV formations to occupy a section of frequency spectrum effectively in the entire network. Therefore, we consider the scenario where UAV formations and satellites share a spectrum. Under the UAV load constraint, the joint allocation of UAV formation power and frequency domain sub-channels is studied, to satisfy the interference constraints of UAV formation on the satellite system and maximize the communication energy efficiency of the UAV formation.

Fig. 2. UAV formation and satellite sharing spectrum in SAG integrated network

Assume that K UAVs form a formation, and there are N frequency domain subchannels available for use. Without loss of generality, it is assumed that each sub-channel

Research on Dynamic Spectrum Allocation

63

is used by a satellite user and a UAV user, that is, there are N UAV users and N satellite users sharing the spectrum in the network. If the m-th UAV user uses the g-th sub-channel, the received signal can be expressed as follows ym,g = Hm,g xg + nm,g ,

(7)

where Hm,g represents the channel matrix and xg represents the transmission signal of the UAV formation on the g-th subchannel. nm,g represents additive white Gaussian noise, each element of which obeys a complex Gaussian distribution with a mean value of 0 and a variance of σ 2 , and the elements are independent of each other. Considering the classic UAV channel model, Hm,g can be expressed as Hm,g = Sm,g Lm,g ,

(8)

where the elements of Sm,g obey the independent and identically distributed standard complexGaussian distribution, which characterizes small-scale channel fading.  Lm,g = diag lm,g,1 , · · · , lm,g,K is a diagonal matrix, where lm,g,K is a large-scale channel parameter, which represents the path loss between the k-th UAV and the m-th UAV user when using the g-th sub-channel. The average R communication rate when the UAV user m uses the g-th channel in T time is.    1 H , (9) Rm,g = ESm,g log2 det INa + 2 Sm,g Lm,g Pg Lm,g Sm,g σ     where Pg = E xg xgH = diag pg,1 , · · · , pg,K is the diagonal matrix, which represents t transmission signal power matrix of the UAV formation. The total data transmission volume of all users in T time is N N αm,g TRm,g . (10) D(P, A) = m=1

g=1

where P = {P1 , · · · , PN } represents a collection of power matrices, and the element αm,g ∈ {0, 1} in A indicates that the m-th UAV user uses the g-th sub-channel for communication. For satellite users, suppose the g-th satellite user uses the g-th sub-channel. Then its channel can be expressed as hg = sg Lg ,

(11)

where the elements in sg obey the standard complex  Gaussian distribution and are inde 2

2

pendent of each other. Lg = diag l g,l , · · · , l g,K represents the large-scale channel information of satellite users. Based on this, the average interference caused by the UAV formation to the satellite user g can be expressed as   K 2 Ig = Esg sg Lg Pg Lg sgH = pg,k l g,k . (12) k=1

64

L. Liu and Q. Liu

7 Conclusion The SAG integrated network is an important public information infrastructure in the future. The concept of dynamic spectral allocation and the related algorithm are analyzed and introduced in the text. This article focuses on spectrum resources of the SAG integrated network in which mobile communication can be matched. The feasibility of the spectrum resource will be obtained according to the ratio of the declarations of different frequency groups of each service. It also believes that the dynamic allocation of spectrum resources in the SAG integrated network, where the UAV formation and satellite sharing spectrum. Related conclusions provide references to improve the feasibility of spectrum resource allocation in the SAG integrated network.

References 1. Qu, H., Xu, X., Zhao, J., Yue, P.: An SDN-based space-air-ground integrated network architecture and controller deployment strategy. In: 2020 IEEE 3rd International Conference on Computer and Communication Engineering Technology (CCET), pp. 138–142 (2020) 2. Liu, J., Shi, Y., Fadlullah, Z.M., Kato, N.: Space-air-ground integrated network: a survey. In: IEEE Communications Surveys & Tutorials, vol. 20, no. 4, pp. 2714–2741. Fourthquarter (2018) 3. Ye, J., Dang, S., Shihada, B., Alouini, M.-S.: Space-air-ground integrated networks: outage performance analysis. IEEE Trans. Wirel. Commun. 19(12), 7897–7912 (2020) 4. Liu, R., Ma, Y., Zhang, X., Gao, Y.: Deep learning-based spectrum sensing in space-air-ground integrated networks. J. Commun. Inf. Netw. 6(1), 82–90 (2021) 5. Yang, H., Xie, X.: An actor-critic deep reinforcement learning approach for transmission scheduling in cognitive internet of things systems. IEEE Syst. J. 14(1), 51–60 (2020) 6. Shi, Y., Xia, Y., Gao, Y.: Joint gateway selection and resource allocation for cross-tier communication in space-air-ground integrated IoT networks. IEEE Access 9, 4303–4314 (2021) 7. Yang, H., Alphones, A., Zhong, W., Chen, C., Xie, X.: Learning-based energy-efficient resource management by heterogeneous RF/VLC for ultra-reliable low-latency industrial IoT networks. IEEE Trans. Industr. Inf. 16(8), 5565–5576 (2020) 8. Li, P., Ren, K., Yang, B., Xia, L., Zhou, G., Liang, M.: Capacity optimization of the hybrid energy storage system for suppressing intermittent loads. In: 2020 12th IEEE PES Asia-Pacific Power and Energy Engineering Conference (APPEEC), pp. 1–5 (2020) 9. Hua, M., Wang, Y., Lin, M., et al.: Joint CoMP transmission for UAV-aided cognitive satellite terrestrial networks. IEEE Access 7, 14959–14968 (2019) 10. Li, X., Feng, W., Chen, Y., et al.: Maritime coverage enhancement using UAVs coordinated with hybrid satellite-terrestrial networks. IEEE Trans. Commun. 68, 2355–2369 (2020) 11. Kong, H., Lin, M., Zhu, W., et al.: Multiuser scheduling for asymmetric FSO/RF links in satellite-UAV-terrestrial networks. IEEE Wirel. Commun. Lett. 9, 1235–1239 (2020) 12. Zhang, S., Liu, J.: Analysis and optimization of multiple unmanned aerial vehicle-assisted communications in post-disaster areas. IEEE Trans. Veh. Technol. 67, 12049–12060 (2018) 13. Liu, C., Feng, W., Wang, J., et al.: Aerial small cells using coordinated multiple UAVs: an energy efficiency optimization perspective. IEEE Access 7, 122838–122848 (2019) 14. Peng, C., Zheng, H., Zhao, B.Y.: Utilization and fairness in spectrum assignment for opportunistic spectrum access. Mob. Netw. Appl. 11(4), 555–576 (2006) 15. Liao, C.L., Chen, J., Tang, Y.X., et al.: Parallel spectrum allocation algorithm in cognitive radio. J. Electron. Inf. Technol. 7, 1608–1611 (2007)

Research on Dynamic Spectrum Allocation

65

16. Wang, W., Liu, X.: List-coloring based channel allocation for open-spectrum wireless networks. In: Proceedings of the 2005 IEEE 62nd Vehicular Technology Conference, VTC-2005-Fall, pp. 690–694 (2005) 17. Liu, P., Zhang, G.Y., et al.: Dynamic spectrum allocation of cognitive wireless network based on graph theory. Telecommun. Eng. 60(6), 625–631 (2020) 18. He, J.Q., Teng, Z.J., Liu, J.: A spectrum allocation algorithm based on improved color sensitive graph theory coloring. Comput. Digit. Eng. 47(8), 1866–1868, 1889 (2019) 19. Neel, J.O., Reed, J.H., Gilles, R.P.: Convergence of cognitive radio networks. In: Proceedings of the IEEE Wireless Communications and Networking Conference, pp. 2250–2255 (2004) 20. Neel, J., Reed, J.H., Gilles, R.P.: The role of game theory in the analysis of software radio networks. In: Proceedings of the SDR Forum Technology Conference (2002) 21. Teng, Z.J., Han, X., Yang, X.: Spectrum allocation algorithm based on game theory in cognitive radio. Appl. Res. Comput. 28(7), 2661–2663 (2011) 22. Xu, C.B., Liu, X.L., Xian, Y.J.: Research on dynamic spectrum allocation technology based on game theory. Appl. Electron. Tech. 38(4), 102–105+109 (2012)

Research on Intelligent Access of Space-Air-Ground Integrated Network Lei Liu1(B) and Qiang Liu2 1 Beijing Research Institute, China Telecom Corporation Limited, Beijing, China

[email protected] 2 University of Quebec, Montreal, Canada

Abstract. With its wide coverage, high throughput, and high robustness, the Space-Air-Ground Integrated Network can be used in a variety of practical fields, including earth observation and mapping, intelligent transportation systems, military tasks, and so on, providing dependable and high-speed wireless access services for a large amount of data generated by new business and new applications. This paper begins by outlining the importance of research on space-earth integrated network access, then goes on to explain the space-earth integrated network and network access technology, and finally provides a network access scheme for the space-earth integrated network access technology that is based on reinforcement learning in order to optimize the network. Keywords: Space-air-ground · Integrated network · Intelligent access · Reinforcement learning

1 Introduction The Space-Air-Ground Integrated Network (SAGIN) has piqued the interest of academics and business in recent years as a novel network design that integrates satellite, air platform, and ground communication systems. SAGIN, on the other hand, provides significant benefits to various services and applications due to its heterogeneity, selforganization, and time-varying characteristics, but it also faces numerous challenges, such as routing, resource allocation and management, power control, end-to-end quality of service (QoS) requirements, and so on. In comparison to typical ground communication systems and satellite networks, SAGIN is constrained by imbalanced network resources in each network segment, making it difficult to achieve optimal network performance during data transfer. As a result, network optimization and system design are critical in the space-earth integrated network. The traditional ground wireless communication network has developed and evolved over the last few decades, from the first generation analog system to the second generation digital mobile communication, and finally to the third generation, fourth generation, or digital mobile communication. Because of the exponential development tendency in both the number of users and the type of service, the wireless communication network has become the fundamental network for human social information exchange. To fulfill the © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 S. Sun et al. (Eds.): ICSINC 2021, LNEE 895, pp. 66–79, 2022. https://doi.org/10.1007/978-981-19-4775-9_9

Research on Intelligent Access of SAGIN

67

growing demand for data, communication networks will need to supply more resources than present systems in the future [1]. The fifth-generation mobile communication system (5G) [2] has drawn attention from all walks of life in order to support new application technologies such as cloud computing, big data, and the Internet of things, as well as meet more wireless access service requirements, and has successively put forward and implemented relevant technologies and standards. Unlike 3G, which places the user at the center [3], and 4G, which places the service at the center [4], the 5G network, which places the user at the center [4], and is application-driven [5], will deploy open, flexible, safe, and reconfigurable network infrastructure, and provide high bandwidth and low delay data access for billions of types and multi-mode user devices, in order to meet the huge demand for data generated by new applications from the ever-growing new business. However, due to network capacity and coverage limitations, relying solely on ground communication systems is unable to provide high-speed and reliable wireless access services anywhere on the planet, particularly in remote and environmentally sensitive areas such as mountains, seas, and other environments. As a result, new network architecture must be designed and developed to suit the various application needs and service quality in future wireless communication. As a novel network design, the Space-Air-Ground Integrated Network (SAGIN) makes extensive use of contemporary information, communication, and computer network technologies. It achieves network resource convergence by the effective integration of air, space, and ground networks, based on the connectivity of satellite, air, and ground heterogeneous networks. Management and dissemination, as well as the gathering, transmission, and analysis of real-time data. Wide coverage, high data transfer rate, and high network dependability are all benefits of SAGIN. Earth observation and mapping, intelligent transportation systems, military missions, homeland security, disaster relief, and other disciplines can all benefit from it. High-throughput satellites can provide global wireless access services, air networks can meet high-quality service needs for the areas they serve, and densely distributed ground network equipment can enable high-speed data access. The combination of space, space, and ground networks can give incalculable benefits for the future 5g wireless communication system, as well as expand the number of applications and services available. In recent years, the United States, Japan, and other developed countries have launched their own space and space integration projects, such as the United States’ Global Information Grid (GIG), Space Communication and Navigation architecture [6], one web [7], space X [8], and so on, as well as the European Union’s Multinational Space-based Imaging System [9], Japan’s Basic Plan on Space Policy [10], and others. Different communication protocols are employed in the satellite, air, and ground network segments of the space earth integrated network to provide high-speed and reliable data transfer as part of a multi-dimensional network design. Unlike a typical single ground network or satellite communication, the data distribution, resource allocation, load balancing, power control, routing strategy, end-to-end quality of service (QoS) needs, and other network segments impact the space earth integrated network. In this approach, network designers must consider how to achieve the greatest network performance in end-to-end data transmission while working with restricted network resources. However, for SAGIN, which is a converged network architecture that includes various

68

L. Liu and Q. Liu

communication systems, it is difficult to use the limited network resources to achieve the best performance of information exchange, particularly cross layer data transmission between different network segments of satellite, air, and ground. As a result, SAGIN’s system design and performance optimization are critical for collaborative control and interconnection of satellite networks, air platforms, and ground communication systems, as well as the development of a real-time, dependable, stable, and efficient integrated information system.

2 Space-Air-Ground-Sea Intergrated Network The goal of the space-earth integrated network is to offer a mobile communication network that may be used by anybody, anywhere, at any time. The space earth integrated network’s handover algorithm seeks to enhance the network’s overall usage efficiency and minimize user access delays. Figure 1 depicts the space-based network’s design. GEO/LEO/MEO

Space network

HAPs

Near-space network

UAVs

Air network

BSs

Terrestrial network

Fig. 1. Space-Air-Ground-Sea integrated network architecture

2.1 Overview of the Research on Space Earth Integrated Network Researchers at home and abroad have worked hard on the satellite network, air network, and satellite ground integration network in the integrated system of space and earth, delving into topics such as switching and mobility management, small satellite systems,

Research on Intelligent Access of SAGIN

69

unmanned aerial vehicle (UAV) communication, and QoS provision and support. K. Chowdhury and colleagues conducted extensive study on the switching mechanism in LEO satellites as early as 2006 [11], comparing the link layer switching and network layer switching mechanisms according to various QoS standards. While Gupta et al. looked at UAV Communication Network research hotspots including high mobility, dynamic topology, link discontinuity, energy constraint, and link quality change [12]. Hayat et al. summarized the features and needs of UAV networks in projected civil applications from 2000 to 2015 [13] from the standpoint of communication and networking. In 2016, Nie phaus and colleagues looked at the state of research on satellite ground integrated networks, focusing on QoS provision and guarantee when satellite and ground links are together [14]. In comparison to a typical ground communication network, the space earth integrated network will be hampered by resource restrictions imposed by multiple network segments, such as restricted spectrum, limited bandwidth, and unstable wireless connections. In this approach, network operators must deliver the highest network performance in order to provide a communication environment with high bandwidth, high dependability, and high throughput. To achieve this aim, extensive study into the performance of the integrated network has been conducted in order to increase system bandwidth, dependability, and throughput. 2.2 Selection of Cross Layer Data Communication Gateway In the integrated network of space and earth, cross-layer data transfer from the ground to the satellite through the air platform is a technological difficulty. The most popular approach for connecting several network domains, similar to multi domain wireless network communication in mobile ad hoc networks (MANETs), is to pick a set number of nodes in each domain, termed gateways. Many strategies for gateway selection in MANETs have been proposed by a lot of research work over the last 10 years. The application situations for these approaches, as well as the challenges to be solved, are summarized in Table 1. Zhioua and colleagues for gateway selection from IOT cluster to LTE advanced infrastructure, proposed a cooperative data transmission method based on fuzzy logic [15], which takes into account received signal strength, load, candidate gateway, and duration of vehicle to vehicle link connection. A. In the integration system of the Internet of vehicles and mobile communication, Alawi et al. designed a simple gateway selection scheme based on three network performance indicators to select multi hop relay nodes, expand network coverage, and keep vehicles connected to the mobile communication network infrastructure continuously [16]. A distributed gateway selection technique is given in reference [17] for obtaining accurate and effective path performance results in hybrid wireless networks. In the Internet MANETs converged network, Zaman et al. designed a gateway selection method based on data priority to minimize the average endto-end delay, packet loss rate, and routing load overhead. Dhaou et al. [19] developed an evolutionary algorithm to tackle the gateway selection problem in MANETs and satellite hybrid networks, with the goal of reducing gateway load and quantity.

70

L. Liu and Q. Liu Table 1. Gateway selection methods

Algorithm

Network scene

Optimization objective

Limitations

Fuzzy logic

Internet of vehicles-LTE

Minimizing average delay; Reduce packet loss rate; Increase throughput

Clustering

Internet of vehicles-3G

Minimizing gateway selection delay Reduce control packet overhead

Only single layer network is considered, and the performance impact of the connected network is ignored

Cooperative detection Hybrid wireless network

Save bandwidth

Traffic priority

MANET-Internet

Reduce average end-to-end delay and packet loss rate

Genetic algorithm

MANET-satellite

Minimize the number of networks and maximize the gateway and link load

Distributed network segmentation

UAV network

Improve gateway stability

Game theory

Multi domain wireless network

Minimize link cost

No constraints were considered

Luo and colleagues to increase the network’s stability, a distributed method was proposed to choose the gateway in an air ad hoc network made up of UAVs. The method splits the whole UAV network into various sub regions based on the network’s application features [20]. Although several successful algorithms for solving the problem of gateway selection have been presented, they all share the same flaw: they only evaluate a single ground or air network and disregard other linked networks. Despite the fact that Zhong et al. devised an optimum gateway selection method based on Game Theory for multidomain wireless networks [21], they failed to account for cross domain communication restrictions. It’s worth noting that the network parameters of the integrated system’s air, sky, and ground segments (layers) can affect the performance of cross-layer data transmission, particularly data volume distribution, cross-layer link quality, and capacity, which are critical for ensuring QoS in cross-layer communication. As a result, optimizing cross-layer data transfer and obtaining the greatest QoS is critical.

Research on Intelligent Access of SAGIN

71

3 Access of Space-Air-Ground-Sea Intergrated Network 3.1 Wireless Access Control Based on Artificial Intelligence Artificial intelligence and big data mining technologies have advanced to the point where intelligent wireless communication network building is on the rise. Leading domestic and international communication companies have realized that the knowledge contained in big data may aid in the operation and efficiency of wireless communication networks. China Mobile has developed a big data-based network operation and maintenance optimization platform [22]. Network performance monitoring, fault detection, intelligent dynamic design of access resources, and other tasks are accomplished through the mining and analysis of signaling data. It also presents a design scheme for an access network with flexible function deployment, user and business situation intelligence perception, such as an access network that anticipates user demand using big data analysis and accomplishes intelligent local content push. Huawei discovers the characteristics of user behavior and network traffic by mining and analyzing user, traffic, network, and other related data, and then uses machine learning to learn the mining features, allowing the network to adaptively manage wireless resources and drive intelligent network planning with user experience. Orange, a French telecommunications company, proposes using big data analysis to achieve intelligent network parameter scheduling and optimization. The European Telecommunication Standardization Association proposes that NWDAFT7 be added to the 5G network in order to provide customized slice level load data analysis for policy control function entities and slice selection function entities to aid in network resource allocation, service orientation, and slice selection. Basic research on wireless network big data aided intelligent access control has recently concentrated on how to utilize big data prediction to distribute active resources and schedule active services. The use of predictive knowledge to plan or undertake actions in the future in advance is referred to as “initiative.” Document [23] extracts the future resource competition relationship of multi-users and analyzes the queuing model of mobile users in active switching mechanism, which is used to coordinate the competition among users in advance, using user mobility prediction information, including the cell and the cell dwell time accessed by the user at the next time. According to literature [24], by using user request content prediction information and channel state prediction information in the prediction window, the user can be accessed to the base station in advance for the requested file transmission, and the longer the prediction window is, the shorter the service delay is. Similarly, the paper [25] analyzes the queuing model of active service scheduling and employs service arrival prediction information to communicate user request data in advance. The literature [26] observes the expected transmission rate and prepares the distribution of the users’ frequency band resources in advance using the deep learning approach. In each frame of the prediction window, the document [27] preplans the user access choices. The article predicts the user’s position in the future and utilizes that information to arrange base station sleep management, resource reservation, and data transmission. The robust optimization approach is presented in literature [28] to combat the uncertainty induced by the prediction error, because the active mechanism’s performance gain is largely dependent on prediction accuracy. The study anticipates the change in traffic flow over time and utilizes that knowledge to model the load on a

72

L. Liu and Q. Liu

geographic location. The prediction error is assumed to have a Gaussian distribution, and an optimization algorithm based on gradient descent is proposed to optimize the user access selection pre-planning to balance the negative base station, with the goal of minimizing the expectation of the sum of regional load squares. Although research has been conducted in the industrial and academic circles on improving the serviceability of wireless communication networks through intelligent means and learning from the communication environment through intelligent methods, research on how to transform wireless data into useful knowledge to solve access control problems, as well as when and how to use the knowledge, is still limited. 3.2 Multiple Access Selection in Heterogeneous Wireless Networks The era of big data has arrived in the mobile wireless network. The rising traffic demand necessitates more severe criteria for future wireless network service capabilities, such as a 10–100 fold increase in network capacity and a 1 ms end-to-end latency. Effective technologies, such as MIMO, millimeter wave, and mobile edge computing, have evolved to attain greater performance metrics. Furthermore, for a long time, a variety of wireless networks based on different standards, such as LTE and WiMAX, will coexist to build heterogeneous wireless networks. Making full use of the coexistence of multiple wireless networks for parallel transmission in this heterogeneous scenario will result in rat multiplexing gain [29], greatly improve network capacity, reliability, and reduce service delay, and become the key and effective scheme to enhance network service capability. Wireless communication equipment may now be fitted with a number of rat interfaces because to advances in electronics. Intelligent user access selection and band resource allocation strategies may substantially enhance throughput and QoS in heterogeneous wireless networks. On the user side, wireless communication network service kinds will become more diverse in the future. Varied services have quite different service needs. Moreover, the variety of terminal equipment, user preferences, and other variables all contribute to varying QoS needs for a same service. Because services are clearly interconnected, a wireless network is required to deliver unique services. On the network side, heterogeneous networks provide a wide range of network characteristics to the access environment, including transmission characteristics, energy consumption characteristics, coverage characteristics, and so on. As a result, the network’s diverse features must be adapted to the wide range of business requirements. The term “context” refers to all of the information that may be used to characterize the properties of things. For users, it may be their location, QoS needs, or other factors; for the network, it could be changes in base station load, interference, network characteristics, or other factors. The adaptability of network and user side conditions is a fundamental need of intellectualization, and failure to address either situation will result in performance loss. For example, current network maximum signal-to-noise (max SNR) access technology overlooks network load, resulting in network congestion, low transmission rate, and frequent handoff. Reference [30] suggests integrating user services through various wireless transmission routes to accomplish load balancing in other research on user multi access choices. The utility function is maximized depending on network capacity to improve bandwidth allocation. The return link latency is also taken into account in reference [31]. However,

Research on Intelligent Access of SAGIN

73

the research above only looks at a single service, ignoring the differences in multiservice requirements in heterogeneous networks. Joint optimization of access selection and resource allocation has been done in the literature [32–34]. The file transmission service is discussed in reference [32]. The best wireless network group for access and the size of file block transmitted by each wireless network in the group are selected in order to reduce transmission costs. However, the cost of interference and congestion induced by user resource competition is not taken into account in this work. The network congestion threshold is utilized to decide user access selection in reference [33], and following the user access, the allocation of time slot and frequency band resources is further optimized. Reference [34] optimizes the best wireless network group for each user, however in practice, each wireless network has numerous APS, and the AP selection is not taken into account in this study. However, there is a scarcity of studies on the multiplexing benefit of multiple access in rats. To summarize, past research on user access and resource allocation in heterogeneous wireless networks has failed to take into account both user and network situations. Furthermore, it does not fully utilize coexisting heterogeneous network resources, only considers the use of a single wireless network for transmission, or only optimizes the wireless network group for parallel transmission, and does not take into account the AP selection problem within each wireless network. In light of the aforementioned flaws.

4 Reinforcement Learning Based Intelligent Access of Space-Air-Ground-Sea Intergrated Network 4.1 Heterogeneous Wireless Network Access Algorithm When a terminal device is in a heterogeneous wireless network environment, it automatically identifies the presence of a range of networks and may access those that match their requirements. The process of deciding which network of access to use is essentially a combination of network resource reassignment and resource scheduling. The user’s terminal continuously creates different services to pick different networks based on network QoS and user QOE, but how to effectively use this information to enable consumers access the best network is a network. The access algorithm’s main flaw. Existing network access algorithms are based on game theory, historical data, historical data from high altitudes, optimization theory, and tactics. The most basic network access method is based on SINR, which compares the SINR values of each network to access the best network, along with RSS to establish the standard access algorithm and its association, although after access performance is frequently poor. MADM is the most researched algorithm in the field of network access algorithms, and multi-attribute decision policies are primarily split into policies and other policies based on cost functions. Network QoS and user QoE tend to establish a utility function as a standard for terminal access networks during terminal selection, and the service type provided by the terminal is an impacted network in the network access algorithm. A typical technique is to use access parameters. Because the terminal’s program communicates via the HTTP/HTTPS protocol, it may be categorized as different service kinds based on the traffic it generates.

74

L. Liu and Q. Liu

Raschellà, et al. propose an algorithm based on load and priority, while classifying the type of service into real-time and non-real-time, the status utility function of each network is represented, and the adaptation factor is calculated in conjunction with the network’s load, and the user final outcome is calculated. To get access to the network, the maximum network of the adaption factor will be chosen, it is excellent for the efficient use of network resources since it not only ensures network load balancing but also reduces the accessibility rate of the access. The influence of user choices on network access is examined by Chen J, et al. An acceptable multi-criteria decision technique is presented in the client, and the flow cost is taken into account, as well as the network traffic being split into three categories: non-elastic, streaming, and elasticity. Different throughput needs are recommended for different types of traffic, and end users can pick the right network to connect to. The user’s speed and channel occupancy time are used by Udhayakumar S, et al. to forecast the user’s residence time in the WLAN. Real-time and non-real-time services are likewise distinguished, as is non-real-time business choices. The strain on the cell and WLAN will be detected via WLAN and real-time traffic. Furthermore, if high-priority users are accessible, traffic is moved, while low-priority users are transferred to other networks to free up resources for advanced users. Under the SDN’s network management architecture, access to the terminal can help better monitor changes in network state. Wu X, et al. present a network access policy for an SDN-based framework that employs the adaptive factor to calculate the QoS needs of the downlink stream. To assist the access terminal to an ideal network, the framework depends on SDN flexibility to implement the functionality and concentration of the monitor and assess network capacity. When compared to addressing difficult problems like resource allocation in heterogeneous wireless networks, fuzzy logical theory may be utilized to solve access alternatives in heterogeneous wireless networks. The computational complexity of fuzzy logic can be reduced. Wu X, et al. utilize fuzzy logic to rate the priority of the user who accesses the WLAN, which decides who is connected to the WLAN. Fuzzy logical theory, on the other hand, solely examines fixed indications and is unable to adapt to complicated and dynamic custom rules. Ontological reasoning approaches have the benefit of allowing the network’s access rules to be modified. Q. Zhou and colleagues propose a semantic access point resource allocation service for heterogeneous wireless networks, based on knowledge-based independent network management systems. Knowledge Base may introduce an access point that delivers the highest level of service quality automatically. With enough flexibility, this body-based knowledge system may also automatically modify the access point selection policy according to customer-defined requirements. The ontology and fuzzy reasoning binding approach may also be utilized for heterogeneous network access, and Al-Saadi A, et al. employ semantic reasoning of cross-layer QoS parameters from the heterogeneous network design to manage and optimize the network’s performance. The fuzzy reinforcer infers the next action in the heterogeneous wireless network using the Knowledge Base’s rule set, and selects the network architecture that can be handled. The decision is sent to the layer in the Internet protocol stack that is in charge of completing the necessary operation. The results demonstrate that in the event of heavy load, the cognitive network architecture based on ontology fuzzy

Research on Intelligent Access of SAGIN

75

reasoning can substantially enhance throughput and packet delivery rate (PDR). In a heterogeneous wireless network, defining the terminal access problem as a classification problem is a sympathy, where each potential network represents a category. The aim of network access is based on previous knowledge or statistical information to identify unknown objects as a known class, and a network user is regarded to have an object that specifies a collection of features of the corresponding decision factor. Machine learning is a field of research that focuses on learning systems and is often used to gain this information from a group of input items (also known as training data). A lot of work has been done using machine learning technologies to explore access strategies for heterogeneous wireless networks. The machine learning algorithm is a powerful tool for resolving categorization problems. Simple Bayes, decision trees, and neural networks, for example, can classify the type of company in the network access process. Lee D K, et al. split user traffic into four types: high-definition video streams, Internet telephony, audio streams, and files, using a decision tree method. Prior to terminal selection AP, first prioritize the user’s request based on the service type, which is transmitted to the controller through the AP, and then propose an access method by the controller, which not only considers the business’s QoS needs, but also the AP’s backhaul connection state, this can finally offer acceptable load balancing between the AP and supply the terminal with the QoS it requires. How to examine these aspects to guarantee better access to users, given that users must take into consideration numerous characteristics during network access, such as QoS, RSS, and user-level user preferences. A critical concern is the network. Artificial neural networks may produce correct answers for inputs that do not exist during training, and neural network algorithms are a viable solution for solving such issues. In recent years, stronger learning technology has been increasingly used in network access algorithms, and it has been coupled with other supervised or non-monitoring learning algorithms to improve algorithm intelligence. The qualities of online learning are reflected in the term “strengthen learning.” If the network deployment changes, the previously learnt method may no longer be the best option, and the cumulative advantages may be reduced. This bias will be detected through online learning, and the optimal policy will be modified by another round of training. Because of its independent characteristics, which are typically used to solve decision-making issues, and because of their constant update features, which allow it to adapt to changes in network environments, the importance of strengthening learning in network access management technology is growing. 4.2 Heterogeneous Wireless Network Access Algorithm Based on Reinforcement Learning The space space integrated network can achieve worldwide network coverage thanks to the presence of a satellite network. Multiple networks (such as satellite, terrestrial LTE, WiFi, and so on) may overlap coverage in some user-intensive locations (such as cities), and user access to diverse networks will have a significant influence on network performance and user experience. Meanwhile, spectrum resources, access methods, and protocols differ between space-based, space-based, and ground-based networks. As a result, in the field of space earth integrated network research, user access selection, or

76

L. Liu and Q. Liu

optimizing the user’s access network to increase network performance, has become a key priority. Unlike conventional network handoff, where the aim is to ensure service continuity, the goal of radio access technology selection is to maximize network performance in real time. As a result, the implementation technique is modified from a location change triggered (passive) switch to an active selection approach, in which the user’s access network is determined for each time period. The user association problem is another name for this type of network access selection challenge. In the integrated Space-Earth network, the conventional and optimization-based network access selection strategies will confront the following two problems. To begin, the majority of user allocation issues will be built as an integer or mixed integer combination optimization problem. This issue is not just non-convex, but it has also been shown to be NP-hard. The optimization approach can result in a significant quantity of computing and a long calculation time, and it can’t handle the air-earth integrated network’s vast-scale and high complexity. The optimization-based approach, on the other hand, is based on prior knowledge and modeling assumptions (such as network topology model, user distribution model, user mobility model, channel characteristic statistical model, service arrival model, and so on), which are not only for modeling network behavior with large particle sizes or for special network scenarios. At the moment, none of them can fulfill the demands of the integrated network of space and space, reducing the efficiency of optimization outcomes. Unlike optimization-based methods, the RL approach uses “observation and trial and error” to learn new network environments without the use of a previous model. Furthermore, after a period of operation, the RL neural network may theoretically suit a very complicated network environment and guarantee to output optimization results at a very rapid speed, allowing for real-time network access selection with little computing complexity. The study of RL-based network access selection is still in its early stages. Network access option may be separated into two categories based on the RL agent’s deployment site. 1) A RL agent is installed on each user, and the user selects pure dispersed access [35]. This approach may swiftly respond to changes in the user’s present environment and decrease data gathering signaling overhead by collecting data and making decisions locally. However, due to a single user’s limited observation ability, joint optimization of multi-user access selection techniques across a broad range is difficult to achieve. 2) To allow many users to share particular access resources, RL agents are deployed on access nodes (base stations, UAVs) or edge controllers [36]. This centralized deployment approach can easily optimize a large number of users, but it is restricted by the wireless transmission data and signaling overhead in the process of user data collecting and decision-making distribution, making real-time reaction to the user environment difficult. Multi wireless access technology is widely employed in diverse services of vehicle users in the space earth integrated network. As a result, one of the most important topics to address in order to enhance network performance is access control. In this case, the simulation platform may simulate several access techniques. The best access strategy for car users is analyzed, and the access scheme with the highest network data rate is selected, thanks to the gathering of global information by cloud controller. The simulation platform is utilized in this study to perform preliminary training of the DRL

Research on Intelligent Access of SAGIN

77

model in order to address the issues of difficult to get training samples and low training effectiveness of the space earth integrated network. Figure 2 depicts the DRL model training and use with the help of the simulation platform.

Fig. 2. Reinforcement learning based intelligent access of Space-Air-Ground Integrated Network

Simultaneously, because the network’s statistical features may vary, network environment data is gathered in the real network, and the model is updated to respond to the dynamic network environment of air space integration. Vehicle users can be serviced by several access modalities via the network’s terrestrial base station, UAV, and LEO satellites. RL is a mechanism for continually learning information from the external environment (such as user channel conditions and other complicated environments) in order to identify the best approach. In the space-earth integrated network, it can adapt effectively to the complex network environment and multi-user scenarios. The actor critical RL algorithm is used to discover the best access method. The parameters of two neural networks, actor and critical, are specified and initialized. Actor network determines each user’s access mode based on current network information, whereas critical network assesses the access mode in order to influence actor network to make a better decision in the future. At each learning moment, the current position information of each vehicle user is fed into the actor network, and each vehicle user’s access strategy is decided based on the probability distribution of the actor network’s output. Then, based on each car user’s access choice and the vehicle’s position information, we can calculate the reward value at any given time, i.e. the vehicle’s average data rate. Simultaneously, the critical network evaluates the error between the real reward value and the estimated reward value at the present instant in order to enhance the accuracy of the evaluation of actor network action output. The performance of an RL-based strategy will approach the optimal through a huge number of data input and iterative learning. Figure 5 depicts the experimental findings of network access selection. Although it is difficult for a learning algorithm based on a neural network to converge to the global optimal owing to the network’s complexity, forward propagation of a neural network requires minimum computing, a short running time, and a quick response, it makes it more appropriate for a wide range of real-time services in the integrated space-and-earth network.

78

L. Liu and Q. Liu

5 Conclusion The intelligent access technology in the Space-Air-Ground Integrated Network is discussed in this article. In light of current research and issues in space earth integrated intelligent access technology, this article analyzes the use of a reinforcement learning computer in a space earth integrated network and proposes a space earth integrated intelligent access technology scheme for network optimization.

References 1. Oh´len, P., Skubic, B., Rostami, A., et al.: Data plane and control architectures for 5G transport networks. J. Lightwave Technol. 34(6), 1501–1508 (2016) 2. Andrews, J.G., Buzzi, S., Choi, W., et al.: What will 5G be? IEEE J. Sel. Areas Commun. 32(6), 1065–1082 (2014) 3. Xu, L.: Context aware traffic identification kit (Tri CK) for network selection in future Het Nets/5G networks: ISNCC, Marrakech, Morocco, 16–18 May 2017. IEEE, New York (2017) 4. Aziz, D., Bakker, H., Ambrosy, A., et al.: Signalling minimization framework for short data packet transmission in 5G. In: IEEE VTC-Fall, Montreal, QC, Canada, 18–21 September 2016. IEEE, New York (2016) 5. Ravindran, R., Chakraborti, A., Amin, S.O., et al.: 5G-ICN: delivering ICN services over 5G using network slicing. IEEE Commun. Mag. 55(5), 101–107 (2017) 6. Sanchez, M., et al.: Results of the MIT space communication and navigation architecture study. In: IEEE Aerospace Conference, Big Sky, MT, USA, 1–8 March 2014. IEEE, New York (2014) 7. Radtke, J., Kebschull, C., Stoll, E.: Interactions of the space debris environment with megaconstellations-using the example of the One Web constellation. Acta Astronaut. 131, 55–68 (2017) 8. Pultarova, T.: Telecommunications-space tycoons go head to head over mega satellite network [News Briefing]. Eng. Technol. 10(2), 20–27 (2015) 9. Oikonomou, I.: The European defence agency and EU military space policy: whose space odyssey? Space Policy 28(2), 102–109 (2012) 10. Government of Japan. Basic plan on space policy. Strategic Headquarters for Space Policy, Government of Japan, Japan (2013) 11. Chowdhury, P.K., Atiquzzaman, M., Ivancic, W.: Handover schemes in satellite networks: state-of-the-art and future research directions. IEEE Commun. Surv. Tutorials 8(4), 2–14 (2006) 12. Gupta, L., Jain, R., Vaszkun, G.: Survey of important issues in UAV communication networks. IEEE Commun. Surv. Tutorials 18(2), 1123–1152 (2016) 13. Hayat, S., Yanmaz, E., Muzaffar, R.: Survey on unmanned aerial vehicle networks for civil applications: a communications viewpoint. IEEE Commun. Surv. Tutorials 18(4), 2624–2661 (2016) 14. Niephaus, C., Kretschmer, M., Ghinea, G.: QoS provisioning in converged satellite and terrestrial networks: a survey of the state-of-the-art. IEEE Commun. Surv. Tutorials 18(4), 2415–2441 (2016) 15. el mouna Zhioua, G., Tabbane, N., Labiod, H., et al.: A fuzzy multi-metric QoS-balancing gate way selection algorithm in a clustered VANET to LTE advanced hybrid cellular network. IEEE Trans. Veh. Technol. 64(2), 804–817 (2015) 16. Alawi, M.A., Saeed, R.A., Hassan, A.A., et al.: Simplified gateway selection scheme for multihop relay in vehicular ad hoc network. Int. J. Commun. Syst. 27(12), 3855–3873 (2014)

Research on Intelligent Access of SAGIN

79

17. Ko, B.J., Liu, S., Zafer, M., et al.: Gateway selection in hybrid wireless networks through cooperative probing. In: IFIP/IEEE IM, Ghent, Belgium, 27–31 May 2013. IEEE, New York (2013) 18. Zaman, R.U., Khan, K.u.R., Waseem, M.A., et al.: Traffic priority based gateway selection in integrated Internet-MANET. In: IEEE iCATccT. IEEE, New York (2016) 19. Dhaou, R., Franck, L., Halchin, A., et al.: Gateway selection optimization in hybrid MANETsatellite network. In: Pillai, P., Hu, Y., Otung, I., Giambene, G. (eds.) Wireless and Satellite Systems, WiSATS 2015. LNICS, Social Informatics and Telecommunications Engineering, vol. 154. Springer, Cham (2015). https://doi.org/10.1007/978-3-319-25479-1_25 20. Luo, F., et al.: A distributed gateway selection algorithm for UAV networks. IEEE Trans. Emerg. Topics Comput. 3(1), 22–33 (2015) 21. Zhong, S., Zhang, Y.: How to select optimal gateway in multi-domain wireless networks: alternative solutions without learning. IEEE Trans. Wireless Commun. 12(11), 5620–5630 (2013) 22. Chih-Lin, I., Liu, Y., Han, S., Wang, S., Liu, G.: Big data analytics for greener and softer RANOn big data analytics for greener and softer RAN. IEEE Access 3, 3068–3075 (2015) 23. Paranthaman, V.V., Mapp, G., Shah, P., et al.: Exploring Markov Models for the allocation of resources for proactive handover in a mobile environment. In: Proceedings of 2015 IEEE 40th Local Computer Networks Conference Workshops, pp. 1–7. IEEE, Florida (2015) 24. Zhao, S., Shao, Z., Qian, H., et al.: Online user-AP association with predictive scheduling in wireless caching networks. In: GLOBECOM 2017–2017 IEEE Global Communications Conference, pp. 1–7. IEEE (2017) 25. Huang, L., Zhang, S., Chen, M., et al.: When backpressure meets predictive scheduling. IEEE/ACM Trans. Networking 24(4), 2237–2250 (2015) 26. Guo, J., Yang, C.: Predictive resource allocation with deep learning. In: 2018 IEEE 88th Vehicular Technology Conference (VTC-Fall), pp. 1–7. IEEE (2018) 27. Xu, Z., Guo, J., Yang, C.: Predictive resource allocation with interference coordination by deep learning. In: 2019 11th International Conference on Wireless Communications and Signal Processing (WCSP), pp. 1–6. IEEE (2019) 28. Liakopoulos, N., Paschos, G.S., Spyropoulos, T.: Robust optimization framework for proactive user association in UDNs: a data-driven approach. IEEE/ACM Trans. Networking 27(4), 1683–1695 (2019) 29. Wang, R., Hu, H., Yang, X.: Potentials and challenges of C-RAN supporting multi-RATs toward 5G mobile networks. IEEE Access 2, 1187–1195 (2014) 30. Bohli, A., Bouallegue, R.: How to meet increased capacities by future green 5G networks: a survey. IEEE Access 7, 42220–42237 (2019)

Spectrum Sensing Based on Federated Learning with Value Evaluation Mechanism Zheng Liu1(B) , Junsheng Mu1 , Fangpei Zhang2 , Xiaojun Jing1 , and Bohan Li3 1 Beijing University of Posts and Telecommunications, Beijing, China

{l_z,mujs,jxiaojun}@bupt.edu.cn

2 Information Science Academy of China Electronics Technology Group Corporation,

Beijing, China 3 University of Southampton, Southampton, UK

[email protected]

Abstract. In the Internet of things (IoT), the extensive use of IoT devices makes the problem of spectrum sharing among devices increasingly prominent. Spectrum sensing is very significant to promote spectrum efficiency in IoT. However, due to network security and industry privacy issues, it is difficult to obtain large-scale data sets needed for spectrum sensing. Therefore, federated learning (FL) is an effective technique to solve the problems that may be encountered in the establishment of data sets and the problem of data leakage. In this paper, FL is utilized to study the problem of spectrum sensing, and a value evaluation mechanism of IoT devices is proposed to improve the performance of FL and resist poisoning attacks. Simulation shows that the proposed value evaluation mechanism can make the global model of FL converge more quickly and stably, and at the same time it is almost unaffected by malicious nodes when poisoning attacks occur. Keywords: Spectrum sensing · Federated learning · Value evaluation mechanism · Poisoning attack

1 Introduction With the popularity of 5G technology, the IoT paradigm and a variety of emerging applications (such as smart home, industrial IoT, etc.) are developing rapidly. During this period, the number of connections between smart devices and terminals increased explosively. In any case, the rapid growth in the number of connections in the IoT is bound to take up all the 5G spectrum. Therefore, both now and in the future, it is an important challenge for the network to improve spectrum efficiency. In this regard, cognitive radio is regarded as a potential solution [1–3]. Cognitive radio technology can monitor the spectrum utilization in real-time and dynamically adjust the devices accessing the spectrum [4, 5]. Before spectrum allocation, it is necessary to determine whether the target spectrum is available or not. Recently, machine learning has been used in spectrum sensing [6]. Sarikhani et al. [7] have proposed Deep Reinforcement Learning based cooperative © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 S. Sun et al. (Eds.): ICSINC 2021, LNEE 895, pp. 80–87, 2022. https://doi.org/10.1007/978-981-19-4775-9_10

Spectrum Sensing Based on Federated Learning

81

spectrum sensing algorithm. Zheng et al. [8] have proposed a sensing method based on deep learning classification. However, in these methods combined with machine learning, it is troublesome to build a centralized dataset containing a large number of samples. At present, some people have proposed methods to expand the data set [9, 10]. However, if the data sets in the IoT devices are required to be transmitted to the cloud to build a large data set, which is then used to train machine learning models, it may lead to serious network security or user privacy problems [11]. FL [12] solved this problem. In FL, IoT devices train the model independently, and then the central node aggregates the local model to get the global model. Google proposed a FederatedAveraging algorithm [13], which averages the neural network parameters of each edge device to improve the global model. However, the average aggregation method can not resist poisoning attacks. Therefore, this paper proposes a value evaluation mechanism, which can accurately evaluate the effectiveness of IoT devices and resist poisoning attacks. This paper is organized as follows. Section 2 introduces the framework of FL and the system model. In Sect. 3, the workflow of the model and the value evaluation mechanism of IoT devices are introduced. The simulation and analysis are conducted in Sect. 4. Finally, conclusions are drawn in Sect. 5.

2 System Work In this paper, the OFDM signal is used as the signal of the primary user (PU). Different Internet of things devices will correspond to different signal acquisition devices, so they will produce their own local data sets that are different from other devices. The device Di regards the spectrum sensing problem as a binary classification problem and uses local data sets to train the local model. The system model is shown in Fig. 1.

Fig. 1. System model

The essence of FL a distributed machine learning. FL mainly includes IoT devices and cloud servers. IoT devices jointly train the model under the coordination of the cloud server (CS). Each of these IoT devices has a copy of the global model, which is called the local model. IoT devices use their local data to update the local model, and the cloud server aggregates all the local models to get a global model ω, which is similar to the

82

Z. Liu et al.

result of centralized machine learning after many iterations. In this way, problems such as data leakage can be effectively avoided. However, if there is a device Dj that maliciously uses the wrong dataset to update the model, the effectiveness of the global model may be seriously affected.   In FL, the collection of devices can be defined as D = D1 , D2 , . . . , DND , where Di (i = 1, 2, . . . , ND ) represents the i-th device, ND = |D| indicates the total number of devices. Each device stores its own local dataset. The local dataset of the device Di is represented as Si , where |Si | = Ni . The model of device Di utilizes an Mi -element antenna system to receive signals based on Ni observation vectors, then get the dataset by the method in [14]. Then we use mathematical methods to calculate the covariance matrix and finally get true color pictures as dataset Si [9].

3 Spectrum Sensing Based on FL 3.1 Work Flow As shown in Fig. 2, the operation at the l-th epoch consists of the following five moves: a. Get and store datasets. Follow the method in Sect. 2 to create a dataset Si for device Di ; b. Global model distribution. The CS sends the global model ωl−1 to each device; c. Edge model update. The device Di updates the edge model based on the global model (m,l) ωl−1 . Then the parameter ωi of the m-th iteration of the local model at the l-th epoch can be expressed as   (m,l) (m−1,l) (m,l) (1) ωi = ωi − γ ∇Fi ωi   (m,l) where γ represents the learning rate, Fi ωi is the loss function. The final parameter is taken as the local model parameter ωil of the l-th epoch. d. Local model upload. the device Di upload the parameter ωil of the updated edge model to the CS. e. Global model aggregation. In FL, the aggregation at the l-th epoch can be expressed as ND αi ωil . (2) ωl = i=1

i where αi = ST ST is the weight of the device Di , ST i indicates the score of the device Di , ST represents the total score of all devices.

Repeat these steps until the global model converges or the model reaches the required accuracy.

Spectrum Sensing Based on Federated Learning

83

3.2 Value Evaluation Mechanism of Parameters Because of the long distance between edge devices, there are many difficulties in the correct dissemination of information. There are even some devices that tamper with an edge or global model during move e. Therefore, it is very significant to make a complete effectiveness evaluation of the parameters uploaded by the equipment. At present, many objective weighting methods are widely used to determine the weight [15, 16]. The CRITIC weight method is an objective weighting method. It is based on the contrast intensity of indicators and the conflict between the indicators to comprehensively measure the objective weight of indicators. In this paper, we use several indicators to evaluate the score of the parameter, such as the size of the dataset, the correlation with the global model, and the accuracy of the local model. The CRITIC weight method is utilized to evaluate the weight of the indicators. The score for the edge device is generated during the global model aggregation. The score affects the weight of the parameters of the local model in FL. Due to the complexity of deep learning, it is difficult to assess the validity of parameters by simply comparing the accuracy of local models. Therefore, we determine the local training performance by calculating the correlation between the parameters of the edge model and the global model. Suppose ωi = {ωi1 , ωi2 , . . . , ωiP }, (i = 1, 2,  . . . , ND ) are all the parameters of the local model uploaded by the device Di , ω = ω1 , ω2 , . . . , ωP are all parameters of the global model. We use Pearson product-moment correlation coefficient (PPMCC) to represent the degree of correlation between the edge model and the global model:    P   ω ω − ω − ω ij i j=1 j ri =

(3)  2 2 P   P   j=1 ωij − ωi j=1 ωj − ω The larger the ri , the greater the correlation between the edge model and the global model. In addition, we can make further improvements to ri , r if rj > 0 (4) rj = j 0 if rj ≤ 0 CRITIC Weight Method The number of dataset in the device Dj is Nj , the accuracy of edge model ωj is Aj , and the correlation between the local parameter ωj and the global parameter ω is rj . In the following sections, we use xij (i = 1, 2, 3, j = 1, 2, . . . , NT ) to denote Nj , Aj and rj , that is, x1j = Nj , x2j = Aj , x3j = rj . Then the proportion of xij can be expressed as xij Pij = N T

j=1 xij

(5)

First of all, we use the standard deviation SDi to express the contrast intensity of  T the i-th indicator. First calculate the mean value xi = 1n N j=1 xij , and then the standard

84

Z. Liu et al.

deviation of the i-th indicator is obtained,

 SDi =

NT  j=1 xij

− xi

2

n−1

(6)

Secondly, the correlation coefficient Ri is used to express the conflict of the indicators. First of all, we need to calculate the correlation degree rik between different indicators. According to PPMCC, ND

− xi )(xkj − xk )

2 ND 2 (x − x ) i j=1 ij j=1 (xkj − xk )

rik =  ND

j=1 (xij

(7)

Then Ri =

3 k=1

(1 − rik )

(8)

Then the amount of information Ci of the i-th indicator is calculated according to the standard deviation SDi and the correlation coefficient Ri . Ci = SDi ×

3 k=1

(1 − rik ) = SDi × Ri .

(9)

So the objective weight of the i-th indicator is Ci Wi = 3 1 Ci

(10)

Therefore, the score of the device Di can be expressed as STj =

3 i=1

Wi × Pij .

(11)

After the above steps, we adjust the weight of each edge device in the model aggregation to prevent the bad model uploaded by malicious nodes from affecting the accuracy of the global model. This method significantly improves the accuracy, convergence and anti-interference of the global model.

4 Numerical Result In this paper, we set up 10 nodes in FL and establish local datasets for each node under different signal-to-noise ratios (SNR). At the same time, two cases are set, one is that there is no malicious node in 10 nodes, and the other is that there are two malicious nodes in 10 nodes, which is called a poisoning attack. The dataset of the malicious node is wrong, and the distribution of the wrong dataset is opposite to that of the normal dataset. As a result, the local model of the malicious node has the opposite effect on the aggregation of FL. At the same time, we compare the performance of average aggregation, called FLavg, and weighted aggregation with value evaluation mechanism, called FLvem, under different

Spectrum Sensing Based on Federated Learning

85

SNR. The probability of detection (PD) and probability of false alarm (PFA) are shown in Fig. 2. When there are no malicious nodes, the performance of FLavg is almost the same as that of FLvem. with the increase of SNR, PD increases and PFA decreases. However, when subjected to poisoning attacks, the performance of FLvem is almost unchanged under most SNR, while the performance of FLavg degrades sharply. At the same time, we can also see the advantages of FLvem from loss function. Figure 3 shows the loss function of FLavg and FLvem when subjected to poisoning attacks under SNR = −2 dB, respectively. It can be seen that the loss function of FLavg can not always decrease steadily, but will increase when it decreases to a certain extent, which shows that the malicious nodes have a serious impact on the global model, and the loss function of the global model is difficult to converge to the lowest value. However, the loss function of FLvem can maintain a steady and continuous decline, and its global model can gradually converge to the lowest value, which indicates that malicious nodes have almost no effect on the global model.

Fig. 2. The probability of detection and the probability of false alarm

In fact, not only in the case of poisoning attack, the performance of FLvem is superior, but also the loss function of FLvem converges faster when there is no poisoning attack.

Fig. 3. The loss function of FLavg and FLvem

86

Z. Liu et al.

5 Conclusion In summary, we design a spectrum sensing framework based on federated learning in IoT. At the same time, in federated learning, we propose a value evaluation mechanism for IoT devices, which can effectively strengthen the positive role of beneficial nodes and weaken the impact of malicious nodes. In federation learning, this mechanism not only plays a significant role in making the model converge more quickly and stably but also can effectively resist poisoning attacks.

References 1. Boccardi, F., Heath, R.W., Lozano, A., Marzetta, T.L., Popovski, P.: Five disruptive technology directions for 5G. IEEE Commun. Mag. 52(2), 74–80 (2014). https://doi.org/10.1109/MCOM. 2014.6736746 2. El Tanab, M., Hamouda, W.: Resource allocation for underlay cognitive radio networks: a survey. IEEE Commun. Surv. Tut. 19(2), 1249–1276 (2017). https://doi.org/10.1109/COMST. 2016.2631079 3. Yang, C., Li, J., Guizani, M., Anpalagan, A., Elkashlan, M.: Advanced spectrum sharing in 5G cognitive heterogeneous networks. IEEE Wirel. Commun. 23(2), 94–101 (2016). https:// doi.org/10.1109/MWC.2016.7462490 4. Mitola, J., Maguire, G.Q.: Cognitive radio: making software radios more personal. IEEE Pers. Commun. 6(4), 13–18 (1999). https://doi.org/10.1109/98.788210 5. Haykin, S.: Cognitive radio: brain-empowered wireless communications. IEEE J. Sel. Areas Commun. 23(2), 201–220 (2005). https://doi.org/10.1109/JSAC.2004.839380 6. Gao, N., Jin, S., Li, X., Matthaiou, M.: Aerial RIS-assisted high altitude platform communications. IEEE Wirel. Commun. Lett. 10(10), 2096–2100 (2021). https://doi.org/10.1109/ LWC.2021.3091164 7. Sarikhani, R., Keynia, F.: Cooperative spectrum sensing meets machine learning: deep reinforcement learning approach. IEEE Commun. Lett. 24(7), 1459–1462 (2020). https://doi.org/ 10.1109/LCOMM.2020.2984430 8. Zheng, S., Chen, S., Qi, P., Zhou, H., Yang, X.: Spectrum sensing based on deep learning classification for cognitive radios. China Commun. 17(2), 138–148 (2020). https://doi.org/ 10.23919/JCC.2020.02.012 9. Davaslioglu, K., Sagduyu, Y.E.: Generative adversarial learning for spectrum sensing. In: 2018 IEEE International Conference on Communications (ICC), pp. 1–6 (2018).https://doi. org/10.1109/ICC.2018.8422223 10. Liu, Z., Jing, X., Zhang, R., Mu, J.: Spectrum sensing based on deep convolutional generative adversarial networks. Int. Wirel. Commun. Mob. Comput. (IWCMC) 2021, 796–801 (2021). https://doi.org/10.1109/IWCMC51323.2021.9498871 11. Zhao, J., Chen, Y., Zhang, W.: Differential privacy preservation in deep learning: challenges, opportunities and solutions. IEEE Access 7, 48901–48911 (2019). https://doi.org/10.1109/ ACCESS.2019.2909559 12. Koneˇcný, J., McMahan, B., Ramage, D.: Federated optimization: distributed optimization beyond the datacenter. arXiv preprint arXiv:1511.03575 (2015) 13. McMahan, H.B., et al.: Communication-efficient learning of deep networks from decentralized data. arXiv preprint arXiv:1602.05629 (2016) 14. Gao, N., Li, X., Jin, S., Matthaiou, M.: 3-D deployment of UAV swarm for massive MIMO communications. IEEE J. Sel. Areas Commun. 39(10), 3022–3034 (2021). https://doi.org/10. 1109/JSAC.2021.3088668

Spectrum Sensing Based on Federated Learning

87

15. Lu, C., Li, L., Wu, D.: Application of combination weighting method to weight calculation in performance evaluation of ICT. In: 2015 IEEE 15th International Conference on Advanced Learning Technologies, pp. 258–259 (2015). https://doi.org/10.1109/ICALT.2015.15 16. Lee, D., Lee, J.: Incremental receptive field weighted actor-critic. IEEE Trans. Industr. Inf. 9(1), 62–71 (2013). https://doi.org/10.1109/TII.2012.2209660

Application of Artificial Intelligence for Space-Air-Ground-Sea Integrated Network Shaowei Zhang1 , Lei Liu1(B) , and Mohamed Cheriet2 1 Beijing Research Institute, China Telecom Corporation Limited, Beijing, China

[email protected] 2 Synchromedia Laboratory, École de Technologie Supérieure, Université du Québec,

Montreal, Canada

Abstract. The spatial scope of information services is expanding, various spacebased, space-based, sea-based, and ground-based network services are emerging, and the need for multi-dimensional comprehensive information resources is steadily increasing. The combined air, space, and sea network can deliver seamless information services to land, sea, air, and space users, as well as satisfy the future network’s demands for all-time, all-space communication, and network connectivity. Firstly, the development trend of the network technology and protocol system of the integration of air, ground and sea is analyzed, and the research progress of LEO satellite communication system and air ground network integration is discussed. Then, the artificial intelligence technology is discussed. Finally, the application of artificial intelligence technology in the integrated network of space, earth and sea is discussed. Keywords: Space-air-ground-sea · Artificial intelligence · Integrated network · Reinforcement learning

1 Introduction The combined air, space, and sea network can deliver seamless information services to land, sea, air, and space users, as well as satisfy the future network’s demands for all-time, all-space communication, and network connectivity. Compared with the communication needs of ordinary people, the communication of the Internet of things will be greatly expanded both in space and content. A variety of IOT devices and services will cover mountainous areas, deserts, oceans, deep earth, sky, space and other broader areas. For Internet of things applications, 5G network specially plans two important service scenarios, namely ultra-high reliability and low delay communication and large-scale machine communication [1]. 5G network technology actively promotes narrow-band Internet of things, beamforming, uplink/downlink decoupling and other technologies, which can solve the key technical problems of Internet of things, such as wide area coverage, energy consumption, large connection and so on. However, large-scale 5G network deployment requires high cost. Intensive base station deployment and backhaul © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 S. Sun et al. (Eds.): ICSINC 2021, LNEE 895, pp. 88–102, 2022. https://doi.org/10.1007/978-981-19-4775-9_11

Application of Artificial Intelligence

89

network construction will result in expensive infrastructure costs, as well as the installation and maintenance costs of optical cable. At the same time, the ground-based network is difficult to cover remote areas, ocean, deep earth, sky and even deep space. As a result, it will be challenging for 5G ground-based network technology to satisfy the ubiquitous communication demands of network space growth. In addition, in the future, the demand for multi-dimensional comprehensive information resources of information services will gradually increase [2]. In this context, the creation of an integrated network of space, earth, and sea, as well as the deep integration of space-based networks, space-based networks, ground-based networks, and sea-based networks, as well as the full play of the functions of different network dimensions, can break down data sharing barriers between independent network systems and achieve full coverage of wide area and network interconnection. The space-based network [3] is based on the ground-based network, supplemented and extended by space-based network and space-based network, and provides ubiquitous, intelligent, collaborative and efficient information support infrastructure for various network applications in wide area space. In the air space integrated network, the groundbased network is mainly composed of ground-based Internet and mobile communication network, which is responsible for the network services in the business intensive areas; The space-based network is composed of high altitude communication platform and UAV ad hoc network, it has the capabilities of improving coverage, providing edge service, and allowing for flexible network reconfiguration. Through the deep integration of multi-dimensional network, the space earth integrated network can effectively utilize all kinds of resources, carry out intelligent network control and information processing, so as to be able to cope with the network services with different needs, and achieve the goal of “network integration, functional service and application customization”. Among them, space-based network (mainly various satellite networks) technology is in the core position, which is the key enabling technology to build omnipresent, connected and omniscient space-based integrated network. In recent years, with the gradual maturity of LEO satellite constellation technology represented by Starlink project of the United States, a large number of LEO satellites will form an Internet infrastructure with global coverage, high-capacity broadband access and low communication delay to provide seamless high-speed Internet access for global users. Low earth orbit satellite constellation, medium and high earth orbit satellite and various navigation, remote sensing, meteorological and other functional satellites will jointly build a heterogeneous spacebased infrastructure network with diverse functions, high intelligence, complementary orbits and convenient expansion. With the importance of space-based network becoming more and more important, the research on the integration of ground-based network and space-based network has attracted great attention. Although the existing 5G network standards and commercial deployment have not yet integrated satellite communication, the standardization work related to them has been advancing. Starting from rel-16, 5G network began to study the technical characteristics of non-terrestrial communication network. The third generation Partner Program in the recent proposal, it defines and discusses the potential technical problems, business characteristics, network structure and deployment scenarios of integrating satellite network into 5G network. With the launch of 6G network research, in order to meet the “5A” ubiquitous communication

90

S. Zhang et.al

demand that “anyone can do any business with anyone at any time and any place”, the space earth integrated network will become an indispensable part of 6G network. The Space Earth Integrated Network [4] is a multi-dimensional heterogeneous network. When several networks are combined, the network structure becomes very complicated, with a wide range of network resources. Because of the various dynamic features of space, sky, and earth networks, the whole network has more complicated mobility properties than a ground-based network. As a result, correctly describing and modeling the network is challenging. Simultaneously, the space-based integrated network delivers network services for a variety of space-based, ground-based, and maritime information services. The diverse service features and service quality standards make allocating network resources and arranging services exceedingly complex. As a result, standard optimization approaches are inefficient and slow to respond, and they are unable to adapt to the complex and dynamic network environment of a space-based integrated network. Artificial intelligence method is considered to be a solution with great potential for complex dynamic problems that are difficult to model. This approach can build the best mapping model of network environment and network control by extracting and analyzing a huge quantity of data, so as to carry out network design, control, management and optimization efficiently and intelligently. RL as an important method of machine learning, can learn the best action strategy through the feedback of agent interaction with the environment, and can deal with the learning decisions in the unknown network environment [5]. It is very suitable for the complex, dynamic and high cost of network data collection characteristics of the space earth integration network, and it is a solution to the optimal network control, resource allocation, resource allocation and resource allocation Service scheduling and other key issues.

2 Space-Air-Ground-Sea Integrated Network This approach can build the best mapping model of network environment and network control by extracting and analyzing a huge quantity of data, large-scale connection, and greatly improve energy consumption and cost, so as to greatly enhance the user experience. On the other hand, communication network is an important means of situation information acquisition and interaction, The future network presents the trend of spacebased integration, which is composed of space-based, space-based, ground-based and ocean networks, and organically combines the space, air, ground and sea independent and scattered sensing nodes, communication nodes and other equipment [6]. The architecture of the space-based network is shown in Fig. 1. 2.1 Geostationary Satellite Constellation INMARSAT is a ship radio communication system using geostationary orbit communication satellite as relay station. The first generation of INMARSAT satellite system was mainly realized by renting satellites. INMARSAT deployed the first dedicated satellite constellation, inmarsat-2, in the early 1990s, which is mainly used for maritime purposes. Inmarsat-3, the third-generation satellite system of INMARSAT, was launched in 1996–1998. The constellation consists of five L-band satellites that are primarily used to

Application of Artificial Intelligence

91

Fig. 1. Space-air-ground-sea integrated network architecture

provide low-bandwidth communication and security services for worldwide shipping. INMARSAT has been developing and launching the first worldwide coverage satellite communication system, the BGAN (broadband global area network), since 1999, with four inmarsat-4 satellites deployed between 2005 and 2013. Since 2010, INMARSAT has started to develop the high throughput satellite constellation Global Express (GX), which works in Ka band. GX system includes five satellites, which will be launched successively from 2013 to 2019 to provide global satellite communication for various markets including shipping and aviation [7]. The satellite in GX system adopts spot beam coverage mode. Each satellite has 89 fixed spot beams, and the downlink rate of each fixed beam can reach 50 Mbit/s. In addition, each satellite is also equipped with six large capacity mobile spot beams, which can flexibly cover any hot spot area, and can provide up to two 100 MHz channels. 2.2 Non Geostationary Orbit Satellite Constellation Iridium satellite system is the first Leo global personal satellite mobile communication system proposed by Motorola. The first generation iridium system consists of 72 satellites (6 spare satellites), which are distributed on six polar circular orbits with an altitude of 780 km. The system mainly provides voice services for global users, and uses on-board processing, on-board switching and interstellar link technology to form a complete spacebased network. Iridium put forward the second generation Iridium satellite system plan in 2007, and launched 75 satellites in orbit from 2017 to 2019. The satellite constellation can provide up to 128 kbit/s data rate for mobile terminal in L-band, 1.5 mbit/s data rate for iridium pilot navigation terminal and 8 Mbit/s data rate for fixed/mobile terminal in

92

S. Zhang et.al

Ka band. Iridium next aims at IP Broadband Networking and scalable and scalable load capacity, which makes it able to adapt to the complex needs of future space information applications. However, for the current increasing demand of mobile Internet, especially in the face of 5G communication era, iridium next data transmission capacity is still insufficient. Starlink is a satellite constellation development project of SpaceX company, which aims to develop a low-cost, high-performance space-based Internet communication system. Starlink will deploy a large number of satellites in ultra-low earth orbit to achieve global network coverage. Starlink’s goal is to provide services to northern United States and Canada by 2020, and expand the service scope to most parts of the world by 2021. The Starlink project will launch about 12000 satellites, first about 1600 in 550 km orbit, then about 2800 Ku band and Ka band satellites in 1150 km orbit, and finally about 7500 V-band satellites in 340 km orbit. As of April 22, 2020, Starlink has launched 422 satellites. SpaceX uses a fast and reusable launch system to reduce the cost of launching a large number of satellites. SpaceX said Starlink will provide users on earth with broadband services with data rates of at least 1 Gbit/s and ultra-high speed broadband services up to 23 Gbit/s. The end-to-end delay range is 25–35 ms, comparable to that of cables and optical fibers. Hongyan global satellite constellation communication system is proposed by China Aerospace Science and Technology Group Co., Ltd. It plans to deploy more than 300 satellites by 2025, and the first batch of 60 satellites will be deployed around 2023. The first test satellite of “Hongyan” global satellite constellation communication system was put into a predetermined orbit of 1100 km in December 2018. The satellite has L/Ka band communication payload, navigation enhancement payload and aviation surveillance payload. After full deployment in the future, the system will be able to realize all-weather, all time and real-time two-way communication under complex terrain conditions through LEO satellites and global data processing center, and can provide users with global real-time data communication and integrated information services. Hongyun project is proposed by China Aerospace Science and Industry Corporation. It is composed of 156 low-orbit WeChat and operates at an orbital altitude of 1000 km. The maximum supported rate is 4 Gbps. After the constellation deployment is expected to be completed in 2022, Hongyun Project will be able to provide broadband mobile communication services with seamless global coverage. Based on the space-based network, it will integrate navigation and remote sensing to realize the integration of communication, navigation and remote sensing, and build a broadband low-orbit satellite communication system for environmental detection. Mobile communications and other fields to provide information transmission services. The user groups targeted by the Hongyun Project are mainly cluster user groups, including airplanes, ships, passenger and cargo vehicles, field areas, operation teams, and villages and islands in remote areas. UAVs and unmanned driving industries are all industries that Hongyun Engineering may serve in the future. With its extremely low communication delay, extremely high frequency reuse rate, and true global coverage, Hongyun Engineering can meet the needs of China and the Internet underdeveloped areas and large-scale user units to share broadband access to the Internet at the same time. It may also fulfill the application criteria of emergency communication at the same time, sensor data collection, industrial Internet of things,

Application of Artificial Intelligence

93

remote control of unmanned equipment, etc., which require high real-time information interaction. In addition to the above satellite systems, a large number of LEO satellite constellation systems are under preparation. For example, Boeing plans to launch 1396–2956 satellites in an orbit of 1200 km by 2022, Samsung plans to launch 4600 satellites in an orbit of 1400 km by 2028, and telesat plans to launch 117–512 satellites in an orbit of 1000 km and 1200 km by 2021.

3 Artificial Intelligence for Space-Air-Ground-Sea Integrated Network Artificial intelligence is one of the hottest topics for giving computers intelligence and allowing them to do jobs better than humans. People believe that integrating artificial intelligence’s benefits is a hard and intriguing concept. Despite the fact that traditional approaches have had a lot of success in tackling many problems in this sector, it’s still worth investigating whether AI may give more powerful and precise answers. We think that intelligent solutions are not necessarily better to traditional ones; in certain situations, traditional approaches may give simple and powerful solutions. This duality is undoubtedly one of the grounds for researching the use of artificial intelligence in a variety of unique challenges relating to the space-air-ground-sea integrated network. However, with the rise of artificial intelligence, intelligent solutions for a space-air-ground-sea integrated network have become fashionable [8].

Fig. 2. Artificial intelligence technology

94

S. Zhang et.al

3.1 Deep Belief Architecture A neural network, Deep Belief Nets are a type of neural network. It is similar to an autoencoder in that it may be used for unsupervised learning; however, it can also be utilized as a classifier in supervised learning. The goal of unsupervised learning is to keep as many of the original features’ properties as feasible while lowering the size of the features. The goal of supervised learning is to reduce the categorization error rate to the lowest achievable level. Whether supervised or unsupervised learning, the process of feature learning, or how to improve feature expression, is at the heart of DBN. Neuron is an important component of a neural network. The limited Boltzmann machine is a component of the DBN, which is made up of multiple layers of neurons (RBM). There are just two layers of neurons in the RBM. The visible layer, which is made up of visible units and is used to input training data, is one of the layers. The hidden layer, which is made up of hidden units and is utilized as a feature detector, is the other layer. Each layer is represented by a vector, and each neuron is represented by a dimension. Keep an eye on the bidirectional link that exists between the two levels. Neurons are self-contained units. The benefit is that the values of each implicit element are irrelevant given the values of all explicit components. A DBN is made up of a succession of RBMs. The previous RBM’s hidden layer is the explicit layer of the following RBM, while the previous RBM’s output is the input of the next RBM. During the training process, the current layer’s RBM can only be trained once the upper layer’s RBM has been entirely taught to the last layer. The process of DBN tuning is a model generation process: 1. Initial stage: The weights of the other RBMs are split into upward cognitive weights and downward generative weights, in addition to the top RBM; 2. Wake stage: External characteristics and upward weights (cognitive weights) produce the abstract representation (node state) of each layer in the cognitive process, whereas gradient descent modifies the downward weights (generation weights) between levels. 3. Sleep stage: The top-level representation (the notion gained during waking up) and the down weight are used to produce the state of the bottom layer, while the up weight across levels is updated at the same time. Using random recessive neuron state values, enough Gibbs sampling is performed in the top RBM; Propagate down to get the state of each layer. There are several approaches to convert DBM to supervised learning, including adding neurons encoding categories in each RBM and a softmax classifier in the final layer. The DBM-trained w may also be considered the pretrain of NN, meaning that fine tuning can be done using the BP method. The forward algorithm is, in reality, the original DBN method, whereas the update algorithm is the BP algorithm. The BP algorithm used here might be the most original BP algorithm or one that we created ourselves. 3.2 Deep Q-network The main idea of the Q-learning algorithm is to construct a Q-table of State and Action to store the Q value, and then select the action that can obtain the greatest benefit based

Application of Artificial Intelligence

95

on the Q value. When the state and action space are discrete and the dimension is not high, q-table can be used to store t. It’s tough to utilize a q-table to store the Q value of each state action pair when the state and action space is high-dimensional continuous. As a result, the q-table update may be recast as a problem of function fitting. Comparable states can receive similar output actions by fitting a function to replace the q-table to create Q value. As a result, we may assume that deep neural networks have a positive impact on the extraction of complicated features, and that deep learning and reinforcement learning can be combined. Deep learning is supervised learning that necessitates the use of a training set. No training is required for reinforcement learning. It solely gives back reward value in the form of the surroundings. At the same time, it has issues with noise and delay. As a result, there are numerous states with a reward value of zero, indicating that the samples are sparse. Each deep learning sample is independent of the others, and the current state value of reinforcement learning is determined by the later state’s return value. There is a risk of instability when using a nonlinear network to represent the value function. The DQN algorithm basically follows the idea of Q-learning, but has made some improvements in order to be able to better integrate with the deep neural network. Two powerful tools in DQN solve the above problems: using reward to construct tags through Q-learning, using experience replay (experience pool) to solve the problem of correlation and non-static distribution, using a Mainnet to generate the current Q value, and using another target to generate target Q. Because Q learning is an off-policy offline learning technique, it may learn from present experiences, prior experiences, and even other people’s experiences, the memory bank in the experience pool is utilized to learn previous experiences. As a result, adding prior experience to the neural network at random throughout the learning process will make it more efficient. As a result, the experience pool overcomes the problem of correlation and non-static distribution. It saves the transfer samples received by the agent interacting with the environment at each timestep in the playback memory network, and then randomly selects some (minibatch) to train while training, breaking the correlation. Q-targets are essentially a technique for disrupting correlation. When you use Qtargets, you’ll receive two networks in DQN with the same structure but different parameters. MainNet, a network for forecasting Q estimate, utilizes the most recent parameters., but the TargetNet parameters of the neural network that predicts Q reality have been in use for quite some time. TargetQ can be solved, and MainNet parameters can be modified based on LossFunction. The parameters of MainNet are transferred to TargetNet after a specific number of iterations. The target Q value remains fixed for a period of time after the introduction of TargetNet, which decreases the correlation between the present Q value and the target Q value to some extent and enhances the algorithm’s stability. 3.3 LSTM The LSTM (long short memory neural network) is a kind of RNN that can learn lengthy dependencies [10]. Hochreiter & schmidhub pioneered them, and many others refined and popularized them. They are currently extensively utilized and function well on a range of difficulties (Fig. 3).

96

S. Zhang et.al

Fig. 3. LSTM architecture

Long-term dependency is avoided with LSTM. It’s not what people study hard for, but it’s how they remember vast periods of history. Recurrent neural networks all take the shape of repeating modular neural network chains. The repeating module in a typical RNN will have a relatively basic structure, such as a single tanh layer. The state of the cell is crucial to LSTM, and the line that reflects the state of the cell runs horizontally across the graph’s top. The cell is in a similar state as the conveyor belt. The state of the cell is maintained over the whole chain, with only a few tiny linear actions allowing information to flow freely throughout the chain. The LSTM can remove or add information to the cell state, which is determined by the gate structure. A gate is a method for information to flow through that is optional. A sigmoid neural network layer and a point multiplication operation are included. The sigmoid neural network layer produces a value between 0 and 1, indicating how much data each component may flow through. No information is transmitted via LSTM if the value is 0, and all information is passed through LSTM if the value is 1. The condition of cells is protected and controlled by three gates. The initial stage in LSTM is to select whatever information from the cell’s state we wish to discard. A sigmoid layer termed “forget gate” implements this choice. It takes HT-1 (prior output) and XT (current input) as inputs and produces a number between 0 and 1 for each number in the CT-1 cell state (previous state). 1 denotes complete retention and 0 denotes complete deletion. 3.4 Convolutional Neural Networks A deep neural network with a convolutional structure is known as a convolutional neural network. The deep network’s memory use can be reduced thanks to the convolutional structure. The local receptive field is one of its three essential activities, and the other is weight sharing. The pooling layer, which lowers the amount of network parameters

Application of Artificial Intelligence

97

and alleviates the model’s over-fitting problem, is the third layer. Pattern identification, image processing, video analysis, and natural language processing are just a few of the disciplines where Convolutional Neural Networks (CNNs) have found widespread use. A multi-layer supervised learning neural network is a convolutional neural network. The convolutional layer and pool sampling layer of the hidden layer are the key modules that allow the convolutional neural network to perform its feature extraction function. The network model uses the gradient descent approach to minimize the loss function in order to reversely change the weight parameters in the network layer by layer, and iterative training is used to enhance the network’s accuracy. The convolutional neural network’s lower hidden layer alternates between a convolutional layer and a maximum pool sampling layer, while the top layer is a hidden layer and a logistic regression classifier, similar to a typical multi-layer perceptron. The feature picture generated by feature extraction of the convolutional layer and the subsampling layer is the input of the first fully connected layer. The last layer’s output layer is a classifier, which may categorize the input picture using logistic regression, Softmax regression, or even support vector machine. As may be seen in Fig. 2. A CNN is made up of convolutional, pooling, and fully linked layers. Unlike DNNs, the units in the convolutional layer are only connected to a portion of the units in the next layer, reducing the number of parameters substantially. In most cases, the pooling procedure occurs between two convolutional layers. The output of the convolution operation is sampled in the pooling process to gradually reduce the feature size. 3.5 DDPG The reinforcement learning process generally follows the Markov Decision Process (MDP). MDP is composed of a quaternary group of , where S (State) is the state space, which represents the set of state descriptions that the agent may exist in the environment. A (Action) is the action space, which represents the collection of action descriptions that the agent may take in the environment. P (Policy) is a transfer strategy. The agent in a certain state will choose action according to P, and then transfer from one state to another. R (Reward) is the reward, which represents the reward value obtained from the environment by the agent taking a certain action in a certain state. The aim of reinforcement learning is to find the best strategy P, perform a series of actions in the environment, and make the agent complete the given task with the best round return. Actor-Critic is an algorithm proposed by Vijay R. Konda and John N. Tsitsiklis to be applied in the Markov decision process. The algorithm consists of two parts, used to generate decision-making actions, the Actor part and used to perform actions. In the Critic part of the evaluation, Actor is an action generator, which takes the current state as input and outputs an action to be executed in the current state. Critic is an evaluator, that is, a value function generator. It uses the current state and the action taken by the actor to create a value amount that is used to weigh the advantages and disadvantages of the action taken by the actor. The DDPG (Deep Deterministic Policy Gradient) algorithm combines many characteristics of AC, PG, and DQN, and is the first to extend deep reinforcement learning to the continuous space domain. DDPG adopts the Actor-Critic frame structure as a whole. Both the Actor and Critic parts of DDPG are constructed by neural networks.

98

S. Zhang et.al

The networks of the two parts adopt the design idea of DQN respectively, which are two time-differential networks. When Critic is updated, the update method of policy gradient is adopted. Unlike the traditional policy gradient, DDPG adopts a deterministic strategy for action selection.

4 Application of Reinforcement Learning for Space-Air-Ground-Sea Integrated Network 4.1 Network Control Based on Reinforcement Learning The control method of the air space integrated network is one of the most critical aspects that directly impact network performance. For RL, the classic single agent RL and multi-agent RL (RL, multi-agent RL) correspond to two classical control architectures: centralized and distributed, respectively [11]. In the space earth integrated network, distributed management architecture can seamlessly match heterogeneous network slices, and configure local control intelligence for each slice or even specific network nodes (such as LEO satellites), so as to reduce control signaling bottleneck and response time. But at the same time, mal must rely on the cooperation and information sharing among agents to ensure the optimization of the whole network. This kind of cooperative operation will increase the complexity of MAL interaction mechanism and network. On the other side, centralized control mode can make network structure and RL algorithm deployment easier, but considering the huge difference between response delay and coverage of different network elements in the space earth integrated network, how to deploy the control center and how to coordinate and synchronize different network elements will significantly affect the network performance. Therefore, in the future, the hierarchical hybrid control architecture which integrates the two control methods will become the mainstream, in order to enhance the adaptability of the integrated network to heterogeneous complex environment. In this paper, the hierarchical controller is used to distribute the control behavior on different time and space scales. The behavior of small-scale fine control (user access to a base station or UAV) is left to the bottom controller to make local decisions, while the behavior of large-scale macro control (how much satellite spectrum resources are allocated to a certain area for user access) is not. The strategy of the bottom controller is directly adjusted by the upper controller according to the user distribution and environmental changes fed back by the bottom controller. This hierarchical hybrid control architecture can decompose the complex strategy into hierarchical DRL of sub objective hierarchical decision control. At present, the most important application of RL is the trajectory control of UAV. Unmanned aerial vehicle (UAV) is an intelligent network device whose mobile mode is completely controlled and whose deployment position and trajectory can be adjusted according to specific needs. RL framework is naturally suitable for UAV control and trajectory planning. RL based trajectory planning for UAVs has been widely studied by relevant scholars in recent years. As a representative, document uses deterministic policy gradient (DPG) method to plan the trajectory of a single UAV to maximize user throughput; In reference, a distributed RL framework was designed for trajectory planning of multiple UAVs based on the self-developed perception and transmission protocol.

Application of Artificial Intelligence

99

Another new research point of RL in the field of space earth integrated network control is to optimize the access and handover mechanism of LEO satellite network. Different from the ground base station, UAV and other access nodes with slow moving speed and limited coverage, LEO satellite has the characteristics of fast moving speed, large coverage, long transmission delay and frequent handoff. Therefore, the access or handoff mechanism of traditional ground-based network can not be applied to users accessing LEO satellite network. On the other hand, the large coverage area and frequent handoff frequency lead to the highly dynamic and changeable environment of LEO satellite network, which is difficult to be modeled as a simple mathematical model. Therefore, RL is needed to optimize the access handoff strategy. As more and more researchers begin to pay attention to LEO satellite network, the research in this area is expected to burst out. For example, in reference, according to the remaining service time and historical signal quality data, the method of Q-learning is used to select the Leo to be accessed in the next hop; According to the current user terminal distribution and channel allocation status, reference learned dynamic satellite channel selection strategy through deep Q-learning (DQN) including convolutional neural network (CNN). 4.2 Resource Allocation Based on Reinforcement Learning As the heart of the research on space earth integrated network, communication, computing resource allocation and resource slicing strategy have been facing the challenge of complex heterogeneous network environment. The introduction of RL method can bring the following two advantages for heterogeneous resource allocation of space earth integrated network [12]. First of all, RL method can learn the accurate trend of traffic or user movement, so as to customize the dynamic resource allocation strategy. Only based on this accurate trend of environmental change can controllers at different levels effectively allocate communication and computing resources or partition network slices. Different from DNN, LSTM (long short term memory) and other supervised learning methods, RL does not directly output the prediction results of the next network environment change, but directly outputs the optimal strategy for the next step after connotative learning of the change trend. Therefore, RL can be said to be a more intuitive and easy to deploy integrated resource allocation solution. In addition, RL can flexibly define the time step and control precision of the next strategy, so as to adapt to the controller in different levels and different scenarios. On the other hand, a RL neural network which has been trained for a long time to reach the convergence state can obtain the resource allocation strategy in complex environment with very low computational complexity. In this highly complex and heterogeneous environment, RL is a fast response and efficient resource allocation scheme compared with the traditional method with super high computational complexity. In the current research of space earth integrated network, there are many achievements in resource allocation based on RL. Most researches usually use RL method to jointly allocate a variety of resources including transmission power, spectrum bandwidth, computing capacity, etc., and some also combine the characteristics of space-based and space-based platforms to jointly optimize UAV trajectory, LEO satellite orbit and switching characteristics. At present, most of the researches are only concerned with the joint

100

S. Zhang et.al

resource optimization of air ground or sky ground integrated network. It is an important research direction in the future to really consider the joint resource optimization of the three kinds of network. 4.3 Network Access Selection Based on Reinforcement Learning This approach, like dispersed deployment, makes it difficult to create real-time responsiveness to the user environment. However, in some densely inhabited district, multiple networks will have overlapping coverage, and the performance of the network and the user experience would be greatly affected by users accessing various networks.. At the same time, space-based, air-based and ground-based networks have different spectrum resources, access technologies and protocols. Therefore, user access selection, that is, to improve network performance by optimizing the user’s access network, has become the top priority in the field of air, space and ground integrated network research. Unlike traditional network handover (the goal of handover is generally to maintain service continuity), the goal of network access selection (RAT selection, radio access technology selection) is to optimize network performance in real time. Therefore, the implementation method is also changed from the (passive) handover triggered by the location change to the active selection, that is, the user’s access network decision is made in each time slot. This type of network access selection problem is also commonly referred to as a user association problem. Adopting the traditional network access selection strategy based on the optimization method in the air, space and ground integrated network will face the following two challenges. First of all, most user assignment problems will eventually be constructed as an integer or mixed integer combinatorial optimization problem. This type of problem is not only non-convex, but also usually proved to be NP-hard. Using optimization methods to solve such problems will result in a large amount of calculation and a long calculation time, and cannot cope with the large-scale and high-complexity environmental characteristics of the air, space and ground integrated network. On the other hand, optimization-based methods highly rely on prior knowledge of network topology and modeling assumptions (such as network topology model, user distribution model, user mobility model, channel characteristic statistical model, service arrival model, etc.). These models Whether it is large-grained network behavior modeling or modeling only for special network scenarios, it cannot satisfy the requirements of the air, space and ground integrated network at this stage, which reduces the effectiveness of the optimization results. Different from the optimization-based method, the RL method provides a way to learn the unknown network environment based on “observation and trial and error” without having to preset any prior models. In addition, the RL neural network that reaches the convergence state after a period of operation can theoretically fit a highly complex network environment and ensure that the optimization results are output at an extremely fast speed, thereby realizing real-time network access selection with low computational complexity. RL-based network access selection is currently in the initial research stage. According to the deployment location of the RL agent, network access options are mainly divided into two types.

Application of Artificial Intelligence

101

1) The RL agent is deployed on each user, and the user makes a purely distributed access selection [13]. This method can quickly respond to changes in the user’s current environment, and reduce the signaling overhead of data collection by performing data collection and decision-making locally on the user. However, due to the limited ability of a single user to observe the environment, this distributed approach is difficult to achieve joint optimization of multi-user access selection strategies on a large scale. 2) RL agents are deployed on access nodes (base stations, drones) or edge controllers to facilitate multiple users to share specific access resources. This centralized deployment method can easily perform joint optimization for many users in a large area, but is limited by the wireless transmission data and signaling overhead in the process of user data collection and decision-making distribution. This approach, like dispersed deployment, makes it difficult to create real-time responsiveness to the user environment.

5 Conclusion This paper introduces the space earth integrated network. The space earth integrated network can integrate large-scale multi-dimensional heterogeneous network, effectively utilize various resources, meet the needs of future network for expanding service space, improving QoS, mass data processing, etc., greatly improve the network information service capability, and enable it to realize wide area high-speed access, disaster prevention and mitigation, aerospace information support and other services, Ensure the strategic needs of the country, people living and working in peace and contentment, as well as promoting rapid progress in all fields. Finally, the application of reinforcement learning technology in the integrated network of space, earth and sea is introduced in detail.

References 1. Shafi, M., Molisch, A.F., Smith, P.J., et al.: 5G: a tutorial overview of standards, trials, challenges, deployment, and practice. IEEE J. Sel. Areas Commun. 35(6), 1201–1221 (2017) 2. Yu, J., Liu, X., Gao, Y., Shen, X.: 3D channel tracking for UAV-satellite communications in space-air-ground integrated networks. IEEE J. Sel. Areas Commun. 38(12), 2810–2823 (2020) 3. Zhang, N., Zhang, S., Yang, P., et al.: Software defined space-air-ground integrated vehicular networks: challenges and solutions. IEEE Commun. Mag. 55(7), 101–109 (2017) 4. Ye, J., Dang, S., Shihada, B., et al.: Space-air-ground integrated networks: outage performance analysis. IEEE Trans. Wirel. Commun. 19(12), 7897–7912 (2020) 5. Lyu, F., Wu, F., Zhang, Y., et al.: Virtualized and micro services provisioning in space-airground integrated networks. IEEE Wirel. Commun. 27(6), 68–74 (2020) 6. Cheng, N., Quan, W., Shi, W., et al.: A comprehensive simulation platform for space-airground integrated network. IEEE Wirel. Commun. 27(1), 178–185 (2020) 7. Liu, J., Shi, Y., Fadlullah, Z.M., et al.: Space-air-ground integrated network: a survey. IEEE Commun. Surv. Tut. 20(4), 2714–2741 (2018) 8. He, X., Wang, K., Huang, H., et al.: Green resource allocation based on deep reinforcement learning in content-centric IoT. IEEE Trans. Emerg. Top. Comput. 8(3), 781–796 (2018)

102

S. Zhang et.al

9. Mao, B., Fadlullah, Z.M., Tang, F., et al.: Routing or computing? The paradigm shift towards intelligent computer network packet transmission based on deep learning. IEEE Trans. Comput. 66(11), 1946–1960 (2017) 10. Tang, F., Mao, B., Fadlullah, Z.M., et al.: On removing routing protocol from future wireless networks: a real-time deep learning approach for intelligent traffic control. IEEE Wirel. Commun. 25(1), 154–160 (2017) 11. Kato, N., Fadlullah, Z.M., Tang, F., et al.: Optimizing space-air-ground integrated networks by artificial intelligence. IEEE Wirel. Commun. 26(4), 140–147 (2019) 12. Wu, H., Chen, J., Zhou, C., et al.: Resource management in space-air-ground integrated vehicular networks: SDN control and AI algorithm design. IEEE Wirel. Commun. 27(6), 52–60 (2020) 13. Cao, Y., Zhang, L., Liang, Y.C.: Deep reinforcement learning for multi-user access control in UAV networks. In: 2019 IEEE International Conference on Communications, ICC 2019. IEEE (2019)

Machine Learning Based 5G RAN Slicing for Channel Evaluation in Mobile State Fangpei Zhang1 , Xiaojun Jing2 , Junsheng Mu2(B) , Jia Zhu2 , and Bohan Li3 1 Information Science Academy of China Electronic Technology Group Corporation,

Beijing 100086, China 2 School of Information and Communication Engineering, Beijing University

of Post and Telecommunications, Beijing 100876, China [email protected] 3 University of Southampton, Southampton, UK

Abstract. The number of mobile broadcast users has increased significantly because of evolved multi-media broadcast multicast services (eMBMS). According to the above reasons, machine learning technology is introduced into 5G RAN slices to forecast the state of communication channel in a mobile scene. We propose a new architecture not only including convolutional neural network (CNN), but also fusing long short-term memory network (LSTM) to implement channel estimation, simultaneously demonstrate the performance of our proposed scheme with simulation results. Keywords: Machine learning · 5G RAN slicing · Channel estimation

1 Introduction Under the background of the increasing demand for mobile communication, how to communicate reliably, quickly, and with a low delay has become an urgent problem to be solved. Because of the flexibility of its function, network slicing is adopted by 5G operators to better adapt to the various needs of mobile networks. We consider two types of segments in 5G RAN, not only including ultra-Reliable low-latency Communications (uRLLC), but also including enhanced Mobile Broadband (eMBB). Based on the obtained channel states, various types of resources can be reasonably allocated to the base station. First, to build a two-dimensional mobile user data set, we should collect multiple signals. Then, CNN-LSTM networks are constructed by combining sequence information and spatial information, on the basis of information, the channel dynamics can be estimated and predicted. Finally, based on the channel information, by using DQN the radio resource allocation can be optimized and the best energy can be obtained. The validity of the proposed scheme is verified through simulation calculation.

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 S. Sun et al. (Eds.): ICSINC 2021, LNEE 895, pp. 103–107, 2022. https://doi.org/10.1007/978-981-19-4775-9_12

104

F. Zhang et al.

The key contributions of the paper are summarized below. (1) We studied 5G RAN network slices for broadcast services. Accordingly, there is currently no solution addressing the physical layer for broadcast applications, which is used to sustain network slicing. (2) A new architecture is proposed, not only including CNN, but also fusing LSTM, and the channel state of mobile users can be estimated. We use the CNN module to extract spatial features in the channels and the LSTM module to extract time series. (3) We develop RRM based on DQN for 5G RAN network slice to employed broadcast service performance in mobile computing environment. In particular, energy efficiency is expressed mathematically under certain data rate and delay constraints. The rest of the paper is arranged as follows. Section 2 presents related work of other researchers and highlights the uniqueness of our contribution to review related work. Section 3 presents the overall of the proposed system model and completes the target problem by mathematical derivation. Section 4 demonstrates the performance of our designed system through simulation studies. Finally, we conclude the paper in Sect. 5.

2 Related Work Based on virtualization technology, network slicing enables multiple logical network projects to share the physical network infrastructure. Network slicing (NS) maintains the independence of each logical network while allowing for network features such as bandwidth, latency and capacity. Network Function Virtualization (NFV) migrates network function from dedicated hardware to virtual machines on general equipment by using virtualization technology. In addition, these independent network functions can be customized to the user’s needs, including network resources, compute resources and storage resources. Network slicing is a key technology for 5G, it can not only be combined with 5G, but also applied to traditional networks. With the advancement of computing ability and NFV technologies, network slicing technology attracts more and more attention. NS can meet the service requirements of different types applications. In [1], the authors introduce NS from the perspective of the development of NS. In [2], the authors propose a new generic framework that NS can adopt diffusely.

3 System Model The main of our work is devoted to the downlink of a 5G broadcasting network in which a number of N base stations and user equipment (UE) are randomly put up. To represent the placement of BSs, A Homogeneous Poisson Point Process (HPPP)  with an arrival rate of λb is implemented. In addition, an HPPP u with a density of λu represents UEs deployment, and it is independent of the process b . B represents the system bandwidth, and Pb represents each BS power. [t, t + 1) defines the time slots where the period of τ for each slot is t. It connects each BS that owns data queue buffer.

Machine Learning Based 5G RAN Slicing

105

Fig. 1. A framework that combines CNN and LSTM.

The network resources will be partitioned accordingly to the demand of each user. We assume that there are two types of slices, respectively named uRLLC and eMBB. uRLLC and eMBB are the different slices in the network, respectively. The bandwidth coefficient assigned to Slice s (s = 1, 2) is represented as ps (t), and the power coefficient assigned to Slice s (s = 1, 2) is represented as bs (t), where 0 < ps (t) < 1, 0 < bs (t) < 1. Suppose the sensed signal y(n) at the base station of user consists original signal s(n) and additive white Gaussian noise (AWGN) x(n) then. y(n) = h(n)s(n) + x(n),

(1)

where h(n) indicates channel gain and |h(n)| is Nakagami distribution. s(n) and x(n) 2 , σ2 are independent identically distribution, and their mean zero and variance are σs−i x−i respectively (iid). For the process of the proposed method, the signal enters the channel and undergoes a series of processes to obtain the corresponding channel characteristics [3]. First, the input two-dimensional signal data is derived from the received time domain one-dimensional signal by frequency transformation. Then, the convolutional neural module is used to extract the spatial information of the input signal data. In addition, the total number of parameters of the global model is reduced by weight sharing to reduce the computational effort. The three convolutional layers can obtain more extensive spatial feature of the signal. The signal processed by the convolution layer can be expressed as nin [Xk ∗ Wk ][i, j] + b, (2) S[i, j] = k =1

where Wk is the k-th kernel matrix of sub-convolution,  Xk indicates the corresponding k-th input of the model, the bias of CNN is b, and S i, j signifies element value of the output matrix.

106

F. Zhang et al.

4 Simulation and Analysis

Fig. 2. The accuracy comparisons under various condition.

The parameters used in our simulations are shown in Table 1, the basis of our simulations is the NR-MBMS system described in [4]. We generate experimental data in the simulation platform, which mainly includes the number of users of each base station, Table 1. The simulation parameters Parameter

Value

Bandwidth

B = 100 MHz

Power

Pb = 40 × 103 mW

Slot

τ = 1 ms

Pass loss exponent

α=4

Noise power

α 2 = 10−17.4 mW

Rate constraint

r1 = 900 Mbps

Delay constraint

D2 = 2.5 ms

Cell radius

R = 500 m

User number of slice 1

N1 = 30

User number of slice 2

N2 = 70

Machine Learning Based 5G RAN Slicing

107

each user’s geographical location, the distance between the base station and the user, and the user azimuth obtained from associated based station. The details of generating data are described as below. The simulation band is performed at ultra-high frequency (VHF), where the setting of carrier frequency is 7.0 × 108 Hz and fs = 1.6 × 109 Hz is the receiver sampling frequency. To deal with the subsequent simulation process, first we generate the OFDM signal, and then we add Gaussian white noise to the signal with mean zero and variance 1. Next, we normalize the received signal in terms of energy and estimate the channel state estimation. Finally, the base stations energy efficiency is optimized by jointly allocating the RPM. We calculate the total number of base station traffic data at any given point in time.

5 Conclusion In this paper, a cascaded CNN-LSTM network is proposed which is able to reckon and forecast the channel state of mobile broadcast users. Also, the RAN slicing function is implemented by using a CNN module to abstract the spatial features of the signal and an LSTM module to extract the temporal features. Then, the spatial and temporal information is integrated into the prediction network based on deep neural networks. In addition, we build a base station model, mainly considering the energy efficiency, and optimize the state information of the channel by using DQN method, and get a better solution. Finally, a simulation study is performed on simulated data to verify the accuracy and efficiency of our proposed algorithm.

References 1. Samdanis, K., Costa-Perez, X., Sciancalepore, V.: From network sharing to multi-tenancy: the 5G network slice broker. IEEE Commun. Mag. 54(7), 32–39 (2016) 2. Foukas, X., Patounas, G., Elmokashfi, A., Marina, M.: Network slicing in 5G: survey and challenges. IEEE Commun. Mag. 55(5), 94–100 (2017) 3. Afolabi, I., Taleb, T., Samdanis, K., Ksentini, A., Flinck, H.: Network slicing and softwarization: a survey on principles, enabling technologies, and solutions. IEEE Commun. Surv. Tuts. 20(3), 2429–2453 (2018) 4. Zhang, S.: An overview of network slicing for 5G. IEEE Wirel. Commun. 26(3), 111–117 (2019)

Use Case Analysis and Architecture Design for 5G Emergency Communications Hongbiao Jiao1(B) , Jijiang Hou1 , and Chengli Mei2 1 China Telecom Corporation Limited, Beijing, China

[email protected]

2 China Telecom Research Institute, Beijing, China

[email protected]

Abstract. 5G has significant advantages in emergency communications. If it is directly deployed in the disaster area, there are still some disadvantages such as large data traffic and long service delays. This paper analyzes the needs of public safety network (PSA), and studies the application of dynamic messaging delivery, network slicing, C-RAN, and D2D in 5G PSA. Then, this paper puts forward two 5G emergency rescue solutions, one is the portable 5G private network and 5G public network collaboration, and the other is the public network UPF sinking. Finally, we compare these two solutions according to the needs of the emergency management department. The conclusion is that the portable 5G private network and 5G public network collaboration solution can be independently deployed locally. It does not have the limitations of the 5G public network UPF sinking solution but has the same network functions, which is more in line with the requirements of the emergency management department. Keywords: Emergency communications · Public safety network (PSA) · Lightweight 5GC · 5G private network · UPF sinking

1 Introduction Emergency communication is generally used in scenarios such as floods, earthquakes, forest protection, urban fire rescue, emergencies, and major security. In order to get timely and scientific rescue of emergency events and protect people’s lives and properties from losses, it is necessary to quickly establish an emergency communication dispatching and command system to help scientific decision-making, multi-party consultation and visual command of emergency events. 5G has a broad application prospect in the field of emergency communications with its larger bandwidth, lower latency, and more stable and reliable network capability. The high bandwidth of 5G network can be used to transmit high-definition video from the emergency site to the command center in real time, and to realize multiparty video of high-definition video through 5G communication network. It is also possible to use the 5G slicing capability to directly download the service stream locally, and the onsite emergency video interaction does not need to interact through the remote core © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 S. Sun et al. (Eds.): ICSINC 2021, LNEE 895, pp. 108–119, 2022. https://doi.org/10.1007/978-981-19-4775-9_13

Use Case Analysis and Architecture Design

109

network, which can make the on-site command more efficient. Using 5G network low latency features to provide end-to-end data transmission services, 5G low latency is conducive to real-time control of rescue robots and rescue UAVs during emergency rescue, and remote control of rescue equipment in the command center through the 5G communication network. Many departments that need to make emergency calls have high requirements for the security of mobile communication network architecture and prefer to use a separate dedicated emergency communication network than the congested and delayed public network. According to the needs of emergency communication, by cutting out unnecessary network elements in the 5G core network of the public network, and connecting their interfaces to the 5G network elements required for emergency communication through direct connection or co-location, a customized 5G lightweight core network for emergency communications can be formed. The ease of deployment and low consumption of the 5G lightweight core network will play a critical role in emergency communications. The customized 5G lightweight core network can quickly build an emergency communication system at the communication site to cover the entire disaster area, thereby ensuring on-site communication scheduling and helping the affected people communicate with the outside world. Emergency communication vehicles play an important role in traditional rescue and disaster relief and emergency handling. It is currently the most typical and most extensive emergency communication system platform. It has the characteristics of quick deployment and opening, long working hours, convenient scheduling, and convenient access to existing communication networks. Emergency communication vehicles will still be a key means of emergency communication guarantee in the future. UAV communication systems have gradually emerged in the process of continuous development of UAV technology. UAVs equipped with “air base stations” have become an important guarantee for emergency communications. When disasters occur, a new communication system can be quickly created through UAV communication systems, so as to quickly contact the disaster area, so that the disaster relief work can be carried out smoothly. In view of the high speed, high reliability, low latency and low power consumption of 5G communication network and the rapid development of drone cluster technology, the intelligent networking technology of drones and 5G communication technology are integrated to achieve rapid air-to-ground Communication to solve the problems of the existing emergency communication system becomes possible. However, the emergency communication vehicle cannot enter the disaster area in time to provide communication services on the spot when the road is interrupted. UAVs can quickly gain access to the network in areas where emergency communication vehicles cannot reach. They are one of the most powerful communications guarantee tools in disaster scenarios with large areas affected by earthquakes, floods, and mudslides and severe road damage. Although UAV communication has the characteristics of rapid deployment and flexible mobility, it also has obvious disadvantages, such as short battery life, small coverage area, and high flight cost. Moreover, UAV communication is also vulnerable to extreme weather such as severe convective weather, icy plateaus, and marine environments. In addition, UAVs can also be used to airdrop 5G base station equipment. By dispatching UAVs to fly to the target area and accurately throw base

110

H. Jiao et al.

station equipment, the emergency communication support capability is significantly improved. Even if the airdrop target area is severely affected, UAVs can also operate at ultra-low altitude and throw base station equipment. The structure of this article is as follows. We introduce the basics of 5G for public safety networks and emergency communications in Sect. 2. In Sect. 3, we focus on China Telecom’s 5G on-site emergency solutions, including two major categories: public network mode and private network mode. We will give specific solutions for each type of mode, and in-depth analysis of the corresponding system capabilities and resource requirements. Finally, we conclude the full text in Sect. 4.

2 Basics of 5G Public Safety Network Emergency communications are an important communication method to ensure emergency relief and essential communications, mainly for major natural disasters and sudden emergencies, and there is a surge in the demand for communications at certain specific times. These networks that function in special times are formally called public safety networks (PSN). In modern society, with the changing emergency emergencies and the rapid development of mobile communication devices, PSNs will provide not only voice services, but also instant video services. In specific emergency situations, it is difficult for the communication devices of rescue teams to meet the communication needs, so continuous and stable communication becomes an important factor limiting the emergency response, and heterogeneous cloud radio access network (HCRAN) has been considered as an important area to facilitate the rapid development of the 5th Generation Mobile Communication Technology (5G). In HCRAN is no longer bounded by the traditional cellular network technology, the introduction of high-power node (HPN) technology allows massive coverage of signals. Also, Radio Remote Units (RRU) will support emergency communications, which allows a much higher local signal coverage and network transmission rate. At the same time, all centralized management units control the allocation of resources centrally. In case of emergency, the H-CRAN can dynamically allocate resources to the required units. This allows a more efficient and flexible use of limited network resources. The stability of emergency communications depends on the communications infrastructure, and natural disasters of varying magnitude will disrupt the infrastructure, causing communications outages and rendering emergency services inoperable. And cyberattacks launched by some malicious actors can also disrupt critical functions or reduce the efficiency of response operations. In emergency situations, first responder teams may rely on the Public Safety Network (PSN). Also, to ensure the stability and reliability of information transmission, different methods can be used such as widening the signal coverage of base stations (BSs), changing the deployment of BSs equipment, and using coordinated multipoint (CoMP) to improve signal quality. These schemes achieve increased coverage area at the expense of energy consumption, at the expense of flexibility, and with limited compensation. There are many other methods to improve communication quality and network coverage, and relaying is often used to extend the coverage of BS, such as airborne base stations (ABS) that can be deployed in large post-disaster scenarios, but it leads to long

Use Case Analysis and Architecture Design

111

deployment times and high network costs. In addition, there is no effective method to meet the needs of the increasingly demanding emergency communication field. Building on existing PSNs, many other types of PSNs have been established to provide first responders with sufficient information and resources to better accomplish rescue missions. In this case, a successful PSN management can provide fast communication services to first responders. Figure 1 depicts a specific example of a PSN where the connectivity of these devices is general in day-to-day situations (see Fig. 1a). while Fig. 1b depicts a PSN deployment in the event of a disaster. The PSN in special times (see Fig. 1b) utilizes the existing communication infrastructure of the PSN (see Fig. 1a), such as vehicles, drones, sensors, existing base stations, etc. In this way, all the communication devices required for the emergency scenario will function in the PSN.

Fig. 1. Using public mobile communications to support emergency management in a Public Safety Network (PSN)

Most of the previously proposed contingency scenarios will be based on already existing network communication protocols for wired or wireless networks, the latter

112

H. Jiao et al.

mainly focusing on Wi-Fi, 4G or LTE technologies. In order to better achieve scenario efficiency, all the above proposals will be based on LTE, a broadband communication technology considered as a 5G pioneer [1], reevaluated in terms of integration. Based on this, this paper proposes a base station deployment scheme based on UAS [2]. From the development of the technology, the heterogeneous system will provide security for the public network, including an LTE core network with multiple subnetworks and layered services [3]. The following uses and requirements will be the main considerations for the authorities in the concrete implementation of building PS-LTE networks [4]. (1) smart device terminals to ensure network security for users and operators; (2) quality of service (QoS) assurance and signal transmission quality; (3) convergence of commercial and private networks; (4) combination of fixed and mobile to ensure service continuity and efficiency; (5) development of a dedicated LTE network construction policy, taking into account the limited budget, to guide the extension of coverage from economically developed cities to economically less developed areas. 2.1 Application of Dynamic Message Provision in 5G Public Safety Network It is well known that effective communication is a key issue for successful management of PSNs. The joint application of Network Function Virtualization (NFV) and Software Defined Networking (SDN) technologies can effectively solve this problem by providing an effective communication capability for managing PSN services [5]. And related fields propose a TCP/IP network based on the converged application of NFV and SDN for resilient operation of 5G networks and protection against malicious network attacks, respectively, as in [6, 7]. In emergency situations, PSNs with strong autonomous capabilities allow the deployment of communication resources as VNF services in case of base station collapse. For example, in [8], an emerging cognitive intelligence engine is proposed to enable real-time adaptation of communication resources through VNF services. In the case of damaged base stations, where virtualization is not possible, this scheme allows sending unmanned aircraft, such as drones (UAVs) to areas in emergency situations to provide deciphering operations [9]. In addition, it is possible to establish device-to-device (D2D) communication between the user’s devices and to restore basic communication by establishing a mobile Ad hoc network (MANET) through a multi-hop connection. This has been extensively discussed in the literature [10] to enable real-time data transmission and feedback with the help of D2D communication and the mobility of UAVs. Security and privacy protection issues must also be considered when deploying PSNs, such as personal information used by search and rescue teams, which cannot be accessed by other unrelated personnel. This point is highlighted in the related agenda [11], which emphasizes that the relevant authorities should further strengthen the regulation to ensure the reliability of data in emergency situations. 2.2 Application of Network Slicing in 5G Public Safety Network Network slicing is a technology specifically designed to customize scenarios with heterogeneous and complex requirements. This technology enables on-demand management of network resources and services to enhance SDN/NFV environments, which allows

Use Case Analysis and Architecture Design

113

operators to create multiple virtual networks as overlay networks from a single communications infrastructure. That is, NFV and SDN-based technologies allow the application of virtual resources to specific networks, and network slicing technology, a new 5G technology, will address potential challenges in 5G networks through the architecture of existing technologies, as detailed in. In addition, enabling and maintaining secure communication between two parties by defining specific network slices is an important application of network slicing in the security domain, for example, using IPsec to protect user-to-user connections, while another network interface will be used to maintain interface connections for mobile devices. Table 1 summarizes the advances and opportunities for PSNs. 2.3 Application of C-RAN in 5G Public Safety Network The architecture of a cloud radio access network (C-RAN) applied to a PSN is shown in Fig. 2. The architecture specifically consists of a radio remote unit (RRU), a baseband unit (BBU) and a central management unit (CMU). The BBU pool centralizes all the network resources and the CMU is responsible for managing them. the CMU directly manages and assigns them to the RRU regardless of the link between the user and the RRU, which can make the network operation more efficient and flexible. However, if an eNB loses service in an LTE or LTE-A network, the resources there will be lost. However, in a C-RAN-based PSN, there are links between BBUs and RRUs are responsible for sharing. Even if the RRU is out of service, other network nodes can continue to use these resources.

Fig. 2. Outage compensation in C-RAN based PSN

As shown in Fig. 2, RRU0 stops working due to various reasons. RRU1, RRU4 and RRU5 are selected to participate in the compensation. RRU1 and RRU4 transmit the same resource to UE1 at the same time and UE will be responsible for receiving the signal. Then the transmitted signals from RRU1, RRU4 and RRU5 will be reinforced to cover other areas.

114

H. Jiao et al.

2.4 Application of D2D in 5G Public Safety Network In the event of a disaster, network communication may be difficult to maintain due to various complex reasons, and using other devices to compensate power usage for this can effectively relieve communication pressure, which will fully utilize the function of the PSN [12], and D2D technology, a common communication enhancement technique, has long been proven in related fields [13–15], and various compensation mechanisms will make D2D communication even better. D2D communication will use area resources to satisfy the terminal communication.D2D technology can implement different types of communication in 5G PSNs: resource communication between HPNs, resource reuse in a single HPN, extended communication between PN overlay and external [16], and self-organized communication without HPNs. The H-CRAN architecture will consist of HPNs, baseband units, central management units, etc., and the specific structure is shown in Fig. 3 错误!未找到引 用源. The CMU manages various resources in the BBU pool in a unified manner. The CMU enables efficient resource allocation. The advantages of D2D are: high speed, low latency, low load, efficient resource utilization, and extended coverage. Since HCRAN has the advantage of efficient resource allocation, the expanded coverage of D2D signal will effectively meet the user network demand. And D2D is used to realize efficient emergency communication scenarios.

Fig. 3. Device-to-Device (D2D) technology and the application in PSN

In the first hour after the disaster, rescue teams and equipment vehicles will arrive at the scene, these equipment and personnel are called User Equipment (UE), after the completion of information collection, these information about the terrain, weather, damage and other information will be fed back to the PSN command and dispatch center, with D2D technology, the signal coverage further expanded, the network load is reduced, signal quality and transmission rate greatly improved thus This helps the command center to complete the initial judgment of the disaster situation. In Fig. 3, the user’s communication service will not only rely on HPN technology, but also D2D technology

Use Case Analysis and Architecture Design

115

will provide emergency services in emergency situations and achieve continuous and stable communication service through related technologies.

3 Emergency Communication Solutions Based on 5G 3.1 Portable 5G Private Network and 5G Public Network Collaboration Solution The basic hardware requirements of the program will include: local emergency configuration 5G RRU, BBU, lightweight core network, emergency service platform, satellite terminal and other equipment. The portable 5G private network and 5G public network cooperation scheme has the following system capabilities: realize local 5G public network coverage; realize local high-speed command and dispatch and audio/video transmission; can connect to the Internet (or point to a private network address) via satellite link; and serve both public and private network users. When conducting emergency communications in the 5G lightweight core network, it is important to prevent the current communication network facilities from being damaged in order to achieve the normal application of special communications in the independent communication network in the temporary emergency communication network. The advantages of easy deployment and low consumption that portable 5G private and 5G public networks possess will play a key role in disaster relief efforts. In the initial moments, the original infrastructure in the disaster area including roads, power facilities, and communication base stations will be damaged to varying degrees, so emergency communication in the initial moments is the key to on-site command and control. The local public network in the disaster area should be restored as soon as conditions permit, and a one-time transmission network should be activated to ensure the transmission of information between the disaster area and the outside world. Due to the uncertainty of the situation in the disaster area, there is a chance that the local 5G public network core base stations in the disaster area will not be able to connect to the outside world, resulting in information interruption. At this time, professional equipment such as drones can be used to build a professional 5G emergency network to quickly restore the local public 5G mobile communication network. The requirements for the portable 5G private network and 5G public network cooperation program are as follows: prototype product upgrade, custom R&D 5G equipment, emergency service platform, interference avoidance with local telecom, high-throughput satellite links, emergency ground dedicated line resources, telecom commercial core network, and technical service contracts with suppliers. In addition, the portable 5G private network and 5G public network synergistic solution can also quickly set up emergency communication system at the communication site. After a disaster, it is highly likely that the local public communication network in the disaster area will be directly paralyzed and cannot assume the role of remote emergency communication dispatch, and the public network facilities cannot communicate, so a temporary communication system that can quickly connect to the outside world is needed, and this communication network should have very high mobility, flexibility and stability. With the help of emergency communication transmission equipment, portable 5G network base stations and tethered drones, a lightweight 5G core network capable of covering the entire disaster area for emergency

116

H. Jiao et al.

information management can be built, thus ensuring on-site communication dispatch and helping the affected people to communicate with the outside world (Fig. 4).

Fig. 4. Portable 5G private network and 5G public network cooperation

3.2 Public Network UPF Sinking Solution The basic hardware requirements of the program will include: local emergency configuration 5G RRU, BBU, UPF, emergency service platform, etc. The portable 5G private network and 5G public network cooperation has the following system functions: realizing local 5G public network coverage; realizing local highspeed command and dispatch, audio and video transmission, which can be connected to telecom commercial core network via satellite link to serve public network users and private network users. Emergency rescue teams should further integrate heterogeneous networks such as public networks, satellites, broadband self-assembling networks, cluster intercom and short-wave communications, etc., and conduct integrated dispatching of multi-source data such as voice, video, SMS, etc. Therefore, the solution is highly adaptable and flexible on site. In summary, the emergency communication and networking problem of large area “air, sky and ground” three-dimensional communication network integration is imminent. Large area “air, sky and ground” three-dimensional communication network mainly refers to the use of space satellite systems, helicopters and UAVs, ground base stations for 5G emergency communications, helicopters can be used for communication relay, UAVs can be used to build 5G/ultra-short wave/self-assembling network communication relay, three-dimensional modeling, disaster assessment, expand the command and control network and search and rescue range, to open up the last mile of rescue. Finally, in the entire communication network, based on 5G communication technology, integrate drones, helicopters, satellites and other emergency rescue space resources, real-time convergence of meteorology, terrain and other types of important information, build interoperable, interoperable and complete emergency rescue command platform, and ultimately achieve a high degree of integration of command and dispatch and information interaction to improve the scientific and intelligent level of rescue. The requirements of the 5G public network UPF sinking program for resources are as follows: prototype productization and upgrade, UPF equipment supporting with the core network.

Use Case Analysis and Architecture Design

117

Emergency service platform, interference avoidance with local telecom, highthroughput satellite links, emergency ground resources, telecom commercial core network, technical service contracts with suppliers, etc. Based on the application of 5G communication technology and digital technology, analysis and research to achieve real-time monitoring of environmental parameters in the field. In the existing communication system, it is difficult to transmit the detection data effectively due to the limitation of bandwidth and speed. After entering the 5G era, the above limitations will be broken, and more detailed data will be transmitted in a timely and effective manner, such as monitoring real-time terrain changes in the disaster area, observing weather conditions at any time, and updating the location of rescue workers. In addition, with 5G communication technology, various types of real-time data at the scene can be transmitted back to the command terminal and form visual images, which can then be used to command and control rescue operations according to the actual situation (Fig. 5).

Fig. 5. Public network UPF sinking solution.

After R&D testing in 2019–2020, China Telecom completed the basic technology verification and principal function test. 2021–2022 is planned to evolve the basic technology to the portable 5G private network and 5G public network cooperation scheme and 5G public network UPF sinking solution. China Telecom’s 5G emergency solution evolution is mainly divided into two directions, private network mode and public network mode. Since the internal application department of China Telecom is the mobile bureau, the 5G public network UPF sink solution is the main one, and the portable 5G private network and 5G public network cooperation solution is the backup. The 5G public network UPF sink solution has the following problems compared with the demand of the Ministry of Emergency Management: the bandwidth and latency of the solution are limited by the satellite link, which cannot take advantage of the bandwidth and latency of 5G (slower than 4G). To solve this problem, the program improved UPF equipment with the supplier strong binding cannot be miniaturized, only vehicle. At the same time the general problem of the program is that as long as the satellite link does not work, the entire system cannot work. The main use scenario of the telecom mobile bureau is the vehicle scene, with the vehicle configuration satellite engineer, so the above problem does not affect the use of mobile bureau.

118

H. Jiao et al.

After in-depth communication between industry experts and scholars and relevant departments, it is concluded that the portable 5G private network and 5G public network cooperative solution is flexible, miniaturized, can be deployed and used locally and independently, and can be connected to the public network via satellite; after the solution is completed, it has the same network functions as the 5G public network UPF sinking solution and does not have the limitations of the 5G public network UPF sinking solution, which is more in line with the use requirements of the Ministry of Emergency Management.

4 Conclusions With larger bandwidth, lower latency, and more stable and reliable network capabilities, 5G has broad application prospects in the field of emergency communications support. In order to enhance the role of 5G network in emergency communications, this paper proposes two emergency rescue solutions based on 5G. The portable 5G private network and 5G public network collaboration requires the development of a lightweight 5G core network for emergency communications, and access to the Internet and 5G public networks through satellite links. The 5G public network UPF sinking solution is to sink the UPF function in the core network to the edge, which can greatly reduce the data traffic transmitted through the satellite link and greatly reduce the end-to-end network delay. For emergency communications after major disasters, a single solution cannot meet multiple needs. A combination of multiple solutions may be required to meet the needs of rapid post-disaster relief and reconstruction. As a cross-border combination of new technologies, emergency communication solutions based on 5G applications still have disadvantages and limitations at this stage. However, with the development of technology, such as new batteries and power supply technologies, emergency communication solutions based on 5G applications will surely become a universal solution for emergency communication, which can better serve the applications such as post-disaster rescue and post-disaster reconstruction.

References 1. Xiang, W., Zheng, K., Shen, X.S. (eds.): 5G Mobile Communications. Springer, Cham (2017). https://doi.org/10.1007/978-3-319-34208-5 2. Sharma, V., Srinivasan, K., Chao, H.C., Hua, K.L., Cheng, W.H.: Intelligent deployment of UAVs in 5G heterogeneous communication environment for improved coverage. J. Netw. Comput. Appl. 85, 94–105 (2017) 3. Ferrús, R., Sallent, O., Baldini, G., Goratti, L.: LTE: the technology driver for future public safety communications. IEEE Commun. Mag. 51(10), 154–161 (2013) 4. Xinjun, M.: An evolution in public safety networks. Huawei Technologies Co. (2015). https:// e.huawei.com/us/publications/global/ict_insights/201608271037/focus/201608271435 5. Yi, B., Wang, X., Li, K., Das, S.K., Huang, M.: A comprehensive survey of network function virtualization. Comput. Netw. 133, 212–262 (2018) 6. Sampaio, L.S.R., Faustini, P.H., Silva, A.S., Granville, L.Z., SchaefferFilho, A.: Using NFV and reinforcement learning for anomalies detection and mitigation in SDN. In: IEEE Symposium on Computers and Communications (ISCC), pp. 432–437 (2018)

Use Case Analysis and Architecture Design

119

7. Bernini, G.: Combined NFV and SDN applications for mitigation of cyber-attacks conducted by Botnets in 5G Mobile Networks. In: 16th International Conference on Networks, pp. 148– 153 (2017) 8. L. Xu: CogNet: A network management architecture featuring cognitive capabilities. In: European Conference on Networks and Communications (EuCNC) (2016) 9. Tuna, G., Nefzi, B., Conte, G.: Unmanned aerial vehicle-aided communications system for disaster recovery. J. Netw. Comput. Appl. 41, 27–36 (2014) 10. Zeng, Y., Zhang, R., Lim, T.J.: Wireless communications with unmanned aerial vehicles: opportunities and challenges. IEEE Commun. Mag. 54(5), 36–42 (2016) 11. Dunaway, M., Murphy, R., Venkatasub, N., Palen, L., Lopresti, D.: Research Agenda in Intelligent Infrastructure to Enhance Disaster Management, Community Resilience and Public Safety, CoRR abs/1705.01985 (2017) 12. Favraud, R., Apostolaras, A., Nikaein, N., Korakis, T.: Toward moving public safety networks. IEEE Commun. Mag. 54, 14–20 (2016) 13. Bhardwaj, A., Agnihotri, S.: A resource allocation scheme for multiple device-to-device multicasts in cellular networks. In: Proceedings of the IEEE Wireless Communication and Networking Conference, Doha, 3–6 April 2016, pp. 1–6 (2016) 14. Ali, K., Nguyen, H.X., Shah, P., Vien, Q.T., Bhuvanasundaram, N.: Architecture for public safety network using D2D communications. In: Proceedings of the IEEE Wireless Communication and Networking Conference Workshops, Doha, 3–6 April 2016, pp. 206–211 (2016) 15. Li., Y., Zhang, Z., Wang, W.: Concurrent transmission based Stackelberg game for D2D communications in mmWave networks. In: Proceedings of the IEEE International Conference on Communications, Paris, France, 21–25 May 2017, pp. 1–6 (2017) 16. Kim, H.: Dynamic resource scheduling algorithm for public safety network. In: Proceedings of the UKSim-AMSS 20th International Conference on Computer Modelling and Simulation, Cambridge, UK, 27–29 March 2018, pp. 127–132 (2018)

A Resource Allocation Method for Power Backhaul Network Based on Flexible Ethernet Dayang Wang1 , Song Jiang1 , Yong Dai1 , Wei Li1 , Huichen Xu1 , Lin Cong2(B) , Ying Wang2 , and Peng Yu2 1 Communication Branch, State Grid Jiangsu Electric Power Co., Ltd., Nanjing, Jiangsu, China

{wangdy1,xuhc1}@js.sgcc.com.cn

2 State Key Laboratory of Networking and Switching Technology, Beijing University

of Posts and Telecommunications, Beijing, China [email protected], {wangy,yupeng}@bupt.edu.com

Abstract. The existing network resource allocation methods pay a greater share of attention to the requirements of low delay and high reliability of service transmission, ignoring the utilization of network resources caused by traffic scheduling and inflexibility. Flexible Ethernet can decouple the interface rate between router and transmission box, which can well meet the flexibility requirements of power backhaul network. This paper fully discusses the feasibility and effectiveness of applying FlexE technology to power backhaul network, and designs the algorithms for FlexE-unaware, FlexE-aware and FlexE-terminating in the FlexE flow scheduling process in the transmission scenario, so that each transmission device can be used more efficiently. Keywords: Flexible Ethernet · Power backhaul network · Flow scheduling process

1 Introduction With the vigorous development of smart grid, the types of power distribution business are increasing, and the requirements for business are becoming higher [1]. Research on a reasonable and effective resource allocation algorithm for power backhaul network can not only provide differentiated, isolated services that meet the quality of service requirements for different services, but also improve the network utilization and reduce the network operation and maintenance cost. However, the power backhaul network still faces the following challenges: (1) Low network utilization: At present, the power communication network is built independently according to different power distribution and consumption services. This mode not only leads to repeated construction of the network and reduces the network utilization, but also requires independent management and maintenance of each network, increasing the cost of network operation and maintenance.

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 S. Sun et al. (Eds.): ICSINC 2021, LNEE 895, pp. 120–128, 2022. https://doi.org/10.1007/978-981-19-4775-9_14

A Resource Allocation Method for Power Backhaul Network

121

(2) Ensure business isolation: The current power communication network can ensure the isolation between businesses by independently building the network according to different power distribution and consumption businesses, but this independent network building mode will bring the problems of rising network construction cost and waste of resources. As the UNI interface between router and optical transmission network equipment [2], FlexE can realize one-to-one correspondence between the data stream bandwidth actually carried by UNI interface and WDM link bandwidth of optical transmission network NNI interface through rate matching [3], so as to greatly simplify the mapping of transmission equipment and reduce equipment complexity, investment cost (CAPEX) and maintenance cost (OPEX). As a technical architecture based on Ethernet and industrial chain expansion, FlexE technology meets the requirements of large bandwidth, flexible rate and channel isolation under the IP/Ethernet technology system, which is in line with the development trend of technology and industry [4]. Based on the above analysis, this paper proposes a resource allocation algorithm for power backhaul network, and proposes a resource allocation algorithm for power backhaul network based on three transmission modes of FlexE. Under the conditions of ensuring service isolation and low delay of intelligent distribution communication network, this paper realizes the optimal resource allocation of intelligent power backhaul network.

2 Related Work Many studies have proposed computing models or optimization schemes for network resource allocation from different aspects. Literature [5] studies the resource allocation of competitive base stations between two different operators, and discusses sharing a common optical backhaul network infrastructure. Evolutionary game theory is proposed and applied to study the interaction between base stations and passive optical networks. Using the replicator dynamic model, the asymptotic stability of the proposed system design is proved. Literature [6] studies the joint routing and resource allocation in software defined backhaul network (SDBN) based on OFDMA. A SDBN system model is proposed, in which the control panel can use high complexity algorithms in the configuration stage to simplify the algorithms in the operation stage. By constructing the interference directed graph of the network and analyzing the vertex degree characteristics of the network, a greedy algorithm based on directed graph (DBGA) is proposed. Literature [7] studies several dynamic bandwidth allocation algorithms for user mobility and fog computing in Mobile Backhaul. The proposed algorithm can reduce the migration delay and jitter by giving high priority to the migration traffic. When slicing resources according to the traffic type, the average packet delay of non-migrated traffic can be guaranteed regardless of the migration traffic load. There are still many deficiencies in the above research, such as: (1) The characteristics of optical network are not analyzed and the problems of service flow transmission in optical network are not solved. The services can’t be logically

122

D. Wang et al.

or physically isolated, and can’t meet the service transmission requirements such as high delay and high reliability. (2) Without considering the requirements of service delay and reliability, the hardware cost can’t be maximized. (3) The impact of services on each other in the transmission process is not taken into account, that is, it can’t achieve good isolation effect and meet the reliability requirements of services. Based on the above research background, this paper proposes a resource allocation method of power backhaul network based on FlexE. Because FlexE technology can flexibly allocate the underlying bandwidth to grid services with high delay and high reliability requirements, it is introduced into the power backhaul network to make the whole backhaul network have the functions of flexible bandwidth allocation and physical layer service isolation. By analyzing the characteristics of three FlexE transmission modes, a traffic allocation mechanism with physical isolation and low delay is designed. It maximizes the carrying capacity of Ethernet technology in differentiated service requirements and network bottom bandwidth, ensures the isolation of service parts and low delay, improves the utilization of network resources and realizes the efficient allocation of power backhaul network resources.

3 Problem Description 3.1 FlexE Transport Mode

Fig. 1. FlexE transport scenarios

The application of the FlexE paradigm in the transmission architecture determines that each router has a FlexE shim to map/demap the Ethernet physical layer between the router and the transmission box [8]. Under this mechanism, three different transmission modes are generated according to whether the transmission box has sensing ability: FlexE unaware, FlexE aware and FlexE terminating [9]. (1) In the FlexE-unaware mode, the flexible Ethernet interface is transparently hosted similar to the optical transmission network [10]. This mode can make full use of the old existing optical transmission network equipment, realize the bearing of FlexE without hardware upgrade, and realize the end-to-end Super bandwidth channel across the optical transmission network based on the FlexE binding function.

A Resource Allocation Method for Power Backhaul Network

123

In this mode, the transmission box is functionally equivalent to the multiplex repeater. Since the transmission device has no visibility to the content of FlexE, the existing transparent transmission device can be used for traffic transmission. (2) FlexE-aware mode is very similar to FlexE unaware mode. The difference is that the transmission box has the ability to perceive time slots. In this mode, FlexE identifies unavailable slots by filling in a special error control block data block. When the flexible Ethernet interface on the UNI side is mapped in the optical transmission network through aware mode, the optical transmission network directly discards unavailable slots, extracts the data to be carried according to the original data stream bandwidth, and then maps it to the DWDM transmission pipeline of the optical transmission network with rate matching [11]. (3) Finally, in the FlexE-terminating mode, the transport box has a shim layer, which means that the transport box can sense the transmitted FlexE group and terminate the FlexE group [12]. In this case, streams with multiple destinations can be carried in a single FlexE group because they can be separated in the transport box. Therefore, the transport box also needs to have the ability to comb and plan the flow, so as to reasonably allocate the mixed flow in the PHY to the optical channel. 3.2 Specific Description of the Problem This paper studies how to divide the flow reasonably on the shim layer of the router into the FlexE group under three different modes, and enter the optical transmission network for transmission under the demapping of the transmission box, so as to minimize the cost of hardware resources. The mapping on the shim layer mainly considers how to logically bind PHY for different transmission modes, so that the ports can be maximized, and reuse the bound router ports and transport box ports to a large extent. In FlexE unaware mode, since the transmission device has no visibility to the client stream, the transmission box will send all time slots in the PHY to the optical path. As shown in the example in Fig. 1(a), two 100GE PHYs are bundled, and 150G A-B client streams are sent to the FlexE group. There is only 150G effective traffic on the corresponding 200G optical path. In FlexE aware mode, the transport box has visibility into whether there is a stream in the time slot of PHY. Therefore, unused time slots in PHY can be discarded, so as to save line end ports and improve equipment utilization [13]. The FlexE-terminating mode takes advantage of the characteristics of FlexE channelization. In the same PHY or FlexE group, multiple client streams to different destinations can be accommodated at the same time. By filling the client streams sent by the source end with PHYs as much as possible, the hardware devices can be minimized, and the binding of FlexE group is also quite different from the first two modes, just bundle all PHYs used into one FlexE group. In Fig. 1(c), the content of the transport box convection has the ability to sense, and multiple streams in a group can be demapped and sent to the appropriate optical path.

124

D. Wang et al.

4 Flow Scheduling Algorithm 4.1 FlexE-Unaware Scheduling Algorithm For the newly arrived flow, the algorithm first attempts to accommodate it by reusing a FlexE group. If there are no reusable FlexE groups, analyze whether existing FlexE groups can be extended to accommodate new flows, and always try to reuse FlexE groups as much as possible [14]. Finally, if the existing FlexE group cannot be reused, a new group is created to accommodate the flow by activating the new port and phy and bundling them. The flow provisioning is described by the pseudocode in Algorithm 1. Algorithm 1: FlexE-unaware Scheduling Algorithm Input: All client stream sets f; all existing FlexE group sets G Output: FlexE group set G 'after a new round of scheduling 1 Select a client stream f from F, whose starting node is s and destination node is D, which is recorded as f (s, d); The traffic of the client stream is recorded as CF; Select all FlexE group sets between nodes (s, d) from G and record them as G (s, d); Note that the capacity of each FlexE group in G (s, d) is CG. 2

Cg in G(s,d)

= Cf goto step 3 goto step 4

3 4

5

6 7 8 9 10 11 12 13

min(F) distribute(f) to G(s,d) there are free groups in G (s, d) goto step5 goto step 14 Cg in G(s,d): Calculate the difference between Cg and Cf of each group in G (s, d), record the difference as D, and select the group G with the smallest D. portNum(g)

= D: goto step8

G(s,d).remove(g) goto step5 Add PHY capable of holding D flow into g; Calculate how many additional optical paths (s, d) need to be added to accommodate the new PHY. !isEmpty(transport box): goto step11 Activate(transport box) and goto step9 Connect a PHY to a client port of the transmission box, and connect an optical path to the line end port of the transmission box according to path P. not assign(PHY): goto step9 not assign(f): goto step1 else: goto step17

A Resource Allocation Method for Power Backhaul Network 14 15 16 17

125

shortestpath(s,d) = Cf.PHY: goto step16 stop(f(s,d)) Put(Cf. PHY) into g and put(g) into G(s,d) goto step9 G

The main difference of FlexE-aware mode is that it has one more step to identify blank time slots than FlexE-unaware mode, so the algorithm flow of this mode will not be repeated. 4.2 FlexE-Terminating Scheduling Algorithm The emphasis of the terminating mode algorithm is no longer on the division of FlexE groups, but on the allocation scheme of data flow at the line end of the transmission box. As long as the allocation of data flow at the line end of the transmission box is confirmed, the usage of data port and the cost required for mode operation can be determined. According to the idea of heuristic algorithm, the greedy algorithm relying on sequential input is generally used to solve the optimal scheme of data flow allocation [15]. However, there is no reasonable allocation process in this way, so that the optimal solution can’t be obtained. Therefore, this paper adopts the idea of dynamic programming algorithm to allocate all data streams on the same data link more reasonably, maximize the utilization of each optical path, and reduce the number of ports used in router data box. The flow provisioning is described by the pseudocode in Algorithm 2.

Algorithm 2: FlexE-Terminating Scheduling Algorithm Input: Flow set

available transmission line capacity cap

Output: All line flow dispatching set 1 Initialize traffic collection line combination All line flow dispatching sets 2 While 3

null do:

for

in F:

4

A.add

5

if

6 7

Transmission line capacity Single line flow collection

break else:

8 9 10 11 R.add

end if A.get(Max(

))

Maximum flow of

126

D. Wang et al.

5 Experiments and Results 5.1 Algorithm Evaluation Index and Test Scheme This paper takes the number of hardware ports used by the sender as the evaluation index of the algorithm. For the multi round scheduling of the three modes, multiple client streams are randomly generated in each round and input into the simulated scheduling scenario under the three modes respectively. After running the algorithm, the scheduling situation of the stream in the hardware device is obtained. According to this situation, the port usage of hardware equipment can be obtained, the usage of router port, transmission box client port and transmission box line end port in each mode can be compared horizontally, and the equipment usage characteristics of FlexE transmission modes can be summarized and analyzed. The experiment is divided into three rounds. Each round uses the FlexE group bound by the previous round as the input. At the beginning of each round of experiment, 10 groups of flows with the same source node are randomly generated, the source node is a, and the target node is randomly generated from B–Z, which may be the same or different. After generation, three algorithms of three modes are input for calculation, and three different scheduling schemes can be obtained. Record the client stream allocation and port occupancy of each round, and start the random generation and scheduling of the next round of streams. 5.2 Horizontal Comparison of Three Modes Comparison of Activated Router Ports. The algorithms of the three modes are tested with the same randomly generated three groups of continuously arriving client streams. The number of router ports used after flow scheduling in each round is shown in Fig. 2.

Fig. 2. Comparison of router ports in three modes

In the three-time scheduling, the number of client ports used by unaware mode and aware mode is exactly the same, because the difference between the two algorithms is only reflected in whether the transport box can be recognized free time slots are allocated and discarded, so the FlexE group bundling and port allocation on the router side are completely consistent. Similarly, because both ends of a PHY are connected to the router

A Resource Allocation Method for Power Backhaul Network

127

port and the transport box client port respectively, the occupation of the transport box client port in these two modes is also the same. For the termination mode, it can be seen that the number of router ports used is significantly reduced because the termination uses the subchannel characteristics of FlexE, so that the mixed flow can exist in the same PHY. As long as the transmission box line end allows, the remaining space in any PHY can be used as much as possible, and the useless time slots are greatly reduced, reducing the total number of PHYs required, Thus, the occupied number of router ports is reduced and the cost of hardware equipment is saved. Comparison of the Number for Transport Box Ports. In the above three rounds of continuous experiments, the number of transmission box line end ports used after flow scheduling in each round is shown in Fig. 3.

Fig. 3. Comparison of line-ports of transport boxes under three modes

During the first scheduling, the number of ports at the line end of the transmission box used in unaware and aware modes is the same, because the previous group has not been multiplexed during the first binding, and there will be no idle PHY. During the third scheduling, the unaware mode flex group will have idle PHY. Because the transmission box has no perception of PHY, it will send all PHY time slots to the line end, resulting in the occupation of redundant ports; at this time, in aware mode, you can identify the idle PHY and discard the time slot, and only transfer the effective time slot to the line end. For the terminating mode, in the subsequent scheduling, because it is not bound by the FlexE group binding, the algorithm of this mode only needs to determine the minimum line end ports that can send client streams, and allocate corresponding hardware device support to these ports from the transmission box to the router, so as to realize the normal transmission of streams.

6 Conclusion In this paper, the algorithm design and experimental comparative analysis of the transmission architecture based on FlexE technology are carried out. Appropriate traffic scheduling algorithms are designed for the three FlexE modes, and several transmission examples are tested to investigate their efficiency in router, transmission box and

128

D. Wang et al.

interface configuration. Through experimental comparison and analysis of simulation results, the advantages and disadvantages of each transmission scenario are revealed, as well as the different application types suitable for each scenario. Acknowledgements. This work is supported by science and technology project from State Grid Jiangsu Electric Power Co., Ltd: “Key Technologies of the application of Flexible Ethernet in power communication networks(SGJSXT00TYJS2000316)”.

References 1. Gupta, A., Jha, R.: A survey of 5G network: architecture and emerging technologies. IEEE Access 3, 1206–1232 (2015) 2. Trowbridge, S.: Ethernet and OTN: 400G and beyond. In: Optical Fiber Communication Conf. (OFC), Paper Th3H.1, March 2015 3. Patel, A., Kanonakis, K., Ji, P.N., Hu, J., Wang, T.: Flexible-client: the missing piece towards transport software-defined networks. In: Optical Fiber Communication Conference (OFC), Paper Th3I.3, March 2014 4. Hofmeister, T., Vusirikala, V., Koley, B.: How can flexibility on the line-side best be exploited on the client side? In: Optical Fiber Communication Conference (OFC), Paper W4G.4, March 2016 5. Loumiotis, I., Kosmides, P., Adamopoulou, E., Demestichas, K., Theologou, M.: Dynamic allocation of Backhaul resources in converged wireless-optical networks. IEEE J. Sel. Areas Commun. 35(2), 280–287 (2017). https://doi.org/10.1109/JSAC.2017.2659023 6. Li, H., Zhang, J., Hong, Q., Zheng, H., Zhang, J.: Digraph-based joint routing and resource allocation in software-defined backhaul networks. In: 2017 IEEE 22nd International Workshop on Computer Aided Modeling and Design of Communication Links and Networks (CAMAD), pp. 1–5 (2017). https://doi.org/10.1109/CAMAD.2017.8031525 7. Ou, J., Li, J., Yi, L., Chen, J.: Resource allocation in passive optical network based mobile Backhaul for user mobility and fog computing. In: 2017 Asia Communications and Photonics Conference (ACP), pp. 1–3 (2017) 8. Vusirikala, V.: FlexEthernet (FlexE) use cases, Ethernet Alliance. http://www.ethernetalli ance.org/wp-content/uploads/2014/10/Flex-Ethernet-TEF-01-Use-Cases.pdf 9. Flex Ethernet 2.0 implementation agreement, Optical Internetworking Forum, June 2018. https://www.oiforum.com/wp-content/uploads/2019/01/OIF-FLEXE-02.0-1.pdf 10. Lu, W., et al.: How much can flexible ethernet and elastic optical networking benefit mutually? In: Proceedings of ICC 2019, pp. 1–6, May 2019 11. Eira, A., Pedro, J.: How much transport grooming is needed in the age of flexible clients? In: Optical Fiber Communication Conference (OFC), Paper W1I.3, March 2017 12. Ofelt, D., Chiesa, L.D., Booth, B., Hofmeister, T.: FlexEthernet—what is it and how can it be used? In: Optical Fiber Communication Conference (OFC), Paper W4G.1, March 2016 13. Eira, A., Quagliotti, M., Pedro, J.: Impact of client- and line-side flexibility in the lifecycle of next-generation trans-port networks. J. Opt. Commun. Netw. 8(7), A101–A115 (2016) 14. Epstein, L., Imreh, C., Levin, A.: Class constrained bin packing revisited. Theor. Comput. Sci. 411, 3073–3089 (2010) 15. Gong, L., et al.: Efficient resource allocation for all-optical multicasting over spectrum-sliced elastic optical networks. J. Opt. Commun. Netw. 5, 836–847 (2013)

Cooperative Routing Algorithm for Space-Based Information Network Based on Traffic Forecast Zhongxiang Jia(B) , Ying Wang, and Hongyang Liu State Key Laboratory of Networking and Switching Technology, Beijing University of Posts and Telecommunications, Beijing, China [email protected]

Abstract. Space-based information network is an important direction for future network. According to the characteristics of space-based information network, we first propose a time-varying topology-oriented space-based information network collaborative routing model, which involves a space-based information architecture, a satellite traffic forecast method and the space-based information network routing planning problem model. On this basis, a space-based information network high and low orbit satellite cooperative algorithm (HLCRA) is proposed. The simulation results show that the proposed algorithm can effectively solve the problems of frequent snapshot generation, excessive rerouting times, and QoS guarantee in the space-based network routing decision, Compared with the traditional algorithms, the proposed algorithm has good results in delay, throughput and rerouting times. Keywords: Space-based information network · LSTM · Time-varying topology · Collaborative routing

1 Introduction With the globalization of society and economy, the requirement of communication quality is getting higher and higher. For some sparsely populated marginal areas, such as mountains, oceans, north and south poles, deserts, etc., the current ground network used to construct signal base stations is too costly and has a low utilization rate [1]. Although traditional satellite communications can solve the above-mentioned problems, there are problems such as single function, small coverage area, and unavailability in case of failure. The space-based information network transmits different business data. Different services have differentiated QoS requirements, which are reflected in aspects such as delay and bandwidth. For example, services such as remote control commands, telemetry information, navigation data, voice services, etc. have higher requirements for delay; while services such as video on demand and data file transmission are not sensitive to service delay. For some large file transfer services, there are also certain requirements for bandwidth.

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 S. Sun et al. (Eds.): ICSINC 2021, LNEE 895, pp. 129–138, 2022. https://doi.org/10.1007/978-981-19-4775-9_15

130

Z. Jia et al.

Literature [2] proposed a low-orbit topology-oriented routing algorithm OPSPF, which uses the regularity of the constellation to perform routing calculations and generate an instant routing table, which handles the periodic topology changes of the satellite constellation well. Literature [3] proposed a multi-layer satellite network routing algorithm. High-orbit satellites collect the routing status of low-orbit satellites and perform routing calculations. Low-orbit satellites are only responsible for forwarding. However, the above low orbit routing method does not carefully divide the plane and orbit relationship of the source and destination satellites in the routing planning, which will lead to too large topological range and affect the routing efficiency; On the other hand, multi-layer routing algorithm does not differentiate different services, so it is difficult to adapt to the differences of delay and bandwidth requirements of space-based information networks. In response to the problems in the above research, we first analyze and propose a time-varying topology-oriented space-based information network collaborative routing model. Then, we propose a time-varying topology-oriented space-based information network cooperative routing algorithms, including high-low orbit coordination algorithms, and low orbit routing algorithm with the different relationship of the plane and orbit where the source and sink nodes are located.

2 Cooperative Routing Model 2.1 Space-Based Information Network Architecture As shown in Fig. 1, the mesh topology of the LEO constellation consists of inter-satellite links. The inter-satellite links meet the following characteristics: Most satellites have four inter-satellite links, including inter-satellite links in two planes and inter-satellite links between two planes. As shown in Fig. 1, A1 ∼ A5 are in the same plane, A1 and A2 is an in-plane inter-satellite link; E5 and F5 are in the same orbit on different planes, and E5 and F5 are an in-plane inter-satellite link.

Fig. 1. Inter-satellite link diagram

Cooperative Routing Algorithm for Space-Based Information Network

131

Assuming that there are M polar orbit planes in the LEO satellite network, and there are N satellites on each plane, the LEO satellite network can be simplified into a graph GL = (VL , EL ). EL is the collection of inter-satellite links between nodes. In Fig. 1, each column represents a plane, and each node represents a satellite logical address Pi, Oi. Where i is the number of the low-orbit satellite, which represents the satellite on the nth orbit of the mth plane i = 11∗(m − 1) + n

(1)

2.2 Satellite Traffic Forecast Method Based on LSTM Recurrent Neural Network (RNN ) is a deep learning method that specializes in processing sequence data. It plays a huge role in the field of data science [1]. It can extract key information from massive amounts of data. Therefore, in this paper, we use a special RNN network-Long Short-Term Memory Network (LSTM ) [4] to predict satellite traffic, which can restore the characteristics of satellite traffic itself, including: non-linearity, self-similarity, long-term correlation, short-term correlation, and burstiness.

Fig. 2. Schematic diagram of LSTM

As shown in Fig. 2,LSTM neural network solves the problem of gradient explosion or disappearance caused by RNN back propagation. Compared with traditional RNN , cell state C is introduced to transmit the previous information. At the same time, the activation output value at the last moment at−1 and the input Xt at the current moment are jointly decided the forget gate and update gate, and then determines the choice of C and Xt information.

132

Z. Jia et al.

The forgetting matrix ft is:     ft = σ1 Wf at−1 , Xt + bf

(2)

    it = σ1 Wi at−1 , Xt + bi

(3)

Update matrix it is:

Finally, the predicted value yˆ t can be calculated according to (4):   yˆ t = σ2 Wy [at ] + by

(4)

The above is the internal information transmission process of LSTM , where Wf , bf ; Wi , bi ; Wc , bc ; Wo , bo ; Wy , by are the forget gate, update gate, output gate, and the weight and bias of the output result value of LSTM , respectively. By controlling the forgetting gate, updating the gate, and outputting the gate, we can flexibly choose between valid information and invalid information. Non-linear functions can be simulated, and the transmission of cell information can retain past information for a long time [5]. This design can solve the data dependency in the timing problem and improve the accuracy of prediction. After that, the initial traffic of each node in the next cycle is predicted by LSTM, and the traffic of the next cycle is obtained, and then the routing simulation is carried out. 2.3 Space-Based Information Network Routing Planning Problem Model The problem model of space-based network routing collaborative planning is as follows.   (Xi,j → Ps,d ∗ Di,j ) min i∈s j∈d , s!=d

s.t. C1 :





(Lli,j ∗ ti,j ) ≤ Ts,d

i∈s j∈d , s!=d

  C2 : min bi,j ≥ Bs,d , ∀i ∈ s, j ∈ d , s! = d The goal of the model is to find the shortest path from s to d . Constraint condition C1 ensures that the total delay of the path is less than the delay threshold. Constraint condition C2 ensures that the remaining bandwidth of each link of the path is greater than the bandwidth requirement.

3 Cooperative Routing Algorithm 3.1 Space-Based Information Network High and Low Orbit Satellite Cooperative Algorithm (HLCRA) Different service of space-based information network have different requirements for QoS. When the amount of data is small, low-orbit satellites with low latency are used

Cooperative Routing Algorithm for Space-Based Information Network

133

directly for transmission. When the amount of data increases sharply, the transmission delay of low-orbit satellites is high, and a blocking queue is formed at this time, which is fatal for data with high QoS requirements. The high-low-orbit satellite coordination algorithm in this paper differentiates data processing. When the traffic is too large, GEO is used to temporarily transmit data with lower QoS requirements, and data with higher QoS requirements are still uses LEO for transmission, and satellite data is differentiated and processed on demand. We assume that the logical address of the  source  satellite P , O i is Pi , Oi , and the logical address of the destination satellite j is  j j . When the angle between i and j is less than 90°, only one high-orbit satellite Pg , Og , g = 100 ∼ 104.between i an j is needed to When the angle is greater than    achieve communication. 90°, two high-orbit satellites Pg1 , Og1 , Pg2 , Og2 are needed to forward data (lines 4, 5). Assuming that the amount of controlled data is G_data, the amount of uncontrolled data is F_data, algorithm is described as follows:

input：
，
， _

,

， _

ℎ

1：

:

2：

:

_ ；

3： if( ℎ 4：

，




5： End if 6： if(ℎ 7：

，


,


Low-orbit satellite routing (LRA) can be divided into three situations: same plane and different orbit, different plane and same orbit, and different plane and different orbit according to the different relationship of the plane and orbit where the source and sink nodes are located, which greatly reduces the time complexity. We assume that the logical address of the source satellite number i is Pi , Oi , and the logical address of the   destination satellite number j is Pj , Oj . When Pi = Pj and Oi ! = Oj , use the same-plane different-orbit algorithm SPDO; when Pi ! = Pj and Oi = Oj , use the different-plane same-orbit algorithm DPSO; when Pi ! = Pj and Oi ! = Oj , Use the different plane and different track algorithm DPDO. Take the same-orbit and different-plane algorithm as an example to introduce the algorithm flow. The same plane and different orbit algorithm (SPDO) is described as follows. Assume that i and j are on different orbits in the same plane, that is, Pi = Pj and Oi ! = Oj . When i and j are in the same hemisphere, then i searches for the shortest path in the same orbit for forwarding; when i and j are in different hemispheres, the traffic increases or decrement along the same hemisphere Oi .

134

Z. Jia et al. ：< ：

>，


)

)

= −1

4： 5：

>

while( ! =

3：

,

，

，




End while

6： END if 7： else 8：

while( ! =

9：

= +1

10：

，

)


11： End while 12： END else 13：END if

3.2 Comparison Algorithm and Time Complexity Analysis In order to evaluate the performance of the proposed algorithm, we choose the following comparison algorithms and analyze their time complexities. DAPR [6]: High-orbit satellites act as controllers to calculate routing, and low-orbit satellites receive routing data from high orbits and forward routing. The time complexity is (N ∗M ). SARA [2]: Low-orbit satellites are responsible for collecting data from ground stations, while high-orbit satellites have full coverage and are responsiblefor transmitting and transmitting data to the ground. The time complexity is O(N ) ~ O N 2 . HLCRA: The algorithm proposed in this paper. High-orbit satellites generally act as controllers. When the bandwidth and delay are not met, data is shunted, and the high-orbit satellites are not sensitive to business delays. The time complexity is O(N ∗ G).

4 Simulation 4.1 Simulation Scenarios and Simulation Parameter Settings The simulation environment and NS3 flow are set according to the literature [2]. We choose Iridium-like constellation. There are a total of 6 orbital planes in low orbit, and each orbital plane has 11 satellites. There are four satellites in high orbit, and the angle between each high orbit satellite is 90°. We use STK simulation software to get the satellite topology, and divide the snapshot according to the logical address method of this paper. We use NS3 to simulate the routing process. The Initial flow parameter settings of NS3 are as Table 2. In a period of 11 min, the request scale is 100–400 Poisson distribution. The proportion of high QoS data for each

Cooperative Routing Algorithm for Space-Based Information Network

135

request is 60%: The proportion of low QoS data is 40%. The initial bandwidth resource of each node is 20 Mbps. Use the onOffApplication mode of NS3 to send traffic to the destination node. OnTime is set to 1, offTime is 0, which means that traffic is always sent, and the sending traffic is CBR. The packet size and rate are as Table 3. The initial bandwidth of each satellite node is set to 20 Mbps, and the initial delay is 2 ms. The traffic of each node in the next cycle is predicted by LSTM, and the traffic of the next cycle is obtained, and then the routing simulation is carried out (Table 1). Table 1. Satellite node traffic parameter settings. Parameter

Numerical value

Flow type

OnOffApplication

DataRate

2000 kbps

PacketSize

50

OnTime

1

OffTime

0

bandwidth

20 Mbps

Delay

2 ms

Number of requests

100–400

Request arrival type

Poisson distribution

4.2 Simulation Results and Analysis Figure 3 is a comparison diagram of the transmission delays of the three high- and lowtrack collaborative routing algorithms. The collaborative routing algorithm HLCRA of this paper distinguishes data that is sensitive to QoS and data that is insensitive to QoS, and plays the role of data shunting. The LRA algorithm is still used for QoS-sensitive data to distinguish tracks and planes. When the number of requests further increases, HLCRA still has a low latency. HLCRA has a time delay of at least 9.3% lower than that of the other two collaborative algorithms. Figure 4 is the throughput comparison of the algorithms. When the number of requests is 100, the throughput of the three algorithms is close. When the number of requests gradually increases, the three cooperative routing algorithms still maintain high throughput. When the number of requests increases to 400, the throughput of low-orbit satellites drops significantly. The HLCRA of this paper is oriented to differentiated traffic processing. The use of high-orbit satellites to shunt low-QoS data and still maintain a high throughput, increasing the throughput by at least 5.4%. Figure 5 shows the comparison of the rerouting times of the algorithms. HLCRA adopts the method of public link set, and the routing link is more stable. When the number of requests increases and the network load increases, the number of rerouting is significantly lower than that of similar algorithms, reducing the number of rerouting by at least 25% .

136

Z. Jia et al.

Fig. 3. Delay comparison

Fig. 4. Throughput comparison

Cooperative Routing Algorithm for Space-Based Information Network

137

Fig. 5. Comparison of rerouting times

5 Concluding Remarks This paper studies the routing algorithm of space-based information network. Considering high-low orbit satellite coordination and differentiated data processing to establish a space-based information network model, and propose a space-based information network satellite routing algorithm. Compared with the traditional algorithms, the proposed algorithm has good results in delay, throughput and rerouting times.

References 1. Liu, J., Luo, R., Huang, T., Meng, C.: A load balancing routing strategy for LEO satellite network. IEEE Access 8, 155136–155144 (2020). https://doi.org/10.1109/ACCESS.2020.301 7615 2. Sun, X., Cao, S.: A routing and wavelength assignment algorithm based on two types of LEO constellations in optical satellite networks. J. Lightwave Technol. 38(8), 2106–2113 (2020). https://doi.org/10.1109/JLT.2020.2965185 3. Qi, X., Zhang, B., Qiu, Z.: A distributed survivable routing algorithm for mega-constellations with inclined orbits. IEEE Access 8, 219199–219213 (2020). https://doi.org/10.1109/ACC ESS.2020.3041346 4. El Alaoui, S., Ramamurthy, B.: MARS: a multi-attribute routing and scheduling algorithm for DTN interplanetary networks. IEEE/ACM Trans. Networking 28(5), 2065–2076 (2020). https://doi.org/10.1109/TNET.2020.3008630 5. Wang, J., Zhang, R., Yuan, J., Du, X.: A 3-D energy-harvesting-aware routing scheme for space nanosatellite networks. IEEE Internet Things J. 5(4), 2729–2740 (2018). https://doi.org/ 10.1109/JIOT.2018.2803111

138

Z. Jia et al.

6. Hu, J., Cai, L., Zhao, C., et al.: Directed percolation routing for ultra-reliable and low-latency services in low earth orbit (LEO) satellite networks. In: 2020 IEEE 92nd Vehicular Technology Conference (VTC2020-Fall). IEEE (2020) 7. Liu, Y., Zhu, L.: a suboptimal routing algorithm for massive LEO satellite networks. In: 2018 International Symposium on Networks, Computers and Communications (ISNCC) (2018)

Exploration on the Practice Teaching of Environmental Design Network Based on Mobile Internet Technology Wei Meng(B) and Hui Liu Academy of Fine Arts, Bohai University, Jinzhou, Liaoning, China [email protected]

Abstract. Internet plus is the evolution of the Internet form driven by the innovation of the knowledge society and the new normal of economic and social development. The use of mobile Internet, cloud computing, big data, Web of Things and other information and communication technologies to transform the original education model will have the profound impact on practical teaching. Based on modern educational technology, in order to adapt to the development trend of the Internet plus is era, this paper designs the practical teaching system for environmental design majors, proposes practical teaching reform measures, cultivates students’ practical ability, and meets the society’s demand for environmental design talents. Keywords: Internet plus · Mobile Internet · Big data · Practical teaching system · Reforming measures

1 Introduction Promoting the application-oriented transformation of universities is an important part of the structural reform of the talent supply side in the education field. Running a good application-oriented undergraduate course, improving the quality of application-oriented talent training, and better serving the needs of regional economic and social development and industrial transformation and upgrading are important issues facing my country’s higher education reform [1]. Cultivating applied talents must go out of the traditional “elite education” school-running concept and “academic” training mode. We should not blindly pursue the transfer of profound theoretical knowledge, but should focus on creating conditions to implement teaching that combines theoretical knowledge and practical ability. Practical teaching is the best carrier for applying theoretical knowledge to practice. Through rigorous training of the practical teaching system, the connection with the work system and work process is strengthened to improve the professional application ability, development and design ability, technological innovation ability of talents. Comprehensive professional quality, and effectively enhance the core competitiveness of professional application of talent training. Through school-enterprise cooperation, reform the practical teaching system, cultivate students’ practical ability, and meet the society’s demand for environmental design talents. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 S. Sun et al. (Eds.): ICSINC 2021, LNEE 895, pp. 139–144, 2022. https://doi.org/10.1007/978-981-19-4775-9_16

140

W. Meng and H. Liu

2 Problems Existing in Practical Teaching of Environmental Design Major Through in-depth investigation and research, it is concluded that the problems existing in the practical teaching of environmental design major are as follows: (1) The practical teaching system is not perfect. According to the viewpoint of system theory, a system generally refers to a whole set of things within a certain range or of the same kind in accordance with a certain order and internal connections. The practical teaching system is an organic whole composed of various elements of practical teaching activities, which specifically includes the objectives, content, management and quality assurance of practical teaching activities. Due to the relatively short development time of the environmental design major, the practical teaching system is not complete, showing fragmentation, abnormality and noninstitutionalization. The proportion of experimental courses, internships, practical training, course thesis and graduation design is inconsistent, which is far from the requirements of continuous development, continuous progress and continuous innovation in modern society, and students cannot effectively acquire professional skills and innovative ability. (2) (The practical teaching environment needs to be optimized urgently. The practical teaching of environmental art design requires certain hardware and software support, as well as corresponding venues and equipment. Many colleges and universities have a lengthy bidding process for equipment and software procurement. The more advanced equipment when the procurement plan is issued is out of date by the time the equipment is put into use. Many teaching-equipment lack the corresponding post-maintenance management, which makes it difficult to carry out some practical teaching effectively. Although some colleges and universities have also established off-campus practice teaching bases, they are limited by funding and management, making it difficult to meet the needs of internships and training. The school-enterprise cooperation stayed at the written cooperation agreement and failed to advance in depth. Some colleges and universities allow students to find internship places on their own, and practical teaching becomes formalistic. (3) The theoretical teaching is out of touch with practical teaching. At present, in many environmental design majors in colleges and universities, theoretical teaching and practical teaching are taught by different teachers, and there is a lack of effective communication between each other, and some practical links are useless. A lot of practice is carried out through a virtual project. Students have no chance to contact the construction site, no actual engineering cases, lack of understanding of the production process, and ignorance of structures, materials and construction techniques. (4) The depth of school-enterprise cooperation is insufficient. Judging from the current practical teaching of environmental design majors, most colleges and universities’ on-campus and off-campus training are still at the initial stage. Even if there is a form of school-enterprise cooperation, the cooperation content lacks depth and the level of cooperation is not high. In order to complete the practical teaching plan, the school sends students to cooperative places for internship training. Some

Exploration on the Practice Teaching of Environmental Design Network

141

companies only regard training students as cheap labor, in order to reduce the cost of human resources, let students do work that has nothing to do with their own majors. Students’ practical skills are not improved, and precious time is wasted. Only by establishing a practical teaching mode of “combination of work and learning” in the true sense can students truly improve their practical skills and give full play to their professional advantages in future employment and the workplace.

3 Practical Teaching System of Environmental Design Major in the Internet Plus Era The environmental design professional practical teaching system in the Internet plus era is composed of content system, management system and guarantee system. The Internet plus era is benefited from the growth of 5G and new network technologies, such as the cell networks [9–11]. (1) Practical teaching content system. The practical teaching content system is guided by market demand, and the practical teaching content is selected in a targeted manner to form a flexible and diverse form of practical teaching. Specifically, it includes course experiment, professional internship, course design, comprehensive training, graduation design, post internship, social practice, and technological competitions. All kinds of practical teaching methods are properly configured and implemented in different experimental environments. According to the basic skills, professional skills and comprehensive skills, the practical teaching content is arranged step by step, and the practical teaching goals are implemented in different practical teaching forms, so that students can master the necessary, complete and systematic skills and techniques. The practical teaching content system under the background of school-enterprise cooperation is shown in Fig. 1 [2]. (2) Practical teaching management system. Effective management can guarantee the efficient implementation of practical teaching. Practical teaching management includes institutions, teaching bases and staffs, as well as off-campus practical teaching rules and regulations, management methods and evaluation index systems. The school has established a practical teaching administrative management agency under the leadership of the teaching steering committee, which is responsible for the planning and organization, management coordination, resource optimization, quality monitoring, and evaluation of practical teaching. For the scientific and standardized development of school-enterprise cooperation, systems and regulations must be improved and professional management must be implemented. Both the school and the enterprise signed a legally regulated cooperation agreement that clarified the rights, obligations and responsibilities of both parties. At the school level, a special school-enterprise collaborative management organization is established to be responsible for the design, overall planning, and organization and coordination of school-enterprise cooperation programs, and invite corporate personnel to participate in management [3]. Industry associations play an important role in formulating, guiding, and implementing industry standards and regulations. Participation of industry association staff may help businesses avoid short-sighted conduct and ensure that talent training is long-term.

142

W. Meng and H. Liu Practical form and practical environment

Ability training

Professional laboratory Course experiment

Confirmatory experiment Design experiment Comprehensive experiment

On-campus training base Course Design Comprehensive training Science and technology activities

Off-campus training base

Teachers' scientific research projects Various technological competitions

Social practice Post internship

Basic professional ability training

Professional innovation ability and expanding innovation ability training

Step by step, from the shallower to the deeper, interpenetra te and merge with each other

Comprehensive vocational training

Fig. 1. The practical teaching content system of environmental design in the Internet plus era

(3) Practical teaching guarantee system. The teaching guarantee system integrates teaching resources by establishing scientific and reasonable teaching goals and standards, quality management procedures, rules and regulations, using monitoring, measurement, and evaluation methods to ensure that the quality of talent training can meet the needs of social development. The practical teaching guarantee system includes many aspects. Among them, the organization guarantee system, the establishment of a coordinated practical teaching organization system, and the job responsibility system clarify the relationship between organizations at all levels and the tasks they undertake; the operation guarantee system is based on the talent training goals. Determine the practical teaching plan, check and supervise the implementation of the plan to ensure the order of practical teaching; the system guarantee system, around the factors that affect the quality of practical teaching, establish rules and regulations, standardize practical teaching activities, and improve the quality of practical teaching; evaluate the guarantee system and check the operation of practical teaching The degree of deviation from the target requirements should be adjusted in time to form a target program that is more suitable for practical teaching. According to the principles of “school-enterprise co-management, target management, process monitoring, and multi-dimensional evaluation”, the construction of a practical teaching guarantee system is shown in Fig. 2 [4].

Exploration on the Practice Teaching of Environmental Design Network

"Multi-dimensional" practical teaching evaluation system

Management by objectives Practical syllabus development practical teaching team in the Internet plus era

Planning ability

Practical teaching standards formulation Practical teaching content development Practical evaluation system development

143

Designing ability

Innovative ability

School practice

Teamwork ability

Management ability

Enterprise practice

Multiple supervision mechanism of "schools, enterprises, and students" Student evaluation

Expert evaluation

Teaching supervision

Enterprise feedback

Tracking survey

All-round, whole-process monitoring and problem feedback mechanism

Fig. 2. Environmental design professional practical teaching guarantee system in the Internet plus era

4 Reform Measures of Practical Teaching of Environmental Design Major in the Internet Plus Era In order to promote the practical teaching of environmental design in the Internet plus era, the following reform measures are proposed: (1) Develop school-enterprise cooperation with characteristic enterprises [5]. The environmental design profession has many processes and types of work, and characteristic enterprises have innovative technologies, innovative ideas and innovative corporate culture. Featured companies can provide corresponding practice places to enable students to master practical work skills such as independent organization design, construction, budgeting, quotation, and procurement in the first line, and improve their comprehensive practical ability. (2) Strengthen the construction of practical teaching resources [6]. Specifically, it includes on-campus training base, off-campus training base, curriculum resource construction, extracurricular resource construction, online resource construction and offline resource construction, etc. The most important thing is to build oncampus and off-campus training bases, so that students can exercise professional skills in practice, cultivate the awareness of innovation and entrepreneurship, and achieve win-win and mutually beneficial cooperation between universities and enterprises.

144

W. Meng and H. Liu

(3) Cultivate the team of “dual-qualified” teachers [7]. “Dual-qualified” teachers refer to the dual qualifications of professional teaching and professional practice. Specific to the environmental design major, in addition to the qualifications of teachers, it should also have industry qualifications such as construction engineers, interior designers, and landscape designers, or engineering practical experience. Teachers are selected for further study or exercise in enterprises to enhance their practical ability. (4) Reach strategic partnerships between enterprises and universities [8]. Universities provide talents and intellectual support for enterprises, and enterprises provide practical bases for students to apply the knowledge learned in schools to production. The company gives students all-round guidance, including the collocation and use of design materials, the reasonable configuration of design structures and design landscapes, and the optimization of design in combination with design themes and intentions.

References 1. Gu, Y.A.: Big application view: New concept and new paradigm of applied talents training (2020). https://reader.gmw.cn/2020-05/18/content_33839283.htm 2. Ren, W.J., Li, M.: Three - tier Practical Teaching System Building by Collaborating Between School and Enterprise (2021). https://www.docin.com/p-1671676114.html 3. Bai, F., Wan, Y.: A practice teaching system based on school-enterprise cooperation: connotation, paths and key problem. Mod. Educ. Manage. 34(10), 85–90 (2014) 4. Hua, G.P.: Discussion on the quality assurance system of college-enterprise cooperative practice teaching. J. Zhejiang Shuren Univ. (Acta Scientiarum Naturalium) 19(1), 61–64 (2019) 5. Zhong, H.: Research on school enterprise cooperation practice teaching of environmental art design major. Art Educ. Res. 11(10), 169–170 (2020) 6. Zhang, T.S., Wang, K.X., Zhou, Y.W.: Reform of practice teaching system of environmental design major under the background of innovation and entrepreneurship education. Western China Qual. Educ. 5(22), 183+185 (2019) 7. Chen, Z.Y.: On the practical education reformation of environment design major in the context of application transformation: taking Putian university as an example. J. Putian Univ. 26(3), 95–98 (2019) 8. Liu, Y.: Research on the practice teaching reform of environmental art design course. Educ. Mod. 6(25), 55–56 (2019) 9. Liu, Q., Sun, S., Rong, B., Kadoch, M.: Intelligent reflective surface based 6G communications for sustainable energy infrastructure. IEEE Wireless Commun. Mag. 28 (2021) 10. Chen, L., Yu, F.R., Ji, H., Rong, B., Li, X., Leung, V.C.M.: Green full-duplex self-backhaul and energy harvesting small cell networks with massive MIMO. IEEE J. Sel. Areas Commun. 34(12), 3709–3724 (2016) 11. Chen, N., Rong, B., Zhang, X., Kadoch, M.: Scalable and flexible massive MIMO precoding for 5G H-CRAN. IEEE Wirel. Commun. Mag. 24(1), 46–52 (2017)

Modern Information Technology Develops Intelligent Elderly Care Service Industry Xiu Sun(B) Dianchi College of Yunnan University, Kunming 650000, Yunnan, China [email protected]

Abstract. With the deepening of population aging, China’s elderly service industry appears some problems, such as lacking large of elder care institutions and talent service staff, more empty nesters elderly and lower medical level, etc. so it is of great urgency to develop elderly service industry. With the help of modern information technology, we should constantly improve infrastructure constructions, strengthen the training of service-oriented talents, courage all parties to participate in, and do a good job in overall supervision, to speed up the integration of modern information technology into the elderly service industry, in order to develop the elderly service industry. Keywords: Aging population · Modern information technology · Elderly service industry · Development

1 Introduction With the social economy and health care developing, people’s life expectancy is increasing, and the degree of population aging is also deepening. Population aging is a concept that the population with high fertility and low life expectancy changes to low fertility and low mortality. According to international standards, country or region is aging and becomes an “elderly country” or “elderly region”, while its aging population number is more than 7% (aged 65 and above) or 10% (aged 60 and above) of the total population. Population aging is a global trend. According to the United Nations Population Division, by 2025, the world’s total population will reach 8.2 billion, of which the elderly population will reach 1.12 billion, above 13.66% of whole world. In this short 75 years, the world’s total population has increased from 2.5 billion to 8.2 billion, more than three times, while the elderly population has increased more than five times [1]. China’s population aging speed is also increasing quickly. The fifth population census at the end of November 2000 shows the number of the elderly aged 60 and above is 130 million, accounting for 10.2% of the total population, and the number of the elderly aged 65 and above is 88.11 million, accounting for 6.96% of the total population. In 2019, aging population is 18.1% (aged 60 and above) and 12.6% (aged 65 and above) of total population. According to Forecast of “2016–2022 China elder care industry market operation situation and investment strategy research report” released by Intelligent and Research Consulting, China’s elderly population will reach 483 million, accounting for © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 S. Sun et al. (Eds.): ICSINC 2021, LNEE 895, pp. 145–149, 2022. https://doi.org/10.1007/978-981-19-4775-9_17

146

X. Sun

34.1% of the total population by 2050. According to the prediction of the World Health Organization, China will be the most aging country in the world by 2050. It can be seen that China’s aging population has become a serious social problem. How to develop China elderly service industry has become an urgent problem to be solved. Starting from the current situation of the elderly service industry, this paper attempts to promote China’s elderly service industry through the use of modern information technology.

2 The Current Situation of China’s Elderly Service Industry 2.1 The Demand of Elder Care Institutions Exceeds the Supply According to civil affairs statistics, 38800 social service institutions provided accommodation in 2010, with 3.208 million beds and 17.8 beds per 1000 elderly people. By September 2015, the number of the above indicators had expanded greatly, reaching 46700, 6549000 and 32.3 respectively. In 2007, the institution of the elder care service nationwide was 155000, a year-on-year increase of 10.7%, but the growth rate decreased by 10 percentage points compared with 2016. The total number of beds in elder care service institutions nationwide in 2007 was about 7.448 million, the growth rate was 6.5% lower than that in the same period of last year. If 10% of the elderly over 65 years old in want to live in nursing homes, the growth rate of bed supply gap will continue to increase. 2.2 Most of the Empty Nesters Are Elderly With young people working in different places, parents living in different places with their children, not having much time and energy to take care of their parents, and the parents of the only children have entered their old age, there are more and more empty nesters in China. Data shows that from 2000 to 2010, the number of empty nesters in urban areas increased from 42% to 54%, and in rural areas from 37.9% to 45.6%; in 2013, the number of empty nesters in China exceeded 100 million [2]. Most of the empty nesters and the elderly living alone have become widowed and left behind. In 2015, empty nesters accounted for 51.3% of the elderly aged 60 and above, indicating that empty nesters have replaced the traditional family living mode of living with children and become the most important way of living for the elderly in China [3]. If calculated according to this proportion, the number of empty nesters aged 60 and above in China will be 128 million in 2019. 2.3 Lack of Service-Oriented Talents The employees of elderly service industry are also in short supply, especially those with professional skills. China Public Welfare Institute of Beijing Normal University released a report on talent cultivation of elderly care service in 2017, which shows that at least 13 million nursing staff are needed in China, but the actual number is less than 500000, and the number of certified personnel is less than 20000, which is a huge gap. In 2018, there

Modern Information Technology Develops Intelligent Elderly Care Service Industry

147

were only 186 colleges and universities in China that established the elderly service and management and other related majors, and the overall education level of nursing staff was low, the age structure was large, and the professional level was low. Among all the respondents, 21.7% were from primary school or below, 45.2% from junior high school, 26.3% from senior high school or technical secondary school, and 6.8% from university or above. The main group of front-line nurses in elder care institutions is women aged 40 to 50, and only 42.12% of the total number directly provide services for the elderly. 2.4 The Medical Level Needs to Be Improved In 2010, the number of disabled and semi disabled elderly in China reached more than 33 million. At the same time, with the growth of age, more and more geriatric diseases emerge, such as hypertension, diabetes, hyperlipidemia and so on. It is difficult to treat these common chronic diseases in the elderly, and they have to rely on drug treatment all the year round. At present, there are fewer doctors serving the elderly, less training for the staff engaged in the work of the elderly, and no special technical guidance. The relevant government departments often ignore the training of nursing staff in nursing homes and hospitals. The low medical level of medical staff directly restricts their mastery and use of medical technology [4].

3 How to Development Elderly Service Industry Modern information technology, which includes big data, cloud computing, Internet of things, artificial intelligence, and other new technologies [5], has brought unprecedented development opportunities to develop elderly service industry, like combining of medical care and elder care, innovating elder care model, developing intelligent medical care in elder service industry, enriching the spiritual life of the elder, and so on. In order to develop the elderly service industry, we should speed up the integration of modern information technology with the elderly service industry. 3.1 Improve Infrastructure Construction The rich of intelligent elder care, smart elder care and elder care model are inseparable from the infrastructure of rest, diet, entertainment, health care, rehabilitation, medical treatment, etc. The following aspects can be taken into consideration to improve the infrastructure construction: first, to build the elderly service institutions. The elderly service institutions shall be established in the community, and the government shall provide the support to establish the old-age club in the community, and improve them. Corresponding services shall be provided around the needs of the elderly in daytime care, including the establishment of elderly dining table, the establishment of elderly activity center and the elderly custody service. Second, improve the level of modern information technology related to the elderly service. Modern information technology such as 5G remote operation, 5G medical monitoring and AI clinical big data cannot be replaced in medical treatment and elderly care. The government, doctors, patients, the elderly and their children can realize data exchange, comprehensive application and

148

X. Sun

inquiry services through modern information technology. Third, it is urgent to expand the coverage of modern information technology. The health information early warning, slow disease and elderly disease auxiliary decision-making, medical and maintenance combined services, health knowledge learning and other functions are integrated into one, forming a comprehensive, professional and comprehensive public service network for medical care, which is a comprehensive, comprehensive public service network of provinces, cities, counties, towns (streets), villages (communities), and realizes the human, intellectual and material links of medical information and medical resources. 3.2 Strengthen the Training of the Aged Service Talents We should strengthen the management and reserve of the elderly care service industry and the personnel training and reserve, gradually improve their professional knowledge and skills training, and gradually implement the examination system of the endowment care qualification certificate. Support vocational colleges to set up nursing major for the aged. On the other hand, qualified and conditional social forces are encouraged to establish specialized training institutions for elderly care. Give excellent management personnel full space for professional development and treatment guarantee, encourage them to sink into the community and serve the community, promote the coordinated development of various elder care services, and improve the overall service quality of community elder care. The existing professionals shall be trained in professional knowledge and skills according to different requirements, and the vocational qualification system and technical grade certification system shall be implemented. The corresponding level of service shall be conducted according to the level of training skill assessment, and the work with certificates shall be achieved. At the same time, we can learn from the relevant experience abroad, establish professional elderly service nursing schools and train special talents. Finally, we should fully integrate and utilize all human resources, vigorously develop the volunteer team of elderly care services, improve the relevant training of volunteers, and strive to build a team of high-quality elderly care service talents combining professional and volunteer. 3.3 Courage All Parties to Participate in Elderly Service The elderly service industry belongs to the social public welfare industry. The government has the obligation to cooperate with relevant departments, to increase investment for the elderly service equipment, medical staff, elder care institutions, etc., from the market demand. And the comprehensive development of the elderly service industry should introduce professional social service enterprises or organizations to participate in the operation, and guide social forces to participate in home-based elder care services through facilities security, financial subsidies, purchase services and other ways, and form a multi-level, multi-channel and multi-channel investment and development system. Using big data management, promoting the system of community helpers for the elderly, and making accurate efforts in community home-based care services [5]. The “public service platform for big data for the aged” is built by the cooperation mode of school and local enterprises and medical research enterprises. It integrates the forces of residential community, elderly relatives, elder care institutions, medical institutions,

Modern Information Technology Develops Intelligent Elderly Care Service Industry

149

medical schools, local governments, science and technology enterprises, and living service institutions to provide professional, efficient, convenient and safe health care services for the aged. We will continue to expand publicity efforts, encourage all sectors of society to pay attention to the cause of the elderly care actively, and encourage social organizations, enterprises and institutions and individuals to donate funds, donations or provide free services to community elder care service institutions1 . 3.4 Do a Good Job in Overall Supervision The development of the elderly care service industry is inseparable from the support of people from all walks of life and the supervision of all walks of life. Whether in the process of infrastructure construction, or in the late use of modern information technology equipment, are inseparable from the supervision of people from all walks of life. First of all, we need to supervise the construction of infrastructure related to the elderly service industry to ensure that the funds are used exclusively and prevent corruption. Secondly, we need to supervise whether the use of elder care service facilities is standardized and reasonable, and whether it can give full play to its effectiveness. In the Internet plus Internet age, the elderly service needs to be supported by the personal information database of the elderly, so that the “Internet plus” elder care service can maximize the ecological function, and ultimately achieve the efficient operation of the “Internet plus” elder care service, providing a reasonable and high-quality service experience for the elderly [6]. Therefore, it is necessary to supervise the correct use of the information of the elderly by the corresponding institutions and organizations to ensure the information security of the elderly. Acknowledgments. This work was financially supported by Yunnan Education Department Teacher Projects Fund, “study on the sustainability of Chinese residents’ pension under the integration of urban and rural areas” (2020J1308).

References 1. 2. 3. 4. 5.

Spiker. China Trends, p. 232. Hualing Publishing House, Beijing (1996) http://www.15lu.com/shijie/6158.html https://www.sohu.com/a/317334834_300488 https://epaper.gmw.cn/gmrb/html/2019-03/25/nw.D110000gmrb_20190325_3-07.htm Xiao, L.: The exploration of the development mode of “Internet plus” pension service industry: based on a case study in Xiamen. J. Jiangsu Ocean Univ. (Humanit. Soc. Sci.) 19(01), 115–121 (2021) 6. Luo, Y., Jia, H.Y., Ma, Y.: Research on the pension service industry under the background of Internet plus. Software 41(11), 165–169 (2020)

1 https://www.sohu.com/a/345098404_762454.

Construction of Piano Live Broadcasting Platform Based on Wireless Network Communication Technology Heda Zhang(B) College of Music, Bohai University, Jinzhou, Liaoning, China [email protected]

Abstract. Information disseminated on the Internet is stored in optical or magnetic storage media in digital form, disseminates at high speed through computer networks or wireless networks, and is read and used by computers or mobile phones. Live teaching is to use the advantages of Internet communication to solve the problem of normalized teaching in the post-epidemic era. The core work of this paper is to provide solutions for the construction of college piano teaching live broadcast platform, including streaming media transmission architecture design, live broadcast platform function design and video live broadcast process design. According to the characteristics of piano teaching in colleges, various methods of online teaching are optimized to meet the diverse learning needs of students, and the practical path of piano live teaching in colleges is proposed to help piano teaching in the post-epidemic era. Keywords: Internet communication · Communication technology · Live video · Research and practice

1 Introduction With the outbreak of the novel coronavirus, the opening of schools was postponed, and the Ministry of Education proposed to use the online platform to “suspend classes without stopping learning and teaching”. Online teaching has become the main method of school education and has attracted the attention of many people. Online teaching is a new form of education that uses the Internet as the media. It has made new breakthroughs in organizational models, teaching models, and service models, and has become an important part of the new educational ecosystem in the digital era [1]. Live teaching has become an important method of online teaching. The use of technology platforms and related conditions are limited. The current application platforms cannot play greater roles and functions, and are mainly restricted by three conditions [2]: First, platform operation with the lack of hardware; second, the defect of the technical conditions; third, teachers and students are not proficient in operation and use. Although the goal of suspending classes but not stopping school is guaranteed, a large number of formalisms has also grown in the process of teaching work, which reduces the effectiveness of teaching. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 S. Sun et al. (Eds.): ICSINC 2021, LNEE 895, pp. 150–157, 2022. https://doi.org/10.1007/978-981-19-4775-9_18

Construction of Piano Live Broadcasting Platform

151

First, the negative treatment and coping attitude in the pre-teaching curriculum; second, the lack of flexibility in the teaching process. Students’ degree of participation is low. Watching related videos during online learning shows that students are only passive in completing tasks, and their enthusiasm has not been mobilized. Third, some teachers have dealt with and ignored the homework after class. The completion of homework basically relies on the autonomy and consciousness of students, and the role of teachers is weakened. Piano originated from western classical music, and is very popular in the music art of various countries in the world, and has a very high artistic status. It is called the “king of musical instruments”. As the cradle of cultivating technical talents, universities should have the concept of innovative teaching, and stimulate students’ potential to learn piano with innovative teaching. This requires reasonable arrangement of teaching content, changing teaching concepts, and improving students’ artistic culture by improving teaching models. In the post-epidemic era, live teaching will become a normalized teaching method. In the process of live piano teaching, teachers should pay attention to presenting the teaching content in blocks, be good at using repetition and summarization; pay attention to interaction, effectively ask questions, and enable students to be affected by the active participation of peers through online discussions; pay attention to promote deep learning and use to provide thinking questions methods such as setting up exploratory tasks; focus on innovative learning technologies, enriching teaching scenes through wearable devices and VR/AR, make live teaching more real and improve students’ enthusiasm [3]. In the era of artificial intelligence and 5G, live teaching has become an important way to solve education problems in the post-epidemic era, and it is an important force in promoting the reengineering of education processes and creating future education.

2 Construction on Piano Live Teaching Platform in Universities The novel coronavirus has promoted the development of live teaching. Currently, live teaching has the following platforms to choose from [4]: Cloud Classroom APP. Beijing Yuxin Technology Co., Ltd. provides online video teaching software including live teaching, recording and student management functions for educational institutions, enterprises, teachers and individuals, and can quickly build their own online education platform. Haoshitong. Shenzhen Huashi Ruitong Information Technology Co., Ltd. products, registered trademarks of video conferencing products. The product focuses on the Internet, 3G mobile Internet business meetings, and collaborative office areas, and is committed to providing Chinese enterprises with international standard multimedia communication services in the form of software as a service. However, in order to adapt to the development of live teaching in the post-epidemic era, universities must get rid of dependence on commercial software and build their own dedicated teaching live teaching platform. Aiming at the characteristics of piano teaching, this article provides solutions to the main content of the construction of piano teaching live teaching platform in universities.

152

H. Zhang

2.1 Streaming Media Transmission Architecture The streaming media transmission architecture is shown in Fig. 1. Terminal interface (4) (2)

(11)

RTSP (1)

Hardware coding decoder

(6) (5)

TCP/IP (21)

(3)

RTP RTCP

(12) (13)

(7) TCP/IP

(22)

Inter-program interface transmission (1) Media data input; (2) Media data output, Video display; (3) Frame output of media data; (4) Interface operation; (5) Parameter setting, behavior control; (6) Data exchange; (7) Data exchange. Streaming media protocol link (11) Request or reply, Transport mode and media format coordination; (12) Media data transmission; (13) Network quality inspection, quality feedback report. IP network transmission link (21) Network layer data transmission; (22) Network layer data transmission.

Fig. 1. Streaming media transmission architecture

2.2 Live Teaching Platform Function The SDK is a specific software package or plug-in package, which assists and expands the existing carrier functions. During the live teaching, AVSDK is a collection of plugins for centralized processing of audio and video, including a series of functions such as camera capture, encoding, decoding, beautification and cuteness. For different platforms, the internal function realization can be shown in Fig. 2 [5]. 2.3 Live Video Streaming Process The live video teaching process is shown in a sequence diagram as shown in Fig. 3 [6].

Construction of Piano Live Broadcasting Platform Android

Android

Rendering

Rendering

Equipment Gravity induction

PC

IE web

Equipment

C++

ActiveX

Gravity

COM

COM

Adaptation Session management

Network control

Mixer

Rendering control

Acquisition control

Audio engine

Video engine Equipment control

Platform / hardware adaptation layer

Fig. 2. Live broadcast platform function Streaming media server

User terminal

(1)

Video capture terminal

Web server

(2) (3) (4)

(5) (6) (7) (8) (9)

(1) Play request (2) Authentication request (3) Authentication channel (4) Return authentication result (5) Return request result (6) Obtaining video (7) Return call result (8) Call successfully returned the video stream (9) Real time video streaming

Fig. 3. Live video process

153

154

H. Zhang

3 Practice Path on Piano Live Teaching in Universities in the Post-epidemic Era The novel coronavirus has stimulated the vitality of re-examining online education, forcing educators to re-examine teaching methods, learning methods and evaluation reforms, optimize various methods of online teaching, and use online live teaching as a normalized teaching method and penetrate into the classroom teaching, adapt to the diverse learning needs of students [7]. In view of the characteristics of piano teaching in universities, the practical path of piano live teaching in universities is proposed to help piano teaching in the post-epidemic era. 3.1 Pay Attention to the Cultivation of Students’ Musical Emotions Emotional communication is indispensable in teaching activities and teacher-student interaction. In education, the spiritual impact, emotional blending and echoing between teachers and students is not only reflected in the cognitive process, but also in the process of emotional communication. Piano performance expresses the player’s own musical emotions, not to highlight the piano or the music itself. The emotional education of students and the emotional communication between teachers and students cannot be ignored in live teaching. Teachers, students and learning resources are the three emotional sources of teacher-student emotional communication, and they are an organic whole that complements each other [8]. Establish classroom teaching support system with the main lecturer as the main teacher and the assistant teacher as the supplement; take the student as the main body to stimulate the student’s subjective initiative and classroom participation consciousness; start from the teaching resources, strengthen the emotional design of synchronized live classroom teaching content. Emotions are incorporated into the performance process, which gives vitality to the performance, which arouses the resonance of students. Only by understanding the connotation and emotion, can you perform art with soul. 3.2 Expand Students’ Imaginal Thinking Imaginal thinking is the thinking process with intuitive image and appearance as the support. In the process of creation, artists are always accompanied by images, emotions, associations and imaginations, using individual characteristics of things to grasp the general laws to create a way of thinking about artistic beauty. Thinking in images can enrich the imagination of students and guide them to grasp and explore the ability of things in the process of exploration and learning. The process of artistic creation is inseparable from rich imagination. The foundation of imagination is thinking in images. The piano playing process carries emotional experience and performs corresponding interpretation and expression. This verbal indirect interpretation and expression of musical emotions stems from the non-verbal nature of musical image thinking and determines the emotional internalization carried by the music external manifestations [9, 10]. Live piano teaching should follow the objective laws of psychology and pedagogy to promote students’ image thinking steadily.

Construction of Piano Live Broadcasting Platform

155

3.3 Adopt Diverse Teaching Methods The biggest difficulty of live teaching is the inability to grasp the learning effect of students. Piano skills training requires repeated practice to achieve proficiency and form solid memory. During the live teaching process, not only the technique but also the practice method must be taught. For the key and difficult points of technology, teachers should teach repeatedly, and use auxiliary practice methods to promote complete technical mastery. Therefore, it is recommended to adopt diversified teaching methods, and adopt flexible teaching methods according to teaching content, teaching situation and teaching objects, which will help maintain a good teaching environment, mobilize students’ enthusiasm, and ensure teaching results. Carry out online MOOC teaching to help students build the overall knowledge structure and systematic thinking of piano, form a more scientific, systematic and comprehensive cognition, promote the organic integration of knowledge, ability and quality, and realize the development of piano teaching ability. Let students think about piano playing technique and musical expressiveness from the perspective of “teaching”, so as to gain a deeper understanding of the charm of piano art [11]. Teachers send relevant course information to students through the teaching platform, WeChat and QQ before class, including course content, assessment methods and grading standards. Students use mobile phones and computers to choose the MOOC they are interested in for learning. 3.4 Further Optimize Piano Teaching Design There are different understandings of piano performance courses from different views. From the perspective of teachers, students can learn more knowledge and the content of the curriculum is relatively rich; from the perspective of musicology, I hope that the design of the course content will be more humane, improve students’ artistic culture, and develop students’ personality and cultivate humanistic feelings. Based on the above factors, live piano teaching needs to highlight individuality and optimize teaching design from many aspects [12]. First is to add a variety of musical elements. Incorporate the characteristic musical elements of various ethnic groups and countries into the live teaching to expand students’ thinking. In the free development activities, students are encouraged to play the songs they have learned, and add different notes or melody according to their hobbies, making them a vehicle for expressing emotions. For example, there is no fixed word limit for each sentence of lyrics in western music, and there is no uniform atmosphere. Students can add Chinese folk music melody in the process of free performance to reconcile the two to create unique music. Second is from a practical perspective. In the live teaching process, teachers should give full play to the beautiful melody of piano performance, guide students to invest in learning, and mobilize the initiative of learning. The design of the course follows the principle of practicality, highlighting students’ performance ability. 3.5 Share Online Piano Teaching Resources Educational technology is the theory and practice of designing, developing, using, managing and evaluating the processes and resources related to learning. Teaching resources

156

H. Zhang

have an important position. Use a new generation of smart technology to expand the open sharing of high-quality educational resources, enrich the supply of high-quality educational services, innovate talent training models and teaching methods, promote the intelligent education process, the intelligent education management, and the precise education services to achieve fairness and high quality The education in the country provides strong support for accelerating the modernization of education. Online piano teaching requires sharing of teaching resources. The large-scale promotion of piano online teaching mode can break the traditional education pattern, allow teachers and students to exchange piano learning skills in an open platform, make full use of learning resources, expand teaching thinking, and make piano online teaching activities a virtuous circle. Online teaching resources are shared, and social forces have become the main source of resource supply, helping college piano education and teaching work, and promoting the steady progress of piano online teaching activities [13]. Relevant departments should formulate piano online education standards, strictly review online teaching institutions, ensure the quality of piano online teaching, and provide quality services for piano live teaching in universities. 3.6 Reasonable Implementation Strategies for Live Teaching Specifically, it includes the following three aspects [14]: first, the choice of live teaching mode. Based on the live teaching model of lectures and deductions, it supports students to complete the whole process of “getting what they have learned, and applying them to use”. Every link implements “targeting, thinking, and evaluating”; the flipped classroom live teaching model based on online courses, in accordance with the process of “selflearning, live-streaming Q&A, organizing exercises, and comprehensive application”. In the live-streaming Q&A session, the goal is to enable all students to achieve “exquisite learning” of the learning content without leaving questions. Second, the choice of live teaching methods. The current teaching live teachings mainly include “voice live teaching, video live teaching, voice + PPT live teaching, video + PPT live teaching”, etc. It is recommended that the two forms of “voice + PPT live teaching or screen sharing video live teaching” should be used first, as there are visual images information, together with auditory voice information, is the most ideal “dual-channel” multimedia teaching form under current conditions. The third is the timing of live teaching. In the live teaching mode of lectures and deductions, live teaching is recommended in the first link “live teaching”, and in the flipped classroom live teaching mode based on online courses, live teaching is recommended in the second link “live Q&A”. If the teacher has the ability to develop micro-courses, it is recommended to choose the flipped classroom live teaching mode based on online courses.

References 1. Fu, W.D., Zhou, H.Y.: Challenges brought by 2019-nCoV Epidemic to online education in China and coping strategies. J. Hebei Normal Univ. (Educ. Sci.) 22(2), 14–18 (2020) 2. Xia Ri: Online teaching under the background of epidemic situation: problems and countermeasures. Pengbai news (2020). https://www.thepaper.cn/newsDetail_forward_606 3562

Construction of Piano Live Broadcasting Platform

157

3. Qu, Q.M., Zeng, J.L., Shang, Z.J.: How to effectively carry out online teaching during the epidemic period. Teach. Educ. Forum 34(4), 21–24 (2020) 4. Wen, C.: Ranking of teaching live streaming platforms in 2020. China Internet Week 23(5), 36–37 (2020) 5. Personal library: network video broadcast system construction (2020). http://www.360doc. com/content/19/0927/17/66395326_863553133.shtml 6. Zhang, Y.: Design on mobile live broadcasting platform for aerobics video teaching based on android. Inf. Technol. 44(5), 41–44 (2020) 7. Shen, X.: The future and prospect of online live teaching in the post-epidemic times. J. Hubei Open Vocat. Coll. 33(21), 40–41 (2020) 8. Liu, J., Zhou, Y.X., Xie, J.L.: An analysis of and countermeasures to emotional communication problems between teachers and students in live broadcasting distance classroom. J. Chuxiong Normal Univ. 29(12), 56–61 (2014) 9. Liu, H.M., Xiao, J.: How to develop students’image thinking ability in piano teaching. Northern Music 31(2), 81–82 (2011) 10. Guo, R.: On the cultivation and training of students’image thinking in piano teaching. Popular Lit. Art 57(5), 260 (2012) 11. China University MOOC: Piano teaching method. https://www.icourse163.org/course/FJNU1206674826?tid=1450691456, (20 Nov 2020) 12. Yin, X.Y.: Discussion on the design path and teaching reform measures of piano playing courses in colleges and universities. Peony 64(8), 199–200 (2020) 13. Zhao, X.: Construction of sharing and evaluation mechanism of piano online teaching resources in colleges and universities. Mod. Music. 36(10), 52–54 (2020) 14. Zhang, L.J., Cao, D.B.: Design and implementation of online teaching based on live broadcast technology. Chin. J. ICT Educ. 26(14), 88–92 (2020)

Value Education System of College Students Based on Mobile Internet Technology Shizhe Zhang(B) School of Marxism, Shandong Management University, Jinan, Shandong, China [email protected]

Abstract. Values are the concentrated expression of contemporary Chinese spirit, condensing the common value pursuit of all people. To actively cultivate and practice values in the whole society, college students should be at the forefront. While mobile Internet technology brings opportunities to the value education of college students, it also brings challenges to the value education of college students. Therefore, giving full play to the advantages of mobile Internet technology, guiding college students to cultivate and practice values, and promote the smooth realization of values education goals are important practical issues that colleges urgently need to solve. This article aims to learn how to use mobile internet technology in students’ educational systems. You need to understand the current state of the mobile Internet based on the analysis of valuable knowledge, the opportunities provided by valuable knowledge of students in mobile Internet technology, as well as dilemmas that lead to valuable student education on the mobile Internet technologies. University technology, this article surveyed two universities and made recommendations for existing topics. The survey results show that on the question of “what do you think the existing methods of value education in colleges have?”, 22.59% and 24.67% of people choose a single form of teaching, which is unattractive, and 26.37% and 25.19% of people choose outdated teaching content, the timeliness is poor, 29.83% and 20.76% of the people choose the lack of campus culture construction and campus management system. Keywords: Mobile networking technology · College students · Values education · Education system

1 Introduction With the rapid growth of the Internet industry and popularization of applications, the Internet has been organically integrated with all walks of life in society, and the industrial structure of traditional industries has been upgraded [1, 2]. The concept of “Internet+” has been included in the government work report since it was put forward. With the advent of the “big data” era, “Internet+” has risen to become a national strategy [3, 4]. The growth of mobile Internet technology has brought unprecedented impact to society. As the main user group of mobile Internet technology, college students can not only receive information at any time, but also have the right to choose information, which can create the content of information and change the form of dissemination [5, 6]. On the © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 S. Sun et al. (Eds.): ICSINC 2021, LNEE 895, pp. 158–166, 2022. https://doi.org/10.1007/978-981-19-4775-9_19

Value Education System of College Students

159

one hand, the unlimited communication and communication space, flexible and diverse carriers and personalized services of mobile Internet technology have enriched the forms of value education for college students, and at the same time expanded the selection of educational content, providing new opportunities and ways for college students’ values education: On the one hand, the complexity of mobile Internet technology content and open information flow will inevitably bring negative effects and challenge college students’ values education. Therefore, studying how to use mobile Internet technology to strengthen the value education of college students has practical value and significance [7, 8]. This article aims to examine the use of mobile network technology in student education systems. You need to understand the use of mobile network technologies, analyze the value of education, mobile network technology provides opportunities for college students to learn, and mobile network technology problems when evaluating students. The high school made appropriate suggestions on the current issue.

2 Application of Mobile Internet Technology in the Value Education System of College Students 2.1 Value Education Value education is an educational activity in which educators exert value influence on the educated and guide the formation of their values to meet the requirements of social growth. Value education includes family, school, and social education. Family education is the cradle of values formation. School education is an important link in shaping students’ values, and social education is an accelerator that affects changes in values [9, 10]. Values education has a wide range. “The purpose is to guide the public to establish ideological concepts that are in line with social growth, use correct values to guide practical activities, and promote economic and social growth and the improvement of personal quality.” Family values education is based on family tradition and family motto. Basic, presenting different educational methods in specific family education. Affected by the internal drive of the attachment, the educated learnt the behaviors, words, and habits of family members in the early stage, and gradually formed opinions on some things and formed a certain value orientation. However, the values formed at this time were vacillating and unsuccessful. The system is extremely susceptible to external influences. School values education is mainly based on classroom education and practical activities, with values running through it. Classroom education infiltrates the rational knowledge of values, and strengthens the correct requirements of values through practical activities. Under the influence of cognitive drive, the educated is eager to obtain further knowledge and guidance on values, and the educator helps the educated to build Value system, educated persons gradually form a correct understanding of values, with clear thinking and firm value stand. Social values education has two sides [11, 12]. On the one hand, positive social phenomena and events have a positive impact on the consolidation and strengthening of the values of the educated. Relevant departments or the media can expand the audience of correct values through wave-like communication and

160

S. Zhang

flat communication, and create a clean and upright attitude. On the other hand, corrupting wrong values and the impact of negative foreign cultures pose challenges to the social education of values, especially threats to the cohesion and sense of identity of values. Relevant departments must strive to optimize the value education environment, weaken or even eliminate negative factors, and provide conditions for social education of values. 2.2 Opportunities Brought by Mobile Internet Technology to College Students’ Values Education (1) Mobile Internet technology has expanded the form of value education for college students With the growth of mobile Internet technology, colleges can actively build an “online” model, set up value columns in WeChat and Weibo official accounts, and link with mobile Internet technologies such as the homepage of college campus networks, and add a “values” learning zone, regularly release learning information, show the activities of colleges, and effectively use the cloud system to move the learning form from offline to online. Screen reading, online reading, and audio-visual integration have become the learning methods for college students. In addition, some colleges use mobile Internet technology carriers to select advanced models such as the top ten moral models, the top ten virtuous youths, and the top ten innovative teams. Announce publicity in the mobile Internet technology carrier, attract the attention and reading of college students, guide the majority of college students to study harder, and provide strong spiritual motivation and rich moral nourishment for filling every corner of colleges with value education content. (2) Mobile Internet technology has enriched the content of value education for college students The purpose of value education for college students is to guide college students to establish correct values in the new situation and new era. The emergence of mobile Internet technology has provided new teaching content for college value educators, applying and developing new teaching methods such as micro-classes, and using mobile Internet technology teaching software to enrich classroom content. Some colleges use mobile Internet technology to carry out value teaching micro-class competitions, establish WeChat contact groups to interact, exchange micro-class production methods and mobile Internet technology skills through the combination of teaching and experience, and use Weibo, WeChat and other carriers to share feedback and comments, opinions, experience, etc., the educational content greatly breaks through the limitations of textbooks. At the same time, with the fast search characteristics of mobile Internet technology, hot topics, social events, focus of public opinion, popular vocabulary, hit videos, etc., can all become the subject of education, enriching the content of college students’ values education, and making the content more lively. (3) Mobile Internet technology has enhanced the effect of college students’ values education Relying on the extremely interactive nature of mobile Internet technology, it has created many mobile Internet technology products that college students like to hear and see. Some colleges make full use of mobile Internet technology carriers

Value Education System of College Students

161

to promote and establish models, set up corresponding sections or topics through Weibo to promote the spirit of the 19th National Congress of the Party, and divide the inner nature of the party and the country’s policies and values. College students can display their ideas and creativity through mobile Internet technology, and through the production of short videos and pictures that promote values, college students can get more exchanges of opinions, get more responses, and eliminate college students’ resistance to boring text. Psychology changes from passive learning of knowledge to active acquisition of knowledge. The ability of students is continuously improved, making value education more convincing and attractive. 2.3 The Dilemma that Mobile Internet Technology Brings to College Students’ Value Education (1) It is easy to cause the recognition of college students’ values College students’ world outlook, outlook on life, and values are still in the forming stage, and their minds are not yet fully mature. There is a lack of initiative for theoretical knowledge, and some students’ cognition of values is mostly superficial, and it is difficult to understand its essence and connotation. And because of the virtuality and openness of “Internet+”, some behaviors that do not conform to the Internet ethics have appeared on the Internet, in various forms such as online violence, pornography, and fraud. Driven by economic interests, some people will To make some behaviors that break through the moral bottom line for personal benefit, and these behaviors just deviate from the values. In the face of this situation, some college students with poor self-control are very easy to get lost and have a tendency to identify with values. (2) It is easy to influence the systematic effect of the value education of college students The change of teaching form easily affects the systematic effect of values. First, the controllability and purity of the public opinion environment of colleges has been affected to a certain extent. Second, it has had a greater impact on the educational environment of colleges, and the educational environment has become more and more complicated. The past education methods were relatively simple and pure. Teachers and schools had a certain mandatory influence on students’ ideology, and the public opinion environment was also relatively pure. In the context of “Internet+”, an open and free environment is self-evident for college students, has also produced a greater dependence on this convenient way of life. Compared with the past, the share of college students receiving ideological education from schools and teachers is also sparse, and instead choose to obtain timely and diverse information from the Internet. This kind of knowledge acquisition system change can easily affect the system effect of college students’ values education.

3 Experiment 3.1 Questionnaire Design According to the research needs of this article, on the basis of analyzing the previous literature and referring to a large number of relevant research reports, questionnaires were

162

S. Zhang

randomly distributed to college students in two universities in the province for questionnaire surveys. The questionnaire was produced by the “Questionnaire Star” software. 1,000 questionnaires were distributed and 895 valid questionnaires were returned. The effective rate of the questionnaire was 89.5%. 3.2 Reliability Test of the Questionnaire To test the reliability and validity of the questionnaire, we first calculated the difference between the results of the questionnaire and then tested the reliability of the questionnaire returned using the “semi-reliability” test. When calculating the reliability coefficient using Eq. (1), the correlation coefficient of the questionnaire r = 0,883. According to modern research theories and methods, a reliability test of 0.80 or higher can be considered the most reliable test. The results of the test confirm that the questionnaire is reliable. S2 =

(M − X1 )2 + (M − X2 )2 + (M − X3 )2 + ... + (M − Xn )2 n r = 1 − S 2 (1 − r1 )/Sn2 r=

2rban 1 + rban

(1) (2) (3)

4 Discussion 4.1 Survey Results Figure 1 shows the survey results of the question “What is your main way to learn values?”.

Fig. 1. What is your main way to learn the core values of socialism

Value Education System of College Students

163

As can be seen from the above figure, 53.26% and 57.42% of the two colleges choose the Internet, 23.71% and 25.62% choose textbooks, 15.33% and 11.45% choose their parents to teach, 7.7% and 5.51% People choose to watch TV. It can be seen that the Internet and schools are the most important platforms for college students to learn values. Taking measures from these two levels can effectively strengthen the education of college students’ values at the root. Table 1. Educational methods of socialist core values in colleges

Ideological and Political Course

College A

College B

60.33%

63.73%

Campus network platform promotion

16.41%

14.26%

Various cultural and sports activities on campus

12.89%

15.12%

4.37%

6.89%

Other

Fig. 2. Educational methods of socialist core values in colleges

According to Table 1 and Fig. 2, on the question of “Which way does your school carry out values education?”, 60.33% and 63.73% of the two colleges choose ideological and political theory courses, 16.41% and 16.41% and 63.73%. 14.26% of people choose the campus network platform for publicity, 12.89% and 15.12% choose various cultural and sports activities on campus, and 4.37% and 6.89% choose other methods (Table 2). Regarding the question “What do you think are the deficiencies in the existing methods of value education in colleges?”, 22.59% and 24.67% of the people choose a single form of teaching, which is unattractive, and 26.37% and 25.19% choose outdated teaching content and poor timeliness. 29.83% and 20.76% of the people choose the lack of campus culture construction and campus management system. 4.2 Targeted Suggestions (1) Strengthen the construction of network platform

164

S. Zhang Table 2. The insufficiency of the education mode of socialist core values in college College A

College B

The teaching method is single and unattractive

22.59%

24.67%

Outdated teaching content and poor timeliness

26.37%

25.19%

The construction of campus culture is not strong and lack of educational atmosphere

29.52%

30.14%

The campus management system is not sound

21.52%

20%

Paying attention to supervision not only requires strict adherence to the bottom line from the legal level, but also the integration of the original hardware resources, strengthening the construction of network platforms, and purifying the network environment from the technical level. The first is to attach importance to the use of information network technology, increase R&D and investment in related technologies, establish a big data database on the network platform, analyze users’ network behaviors under the premise of ensuring user privacy, and provide early warnings for high-risk users and high-risk behaviors. The second is to speed up the establishment of an information tracing system, establish a complete identity authentication system, automatically back up suspected negative information, prevent the destruction of evidence through deletion operations or modify platform ID operations, and strictly implement online platform sanctions on criminals. According to the identity authentication system, the sanctions are held to individuals, not just accounts, so that people are in awe and dare not browse, spread, or fabricate bad information. The third is to develop more excellent educational resource platforms such as MOOC and “Church Network”, and establish an educational resource database to make full use of massive educational resources. (2) Attach great importance to classroom teaching The living habits and growth environment of college students are affected by the Internet, and they are accustomed to communicating with people through the interactive mode of the Internet to obtain fragmentary information and knowledge. To improve the ability of college students to discriminate Internet information, classroom teaching must be highly valued. On the one hand, we need to innovate the classroom teaching model, realize the innovative growth of traditional classroom teaching, give the ideological and political theory courses the charm and affinity of the new era, and actively construct a competitive interactive teaching model under the background of the mobile Internet. On the other hand, it is necessary to update the content of classroom teaching, combine with the requirements of the times, and continuously integrate values into the process of students’ ideological and political education. (3) Innovative classroom teaching mode Colleges should combine the advantages of mobile Internet technology to optimize classroom teaching methods, make deep use of mobile Internet technology to give colleges the connotation of a new era of value education, improve education

Value Education System of College Students

165

methods, and form a rich, scientific, people-oriented, and innovative education network that cultivates sentiments. The continuous growth and application of Internet technology has made learning no longer limited to the content of paper media. The unique interactivity of online courses and the richness of learning resources have prompted colleges to transform into diversified educational models. Therefore, colleges must try to optimize the way of value teaching under the background of mobile Internet, but still maintain the dominant position of teachers’ teaching, and improve and promote related education models. (4) Speed up the construction of campus culture Colleges must adhere to the core of establishing morality and fostering people, and the construction of campus culture is a long-term plan. Therefore, colleges need to carry out top-level design and create the spiritual home of college students in the new era with goals and plans. The first is to accelerate the construction of campus network culture, and establish a cultural construction platform based on campus network through the campus network covering the whole school to promote values. Conduct psychological interventions and answers to college students through the network platform, adhere to the people-oriented approach, and better protect the privacy of students. At the same time, you can also use the campus network micro-platform to promote the style of teachers and students, mainstream culture, and broadcast news and various activities on campus. The second is to speed up the construction of campus spiritual culture. The construction of campus spiritual culture is the core content of the construction of campus civilization. It should be done from the small to the big, and through various cultural and sports activities to help college students shape good character, help them out of the network world, cultivate values in practice, and promote the construction of campus spiritual civilization.

5 Conclusions With the rapid growth of information technology, mobile Internet technology has profoundly shaped a new dissemination pattern since its inception, changed the past channels of information dissemination, and greatly enriched people’s spiritual and cultural life needs. In the process of accelerating the global transformation to the Internet society, the value education of college students in my country has also kept pace with the world and walked with the times. In order to adapt to the physical and mental characteristics and growth laws of college students, the use of mobile Internet technology to innovate the education carrier of college students’ values and expand educational forms has become a hot issue in academic research.

References 1. Yu, C.: Development of a smart home control system based on mobile internet technology. Int. J. Smart Home 10(3), 293–300 (2016) 2. Purba, M.J., Manurung, S.: Analysis of 4G internet technology quality in medan city with mobile communication system. J. Phys. Conf. Ser. 1361(1), 012030 2019

166

S. Zhang

3. Tetarave, S.K., Tripathy, S., Ghosh, R.K.: GMP2P: mobile P2P over GSM for efficient file sharing. In: Negi, A., Bhatnagar, R., Parida, L. (eds.) ICDCIT 2018. LNCS, vol. 10722, pp. 217–231. Springer, Cham (2018). https://doi.org/10.1007/978-3-319-72344-0_18 4. Wang, Y., Li, S., Jie, H.: Hierarchical medical system based on Big Data and mobile internet: a new strategic choice in health care. JMIR Med. Inform. 5(3), e22 (2017) 5. Milla, J., Martín, E.S., Bellegem, S.V.: Higher education value added using multiple outcomes. J. Educ. Measure. 53(3), 368–400 (2016) 6. Eyuuml, B.P., Tel, M.: Opinions and perceptions of physical education students about value education. Educ. Res. Rev. 11(20), 1918–1924 (2016) 7. Altikulac, A.: Values and value education according to students of education faculty. Int. J. Educ. Technol. Sci. Res. 5(13), 1881–1932 (2020) 8. Hajaroh, M., Rukiyati, R., Pamungkas, J.: The diffusion of value education model in early childhood through traditional songs and games. Jurnal Kependidikan Penelitian Inovasi Pembelajaran 3(1), 1–14 (2020) 9. Szbir, S.A., Akmak, Z.A.: Analysis of children’s songs by Muammer Sun in terms of value education. Int. Online J. Educ. Sci. 12(5), 205–222 (2020) 10. Kemal, K.: Comparison of seventh-grade Turkish and Iranian social studies textbooks in terms of value education. Educ. Res. Rev. 14(17), 595–607 (2019) 11. Sulasman, S.: The value education of Qosidah Burdah arts in boarding school in The Land of Sunda. Al-Tsaqafa Jurnal Ilmiah Peradaban Islam 15(2), 139–154 (2019) 12. Aslan, S.A.: Internalization of value education in Temajuk-Melano Malaysia Border School. Edukasia Jurnal Penelitian Pendidikan Islam 14(2), 419–436 (2019)

The Application of Computer Virtual Reality Technology in the Athletic Training of Colleges and Universities Juan Yin(B) Jiangsu University, Zhenjiang, Jiangsu Province, China [email protected]

Abstract. Computer virtual reality (VR) is a technology that generates a 3D environment with the help of a computer, which allows the user to navigate and interact with it, thus simulating one or more of the user’s operations in real time. As the technology has a good migration, it is used in many fields, and in recent years, the application of VR technology has also received attention in higher education sports training. With its significant advantages of immersion, visualization and interactivity, VR technology provides an efficient way to train in sports, further improving students’ athletic performance while recording a large amount of highly accurate and reliable training data, further promoting the development of sports training in higher education. Keywords: Computers · Virtual reality · Colleges · Sports training

1 Introduction With the continuous development of electronic information and communication technology, computer virtual reality (VR) technology has been equipped with the function of establishing higher precision virtual 3D scenes. In virtual scenes, users can use their senses to fully experience the information stimuli from the environment and give feedback in real time, which makes this technology widely used in the field of information services, education, etc. In university sports training, VR technology allows students to achieve different intensity of sports training in different scenarios, thus further improving their athletic ability.

2 The Technical Advantages of Computer VR Technology 2.1 Information Interaction Characteristics This technology can fully satisfy the integration of different information models, and effectively construct a spatial artistic conception required for teaching through the twoway information interaction law [1]. In this process, teachers can create an effective © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 S. Sun et al. (Eds.): ICSINC 2021, LNEE 895, pp. 167–173, 2022. https://doi.org/10.1007/978-981-19-4775-9_20

168

J. Yin

space model with the help of designated operating commands, which can systematically reflect different landscape characteristics. At the meantime, the technology links the real scene and the virtual scene, and summarizes the information of the two in a corresponding method, allowing students to transform the atmosphere in different scenes, thereby clarifying the central value of the corresponding scene. 2.2 Perceived Characteristics VR technology (VR) technology uses an integrated information management model to present different scenes in the form of images. At the same time, VR technology integrates theories from different perspectives. With the help of content such as induction mechanics and electromagnetics, different action forms are processed and judged in specific information models, and different signal models are used for presentation and guidance. Students conduct perceptual learning in a holographic spatial environment. 2.3 Features of Immersive Experience VR technology can simulate situations according to people’s needs for different scenarios, and touch the value of physical perception with spiritual concepts, thereby fully satisfying students’ basic needs for situations. This technology can make the situation more realistic and guide students to deeply experience the charm of physical education courses by constructing an intended spatial form [2].

3 The Characteristics of VR in College Sports Training 3.1 Motor Skill Training In university sports training, teachers will first teach students basic movements and demonstrate mistakes, after which students will practice on their own and gradually improve. However, as different students have different physical qualities and learning abilities, it is difficult for teachers to provide one-to-one guidance to each student, and it is also difficult for students to find a suitable training programmer for themselves, while VR technology will develop suitable training scenarios for each student, conduct targeted training and improve students’ sports skills. 3.2 Psychological Training Due to the variability and complexity of the sports competition environment, students often find it difficult to play at their normal level. With the different competitive scenarios constructed by VR technology, students can experience different playing field environments in all aspects to ensure they have strong adaptability and achieve the best level of competition [3].

The Application of Computer Virtual Reality Technology

169

3.3 Theoretical Knowledge Learning The theoretical system of sport is a huge body of knowledge, which contains many aspects of sports training, tactical methods, physical fitness and so on. The original teaching method can only rely on teachers to explain the basic theory and students to practice and experience it on their own, which is less efficient and slower to improve [4]. VR can build a highly realistic and immersive training environment based on scientific sports training theories, so that students can target the strengthening and enhancement of all senses, thus more vividly and deeply understanding and remembering sports theory knowledge, improving training efficiency and promoting the improvement of sports level. 3.4 Virtual Exercise Experiment The sports training environment built by VR technology also has real-time feedback. The system will monitor students’ various physical indicators in real time and make timely targeted adjustments according to the training plan and relevant theories, while at the same time, the system automatically collects training data from each trainer and uploads it to the backstage records to help teachers adjust their teaching plans and improve training efficiency in a timely manner.

4 The Importance of VR Technology in College Sports Training 4.1 Stimulate Students’ Enthusiasm for Learning Computer VR technology, is a deep virtualization technology based on artificial intelligence. Thanks to the high transmission rate, low latency and high reliability of 5G and 6G communication technologies and the increasing development of computer technology [5–7], VR technology can be applied in more and more fields. The VR technology-based university sports training system has changed the original training mode of low efficiency and low autonomy, and has built a 3D training environment with relevance, variability and diversity for students, shortened the average time spent on sports learning, reduced the probability of training injuries, and stimulated students’ enthusiasm for learning. The introduction of VR technology has also reduced the cost of teaching sports in colleges and universities by eliminating the need to use existing training equipment [8]. 4.2 Guarantee the Safety of Students Sports training in colleges and universities is often highly competitive, but because of the fierce competition in sports competition, students often operate improperly for better results, triggering a greater chance of injury to students, and schools generally have greater concerns about this [9]. The use of VR technology in sports training can effectively reduce the safety problems caused by students’ training errors, and at the same time, VR technology will get rid of the constraints of the training venue, so that students do not have to be bound to the existing equipment, providing students with a better training environment and reducing training costs.

170

J. Yin

4.3 Stimulate Students’ Innovative Ability With the help of VR technology, sports training will no longer be limited by equipment and time, and will effectively enhance and improve the teaching and learning environment in schools. The use of VR technology in the computer world has revolutionized the sports training model in universities and institutions. Student autonomy has always been a large part of sports training. In order to address the higher-level training needs of students, high-tech teaching concepts are gradually entering the sports training classroom in colleges and universities [10]. As shown in Fig. 1, students are motivated by the need to be creative. In addition, the use of computer VR technology allows students to use advanced equipment and training grounds that are not available in the college sports classroom. It allows students to truly enjoy learning, thus achieving the school’s teaching effectiveness and realizing the goals of contemporary college sport training.

Fig. 1. Composition and application model of VR sports training system

5 The Application Model of Virtual Technology in University Sports Training The integration of VR technology in the process of sports training is also a very systematic work, so various problems will inevitably be encountered in the implementation process. If you want to improve the effectiveness of sports training, you can do the following links Research on the issue: 5.1 Motion Analysis System Based on VR Technology The computer virtual sports analysis system mainly integrates VR technology with sports competitions to form data information related to the competition on the screen, and

The Application of Computer Virtual Reality Technology

171

display the situation and results of the competition to coaches and athletes. The virtual sports analysis system can be applied to various sports competitions. It can appropriately insert data analysis in the links that require data analysis, to explain and analyze the competition events [11]. The data information in the process of sports games can make the games more authoritative and scientific, and can realize the live analysis of sports games broadcast and the real-time calculation of game results. 5.2 Virtual Action Comparison System Based on VR Technology Taking football, basketball and other ball sports as an example, in the traditional teaching mode, physical education teachers should explain in detail the theory, action points and common misunderstandings of each training item. After adopting VR technology, students will use VR technology to follow up the process of the teacher’s explanation in real time, and feel the standardized serving posture, force and amplitude through computer virtual reality technology, and the VR system will also correct the posture for students in real time. Recommendations for VR systems for self-perception and correction [12]. Not only that, but interactive virtual reality technology can also organize different students, connecting each student together to form a group network. Colleges and universities only need to provide students with an environment of high-tech virtual technology equipment. It will realize practical teaching that simulates reality, as shown in Fig. 2.

Fig. 2. Real-time comparison and correction of actions based on VR technology

The application of computer VR technology in college sports training will bring about a huge change in college sports training, subverting the traditional teaching mode, and gradually transforming the traditional singular teaching mode to a modern and pluralistic teaching mode.

172

J. Yin

5.3 Use Virtual Display Technology for Remote Interactive Training Through computer VR technology, a sports competition device can be set up. This kind of competition device can substitute the competition situation and athletes into the simulation world. The main purpose is to use computer VR technology to turn the training room used for training into a competition site. Athletes can clearly feel the atmosphere of the game on the competition device, enhance the athlete’s sense of tension, and help the athletes to realize their potential [13]. This kind of competition device is mostly used to simulate competition training, which can improve the training level and enhance the athlete’s competition experience. The computer VR technology can detect the physical condition of the athlete at any time, analyze and evaluate the athlete’s technical level accurately through the computer, and diagnose the problem. Computer VR technology can predict the problems that appear in the game, and formulate corresponding solutions according to the problems, to train for the game and improve the athlete’s performance, as shown in Fig. 3.

Fig. 3. VR display technology for remote interactive training

6 Conclusion All in all, with the continuous advancement of technology, virtual reality technology is also affecting all aspects of life, and has a substantial and beneficial impact on education, especially physical education and physical training. With the help of science and technology, the quality of training will be greatly improved, and my country’s sports development level will achieve more significant results.

References 1. Bai, H.J., Gao, Y.L.: Research on the application of computer virtual reality technology in college sports training. J. Heilongjiang Agric. Reclam. Univ. 31(3), 105–107 (2013) (in Chinese)

The Application of Computer Virtual Reality Technology

173

2. Wang, J.H., Yang, J., Sun, L.X.: Research on the application of computer “virtual reality” technology in university sports training. J. North China Inst. Aeronaut. Technol. 22(02), 56–59 (2018) (in Chinese) 3. Xu, D.M.: Research on the application of virtual reality technology in university physical education teaching. J. Lanzhou Univ. Arts Sci. (Natural Sci. Ed.) 32(01), 120–124 (2018) (in Chinese) 4. Zhang, Z.G.: Application analysis of computer virtual reality technology in college physical education. Wireless Internet Technol. 16(19), 147–149 (2019) (in Chinese) 5. Nessa, A., Kadoch, M., Rong, B.: Fountain coded cooperative communications for LTE-A connected heterogeneous M2M network. IEEE Access 4, 5280–5292 (2016) 6. Liu, Q., Sun, S., Rong, B., Kadoch, M.: Intelligent reflective surface based 6G communications for sustainable energy infrastructure. IEEE Wireless Commun. Mag. 28, 49–55 (2021) 7. Chen, L., Yu, F.R., Ji, H., Rong, B., Li, X., Leung, V.C.M.: Green full-duplex self-backhaul and energy harvesting small cell networks with massive MIMO. IEEE J. Sel. Areas Commun. 34(12), 3709–3724 (2016) 8. Chi, Y.H.: Combination analysis of virtual reality technology (VR) and college sports training. J. Chifeng Univ. 35(03), 151–153 (2019) (in Chinese) 9. Sun, M.: The application of computer “virtual reality” technology in college sports training. Xi’an Technol. Univ. 7(22), 148–149 (2016) (in Chinese) 10. Wang, J.H., Li, L.M.: Discussion on the use of “virtual reality” technology to assist college sports training. Educ. Teach. Res. 9(06), 171–172 (2016) (in Chinese) 11. Liu, G.W.: Research on sports training mode based on virtual reality technology. Netw. Inf. Eng. 10(18), 163–167 (2017) (in Chinese) 12. Song, L.T.: Discussion on the simulation training of competitive sports by using virtual reality technology. J. Xi’an Univ. Posts Telecommun. 10(12), 170–173 (2017) (in Chinese) 13. Shi, R.F.: Analysis of computer virtual reality technology and its application in university teaching. China Manage. Inf. 18, 223–224 (2017) (in Chinese)

The Application of Intelligent Mobile Internet Methods in the Development of Smart Physical Education Yongcai Zheng(B) Xi’an Medical University, Xi’an, China [email protected]

Abstract. With the growth of the level of science and technology, China’s information mobile Internet wisdom process is accelerating the pace of construction, “Internet +” technology is integrated into all occupations in China. In the current development of smart sports, it is indispensable to improve the quality of development and construction so that the sports industry can better integrate “Internet +” resources. It is essential to improve the level of smart sports applications. Therefore, the thesis will conduct in-depth research on how to apply advanced mobile Internet smart methods in the development of smart sports to bring high-quality development prospects to the construction of smart sports. Keywords: Mobile · Internet · Smart sports · Sports development

1 Introduction In the era of industry-wide informatization, the in-depth development of “Internet +” technology has brought great convenience to our lives and greatly benefited China’s sports industry [1]. The connection between the sports industry and mobile Internet smart technology is still accelerating. Most countries have gradually realized the importance and necessity of integrating “Internet +” into the sports industry. The “Internet +” application process in the sports industry has also greatly enhanced the scale of sports economic efficiency, bringing a good model experience and vitality to the sports industry [1].

2 The Status Quo and Defects of China’s Sports Industry The current “Internet +” thinking and the increase in the scale of the national economy have made the sports industry gradually attract the people’s attention. The following are some of the current development status of China’s sports industry.

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 S. Sun et al. (Eds.): ICSINC 2021, LNEE 895, pp. 174–181, 2022. https://doi.org/10.1007/978-981-19-4775-9_21

The Application of Intelligent Mobile Internet Methods

175

2.1 The Development Scale and Industrial Chain of the Sports Industry Have Limitations With the improvement of China’s economic and living standards, the people pay more attention to their own health, and the demand for sports products in daily life continues to increase [2]. The scale of China’s sports industry has also increased, and it has reached its peak driven by the market economy. However, due to the relatively imperfect industrial chain and short development cycle, it has not obtained complete industrial construction and large-scale facilities. Compared with the development of the sports industry in developed countries, China’s industry is relatively limited, and various sports cultures are relatively poor, not as diverse as other countries’ sports cultures. At the same time, the economic scale of the industry is insufficient, which is far below the level of sports economy in developed countries [2]. At the same time, the structure of the sports industry is relatively simple and does not better meet the needs of the people. Most of China’s sports industry draws heavily on foreign experience and is not innovative enough. The scale of the industrial chain still needs to be substantially improved [1]. 2.2 The Sports Industry is Weakly Connected to Various Industries The sports industry is in the emerging stage, occupying a larger development advantage in China’s economic scale, and providing China with a potential economic growth point. By cooperating and binding with other industries, various industries can combine and cooperate with each other to develop and contribute to a greater economic growth rate [3]. At the same time, the connection between the sports industry and other industries can increase the emergence of new occupations and promote a more healthy development of the industry. At the same time, it strengthened the training of compound talents, increased employment positions, deeply integrated the sports industry with the health, education, and health care industries, and launched a full-scale sports industry chain. At the same time, integrate sports elements into local tourist attractions, such as marathons and longdistance walking. For the holding of sports events, it can also bring great stimulus to the local economy, promote the close integration of catering, accommodation, travel and other industries, at the same time promote the national sports health atmosphere, and promote the physical and mental health of Chinese people [3]. 2.3 There are Few International Brands in China’s Sports Industry Now is the era of economic globalization, and the degree of internationalization of various industries is different, but for the current situation of the health sports industry, its degree of internationalization is getting higher and higher. Many competitive sports products in China shine in the international market and occupy a certain economic share. At the same time, the competitiveness of health sports companies in the international market is gradually deepening, but it is still far behind the level of companies in developed countries. For example, sports brands such as Adidas, Nike, and Puma have developed in the global sports industry for a long period of time, the industrial chain is very complete, and the annual profit income is very considerable [4].

176

Y. Zheng

3 Mobile Internet “Smart Sports” “Smart sports” is a development form that combines sports and information technology. The emergence of smart sports has comprehensively improved the service level and management capabilities of the sports industry, and promoted the development of sports informatization [4]. At present, the main problems in the development of sports informatization in China are: 3.1 Multi-agent Sports information resources with multiple subjects and multiple structures are too scattered, and there is a lack of ways and platforms to efficiently integrate massive information resources. If the data format is not unified and the data quality is uneven, it is impossible to develop deeper value. 3.2 Non-disclosure of Sports Information The content of the sports information public service platform is not open enough, development methods are insufficient, interoperability is not strong, and updates are not timely [5]. The process of sports informatization lags far behind the development level of the modern sports industry, and a large amount of game information cannot be pushed to interested audiences. “Smart Sports” is a new form of sports informatization and a further expansion and improvement of digital sports. It comprehensively uses mobile Internet intelligence, Internet of Things, cloud computing, big data, intelligent perception, social networking and other information technologies, and has the characteristics of extensive coverage, collaborative operation, intelligent processing, and sustainable innovation [5]. These characteristics effectively solve the development problems of the sports industry, and build a new service system and service model through “smart sports” to meet the needs of different groups of people. High-efficiency and low-cost intelligent services enable sports participants to carry out efficient and coordinated operations and promote the overall development of sports, as shown in Fig. 1.

Fig. 1. Theoretical framework of sports resource sharing

The Application of Intelligent Mobile Internet Methods

177

In non-technical language, when smart sports is completed, sports industry operations will be automated; at the same time, service quality will be doubled. For example, smart sports facilities will be unattended and equipped with self-service wristband collection machines [6]. After verifying the identity by means of QR code, face recognition, etc., you can receive the bracelet and return it with the same method after the end. The light control system can intelligently adjust the lighting; ordinary people can also complete a series of operations such as online reservation, credit payment, ball appointment, match registration and fee AA online [5].

4 Research on the Development of China’s Smart Sports Under the Background of Mobile Internet In the development of the modern era, China’s industrial training mechanism has gradually strengthened, and the application of “Internet +” has been widely used in various industrial operations. The following are some researches on the development of China’s smart sports under the background of the Internet. 4.1 Use Emerging New Media to Promote Smart Sports For the construction and development of smart sports, it is necessary to combine the current emerging new media industry to carry out extensive publicity, establish a comprehensive sports website, incorporate extensive sports resource information, and establish a platform for sports enthusiasts. Cultivate and support the development of sports self-media accounts, so that various new media technologies under the Internet can effectively bring traffic exposure to the sports industry and attract more people’s attention and interest. Establish a platform reward mechanism, look for high-quality sports content exporters, and at the same time improve the platform’s operational capabilities, which can carry a larger number of sports netizens and enrich the sports life of the Chinese people [6]. 4.2 Establish a Smart Sports Venue The speed of information transmission under the Internet is rapid, which brings good transmission and integration benefits to the information resources of sports venues. Through the application of Internet technology, combined with the big data cloudcomputing platform, the stadium information is entered and classified, and delivered to the hands of those who need it [7]. Increase the efficiency of information dissemination and the construction of sports projects. For example, during sports events, event information and content, various price service equipment, etc. can be input into the big data platform, so that people have more channels for obtaining sports information, and they can learn about the progress of sports events and the vacancy of venues in a timely manner.

178

Y. Zheng

4.3 Establish a Full-Featured Smart Sports App The popular APP is an Internet product with a high utilization rate of people from all levels. Through the establishment of a fully functional sports APP, people can access the APP anytime and anywhere, collect sports information and understand sports items, and stimulate the atmosphere of national fitness [7]. App should include functions such as recording and sharing, so that it has social benefits, expands the scale of users, allows users to intuitively feel their sports results, and gain a sense of accomplishment, and enhance the utilization rate and satisfaction of the APP. Let more and more people join in physical fitness and pay attention to their own health. In promoting the development of smart sports, it has also successfully improved the physical fitness and health of the people [8]. At the same time, APP should include shopping mall functions, equipped with complete logistics services, improve service satisfaction and increase user stickiness. Big data calculations can be used to push content that users are interested in and increase usage. 4.4 Building a High Quality Sports Smart Product Brand China’s sports brand building capacity is relatively weak, unable to compete with established companies in developed countries. Nowadays, live broadcast sales are emerging. Companies should seize the opportunity of the Internet to build smart sports brands in China and go abroad through internationalization. The development of smart sports in China has been accelerating, but various cost research and development expenditures are too high, resulting in low consumer willingness to consume [8]. At this time, sports companies should carefully study the key points of brand building, reduce costs, grasp product quality control, understand user needs through the Internet, and accurately place advertising resources. Improve after-sales service; improve the praise rate and user stickiness. Speed up the pace of new technology research and development, expand international influence and competitiveness, and build a high-quality Chinese sports brand.

5 The New Trend of Sports Development in the Era of Mobile Internet 5.1 Mobile Internet is the Mainstream of Sports Development In the Internet age, we are pursuing the development of the sports industry. From the perspective of the future development of the sports industry, “innovation” will become the mainstream of the development of the sports industry. The 5G and cell networks put new energy into the smart sports industry [11–13]. In today’s consumer market, people are all pursuing breakthroughs, leaning towards the front in order to get the most profit in the market [9]. In the era of Internet + sports, the sports industry is also carrying out technological innovations, breaking through traditional business models, establishing its own unique and clear industrial form, innovating sports clients, and realizing online and offline interactive experiences, as shown in Fig. 2.

The Application of Intelligent Mobile Internet Methods

179

Fig. 2. Offline sports experience

On the other hand, people will pursue spiritual enjoyment even when their material life is satisfied. Therefore, in the future development of the sports industry, the service concept also needs to be improved. The existing sports resources are optimally allocated and sports activities are organized in a reasonable and humane manner. Guided by sports elements, sports organizations as the key, and products and services as the focus, we will build a new type of sports industry. At the same time, real-time attention to the feedback information of client users, earnestly grasp the first-hand demand resources, clarify the current consumer population’s dominant consumption direction, and build a stable profit method and business model [9]. In the context of the widespread penetration of Internet thinking, with the acceleration of globalization and the rapid development of sports culture, the sports industry and the economic field are constantly moving towards in-depth integration, showing a strong industrial linkage effect. Favorable policies, coupled with the technical support of mobile Internet, will form a huge market space for China’s sports industry and will attract more strategic investment [10]. The “Internet +” era is a very new and important period. More integration between sports and mobile Internet wisdom will also become an inevitable trend in the development of the sports industry in the future. 5.2 Mobile Internet Smart Sports Basic Services The advancement of mobile Internet smart technology allows various sports data to be effectively recorded, such as users’ viewing records, athletes’ sports and health records, athletes’ performance on the competition field, coaches’ tactical arrangements, referees’ enforcement, etc. Change the presentation mode of traditional sports [10]. Big data and cloud computing can be applied in the following areas: stadium construction and management, marketing, competition organization and arrangement, sports industry planning, national health management, e-commerce, etc., as shown in Fig. 3. Therefore, with the

180

Y. Zheng

continuous penetration and reconstruction of Internet technology into the sports industry, those companies with sports big data and cloud computing capabilities will gain greater competitiveness, which is also the future development direction of Internet sports [10].

Fig. 3. Types of data included in sports big data

6 Conclusion The innovative achievements of mobile Internet wisdom will be deeply integrated with various fields of sports and related industries, promote technical progress, efficiency improvements and organizational changes in the sports industry, enhance the innovation and productivity of the sports industry, and build infrastructure and innovation based on mobile Internet wisdom elements of a new form of mobile Internet smart sports economic development. The era of mobile Internet smart sports has just begun. In the future, with the continuous advancement of mobile Internet smart technology, it will be more widely used in various fields of the sports industry, and mobile Internet smart sports will appear in new commercial forms.

References 1. Cai, N., Wang, J.X., Yang, D.P.: Platform envelope strategy selection and competitive advantage construction under the background of sports integration. China Ind. Econ. 8(05), 96–99 (2015) 2. Liu, J.P.: Smart Sports under the background of “Internet +”. China Sports 12(06), 148–160 (2015) 3. Luo, M., Li, L.Y.: Business model innovation in the internet Era: a perspective of value creation. China Ind. Econ. 10(21), 95–97 (2015) 4. Xiao, R.A.: Shared value, business ecosystem and sports transformation. Reform 9(07), 129– 131 (2015)

The Application of Intelligent Mobile Internet Methods

181

5. Yang, Q.: Research on the path mechanism and reconstruction model of the integration and development of sports and related industries. Sports Sci. 11(07), 13–17 (2015) 6. Liu, T.: Big data and cloud computing highlight smart sports. Internet Weekly 5(13), 22–23 (2013) 7. Wu, J.: Research on the development of China’s smart sports under the background of mobile Internet. Chin Foreign Entrepr. 12(09), 55–56 (2017) 8. Chen, X.: Investigation and research on the development of “Smart Sports” industry in Wuxi based on “Internet+.” Sports Sci. Technol. Lit. Bull. 25(12), 19–23 (2017) 9. Han, S.: Thoughts on the construction of China’s smart sports based on the mobile Internet. Sports Sci. Res. 11(03), 36–42 (2016) 10. Chen, J., Zhang, Y.Y.: Research on the design of smart sports platform under the background of “Internet +”. J. Nanjing Inst. Phys. Educ. 16(03), 125–127 (2017) 11. Chen, N., Rong, B., Zhang, X., Kadoch, M.: Scalable and flexible massive MIMO precoding for 5G H-CRAN. IEEE Wirel. Commun. Mag. 24(1), 46–52 (2017) 12. Nessa, A., Kadoch, M., Rong, B.: Fountain coded cooperative communications for LTE-a connected heterogeneous M2M network. IEEE Access 4, 5280–5292 (2016) 13. Chen, L., Yu, F.R., Ji, H., Rong, B., Li, X., Leung, V.C.M.: Green full-duplex self-backhaul and energy harvesting small cell networks with massive MIMO. IEEE J. Sel. Areas Commun. 34(12), 3709–3724 (2016)

Survey on Wireless Power Transfer in Future Mobile Communication Network Heng Wang(B) , Xu Xia, Wen Qi, and Yanxia Xing China Telecommunication Research Institute, Beijing, China [email protected]

Abstract. With the developing of mobile communication network and the growing anxiety for battery supplies with Internet of Things (IoT), the combination of advanced mobile communication technologies and Wireless Power Transfer (WPT) have become the key research prospect of the future 6 G network. This paper investigates the current development status of two forms of joint transmission of information and energy, discusses the combination of key technologies in EH and mobile communication as well as the opportunities and challenges it brings. Keywords: Mobile communication · IoT · Wireless Power Transfer · 6G

1 Introduction By 2025, 5G network is expected to support 30 billion devices [1] as well as meet critical requirements including high data rate, massive access, ultra reliability with low latency, etc. Motivated by the developing and upgrading of mobile network technology, the IoT is springing up and spreading rapidly, which can influence on many aspects of future social life, such as smart logistics, smart grid, smart city and so on. Considering enormous accesses of large-scale IoT devices, the traditional mode of power supply, usually known as battery supply, has two basic shortages of short battery life and high initial cost [2] that limit its popularity. In contrast with normal rechargeable batteries and super capacitors, the surrounding environment contains wasted green energy. As a solution to promote the life cycle of wireless device (WD), Energy Harvesting (EH) [3] captures these energy, e.g. sound or radio frequency (RF) signals, and converts them into power which can be used for the regular work of WDs even in the mobile state. WPT, one of EH technologies, is attracting more and more attention as a new mode of power supply to bypass the current technical bottleneck of battery. There are mainly two sources providing green energy in WPT. One is signals from environment, the other is signals from a dedicated and fully controlled power supply, such as a base station (BS), which will be detailed introduced.

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 S. Sun et al. (Eds.): ICSINC 2021, LNEE 895, pp. 182–190, 2022. https://doi.org/10.1007/978-981-19-4775-9_22

Survey on Wireless Power Transfer in Future Mobile

183

2 Joint Transmission in WPT 2.1 SWIPT By superimposing information and energy transmission, SWIPT can bring significant benefits of spectrum efficiency and energy efficiency as well as time delay and interference management. WDs supporting SWIPT can be simultaneously charged and communicate with the same signal they receive from the access point [4]. As the blooming of IoT, SWIPT is desirable for energy supply and information interaction of enormous ultra low power WDs supporting heterogeneous applications.

Fig. 1. Integrated receiver schemes of SWIPT: a) SR; b) TS; c) PS; d) AS

As shown in Fig. 1, there are 4 available SWIPT receiver schemes, including: a) Separate Receiver (SR). As shown in Fig. 1a), both Information Decoding (ID) and EH modules contain two separate receivers with independent antennas observing different channels, and their serving transmitter is implemented with multiple antennas. The SR architecture can be easily implemented using existing components of EH and ID receivers. It allows simultaneous and independent execution of EH and ID. b) Time Switching (TS). As shown in Fig. 1b), EH and ID shares the same antenna. The receiver here includes an information decoder, a RF energy collector and a switch for changing the service type of the antenna. Based on TS sequences, the receiver antenna changes periodically between the EH and ID modules. c) Power Splitting (PS). As shown in Fig. 1c), the power splitting receiver separates two power streams with a configured splitting ratio from the original signal. After that, both power streams are delivered to EH and ID modules, so EH and ID can operate at the same time. Furthermore, the splitting ratio configured in each antenna can be optimized to meet the balance of energy and information. d) Antenna Switching (AS). As shown in Fig. 1d), when some antennas of the receiver works on the ID, other antennas can work on the EH. Simple antennas can serve both EH and ID modules respectively that can be easily extended from dual to more antennas in view of appropriate antenna switching protocols. 2.2 SWIPT The latest development of WPT technology enables WPCN, in which the power supply for the WD communication is from dedicated wireless power transmitter. Especially, the power transmission of WPCN, including transmitting power, waveform, time and

184

H. Wang et al.

frequency occupation, are fully controlled and can be flexible scheduled for stable energy supply in a complex practical environment. The current WPT can transmit tens of microwatts of RF power to the WD more than 10 m and it can be further improved on amplitude and range. Therefore, WPCN can be applied to serving various low-power applications, e.g. Wireless Sensor Network (WSN).

Fig. 2. Schematic diagram of basic models of WPCN: a) Separate energy transmitting and information receiving sources; b) Common energy transmitting and information receiving source; c) Full duplex information and energy transmission in different spectrum bands; d) Full duplex information and energy transmission in the same spectrum band

As shown in Fig. 2, there are 4 available WPCN schemes, including: a) Separate energy transmitting and information receiving sources. As shown in Fig. 2a, the energy source and the information source are independent. According to Harvest Then Transmit (HTT) protocol, EH is executed first, then ID using the collected energy is executed after. b) Common energy transmitting and information receiving source. As shown in Fig. 2b, Hybrid access point (HAP) is considered as energy source as well as information source. The HAP first broadcasts energy to WDs, then WDs send information to the HAP after energy collection. Long distance leads to more energy consumed in transmission. Also, less energy can be collected by far WDs, so less information will be sent to the HAP. c) Full duplex information and energy transmission in different spectrum bands. As shown in Fig. 2c, information and energy are transmitted in different spectrum bands to avert co-channel self-interference. The information and energy full duplex operation of the WDs realizes an additional benefit called self energy recovery, in which WDs can also collect RF energy from information signals transmitted by itself. As a result, the antenna of the receiver should be positioned as close to the antenna of the transmitter as possible and the latter must not interfere the radiation pattern of the former. d) Full duplex information and energy transmission in the same band. As shown in Fig. 2d, energy and information can be simultaneously transmitted in the same spectrum to improve spectrum efficiency. However, co-channel self-interference will occur especially when the energy transmitter and the information receiver are integrated in one HAP. Low loop link channel gain can be realized by directional antenna design or large antenna spacing to mitigate harmful co-channel self-interference. Also, interference power can be further reduce by Self-Interference Cancellation (SIC) functions in the full duplex HAP.

Survey on Wireless Power Transfer in Future Mobile

185

3 Combination of WPT and Mobile Communication Technologies The latest mobile communication system, 5G, defines three major scenarios, including enhanced Mobile Broadband (eMBB), Ultra-reliable and Low Latency Communications (URLLC) and massive Machine Type Communications (mMTC). This section introduces the combination of WPT and key mobile communication technologies for better fulfilling requirements of three scenarios with energy considerations. 3.1 Millimeter Wave (mmWave) Aided WPT mmWave can provide larger the achievable signal bandwidth because its frequency is approximately between 30 GHz and 300 GHz. Due to high frequency, mmWave can achieve ultra-high-speed wireless data transmission by increasing the spectrum bandwidth. In addition, transmission efficiency can be enhanced because mmWave has a narrow beam, excellent directivity and extremely high spatial resolution. mmWave WPT: mmWave WPT was originally used for space power transmission [5], mainly to study the design of rectenna. A 24 GHz rectenna, which was first developed to prove the feasibility of the combination of WPT and mmWave technologies [6]. Besides mmWave rectenna, transmission and rectifiers are also hotspots in this direction. mmWave SWIPT: Considering the SWIPT of power distribution in mmWave, mmWave SWIPT is implemented by a low-power receiver architecture that uses antenna switches [7], which provides a good gain for the overall mmWave EH performance. mmWave WPCN: The BS sends energy to the user equipment and then the user equipment sends information upstream in the mmWave WPCN [8]. By analyzing the collected average energy and data rate, it shows that the serving BS can effectively transmit energy.

3.2 Cognitive Radio (CR) Aided WPT CR allows secondary users (e.g. unauthorized users) to share the spectrum of the primary user (e.g. authorized users) without expensive spectrum licenses, as long as secondary users can ensure the communication of primary users without any impacts. Therefore, CR combined with WPT can bring out their potential to enhance the efficiency of both spectrum and energy. EH/RF-Powered Cognitive Radio Network (CRN): EH/RF-powered CRN mainly considers the scenario that the secondary sender collects energy used for spectrum sensing and inform EH/RF-powered Cognitive Radio Network (CRN): ation transmission to other secondary users. A specific CRM model is shown in Fig. 3: (1) When the spectrum of the primary user is occupied, the secondary sender is in a waiting state; (2) When the primary user’s spectrum is free, the secondary sender first performs spectrum sensing to confirm that there is no signal of the primary user, and then uses the spectrum of the primary user to send information to other secondary users.

186

H. Wang et al.

Fig. 3. CRN model

The sensing threshold can be determined by maximizing the instantaneous expected throughput so as to optimize the opportunistic spectrum access strategy. For the same purpose, the statistical description of the primary user’s activities and the energy arrival rate may also be analyzed together under the energy causal constraint (the total energy consumed cannot exceed the total energy collected) and the collision constraint (ensuring the primary user) [9]. Further on maximizing the average throughput of secondary network, the best pairing of sensing threshold and duration can be determined [10]. In contrast with traditional CRNs, a scenario that secondary users collect energy when the primary user is busy and secondary users transmit information when primary user is idle is studied [11]. Under this circumstance, for studying the balance of sensing, information transmission and energy collection, a channel selection strategy based on Markov decision process was proposed and it can be optimized by maximizing the average throughput. Cognitive SWIPT Network: Cognitive SWIPT network is different from normal EH/RFpowered CRN that secondary users can receive information and collect energy at the same time. In cognitive SWIPT, the system needs to make a choice between EH and spectrum access. Therefore, it is important to research on the optimal selection method in different SWIPT models that achieves a balance between energy efficiency and spectrum efficiency. In wideband CRN, the best power and sub-channel allocation ratio can be calculated by algorithms of maximizing the sum of collected energy [12]. The literature [13] establishes a multi-hop CRN model for EH in time slot switching and proposed two charging schemes including an independent charging channel and a variable charging channel. It analyzes interruption probability under the condition of interference as well as maximizes the data rate by optimizing the time slot factor, and the results show that the time slot factor has a significant impact on system performance.

3.3 Multiple-Input Multiple-Output (MIMO) Aided WPT MIMO is a multi-antenna technology, which can significantly promote the reliability and capacity of wireless access networks [14]. The beginning research aimed to point-topoint MIMO links making use of two multi-antenna WDs for information interactions. The focus has now turned to multi-user MIMO systems, where single-antenna WDs can be simultaneously served by a multi-antenna BS. Furthermore, MIMO combined with SWIPT becomes a tendency.

Survey on Wireless Power Transfer in Future Mobile

187

MIMO SWIPT: The basic model of MIMO SWIPT is normally served by two user groups, one group is used to receive information, while the other group is used to collect energy for power supply. When SWIPT is applied to MIMO channels, e.g. a time-slotted SWIPT, the HAP transmits information to one WD in each time slot, while the remaining idle WDs perform EH from the received signal [15].

3.4 Non-orthogonal Multiple Access (NOMA) Aided WPT NOMA, which is being discussed for 5G standards, transmits multiple information streams in overlapping channels in time domain or frequency domain or code domain with different power, and provides wireless services for multiple users with same wireless resources simultaneously, which is expected to further improve the spectrum efficiency and user fairness. NOMA supports large-scale connections with lower delay and less signaling overhead, and does not need accurate channel state information. NOMA Aided SWIPT: In the SWIPT network of two users, without considering decoding energy consumption, it can be proved that NOMA provides a higher data rate than Orthogonal Multiple Access (OMA) [16]. In a SWIPT network of a HAP, a WET WD and multiple WIT WDs [17], where modulation scheme has a significant impact on the efficiency of energy collection, the HAP sends information to multiple WIT WDs by NOMA, while the WET WD utilizes the superimposed signals for energy collection. Furthermore, considering the constraints of power and energy collection threshold, the slot-switching SWPIT was studied with an efficient path following algorithm to maximize the worst user throughput and energy efficiency [18]. NOMA Aided WPCN: NOMA can improve energy efficiency, throughput and fairness by minimizing throughput in WPCN [19], while decoding sequences influence on the performance of the NOMA aided WPCN. When jointly optimizing the transmit power of BS and durations for energy and information, a higher data rate and system fairness level can be achieved [20]. Furthermore, a WPCN model with hybrid NOMA, which is the combination of NOMA and OMA, can help minimize the total system delay considering both circuit and decoding energy consumption of the system [21].

4 Opportunities and Challenges As illustrated above, the combination of WPT and mobile communication technologies is attractive and feasible. However, there are still some issues to overcome. Basic System Design: The WD supporting RF EH usually has strict operating power limits and must have enough power to execute basic communication functions such as modulation, coding, receiver operation strategies and routing protocols. When applying WPT with NOMA, the overall system performance mainly depends on the interference cancellation ability and complexity of the receiver. Therefore, especially in far-field WPT scenario, low computation, sufficient energy collecting efficiency and low power

188

H. Wang et al.

consumption need to be simultaneously achieved by WDs integrated with WPT functions as well as simplified communication functions. Furthermore, a multi-cell multi-user MIMO scenario with symbol-level precoding need a high degree of cooperation between BSs so that the knowledge of user symbols can be obtained timely from each other. When introducing IRS, there are usually a large number of reflection elements, and the pilot and training overhead is very high. Therefore, it is necessary to further reduce the overhead of channel estimation. In addition, the synchronization and security requirements, the practical system impairment and the simplification of communication resource overhead should also be further studied. Resource Allocation: The practical energy efficiency of the system is still lack of assurance, since the remote power transmission leads to significant reducing of the signal strength due to path loss. In order to overcome this bottleneck, it is necessary to optimize the ratio of energy and information transmissions among WDs in different dimensions, such as time, frequency, carrier, to meet the best resource allocation strategy. Particularly in CRN, the energy source is inherently random, which will affect the performance of the node for energy collecting. Besides, the secondary user transmits from the environment of the main network or receives RF energy from a specific primary user with known activities. In this circumstance, the cognitive process of the secondary user is only powered by the RF signals from the primary user. Secondary users need operations for both occupying spectrum and idling spectrum, so the overall performance of the CRN is limited by the collision constraint, whose probability should be always sustained below a predefined threshold for the primary transmission. As a result, energy constraints and collision constraints are the fundamental limitation of EH on the throughput of CRN. Therefore, when introducing WPT CRNs, the mobile networks is challenged to optimize the resource allocation strategy in order to minimize the impact of two constrains. Multiple Access Scheme: WPT can use different multiple access schemes, such as time division multiple access (TDMA), NOMA and space division multiple access (SDMA), for uplink information transmission. Especially when combined with IRS in different schemes, not only different active or passive beamforming and resource allocation strategies are required, but also whether use the same phase shift for both the uplink and downlink channel need to be determined. Therefore, different multiple access schemes on IRS channel acquisition methods and deployment strategies are important directions in future work. However, the design of their joint operation become important, especially the problems brought by IRS, such as IRS reflection coefficient design, IRS channel estimation, IRS deployment, beam design.

5 Conclusion In this article, we have first comprehensively described basic models of two kinds WPT, SWPIT and WPCN, providing information and energy transmission in different ways. Integrating WPT with most recent mobile communication technologies associatively, a large number of studies pursue the goal of improving both spectrum efficiency and energy efficiency from different perspectives. However, to take advantages of WPT in

Survey on Wireless Power Transfer in Future Mobile

189

future 6G mobile communication network, significant further research is needed on several issues, including basic system design, resource allocation and multiple access scheme. Acknowledgment. This work was supported by National Key R&D Program of China under Grant 2020YFB1806700.

References 1. Zhang, S., Wu, Q., Xu, S., Li, G.Y.: Fundamental green tradeoffs: Progresses, challenges, and impacts on 5g networks. IEEE Comm. Surv. Tutorials 19(1), 33–56 (2017) 2. Zhang, Z., Pang, H., Georgiadis, A., Cecati, C.: Wireless power transfer-an overview. IEEE Trans. Industr. Electron. 66(2), 1044–1058 (2019) 3. Ponnimbaduge Perera, T.D., Jayakody, D.N.K., Sharma, S.K., Chatzinotas, S., Li, J.: Simultaneous wireless information and power transfer (SWIPT): Recent advances and future challenges. IEEE Commun. Surv. Tutorials 20(1), 264–302 (2018) 4. Krikidis, I., Timotheou, S., Nikolaou, S., Zheng, G., Ng, D.W.K., Schober, R.: Simultaneous wireless information and power transfer in modern communication systems. IEEE Commun. Mag. 52(11), 104–110 (2014) 5. Koert, P., Cha, J.: Millimeter wave technology for space power beaming. IEEE Trans. Microw. Theory Tech. 40(6), 1251–1258 (1992) 6. Ladan, S., Guntupalli, A.B., Wu, K.: A high-efficiency 24 GHz Rectenna development towards millimeter-wave energy harvesting and wireless power transmission. IEEE Trans. Circuits Syst. I Regul. Pap. 61(12), 3358–3366 (2014) 7. Khan, T.A., Alkhateeb, A., Heath, R.W.: Millimeter wave energy harvesting. IEEE Trans. Wireless Commun. 15(9), 6048–6062 (2016) 8. Wang, L., Elkashlan, M., Heath, R.W., Di Renzo, M., Wong, K.-K.: Millimeter wave power transfer and information transmission. In: 2015 IEEE Global Communications Conference (GLOBECOM), pp. 1–6 (2015) 9. Park, S., Kim, H., Hong, D.: Cognitive radio networks with energy harvesting. IEEE Trans. Wireless Commun. 12(3), 1386–1397 (2013) 10. Chung, W., Park, S., Lim, S., Hong, D.: Spectrum sensing optimization for energy-harvesting cognitive radio systems. IEEE Trans. Wireless Commun. 13(5), 2601–2613 (2014) 11. Lu, X., Wang, P., Niyato, D., Hossain, E.: Dynamic spectrum access in cognitive radio networks with RF energy harvesting. IEEE Wirel. Commun. 21(3), 102–110 (2014) 12. Hu, Z., Wei, N., Zhang, Z.: Optimal resource allocation for harvested energy maximization in wideband cognitive radio network with SWIPT. IEEE Access 99, 1 (2017) 13. Ge, L., Chen, G., Zhang, Y., Tang, J., Wang, J., Chambers, J.A.: Performance analysis for Multihop cognitive radio networks with energy harvesting by using stochastic geometry. IEEE Internet Things J. 7(2), 1154–1163 (2020) 14. Lu, L., Li, G.Y., Swindlehurst, A.L., Ashikhmin, A., Zhang, R.: An overview of massive MIMO: Benefits and challenges. IEEE Journal of Selected Topics in Signal Processing 8(5), 742–758 (2014) 15. Morsi, R., Michalopoulos, D.S., Schober, R.: Multi-user scheduling schemes for simultaneous wireless information and power transfer over fading channels. IEEE Trans. Wireless Commun. 14(4), 1967–1982 (2015) 16. Hedayati, M., Kim, I.-M.: On the performance of NOMA in the two user SWIPT system. IEEE Trans. Veh. Technol. 67(11), 11258–11263 (2018)

190

H. Wang et al.

17. Zhao, Y., Hu, J., Ding, Z., Yang, K.: Constellation rotation aided modulation design for the multi-user SWIPT-NOMA. In: 2018 IEEE International Conference on Communications (ICC), pp. 1–6 (2018) 18. Nasir, A.A., Tuan, H.D., Duong, T.Q., Debbah, M.: NOMA through put and energy efficiency in energy harvesting enabled networks. IEEE Trans. Commun. 67(9), 6499–6511 (2019) 19. Diamantoulakis, P.D., Pappi, K.N., Ding, Z., Karagiannidis, G.K.: Wireless-powered communications with non-orthogonal multiple access. IEEE Trans. Wireless Commun. 15(12), 8422–8436 (2016) 20. Chingoska, H., Hadzi-Velkov, Z., Nikoloska, I., Zlatanov, N.: Re source allocation in wireless powered communication networks with non-orthogonal multiple access. IEEE Wireless Communications Letters 5(6), 684–687 (2016) 21. Lu, M., Li, N., Tao, X., Li, M.: Time minimization in downlink hybrid NOMA wireless powered communication networks. In: 2021 IEEE Wireless Communications and Networking Conference (WCNC), pp. 1– 6 (2021)

Application of Modern Information Technology in Promoting the Reform of Art Teaching Huiying Chen(B) Department of Art Designing, Shandong Vocational College of Light Industry, Zibo, Shandong, China [email protected]

Abstract. As technology advances, information technology exerts more and more important functions in art teaching in universities. Especially in the new environment aimed at cultivating applied talents, classroom teaching reform in universities has become a major means to achieve teaching objectives, among which the investment and application of information technology is of primary importance. It is an inevitable trend for universities to optimize teaching contents and innovate teaching methods by using modern teaching methods. The combination of information technology and the discipline of fine arts integrates the advantages of traditional art teaching and information technology assistance, which is more in line with the requirements of modern teaching. Keywords: Information technology · Art teaching · Reform

1 Introduction As the technology is racing ahead, information technology has penetrated the teaching of various subjects in the school, affecting the teaching quality, as well as art teaching [1, 2]. The reform and innovation of art teaching in universities put forward new requirements for the teaching concept of art teachers [3, 4]. The previous teaching methods can no longer meet the needs of modern teaching. In the new era, teachers should not only have solid professional technology, but also keep pace with the times, have multimedia teaching skills and master the use of modern teaching equipment [5, 6]. In the process of classroom teaching reform, information technology plays a positive role in designing teaching methods, improving teaching methods, optimizing resources, and improving efficiency [7, 8]. Information technology products, such as network teaching platform, break the time and space constraints by means of modern science and technology, expand students’ learning space, and improve learning efficiency through network teaching [9, 10].

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 S. Sun et al. (Eds.): ICSINC 2021, LNEE 895, pp. 191–195, 2022. https://doi.org/10.1007/978-981-19-4775-9_23

192

H. Chen

2 Significance of Information Technology to the Reform, Innovation and Development of Art Teaching 2.1 Reduction of Burden of Teachers in the Process of Art Teaching by Information Technology In traditional teaching, writing on the blackboard is the mainstream teaching method, which not only makes teachers’ workload heavy, but also slows down the teaching progress due to the long writing cycle. Moreover, some teachers’ blackboard writing is unattractive, which is not only a great challenge to students, but also an important reason for understanding deviation. However, nowadays, most of the contents of all classes can be moved into PPT by teachers before class, which not only greatly facilitates teachers, but also makes the teaching contents clearer and complete and the progress faster. 2.2 Information Technology Makes Art Teaching More Convenient Information technology has reached the upper limit of what is physically unattainable. In the teaching, there are always requirements for drawing, modeling and so on. However, when students draw circles and triangles, they must use tools such as compasses and rulers, which not only slows down the progress and reduces the efficiency, but also poses great challenges to their eyesight when drawing too many complex graphics, which is very easy to make mistakes. In addition, the art teaching is benefited from the 5G and future 6G technologies [11–13]. However, with the use of information technology, the pattern can be layered. Circles, triangles, and rectangles only need to be clicked and lengthened, and can be modified at any time, which not only greatly improves the fault tolerance rate, but also increases the beauty. In this way, students will not be fidgety, and can be handier, to enhance self-confidence. 2.3 Information Technology Can Better Show the Charm of Fine Arts There will be images, videos, animation, and other forms in art class by using information technology. Throughout history, it is not difficult to find that one of the main reasons why art education can shine is because of its unique charm, which needs to be displayed, rather than the “dogmatic” preaching. The use of Internet, new media technology and video broadcasting has replaced boring words with vivid historical stories. Moreover, the rich colors of the art pictures also make students deeply attracted.

Application of Modern Information Technology

193

3 Application of Information Technology in Art Teaching in Universities 3.1 Using Information Technology to Expand Students’ Thinking Space and Cultivate Students’ Innovative Ability Applying information technology in art teaching can effectively stimulate multiple sensory functions including auditory sense and visual sense, combine various senses, break the limitation of time and space, effectively expand students’ thinking and actively cultivate their innovative thinking and innovative ability. Students are usually familiar with the subject matter of trees, but will find it difficult to represent them through the brushstroke techniques of landscape painting. If teachers want students to open their minds and break through the difficulties successfully, they need to provide students with materials of association, feeling and experience to help them recall their feelings about trees. Therefore, in the teaching process, teachers can guide students to the special learning website for learning. While enjoying music, students can enjoy the paintings of famous landscape artists, and experience the basic expression methods (with pen, ink, color, etc.) of tree trunks, branches and leaves in landscape paintings, as well as the artistic features. After that, students can close their eyes and listen to the music while recalling the familiar patterns on the leaves and bark, and experience the feeling of various trunks and leaves. At this time, their imaginary wings can be infinitely extended. By making effective use of the stimulation, observation, feeling and experience of various senses, students will no longer simply understand the artistic conception of trees through teachers’ explanation and appreciation of famous paintings, which greatly arouses their creative desire and fully stimulates their creative inspiration. 3.2 Using Information Technology to Catch Students’ Eyes and Stimulate Their Interest in Learning In the cognitive process of students, interest plays an indispensable role, just like salt and food. If students lack interest in learning, their enthusiasm for learning and participation in teaching will be greatly reduced, and the excitement of the cerebral cortex will also be suppressed, so that their normal intelligence can’t be played. Therefore, it can be seen that interest is crucial in the whole cognitive activities. One of the basic conditions to ensure the success of teaching is to stimulate students’ interest in learning. The application of modern information technology in art teaching can inject vitality into teaching. Modern information technology integrates pictures, words, sound, and videos. It has changed the single output mode of art teaching in the past. With the help of various forms, such as audio forms and video forms, the content to be explained is displayed. The types and forms of information are richer and more diverse, which also increases the interest of teaching, makes it vivid and greatly attracts the attention of students. The static boring textbooks are transformed into vivid pictures, beautiful music, and wonderful videos. This combination of pictures, words, sound, and videos can catch students’ eyes and stimulate their strong curiosity and thirst for knowledge.

194

H. Chen

3.3 Using Information Technology to Enrich Teaching Content and Improve Teaching Quality Influenced by the widely used information technology, the teaching content of art course has been greatly simplified, including collection, sorting out of information, collection of resources, etc. Information technology helps teachers and students to collect art learning materials including words, pictures, videos, audio materials, etc. In addition, in terms of teaching forms, information technology can provide a variety of teaching forms for teaching, such as audio forms and video forms. Using information technology, classroom teaching can become more intuitive, visual, and flexible. However, what can’t be ignored is that the application of information technology in the process of art teaching should strictly follow the “appropriate” teaching principle and give full play to its guiding and auxiliary teaching role. Instead of relying too much on information technology, teachers should highlight students’ dominant learning status, link the teaching advantages of traditional classrooms, and select the learning needs of students at different learning levels. For example, when teaching the course of “Computer Art”, the operation process and drawing essentials of computer graphics should be displayed in the form of video, so that students can learn the basic skills of computer graphics. By displaying the magnificent paintings in vivid videos and enriching the teaching content in the form of video teaching, students can have a deeper understanding of the unique charm of art teaching and thus improves the quality of classroom teaching.

4 Conclusions The all-round development of the new media era promotes the use of information, a new educational and teaching means, in education and teaching by most schools. Applying information technology in art teaching can effectively collect art teaching resources and broaden the space of art education. Information technology integrates a variety of information, such as words, pictures, videos, audio forms, etc., with strong interaction and integration. The introduction of information technology into art teaching can promote the effective improvement of students’ aesthetic ability. Using information technology in art teaching not only makes students have a strong desire for art learning, but also effectively changes teaching methods and increases the vividness and interest of classroom teaching. Modern information technology provides ideas and ways for art education and teaching reform. Art teachers should take full advantage of modern information technology to speed up the pace of art education and teaching reform, thus promoting the healthy development of art education. Acknowledgements. This work was supported by 2020ZJJGLX062, 2020RB08, 2020JY21 and Zibo Vocational Education micro project “Research on the integration of Lu Xiu skill inheritance and cultural product innovation in Higher Vocational Colleges under the background of cultural empowerment (2020ZJGW02)”.

Application of Modern Information Technology

195

References 1. You, K.W.: Discussion on the development of digital art education in the era of big data. New Curricul. Res. 7, 81–82 (2020) 2. Li, L.C., Wang, S.H.: Application of information technology in art teaching of colleges and universities. Heilongjiang Sci. 11(19), 106–107 (2020) 3. Fang, Y.L.: On the importance of information education in Higher Vocational Education. Kexue yu Xinxihua (33), 137–139 (2018) 4. He, Y.T.: Development strategy of art education under the background of informatization. Invention Innovat. 7, 21 (2018) 5. Wang, X.Y.: Improvement scheme of educational management in art education in colleges and universities. Hebei Farm Mach. 1, 148–149 (2021) 6. Wang, N., Daisha, R.N.: Research on the curriculum reform of Fine Arts in Higher Education under the background of Internet plus –Taking the course of appreciation of western modern art as an example. Hundred Home Prose 11, 131 (2020) 7. Liu, H.T.: Innovative development of art education in the era of big data. Shanxi Youth 15, 165–166 (2020) 8. Chen, L.H.: Discussion on teaching methods of art education in Colleges and universities in the network era. Youth Journalist 14, 130–131 (2016) 9. Jiang, R., Guo, W.R., Sun, M.Y.: Application of information technology in basic course of art painting in secondary vocational school. Modern Vocat. Educ. 17, 74 (2017) 10. Cheng, P.: On applying multimedia technology to optimize children’s art education. Chin. Foreign Commun. 30, 114 (2017) 11. Chen, N., Rong, B., Zhang, X., Kadoch, M.: Scalable and flexible massive MIMO precoding for 5G H-CRAN. IEEE Wirel. Commun. Mag. 24(1), 46–52 (2017) 12. Chen, L., Yu, F.R., Ji, H., Rong, B., Li, X., Leung, V.C.M.: Green full-duplex self-backhaul and energy harvesting small cell networks with massive MIMO. IEEE J. Sel. Areas Commun. 34(12), 3709–3724 (2016) 13. Liu, Q, Sun, S., Rong, B., Kadoch, M.: Intelligent Reflective Surface Based 6G Communications for Sustainable Energy Infrastructure. IEEE Wirel. Commun. Mag. 28(6), 49–55 (2021)

Application of Digital Media Technology in Modern Art Design Huiying Chen(B) Department of Art Designing, Shandong Vocational College of Light Industry, Zibo, Shandong, China [email protected]

Abstract. With the advancement of sciences and technology in recent years, China’s technological level of hardware and software applications has significantly advanced. Digital media art combines digital media technology with art design to better serve society, to form a new form of media art. The continuous development of digital media not only affects people’s life and daily consumption, but also affects the survival and development of various industries in China. Digital media art is now commonly employed in the realm of current art design. Under the function of digital media art, in addition to effectively strengthening the expressiveness of design, the connotation of art has also been expanded and enriched. Keywords: Digital media technology · Art design · Art form · Application

1 Introduction People continually change their design schemes and concepts in modern art design as a result of the use of digital media art. In addition, the application of digital media technology integrates people’s thinking into the process of art design, this contributes to the concept of art design with fresh ideas, and makes it get high-quality innovation [1, 2]. Due to the influence of network technology, a series of new art forms in art design expand the development space of art design, thus realizing the organic combination of art design and digital media [3, 4]. The effective application of digital media technology provides dynamic and static transformation for art design and realizes the mutual transformation of entity and virtual [5, 6]. As the digital media technology develops, art design is moving towards the development direction of virtual display [7, 8]. Using advanced and scientific network technology, it breaks through the regional restrictions, to build a new, scientific, and reasonable design space for diversified and personalized display [9, 10]. The practical the use of digital media art enriches creative thinking and correspondingly stimulates creative inspiration [11–13].

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 S. Sun et al. (Eds.): ICSINC 2021, LNEE 895, pp. 196–200, 2022. https://doi.org/10.1007/978-981-19-4775-9_24

Application of Digital Media Technology in Modern Art Design

197

2 Integrated Development of Digital Media Technology and Modern Art Design At present, the employment of digital media technologies in the creation of contemporary art has shown to be successful, which is well reflected in the fields of house construction design, art design and exhibition. 2.1 Digital Media Technology Enriches Modern Art Design Forms Modern art design is a prominent trend, because to the ongoing updating and growth of digital media technologies. The process of art design is for art designers a means of expressing their own feelings and ideas. The designed works can convey their real thoughts, which also reflects the core of art design. With the use of digital media technologies, art design may also be more diversified. It should be noted that based on bringing convenience to us, digital media technology will also make artistic works monotonous. Although digital media technology can bring some convenience to the process of art design, designers should use it as little as possible. Actually, with 5G and some new internet technologies make huge headway, digital media develop rapidly [14–16]. Relying too much on it will only make designers lazy in innovation and gradually lose their design inspiration, eventually making all the design works alike. As a result, relevant designers must consider how to employ digital media technology in the process of contemporary art design in a reasonable manner. Designers should not only retain the original essence of works, but also use advanced technology to improve work efficiency, so that modern art forms can be more colorful. 2.2 Digital Media Technology Promotes the Innovative Development of Modern Art Design At present, digital multimedia technology needs the effective support of computer system. Therefore, digital media technology is also a cutting-edge technology in the society. If we can successfully connect digital media technology with current art technology, we will be able to achieve our goals, the art design process will follow the trend of the times and adapt to the development of society. In the process of traditional art design, relevant designers need to conceive the works in their mind, complete the content of their imagination driven by inspiration, and conduct reasonable and effective analysis of the whole works, to finally create a complete art works. Designers may better finish the intermediate design conceptualization link with the use of digital media technologies. They can draw rough graphics on computer drawing software, and then modify them on this basis. Traditionally, people tend to show the results of art creation, but the use of modern digital media can help show the entire process of art creation. For the appreciation of works of art, the overall appreciation value of works of art will be improved. Of course, in the process of designing works, designers need to adhere to the essence of artistic creation and give full play to the beauty of artistic works.

198

H. Chen

3 Influence of Digital Media Technology on Modern Art Design With the effective integration of digital media technology and modern art design process, modern art works show new characteristics and effects. At present, the use of digital media is a relatively new technology, which shows high application value in all walks of life. The expressiveness of digital media technology is primarily represented in various distinct parts of the design process for contemporary art works, such as language expression, communication mode and spatial structure. 3.1 Promoting the Infinite Expansion of Artistic Design Thinking Under digital technology, Designers under the digital technology can not only complete the related work of creative ideas, but also bring new changes to working methods, design ideas, expression methods and organizational forms, to break the limitation that traditional art design can only be realized by professional designers, and make everyone possible to become designers. Nowadays, a variety of sophisticated works are constantly appearing in the network world, and it can be found that many of them are completed by non-professionals, and the designed works are more humanized, intelligent, and popular. Not only may digital technology boost creative efficiency, but it can also demonstrate the fundamental role of thinking expansion. It enables designers to broaden their thinking and design in a more complex style. At the same time, it expands the edge of designers’ solidified inspiration to a certain extent, opens the previously inaccessible space, and thus makes their thinking move to a higher stage. 3.2 Making the Content of Art Design Richer Digital technology promotes the innovative development of media and forms of communication, and expands and enriches the visual and imagination space of art design. Digital technology-based product design, such as digital television, interactive media, interactive games, and virtual space, has brought a strong impact on the traditional visual expression carriers, methods, and theories. At present, the theoretical framework, aesthetic content, and design basis of new media art design have evolved from traditional visual design. Therefore, aesthetics and aesthetic foundation are completely interconnected. At the same time, the basic theory and visual design rules are similar, so the final design content is similar. Therefore, no matter how the design means have changed, what they reflect are real life and psychological activities, and there is no art design completely divorced from reality. Digital technology is highly advanced, and is the inevitable a result of scientific and technological progress. Its information content is great with diversified forms and rich content. With this technology, the art design can reach the height that could not reach before, to realize the innovative development of the art form. 3.3 Promoting the Diversification of Art Design Means The original design process basically relied on professional designers. They designed and created by hand. The whole creation process includes basic sketch, plan, basic

Application of Digital Media Technology in Modern Art Design

199

sample drawing, elevation, and final effect renderings. This design method requires a lot of work and takes a long time, and a design drawing will basically be completed by different people, and then submitted to the user for discussion and change. Once the error is large, it must be overthrown and recreated. However, the whole creative process of new media art design based on digital technology can be repeated and reversed, and the relationship between people and works is dynamic and interactive.

4 Significance of the Progress of Digital Media Technology to Modern Art Design 4.1 Promoting the Development of Modern Art Design Forms With the convenience of technology for digital media, modern art design forms are no longer limited to traditional forms. It can transform a traditional art form into a more three-dimensional one, and even combine several different art forms to form new art forms. For example, in film production, many scenes often need to use CG technology, virtual technology, special effects and other more complex digital media technology. In the use of these technologies, but also need to use more advanced hardware equipment. While using these technologies, more advanced hardware equipment also needs to be used. At the same time, designers need to use these technologies reasonably with rich imagination in order to finally achieve the expected artistic effect. This not only broadens the audience’s horizons, but also gives them a fresh way to shift their aesthetic viewpoint. 4.2 Expanding Modern Art Design Ideas If digital media technology is used correctly in the area of contemporary art design, it can promote designers’ design ideas to be more open. Specifically, digital media technology includes multimedia technology, which can enrich the artistic forms contained in artistic works. In addition, such works of art can further enhance their artistic value if they are processed by reasonable information technology, the designer’s design concept can also be more prominent, and the viewer of artistic works can also get a more intuitive understanding of the designer’s thinking. Digital media technology can also provide a solution for the innovation of design of contemporary art. It is worth noting that such innovation does not mean the rigid splicing of art and technology, but the application of digital media technologies in the information age to convey the designer’s concept of artistic innovation.

5 Conclusions Today’s world is a world of information. As information technology develops rapidly, new communication means and production technology have gradually emerged in public. This kind of design method broadens the design space and makes human beings gradually contact with new technology and new society. What should be paid attention to is the quality of designers, focusing on improving their comprehensive practical ability and innovation level, to create a steady stream of wealth for contemporary society

200

H. Chen

by constantly adapting to the actual needs of the evolution of the information epoch. Currently, the country is gradually moving towards the digital information age. However, while paying attention to the information age, people should not ignore the actual situation. People should pay more attention to the reality and firm faith in future development. Applying digital media art to art design can not only improve the overall effect of modern art design, but also better serve the public. Acknowledgements. This project received funding by 2020ZJJGLX062, 2020RB08, 2020JY21 and Zibo Vocational Education micro project “Research on the integration of Lu Xiu skill inheritance and cultural product innovation in Higher Vocational Colleges under the background of cultural empowerment (2020ZJGW02)”.

References 1. Liu, H.T.: Innovative development of art education in the era of big data. Shanxi Youth 15, 165–166 (2020) 2. Hao, S.N., Zhao, M.M., Luo, M.H.: Research on the influence of digital media art design on traditional art design. Dyn. Fashion 1, 76–77 (2021) 3. Ge, Q.Q., Kang, Y.: Research on digital media art in interaction design. Fashion Colour 3, 60–61 (2021) 4. Yu, Z.W., Zhou, S.M.: On the integration of digital media technology and art design. Voice Screen World 1, 56–57 (2021) 5. Wang, Y.T.: On the specific influence of digital media technology on modern art design. Digit. Space 1, 85–86 (2021) 6. Hou, Z.Y., Liu, F.F., Zhou, C.H.: On the innovative application of digital media art in exhibition design. Panorama Chin. Nationalities 4, 166–168 (2021) 7. Sun, X.C.: The influence of digital media art design on traditional art design. Shen Zhou 8, 243–244 (2021) 8. Ma, J.M.: Analysis of the influence of digital media technology on modern design art. Inf. Recording Mater. 22(2), 164–165 (2021) 9. Wu, X.F.: Application of digital media art in modern advertising design. Art Educ. Res. 3, 96–97 (2021) 10. Zheng, Q.F.: Application of digital media art in exhibition space design. Dyn. Fashion 2, 74–75 (2021) 11. Liu, A.J.: On the influence of digital media art on domestic animated film design. Art Educ. Res. 11, 110–111 (2021) 12. Yang, Z.N.: On the innovation and development trend of modern display art design under the digital media environment. Peak Data Sci. 10(5), 314 (2021) 13. Zhu, J.X.: Application analysis of digital media art design in urban cultural and creative industry. J. Beijing Inst. Graph. Commun. 29(4), 68–71 (2021) 14. Wang, Z., Liu, R., Liu, Q., Thompson, J.S., Kadoch, M.: Energy-efficient data collection and device positioning in UAV-assisted IoT. IEEE Internet Things J. 7(2), 1122–1139 (2020) 15. Chen, N., Rong, B., Zhang, X., Kadoch, M.: Scalable and flexible massive MIMO precoding for 5G H-CRAN. IEEE Wirel. Commun. Mag. 24(1), 46–52 (2017) 16. Chen, L., Yu, F.R., Ji, H., Rong, B., Li, X., Leung, V.C.M.: Green full-duplex self-backhaul and energy harvesting small cell networks with massive MIMO. IEEE J. Sel. Areas Commun. 34(12), 3709–3724 (2016)

Testing Method of Shipborne Radar in Virtual Verification System Fangning Tian(B) No. 38 Research Institute of CETC, Hefei 230088, Anhui, China [email protected]

Abstract. Radar plays an irreplaceable role in both military and civilian life. Radar testing runs through the entire life cycle of the radar from the initial design to the actual development, and then to the delivery of the finished product. In recent years, the competition for maritime rights has become more and more fierce, and it is imperative to establish a complete, powerful and effective coast guard and inspection system. In view of the above, the purpose of this document is to study the radar technology test method transmitted from the ship to the virtual verification system. This document first summarises the basic theory of the radar carried by the ship, and then extend the composition and basic working methods of the radar test system carried on board. This paper systematically explains the model, the production of a sample of errors, the extraction and analysis and software design of the virtual verification system And use comparative analysis method, observation method and other research forms to investigate the subject of this article. Experimental research shows that the performance of the virtual validation system of the radar studied on this paper is better, in particular, the accuracy of the test is improved by more than 15%, which fully reflects the usefulness of the virtual radar test verification system based on the radar technology transferred from the ship. Keywords: Machine intelligence · Radar test · Virtual verification · Test method · System research

1 Introduction The development process of the radar system mainly includes five stages: design, debugging, inspection, production and delivery to users. Radar testing occupies a very important position in the current radar system development [1, 2]. The reason is that the radar system needs to be controlled at all times during the design, generation, debugging and use of the radar [3, 4]. The radar test system can make the design of the radar development work synergistic, can also reduce the production cost, and can also improve the guarantee capability of the radar system itself, so that the radar system can work normally, and the combat effectiveness of the radar system can be maximized [5, 6]. Testability Virtual Verification is the first concept proposed by domestic scholars, and there is no such word as Testability Virtual Verification abroad. Testability virtual verification is a specific application of modeling simulation technology in the field © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 S. Sun et al. (Eds.): ICSINC 2021, LNEE 895, pp. 201–208, 2022. https://doi.org/10.1007/978-981-19-4775-9_25

202

F. Tian

of testability verification. The idea of combining modeling simulation and testability verification (M&SinT&E, Modeling and Simulation in Test and Evaluation) has been germinated very early abroad [5, 6]. SujoySen puts forward the idea of testability analysis and fault diagnosis based on simulation, and uses simulation to automatically generate the system’s dependency model, and the system’s testability index is obtained from the dependency model [7]. MajorDonaldPaulWaters proposed the M&SinT&E technical process for the first time, and believed that modeling simulation technology can be applied to testability verification, and testability verification can also be applied to modeling simulation. The two are integrated and inseparable [8]. The purpose of this document is to improve the performance of the radar and to study the method of testing the intelligent reliability of the radar in the virtual verification system. The test method of the test system based on the factor studied on this paper is the same as the traditional test method and uses the same set of data to perform simulation software to assess the suitability of this subject.

2 Research on the Test Application of Shipborne Radar Testability in Virtual Verification System 2.1 Analysis of Shipborne Radar Test System (1) System function design The functional structure of the system is mainly divided into three parts: the initialization module, the parameter measurement module, and the measurement display module. The hardware configuration is initialized by loading the radar test system hardware configuration file. The main parameters of the configuration file include channel path, adjustable step attenuation, gain, local oscillator frequency, trigger mode, sampling frequency, sampling depth, test object, etc. After loading, the test is initialized. The system interface makes the radar test system in the stage to be tested [10]. (2) Software overall architecture design The design idea of the radar test system software platform is to design a convenient and reliable radar test system. Therefore, the software’s ability to analyze and process radar signals is very important. While processing the radar signal, it is necessary to take into account the user’s operating experience and facilitate the user to observe the radar signal waveform. 2.2 Analysis of Main Indicators of Shipborne Radar Test System (1) Analysis of main technical indicators of radar launch The main measurement parameters of radar emission are: emission pulse repetition period; emission pulse repetition period ratio; emission pulse center frequency; emission pulse bandwidth; emission pulse intrapulse modulation characteristics; emission pulse envelope; emission pulse timing relationship; emission pulse agility characteristics; Transmit pulse power. (2) Main technical indicators of radar receiver The advantages of superheterodyne radar receivers are: high sensitivity, high gain, good selectivity and wide application range.

Testing Method of Shipborne Radar in Virtual Verification System

203

The main parameters for measuring the radar receiver are: the central frequency of the signal sound received; the pulse period of the signal received; its pulse width sound signal received; characteristics of intra-acoustic differentiation of the sound signal received; time relation of the signal received; level of validity of the signal sound received; LO frequency; receiver frequency is stable. 2.3 Analysis of Technical Parameters of Shipborne Radar The beamforming process is to form the main lobe of the pattern in the set direction, so that the signal gain in the specified direction is maximized, and the signal gain in the other directions is relatively reduced, to achieve the purpose of spatial selection. According to the different environment of the external radiation source radar and the change of the spatial position of the target, the beamforming technology can quickly change the beam direction to adapt to the new scene [11]. Take the linear antenna array of discrete elements as an example. It is assumed that each element is an ideal omnidirectional element. There are N in total, and the interval between each element is d. When the angle between the incident signal and the normal direction of the array antenna is θ, it is assumed that the element the signal received at 1 is s1 (t) = s(t)ej(ω0 t+ϕ(t))

(1)

Among them, s(t) is the original baseband signal, ω0 is the carrier frequency, and ϕ(t) is the phase. Then the model in the above figure can be calculated to obtain the expression form of the received signal of all the array elements at the same time: si (t) = s(t)ej(ω0 t+ϕ(t)) +

2π(i − 1) · d · sin(θ ) λ

(2)

where λ is the wavelength corresponding to the carrier frequency of the incident signal. Different arrays are at different positions in the array, which results in a wave path difference between each array that is related to the angle of the incident wave. 2.4 Virtual Verification Modeling of Radar Testability Based on Agent (1) Analysis of testability virtual verification modeling method This article believes that the test virtual verification model needs to have the following functions: This paper believes that the testable virtual verification model needs to have the following functions: quantitatively describe the functional behavior of the system equipment; qualitatively describe the structural composition of the system equipment; quantitatively describe the failure behavior of the system equipment; quantitatively describe the test behavior of the system equipment; The impact of the test [12]. Therefore, this article proposes to establish a five-element hybrid model including function, structure, fault, test, and environment. For the convenience of description, this article names the five-element hybrid model of function-structure-fault-test-environment “FSFTC model”.

204

F. Tian

(2) System level modeling System hierarchical modeling includes two aspects: one is the division of system hierarchy; the other is the division of system modules. The core purpose of dividing the system is to facilitate the maintenance of the system. Therefore, the system hierarchy and system module division proposed in this article are also developed around this perspective. The principle of division is to facilitate the maintenance of the system as the goal, and comprehensively consider the system. Factors such as the functional composition and physical structure of the equipment. 2.5 Generation of Fault Sample Set and Fault Feature Extraction and Analysis (1) Generation of fault sample set The fault sample set contains three parts, which are the number of fault types, the total number of fault sample sets, and the fault injection sequence. In order to produce the whole error sample, first determine the type of error, then thoroughly examine the validity index and the simulation time to determine the total number of error samples and, finally, take the error infusion sequence according to the error rate and the environmental factors of each failure. (2) Fault feature extraction and analysis When the system is performing fault injection, the first simulation is a non-fault simulation. This paper defines the non-fault injection as F0. From the fault model, we know that the system has N faults {F1 , F2 ,…, FN }, plus F0 is {F0 , F1 ,…, FN }, a total of N + 1 types. Knowing from the test model in the FSFTC model, the system has a total of M tests {T1 ,…, T2 , TM }, and YT is the output of the test. Suppose the test output corresponding to the fault Fx is: Yx = (Yx−1 , Yx−2 , ..., Yx−i , ..., Yx−M )

(3)

3 Experimental Research on the Testability of Shipborne Radar in the Virtual Verification System 3.1 Experimental Protocol To make this experiment more scientific and effective, This experiment compares the virtual radar verification system studied in this document with the traditional radar test methods to assess its appropriateness the research content. On this basis, five sets of experiments are set up for comparison, and the repetition period is input into the simulation software to compare and analyze the testability virtual verification system of the shipborne radar studied in this paper. And use the analytic hierarchy process to analyze the obtained results.

Testing Method of Shipborne Radar in Virtual Verification System

205

3.2 Research Methods (1) Comparative analysis method In this experiment, the radar based virtual verification system studied on this paper is compared to the traditional radar test and verification system to assess its appropriateness; the research content. (2) Observation method In this study, the efficiency and error of the test system running on the simulation software were observed and the data was recorded, and the recorded data was sorted and analyzed. These data provide reliable support for the final research results of this article. (3) Ahp The analytical hierarchy procedure shall be used for the statistical analysis of the research results of this paper.

4 Analysis of the Test Experiment Based on the Testability of the Shipborne Radar in the Virtual Verification System 4.1 Test System Performance Comparison Analysis In order to make the experiment more scientific and effective, this experiment uses the same data set to perform the simulation software, and the results are given in Table 1. Table 1. Test system performance comparison analysis Accuracy

High efficiency

Scalability

Others

Agent

71.2%

68.1%

67.9%

61.8%

Traditional

55.9%

56.7%

55.4%

50.3%

As shown in Fig. 1, compared to the traditional radar test system, the performance of the virtual radar verification system studied on this paper is better, in particular the test accuracy is improved by more than 15%, which fully reflects the ability to conduct radar tests carried out from the ship to the virtual verification system.

206

F. Tian

Percentages

Agent

Traditional

56.70% 55.90% 71.20%

55.40% 67.90%

68.10%

Accuracy

High efficiency

50.30% 61.80%

Scalability

Others

Categorys Fig. 1. Test system performance comparison analysis

4.2 Signal Pulse Width Test Results

Table 2. Signal pulse width test result 1

2

3

4

5

Test data 1 (μs)

0.399

1.003

50.002

100.001

200.002

Test data 2 (μs)

0.400

1.002

50.001

100.002

200.001

Test data 3 (μs)

0.4015

1.001

50.002

100.002

199.995

Test data 4 (μs)

0.4005

1.002

49.997

100.002

200.002

Test data 5 (μs)

0.400

1.003

50.001

100.003

200.001

In order to further study the subject of this article, this experiment will run the shipborne radar test virtual verification system on the simulation software and test the signal pulse width. The results are shown in Table 2. It can be seen from Fig. 2 that the average error of group 1 is 0.0002, group 2 is 0.0022, group 3 is 0.0006, group 4 is 0.002, and group virtual verification 5 is 0.0002. Both are less than 0.01, it fully reflects the outstanding performance of the virtual verification system of virtual radar tests being studied on this paper.

Testing Method of Shipborne Radar in Virtual Verification System

Test data 5 (μs) Test data 2 (μs)

Test data 4 (μs) Test data 1 (μs)

207

Test data 3 (μs)

5

Groups

4 3 2 1 0

50

100

150

200

250

us Fig. 2. Signal pulse width test result

5 Conclusion This paper briefly summarizes the basic processing algorithms of the shipborne radar radar test virtual verification system. The performance of the digital beamforming algorithm is directly related to the spatial filtering effect and the system angle measurement accuracy. The performance of the clutter cancellation algorithm directly determines the detection range and detection of the system. Probability, so these two aspects are the key links in the radar system, and the realized performance determines the detection and measurement performance of the radar test virtual verification system. Therefore, this article mainly discusses the testing methods of the radar system’s digital beamforming, clutter cancellation, and system detection performance. The experiment shows that the virtual verification method designed in this article based on airborne radar testing has higher accuracy, the error is smaller.

References 1. Tian, J., Liu, J., Yan, D., et al.: An assimilation test of Doppler radar reflectivity and radial velocity from different height layers in improving the WRF rainfall forecasts. Atmos. Res. 198(12), 132–144 (2017) 2. Wang, G., Wu, W., Chen, X., et al.: A radar parameter test and examination system based on job path evaluation. Firepower Command Control 044(004):131–135, 141 (2019) 3. Zheng, S., Wang, G., Huang, X., et al.: Design and analysis of transmitter charge switch assembly load on weather radar test platform. J. Geosci. Environ. Prot. 06(10), 51–58 (2018) 4. Li, X., Tao, X., Zhu, B., et al.: Research on a simulation method of the millimeter wave radar virtual test environment for intelligent driving. Sensors 20(7), 1929 (2020) 5. Wang, Q., Zha, Y.: Research on measurement radar distribution and radar accuracy test route design. Radar Sci. Technol. 014(006), 675–680 (2016) 6. Xu, X., Xu, T., Wang, R.: Technical status evaluation of missile radar seekers based on test data. J. Naval Aeronaut. Eng. Inst. 033(006), 553–559 (2018)

208

F. Tian

7. Sen, S.: Classification of low RCS targets based on PCL radar network using statistical hypothesis test. J. Korean Inst. Electromagn. Eng. Sci. 30(11), 911–921 (2019) 8. Donald, M., Water, P.: Simulation test environment of radar seeker under system background. Syst. Simul. Technol. 014(004), 297–303 (2018) 9. Ke, K.: Simulation on minimum jamming distance test of radar countermeasure equipment. Syst. Simul. Technol. 014(002), 105–108 (2018) 10. Shi, A.: A L Band target instruction radar maintenance integrated test instrument——based on virtual instrument technique. Electro. World 000(010), 196–197 (2017) 11. Shao, N., Zhi-Chao, B.U., Bai, L.I., et al.: The technology evaluation of S-band business test dual polarization radar (CINRAD/SA-D). Beijing Ligong Daxue Xuebao/Trans. Beijing Inst. Technol. 38(9), 953–958 (2018) 12. Li, H., Li, H., Xia, H.: Secondary radar antenna test method based on mobile test vehicle. ordnance industry automation 037(012):39–42, 57 (2018)

Coexistence Analysis Between HIBS System and IMT System Below 3GHz Band Li Wang1,2(B) , Cheng Wang1 , Weidong Wang1 , and Zhiyan Fan2 1 Beijing University of Posts and Telecommunications, Beijing 100876, China

[email protected] 2 China Academy of Information and Communication Technology, Beijing 100191, China

Abstract. In order to achieve spectrum sharing between HIBS (high altitude platform station as IMT base stations) systems and existing ground communication systems, it is necessary to study interference assessment. In this study, considering the coexistence scenarios of HIBS and IMT systems, the coexistence topology and parameters are given, the evaluation criteria, propagation model and power control model are studied, and the ACIR of coexistence between the two systems is derived. Finally, the throughput of coexistence between the two systems is given through simulation. Keywords: HIBS · IMT · Coexistence analysis · Throughput

1 Introduction With the rapid development of wireless communication technology, more attention is payed to investigate the broadband communication of high-altitude platform stations (HAPS). HIBS is a kind of specific HAPS. Its platform bears International Mobile Telecommunication (IMT) base station and provide terrestrial mobile broadband services. HIBS has many advantages. It can construct network in rural and remote areas easily and provide broadband services with much lower price; it can also be used for emergency communications in disaster scenarios. Therefore, the World Radiocommunication Conference (WRC) carried out a series of studies for spectrum identification of HIBS. This paper mainly focuses on the coexistence study of HIBS and IMT systems in the frequency band below 3 GHz, like 694–960 MHz, 1710–1980 MHz and 2110– 2170 MHz. The topology of coexistence systems is structured in this paper. System parameters, evaluation criteria, propagation model and power control scheme are given. The deduction of adjacent channel interference power ratio (ACIR) between HIBS and IMT systems is also conducted in this paper. Monte Carlo simulation is adopted for coexistence performance of HIBS and IMT systems. This paper studies the coexistence between HIBS and IMT in adjacent frequency bands in the same region. The following evaluation scenarios are considered: (1) HIBS base stations interfere with IMT terminals (downlink). (2) HIBS terminals interfere with IMT base station (uplink). © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 S. Sun et al. (Eds.): ICSINC 2021, LNEE 895, pp. 209–215, 2022. https://doi.org/10.1007/978-981-19-4775-9_26

210

L. Wang et al.

(3) IMT base stations interfere with HIBS terminals (downlink). (4) IMT terminals interfere with HIBS base station (uplink). The rest of this paper is organized as follows. In Sect. 2, the topology, system parameters, propagation model and power control model required by system coexistence are described in Sect. 2. Section 3 discusses the calculation method of ACIR between HIBS and IMT systems. The simulation steps of uplink and downlink and the simulation results based on the Monte Carlo method are given in Sect. 4. Finally, conclusions are drawn in Sect. 5.

2 System Model 2.1 System Topology The topology of HIBS system is shown in Fig. 1. It is assumed that the coverage area of a cell is hexagonal, where A represents the coverage radius of HIBS and B represents √ the distance between two HIBS nodes, B = A 3 [1].

Fig. 1. HIBS system topology.

Fig. 2. HIBS multi-sector deployment topology.

A multi-beam configuration scheme is adopted for HAPS system in order to divide its coverage area into multiple sectors (e.g. 3 or 7). The multi-sector deployment is shown in Fig. 2. In this paper, the 7 cells deployment is considered. IMT system adopts

Fig. 3. IMT system topology.

Fig. 4. HIBS and IMT coexist topology

Coexistence Analysis Between HIBS System and IMT System

211

a 3-sector topology in order to eliminate the edge effect. Cell radius R = ISD/(ISD is station spacing), which is shown in Fig. 3. The topological structure of HIBS and IMT system in the same regional deployment is shown in Fig. 4. HIBS adopts 7-sector structure, while IMT system adopts 19-cell three-sector structure, where D is the distance between ground projection point O of HIBS node and IMT network center. 2.2 System Parameter According to the conclusion of ITU-R WP5D, HIBS system parameters are shown in Table 1. The detailed parameters can be found in [2]. Table 1. HIBS system parameters Parameter name

HIBS base station

Frequency band

Band 1: 694–960 MHz Band 2: 1 710-1 980 MHz, 2 010-2 025 MHz, 2 110-2 170 MHz

HIBS terminal

Channel bandwidth

20 MHz

Duplex method

FDD

Transmission Power

55 dBm (1st layer), 58 dBm (2nd layer)

23 dBm

ACLR

45 dB

30 dB

The deployment parameters of HIBS base station and terminal are in Table 2. Table 2. Deployment parameters of HIBS base station and terminal. Parameter name

Band 1

Band 2

HIBS Platform Antenna pattern

Recommendation ITU-R M.2101

Element gain

8 dBi

Horizontal/vertical 3 dB beamwidth of single element

65º for both H/V

Horizontal/vertical front-to-back ratio

30 dB for both H/V

Antenna array configuration (Row × Column)

2 × 2 elements (1st layer cell), 4 × 2 elements per cell (2nd layer cell)

Horizontal/Vertical radiating element spacing

0.5 of wavelength for both H/V

HIBS Platform Antenna tilt

90º (1st layer cell), 33º (2nd layer cell)

UE density for equipment that are transmitting simultaneously

3 UEs per cell

90º (1st layer cell), 23º (2nd layer cell)

212

L. Wang et al.

AAS antenna is not considered for base stations or terminals below 1 GHz, but only for those between bands 1 and 3 GHz. 2.3 Propagation Model ITU-R Recommendation P.1409-1 is a newly published version which revised HIBS propagation prediction method and extended the applicable frequency range down to 0.7 GHz. For the propagation link between HIBS node and its ground station, the basic propagation loss is calculated by [3]. The impact of clutter loss [4] is also considered in this paper, which is calculated according to the liaison statement [5] sent by SG3 to 5D. The application scenarios include urban and suburban. The calculation of clutter loss needs some input parameters, such as frequency band, elevation angle, and position percentage, height of ground reference object and height of ground station. 2.4 Power Control Model The power control algorithm [6] for user equipment for common research is shown as follows: PPUSCH (i) = min(PCMAX , 10 log10 (MPUSCH (i)) + PO_PUSCH (j) + a(j) · PL) where PPUSCH is the transmission power, PCMAX is the maximum power, MPUSCH is the resource block number, PO_PUSCH is the target power for each resource block, a is the balance factor, PL is the pathloss.

3 ACIR (1) HIBS node interfere with IMT terminals (downlink) Both downlink transmissions of HIBS and IMT systems have only one user at the same sub-frame, who occupies all the bandwidth. The ACIR of IMT FDD terminal at 10 MHz bandwidth is 33 dB [7]. (2) HIBS terminals interfere with IMT base station (uplink). The ACIR of the uplink transmission is determined by the ACLR of the terminal [8]. The bandwidth of HIBS system is 20 MHz. Each sub-frame scheduled 3 users. Each user occupies 16RB, and the RB bandwidth is 375 kHz. The bandwidth of IMT FDD system is 10 MHz, each sub-frame has 3 users, each user occupies 16RB, and the RB bandwidth is 180 kHz. Figure 5 shows the frequency usage of terminals. For HIBS terminal which is adjacent to the frequency, the FACLR of the IMT user is = 10 × LOG10(Bvictim /BAggressor ) = 10 × LOG10(2.88/6) = −3.18 dB. The ACLR of 1 the third user is 10 ∗ log10 ( 0.24/6·10−3 +2.64/6·10 −4.3 ) = 42.07 dB. (3) IMT base stations interfere with HIBS terminals (downlink) Each downlink HIBS and IMT sub-frame has one user occupying all bandwidth. The ACIR of HIBS FDD terminal at 20 MHz bandwidth is 27 dB.

Coexistence Analysis Between HIBS System and IMT System

213

Fig. 5. HIBS terminals interfere with IMT base station (uplink)

(4) IMT terminals interfere with HIBS base station (uplink) The working bandwidth of HIBS system is 20 MHz, each sub-frame has 3 users, each user occupies 16RB, and the RB bandwidth is 375 kHz. The operating bandwidth of IMT FDD system is 10 MHz, each sub-frame has 3 users, each user occupies 16RB, and the RB bandwidth is 180 kHz (Fig. 6).

Fig. 6. IMT terminals interfere with HIBS base station (uplink)

The ACLR of 1st IMT terminal to 1st HIBS terminal is 1 10 ∗ log10 ( 10−3 +10−4.3 +0.24/2.88·10 −5 ) = 29.78 dB, to 2nd and 3rd HIBS terminal is

1 nd IMT terminal to 1st HIBS termi10 ∗ log10 ( 6/2.88·10 −5 ) = 46.81 dB. The ACLR of 2 1 nal is 10 ∗ log10 ( 2·10−4.3 +0.24/2.88·10 −5 ) = 39.95 dB, to 2nd and 3rd HIBS terminal is

1 rd 10 ∗ log10 ( 6/2.88·10 −5 ) = 46.81 dB. The ACLR of 3 IMT terminal to all HIBS terminal

1 is 10 ∗ log10 ( 6/2.88·10 −5 ) = 46.81 dB. The ACS of HIBS is 37.6 dB, therefore the ACIR is ACIR = {29.11 + X, 37.10 + X, 37.10 + X;35.61 + X, 37.10 + X, 37.10 + X;37.10 + X, 37.10 + X, 37.10 + X;}.

4 Simulation Results Figure 7 shows the coexistence of the HIBS system and IMT system in the 900 MHz frequency band. It can be seen that for HIBS base stations that interfere with IMT terminals, the rural scenario has the most significant interference. The reason is that the received signal power of IMT terminals is weak in the rural scenario, which is easy to be interfered by other systems. In addition, it is shown that the IMT base station also has great interference to the HIBS terminal. Since the coverage areas of HIBS and IMT systems are overlapped, and HIBS terminals will receive interference signals from surrounding IMT base stations. We can also see that the IMT terminal has significant interference to the HIBS base station. Therefore, in this frequency band, it is necessary to research the protection strategy

214

L. Wang et al.

Fig. 7. 900 MHz frequency band IMT and HIBS spectrum coexistence results

between the HIBS system and IMT systems to avoid excessive coexistence interference. Figure 8 and Fig. 9 show the coexistence of the HIBS system and IMT system in the frequency band 1800 MHz and 2100 MHz separately. The interference trend of the two bands is the same. The interference of the HIBS base station to the IMT terminal and the IMT terminal to the HIBS base station should be considered. The former is due to the weak IMT coverage capability in suburban scenarios, while the latter is because the number of IMT terminals is much more than that of HIBS terminals. Therefore, interference mitigation scheme is an important research field in the future.

Fig. 8. 1800 MHz frequency band IMT HIBS spectrum coexistence results

Fig. 9. 2100 MHz frequency band IMT and HIBS spectrum coexistence results

5 Conclusion In this paper, we investigate the coexistence of HIBS and IMT systems. Firstly, the frequency usage of the HIBS system is introduced. Secondly, we give the system model and analyze the system topology, system parameters, propagation model and power control model. Once again, we provide the calculation process of ACIR between systems when the HIBS system coexists with the IMT system. Finally, the coexistence situation

Coexistence Analysis Between HIBS System and IMT System

215

of the HIBS system and the IMT system is presented through simulation. It can be seen from the simulation results that there is a certain degree of mutual interference between the two systems, mainly in the scenario where the HIBS base station interferes with the IMT terminal, and the IMT terminal interferes with the HIBS base station. It is necessary to introduce reasonable and effective interference mitigation measures.

References 1. Chapter 4 of Document 5D/716. Annex 4.19 - Working document towards a preliminary draft new Report ITU-R M. [HIBS-CHARACTERISTICS]/Working document related to WRC-23 agenda item 1.4. https://www.itu.int/dms_ties/itu-r/md/19/wp5d/c/R19-WP5D-C0716!H4-N4.19!MSW-E.docx 2. Chapter 4 of Document 5D/716.Annex 4.20 - Working document towards sharing and compatibility studies of HIBS under WRC-23 agenda item 1.4 - Sharing and compatibility studies of high-altitude platform stations as IMT base stations (HIBS) on WRC-23 agenda item 1.4. https://www.itu.int/dms_ties/itu-r/md/19/wp5d/c/R19-WP5D-C-0716!H4N4.20!MSW-E.docx 3. Annex 03 of WP 3K 178: Updated supplemental digital products for the draft revision of Recommendation ITU-R P.528-4 - A propagation prediction method for aeronautical mobile and radionavigation services using the VHF, UHF, and SHF bands. https://www.itu.int/dms_ ties/itu-r/md/19/wp3k/c/R19-WP3K-C-0178!N03!MSW-E.docx 4. ITU-R recommendation P.2108-1: Prediction of clutter loss. https://www.itu.int/rec/R-REC-P. 2108-1-202109-I/en 5. Document 5D/723-E: Propagation information requested from Working Party 5D. https://www. itu.int/md/R19-WP5D-C-0723/en 6. ITU-R recommendation M.2101-0: Modelling and simulation of IMT networks and systems for use in sharing and compatibility studies. https://www.itu.int/rec/R-REC-M.2101-0-201702I/en 7. 3GPP TS.38.101:User Equipment (UE) radio transmission and reception (Release 15). https:// www.3gpp.org/ftp/Specs/archive/38_series/38.101-1 8. 3GPP TS 36.942: Radio Frequency (RF) system scenarios (Release 10). https://www.3gpp. org/ftp/Specs/archive/36_series/36.942/

Photovoltaic Power Prediction Based on Wavelet Analysis Lianhe Li1 , Jihan Cao2 , Tao Hong2(B) , Mingshu Lu3 , Weiting Zhao4 , and Linquan Fang5 1 Henan Logistics Vocational College, Zhengzhou 453514 , Henan, China 2 Beihang University, Beijing 100191, China

[email protected]

3 University of California Irvine, Irvine , USA 4 State Grid Shandong Electric Power Company, Jinan, China 5 Yunnan Innovation Institute·BUAA, Kunming 650233, China

Abstract. The randomness and volatility of photovoltaic power have negative effect on its application, power prediction is the key to solve this problem, however, existing photovoltaic power prediction methods have the problem such as single model, insufficient parameters and large error. Based on 5G and Internet of things technology, real-time monitoring of photovoltaic equipment and weather conditions can be carried out, so as to provide data support for power prediction. According to the collected data, this paper proposes a hybrid photovoltaic power prediction model based on discrete wavelet transform, convolution neural network, Long Short-Term Memory and Numerical Weather Prediction (DWTCNN-LSTM-NWP Model) to reduce the prediction error, then the effectiveness of the new model is verified by simulation. Keywords: 5G · Internet of Things · Power prediction · Green energy

1 Introduction With the development of 5G and Internet of Things (IoT) technology, we can track the production process in a more comprehensive and real-time way, which can improve the fault diagnosis ability and productivity level, realize the intelligent development of society. With the aggravation of resource depletion and carbon emissions, the traditional energy structure has been unable to meet the needs of social development, and energy transformation is imperative. Green renewable energy such as photovoltaic (PV) power is an important force in energy transformation, and its cost has been significantly reduced [1, 2]. Statistical data show that as of November 2020, China’s PV installed capacity ranks first in the world. However, the inherent discontinuity, volatility and randomness of PV lead to its unstable output, therefore, PV grid-connected will bring severe challenges to the stability and even security of power grid [3]. The prediction of PV power can provide effective decision-making information for power dispatching, reduce PV power’s adverse © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 S. Sun et al. (Eds.): ICSINC 2021, LNEE 895, pp. 216–222, 2022. https://doi.org/10.1007/978-981-19-4775-9_27

Photovoltaic Power Prediction Based on Wavelet Analysis

217

impact, reduce the operation and maintenance cost of power system, and hasten the development of renewable energy [4]. However, under the influence of weather, cloud, humidity, season and other factors, PV power shows complex nonlinear characteristics, which is difficult to accurately predict. The IoT technology can be applied to the PV power prediction. Specifically, the distributed sensors are used to monitor the status of power generation equipment, in addition, with the development of satellite and radar detection technology, the accuracy of weather prediction is greatly improved, which can provide more basis for power prediction [5]. Then, the sensor data and weather forecast data are transmitted to the computing centre through the new generation communication technology, the computing centre carries out power prediction by analysing and processing the data [6]. Compared with the traditional PV power prediction method, the method based on the IoT can provide more data support, which makes it possible to reduce the prediction error.

2 Related Work From the time scale, PV power prediction can be categorized into: ultra-short-term, short-term, medium-and-long-term. Ultra-short-term refers to predict the output power in the next 0–6 h, short-term means to predict the next 1–2 days’ output power, mediumand-long-term generally is to calculate the output power in the next few months or one or two years. The ultra-short-term and short-term power prediction strive to obtain the output power in a period of time accurately, which is often used in power grid dispatching. Therefore, short-term PV power prediction has always been the focus of research. From the perspective of method, the main methods of PV power prediction include: physical model, statistical model and machine learning [7]. The physical model method first models the power generation equipment, and then takes the numerical weather prediction (NWP) information as the input to predict the output power in the future; this method can’t be directly transplanted between different equipment, and the error is large, either [8]. The statistical model method usually analysis the autocorrelation between output power by statistical methods [9], this input data of this method is few and the calculation is relatively simple, but it can only learn the linear relationship between the input data. The prediction method based on machine learning takes historical sample data as the research object, considers the hidden characteristics of the data itself, and learns the mapping relationship between vectors from it, due to it’s good nonlinear modelling ability, PV power prediction based on machine learning is the current research focus. A PV power prediction model combining Artificial Neural Network (ANN) and genetic algorithm is proposed in [10], which effectively reduces the error. Another PV power prediction method combining convolutional neural network (CNN) with Long ShortTerm Memory (LSTM) was proposed in paper [11], which can effectively save the training time. Currently, some green energy power generation power prediction systems have been established, such as Zephry in Denmark and WPMS in Germany. However, they do not overcome the problem of large errors in the single system method [12].

218

L. Li et al.

3 Methods It can be seen from the second chapter that it is difficult to find out all the mapping relationship between power and other factors through a single method. Therefore, the combination of different predicting methods is expected to achieve better results. With the development of the IoT technology, the data collection for all aspects of PV power generation is more and more comprehensive, which provides a solid data foundation for power prediction. Aiming at the problems of the single input data and large output error of existing PV power prediction methods, this paper constructs a new hybrid power prediction model by combining physical model with machine learning method, in order to reduce the prediction error. A. Discrete Wavelet Transform Discrete Wavelet Transform (DWT) is a wavelet transform for discrete signals, which is very effective for complex sequence analysis. After DWT, the original data will be decomposed into approximate components and detail components. The approximate components correspond to the steady-state characteristics of the original data, while the detail components are the transient characteristics of the original data. The principle is as follows:   j (1) ψj,k (t) = 2 2 ψ 2j t − k   j ϕj,k (t) = 2 2 ϕ 2j t − k

(2)

where ψ(t) is the mother wavelet function, ϕ(t) is the scale function, t represents the time index, k and j are the translational scale variable and scaled scale variable, respectively. The DWT is used to decompose the historical power data into a stable part and a fluctuating part. Compared with the original input, the frequency components of each sub-sequence obtained after decomposition are more unitary, which is beneficial to improve the predict accuracy. B. Machine Learning ANN can learn the mapping relationship between input and output values automatically, and it has good nonlinear characteristics. So ANN is used to process the input the NWP information in our model, aiming to find the relationship between the input NWP and the output power. CNN has strong feature extraction capabilities, which is used in this paper. The historical power sequence is regarded as grid data sampled at fixed time intervals, and the local features of the input information are extracted through convolution operation, which can dig deeper into the information implicit in the input data. PV power not only has a mapping relationship with weather factors, but also has an implicit autocorrelation relationship. ANN has no time memory characteristics, so it cannot effectively process time series depend on ANN alone. Compared with ANN, Recurrent Neural Network (RNN) has one more backward feedback connection, the function of feedback connection is to establish a connection between adjacent time steps. That is, the output at each moment is not only affected by the current input, but

Photovoltaic Power Prediction Based on Wavelet Analysis

219

also by the output of the previous moment. Influence, its mathematical principle can be written as:   y(t) = φ x(t)T · wx + y(t − 1)T · wy + b (3) where, wx is the weight matrix of forward propagation, wy is the weight matrix of back propagation, b is the bias, and φ is the activation function. Since LSTM solves the problem of gradient disappearance of RNN while maintaining temporal memory, in this article, we use LSTM to process the power sequence. C. Architecture of Hybrid Model Based on the above analysis, we propose a hybrid model combining DWT and machine learning (ML) to predict PV power, the architecture of the new model is shown in Fig. 1. As can be seen from the figure, the model is divided into two modules: (1) neural network based on DWT-CNN-LSTM, the input of which is the historical power information; (2) ANN, which uses the idea of physical model for reference, and the input of it is the NWP of future time.

Fig. 1. The new model for PV power prediction.

220

L. Li et al.

4 Experiments A. Data Set The data set used in this paper is from a PV power station in China. The data items collected include PV power and weather information, wind speed, temperature, solar radiation, wind direction, precipitation and air pressure included. The time interval of data collection is 15 min, that is, 96 records are recorded every day. In order to eliminate the influence of each attribute dimension, this paper uses the maximum minimum normalization method to process each data in the data set, and reduces it to [0,1]. The calculation formula is: xt  =

xt − xmin xmax − xmin

(4)

In the above formula, xmax and xmin are the maximum and minimum values respectively. After the prediction, it is necessary to carry out the inverse normalization operation, the formula is as follows. xt = xt  ∗ (xmax − xmin ) + xmin

(5)

B. Experimental Results In this paper, the PV power of the next day will be predicted. The data set is divided into training set, validation set and test set in a ratio of 7:1:2. Using the training set to train the model, and the validation set to select the model, the test set is used for performance evaluation. If the predicted data can be expressed as [yt , yt+1 , . . . , yt+95 ], then input of the DWT-CNN-LSTM module and the ANN module can be expressed as [yt−96 , yt−95 , . . . , yt−1 ] and [xt , xt+1 , . . . , xt+95 ] respectively, xi represents the weather information at the ith moment. Use the mean square error (MAE) and square absolute error (RMSE) to evaluate the effects of different models, the smaller the value of RMSE and MAE, the better the performance of the model. N   

MAE =

RMSE =

t=1

 yt − yt  

(6)

N   N  2  t=1 yt − yˆ t N

(7)

Predict the PV power by different models, and the results are as follows (Fig. 2 and Table 1).

Photovoltaic Power Prediction Based on Wavelet Analysis

221

Fig. 2. The prediction results of different models of PV power.

Table 1. Comparison of prediction effects of different models of PV power. New model

ANN

LSTM

Physical

MAE

39.05

55.90

90.67

48.29

MRSE

69.07

89.30

147.15

83.74

5 Conclusion In view of the problem that photovoltaic power is volatile and difficult to predict, combined with the new technology of the IoT era, this paper proposes a new photovoltaic power prediction model (DWT-CNN-LSTM-NWP Model). The new hybrid model used discrete wavelet transform to decompose the non-stationary time series into relatively stationary sub-sequences, then ANN, CNN and LSTM are used for feature learning. Experiments show that, compared with the traditional model, the new model can effectively reduce the MAE and RMSE of the prediction results. Acknowledgement. This work was supported by the National Key Research and Development Program of China (Project No. 2018YFB0505100).

References 1. Sithan, M., Lai, L.L.: Application of green technologies in developing countries—reduced carbon emission and conservation of energy. In: 2011 IEEE Power and Energy Society General Meeting, pp. 1–7. IEEE (2011) 2. Hui, H., Ding, Y., Shi, Q., et al.: 5G network-based Internet of Things for demand response in smart grid: a survey on application potential. Appl. Energy 257, 113972 (2020)

222

L. Li et al.

3. Sun, Y., Wang, F., Zhen, Z., et al.: Research on short-term module temperature prediction model based on BP neural network for photovoltaic power forecasting. In: 2015 IEEE Power & Energy Society General Meeting, pp. 1–5. IEEE (2015) 4. Wang, S., Wang, Y., Cheng, Y., et al.: An improved model for power prediction of PV system based on Elman neural networks. In: 2020 Asia Energy and Electrical Engineering Symposium (AEEES). IEEE (2020) 5. Cammarano, A., Petrioli, C., Spenza, D.: Pro-Energy: a novel energy prediction model for solar and wind energy-harvesting wireless sensor networks. In: 2012 IEEE 9th International Conference on Mobile Ad-Hoc and Sensor Systems (MASS 2012), pp. 75–83. IEEE (2012) 6. Banerjee, T., Sheth, A.: Iot quality control for data and application needs. IEEE Intell. Syst. 32(2), 68–73 (2017) 7. Yu, F., Dong, C., Jiang, J.: Summary of research on power forecasting technology of new energy generation. In: 2019 2nd Asia Conference on Energy and Environment Engineering (ACEEE), pp. 49–53. IEEE (2019) 8. Biyun, C., Suifeng, W., Yongjun, Z., et al.: Wind power prediction model considering smoothing effects. In: 2013 IEEE PES Asia-Pacific Power and Energy Engineering Conference (APPEEC), pp. 1–4. IEEE (2013) 9. Li, Y., Zhang, J., Xiao, J., et al.: Short-term prediction of the output power of PV system based on improved grey prediction model. In: Proceedings of the 2014 International Conference on Advanced Mechatronic Systems, pp. 547–551. IEEE (2014) 10. Watson, J.D., Crick, F.H.C.: A structure for deoxyribose nucleic acid. Nature 171(4356), 737–738 (1953) 11. Wang, K., Qi, X., Liu, H.: A comparison of day-ahead photovoltaic power forecasting models based on deep learning neural network. Appl. Energy 251, 113315 (2019) 12. Xu, M., Lu, Z., Qiao, Y., et al.: Study on the adaptability of day-ahead wind power forecast system for on-site use. In: 2013 IEEE Power & Energy Society General Meeting, pp. 1–5. IEEE (2013)

Energy-Efficient Networking for Emergency Communications with Air Base Stations Zifan Li1 , Bozhong Li1 , Hongxi Zhou1 , Yuanlong Peng1 , Fang Chen1 , Jingyue Tian2(B) , and Peng Yu2 1 State Grid Information and Telecommunication Branch,

Beijing 100761, People’s Republic of China 2 Beijing University of Posts and Telecommunications, Beijing 100876, China

[email protected]

Abstract. With the development of 5G technology, a convenient and fast emergency communication solution is needed when the local ground base station is unavailable for disaster. This paper put forward a method of high throughput and low energy 3D position of air base station by considering the users’ service quality and the energy consumption of the air base station: first, cluster the users, then build the system optimization model by considering interference model, path loss model, energy consumption model and resource allocation model, and finally find the 3D position by using the network coverage compensation algorithm based on PSO. Finally, evaluate the simulation experiments. We found this method can effectively meet the emergency communication needs, maximize the energy efficiency ratio of the air base station, qualify the user’s communication quality needs. Keywords: Emergency communications · 5G · Air base station

1 Introduction Air base station, different from fixed location of the traditional ground base station, can move freely and be deployed flexibly at a low cost. The flexibility of the drone can effectively reduce the loss of the signal and complement the coverage of the signal [1]. And it also allows the network to quickly adapt to the current scenario to provide uninterrupted and high-quality communication services to users in poorly covered areas [2] and traffic burst areas [3] after the disaster. The air base station discussed in this paper is to use the drone as the carrier for emergency communication. Based on the 5G mmwave system, the discussion process of this paper is first clusters users, then builds the system optimization model, and then designs the location planning algorithm of multi-air base station, and finally carries on the simulation experiment. The research on the location deployment of air base station can effectively enhance the flexibility, real-time and adaptability of the network, and get full use of the energy, and provide new solutions for emergency communication.

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 S. Sun et al. (Eds.): ICSINC 2021, LNEE 895, pp. 223–231, 2022. https://doi.org/10.1007/978-981-19-4775-9_28

224

Z. Li et al.

2 Related Work The problem studied is based on the 5G network Cell Outage Compensation (COC) problem and is to propose a simple and fast self-healing technique. The conventional interrupt compensation method is to adjust network parameters, such as antenna power and antenna angle of adjacent cell base stations [4]. The main idea is to expand the coverage area and network capacity of the adjacent cell as much as possible [5], but it will affect the neighboring cell itself. So we need to consider new technologies or devices to solve COC problems in 5G network. In the paper [6], the author uses the model based on LSTM and DRL for dynamic traffic model construction and allocation of bandwidth and power resource. And results shows it can accurately predict the multicast needs and improve energy efficiency. The paper [7] gives a method of maximizing the coverage area while meeting the users’ data rate requirements and the capacity of the drone base station, but it doesn’t consider height a lot. The paper [8] proposes a three-dimensional deployment method of distributed UAVs covered in downlink network. This approach takes the dynamic changes and maximized coverage of the users into account. Paper [9] proposes a green resource allocation mechanism driven by lloT intelligence under 5G heterogeneous network, and the result shows that it is superior to other DL methods and the QoS is kept above acceptable level. Paper [10] proposes a method based on Effective Differential Evolution (DE) to minimize the maximum energy consumption for all devices by jointly optimizing drone trajectories and transmission schedules, while ensuring the reliability of data collection and the required 3D positioning. Paper [11] designs a novel precoding scheme for the power supply and backhaul of SBS, which effectively eliminates inter-layer and multi-user interference, improves energy efficiency, and adapts to various of scenarios. To sum up, the COC problem under 5G network does not consider enough about the energy consumption of drones, multi-drone joint deployment, user distribution and other factors.

3 Optimize Model Building 3.1 System-Related Models Interference Model. We assumes that all drones share the same band for a continuous period of time, with a channel set C = {1, 2, ..., c}. Interference Between Base Stations. UAV j traverses to detect whether channel c is idle before communicating with the user. If channel c is idle, it is marked as used and served to the user. Interference Between Users. Within channel c, UAV j serves ground users within the time range T in the form of TDMA. Suppose that the continuous time T is divided into N time slots, each of which is δ = NT .

Energy-Efficient Networking for Emergency Communications

225

Path Loss Model. When UAV communicates with users, it typically includes two types: Non-Line-of-Sight links (NLoS) and Line-of-Sight (LoS).Within time slot t, the probability of LoS propagation of communication between UAV j and User i is LoS (t) = Pi,j

m n=0

[1 − exp(−

(hj (t) − (n + 1/2)(hj (t) − hi (t))/(m + 1))2 )] 2γ 2

(1)

represent the height Especially of UAV j and User i in the time slot t, α, β is the environmental constant. So in time slot t, the probability of NLoS propagation between UAV j and User i is NLoS LoS Pi,j (t) = 1 − Pi,j (t)

(2)

Within the time slot t, the road loss model between UAVj and User i is ⎧ ⎪ ⎨ Li,j (t) = 20 log( 4π fc di,j (t) ) + ηLoS LoS c ⎪ ⎩ Li,j

NLoS (t)

=

4π f d (t) 20 log( c c i,j ) + ηNLoS

(3)

where the fc is carrier frequency, c is the speed of light, di,j (t) is the distance between UAV j and User i in the time slot t, assuming that the 3D position coordinates of UAV j within the time slot t and user i is (xi (t), yi (t)) so ·ηLoS and ηNLoS are the averages of overpath losses at the top of the free space of the environment-determined LoS and NLoS links. Therefore, the average link loss for the air-to-ground model is i,j

i,j

LoS NLoS (t)LNLoS (t) Li,j (t) = Pi,j (t)LLoS (t) + Pi,j

(4)

Within the time slot t, the loss of the entire link from UAV j to User i is path

ηi,j (t) = Li,j (t) + 10δpath lgdi,j (t)

(5)

Energy Consumption Model. The energy consumption of UAV consists of three parts. The first part is the communication energy caused by radio radiation and signal processing. The second part is the hover energy consumed by the air base station when hovering in the air. The third part is the propulsion energy consumed by UAV moving at high altitude. By considering the above three power, the total energy consumption of the drone is calculated [12]. Suppose Pp is the power to overcome the aircraft surface and UAV j consumes a total of Ej (t) in the time slot t, expressed as (6)

226

Z. Li et al.

(7) where M is the number of rotors, G = (W + m)g is Newton force, where W is frame weight (kg), m is battery and payload weight (kg), g is gravity (N /kg), ρ is the fluid density of the air (kg/m3 ), β is the rotor disc radius (m), CD0 represents the drag coefficient, the cb is the rotor string, S is the reference area, and w is the angular velocity vh , va , and vd represent velocity in horizontal directions, rising velocity, and falling velocity. is an indicator function. (8) The total energy consumption within the continuous time range T of all UAVs is E, expressed as  N  N  T E= Ej = EjC (t) + EjH (t) + EjM (t) (9) j=1

j=1 t=0

Resource Allocation Model. We choose to use the equal allocation model to allocate the total bandwidth for fair communication. The bandwidth allocated to each user is: bi = B/N, where B represents the total bandwidth of the air base station and N represents the total number of users in the coverage range [13].

3.2 Problem Optimization Model Assume that user i for Class a QoS requirements allocates a bandwidth of bi,k (t) in time slot t, and the total bandwidth for each UAVj is B, so User i for Class k QoS requirements path receives Si,j,k (t) in time slot t. path

path Si,j,k (t)

−ηi,j bi,k (t) TX pi,j,k (t)10 10 = B

(t)

bi,k (t) TX −δ pi,j,k (t)Apath di,j path (t) B

=

(10)



TX (t) represents the transmission power UAVj to Class k where Apath = 10−Apath /10 , pi,j,k QoS required User i. The total noise power of User i for Class k QoS requirements is

Ni,k (t) = 10

−174+ρi (t) 10

bi,k (t)10−3

(11)

The SINR obtained by UAVj is the user i for Class k QoS requirements in the time slot t is path

SINRi,j,k (t) =

Si,j,k (t) Ni,k (t)

(12)

Energy-Efficient Networking for Emergency Communications

227

UAVj and User i for Class A QoS requirements is transmitted over channel bi,a (t). ci,j,a (t) = bi,a (t)log2 (1 + SINRi,j,a (t))

(13)

The throughput of User i for Class k QoS requirements within the continuous time range T is  Tri =

T

 Ii,j,k (t)ci,j,k (t)dt =

t=0

T

t=0

Ii,j,k (t)bi,k (t)log2 (1 + SINRi,j,k (t))dt

(14)

Therefore, the total throughput of all users is  Tr =

M

 Tri =

i=1

M



T

Ii,j,k (t)ci,j,k (t)dt

(15)

i=1 t=0

The optimization model is shown below maxA,BB,P  = di,j (t) =



∫M Tri Tr = i=1 E ∫N j=1 Ej

(xj (t) − xi (t))2 + (yj (t) − yi (t)hj (t)2

(16) (17)

∀j, xj (t) ∈ [xmin , xmax ], yj (t) ∈ ymin , ymax , hj (t) ∈ [hmin , hmax ]

(18)

∀i, j, k Ii,j,k (t) = {0, 1}

(19)

∀i, k, t

N

Ii,j,k (t) ≤ 1

(20)

∀i, k, t SINRi,j,k (t) ≥ σmin

(21)

path

(22)

j=1

∀i, k, t Si,j,k (t) ≥ μmin M T i=1

M T i=1

Ii,j,k (t)bi,k (t) ≤ B, bi,k (t) ≥ 0

(23)

TX TX Ii,j,k (t)pi,j,k ≥0 (t) ≤ Pmax , pi,j,k

(24)

t=1

t=1

∀i, t bi,k (t) log2 (1 + SINRi,j,k (t)) ≥ Rm

(25)

∀j Ej (t) ≥ 0

(26)

228

Z. Li et al.

4 Algorithm Design and Implementation 4.1 Algorithm Design The maximizing optimization of the model is a difficult mixed integer non-convex and non-concave optimization problem, so the problem is solved by two steps: first, cluster the users by clustering algorithm, then employ the air base station deployment algorithm for each cluster. Zoning Algorithm. Taking users’ distribution density into account, density-based clustering algorithm is used to divide the users into clusters. After that the number of air base stations and the coverage of each air base station is determined. Network Coverage Compensation Algorithm. Considering the characteristics of this problem, we propose the energy efficiency 3D position deployment algorithm of the air base station based on the Particle swarm optimization (PSO).Suppose that in Ndimensional space, a group of M particles looks for the optimal position, and each

particle has a position: Xi = xi1 , xi2 , . . . , xiN , i = 1, 2, 3, . . . M , which represents of the UAVj currently explored by the particle. Speed of the i particle: Vi = the1 height vi , vi2 , . . . , viN , i = 1, 2, 3, . . . M . Particles update their positions and speed each time they reach a new location during the iteration. Speed and location updates are as follows:     (27) vid = wxi2 + c1 r1 pid − xid + c2 r2 pgd − xid xid = xid + αvid

(28)

d , vd where i = 1, 2, 3, . . . M , d = 1, 2, 3, . . . N , vid ∈ −vmax max , w is inertial factor, non-negative, which helps find the global optimal solution. Pi is the local optimal solution of particles, Pg is the global optimal solution. c1 and c2 are accelerated nonnegative constants of velocity. r1 , r2 ∈ [0, 1] is a random number. α is a constraint factor used to control speed weights. Solve Function: Define r(T ) as the reward for UAV after the continuous time T.

r(T ) =

Tr E

(29)

Energy-Efficient Networking for Emergency Communications

229

Algorithm 1: The best 3D position planning algorithm for UAV based on PSO Input：clusters and centers Output：all UAVs’ best positions 1. Initialize the location coordinates of all UAVs as cluster center point coordinates 2. for each UAV in UAVs do 3. for each particle i do 4. Initialize each particle's velocity Vi and height Hi 5. Calculate the energy efficiency ratio of particle i at this height 6. pBesti = Hi 7. end for 8. gBest = max{pBesti} 9. while iteration < ITERATION 10. for i in (1 , N) do 11. Update Particle i's velocity Vi and position Hi 12. Calculate the energy efficiency ratio of particle i at this height 13. if reward(Hi) > reward(pBesti) 14. pBesti = Hi 15. if reward(pBesti) > reward(gBesti) 16. gBesti = pBesti 17. end for 18. end while 19. UAV.height = gBest 20. end for

Because particle group algorithm is easy to get caught up in local optimal solution, we add greedy algorithm for comparative verification.

4.2 Simulation Experiments This paper designs the simulation environment area of 300 m × 300 m, such as stadium grandstands, in which the multiple hot spots distributed randomly (see Fig. 1(1)).

Fig. 1. (1) Simulation-generated users map. Blue dots representing the users. (2) User partition diagram. Users are divided into blue, green and yellow.

4.3 Results and Evaluation Result. First, divide the users by DBSCAN (see Fig. 1(2)). For each cell, initialize UAV’s location as clusters centers and find the height of the maximum energy efficiency

230

Z. Li et al.

by Algorithm 1. Figure 2(1) shows how the height and energy efficiency changes during the iteration. Figure 2(2) shows how global optimal solution and local optimal solution change. It shows that the algorithm converges rapidly, and the height value and global optimization tend to stabilize in about 20 generations.

Fig. 2. (1) Iterative progress of heights. (2) The converge progress of global and local results

Evaluation. Figure 3(1) shows the correlation between the energy efficiency ratio change of the three user clusters and their own height under the greedy algorithm. Compared with Fig. 2(1), the particle group algorithm obtains the global optimal solution. Figure 3(2) shows the final 3D deployment results.

Fig. 3. (1) Energy efficiency ratio and height of three user cluster air base stations in the greedy algorithm. (2) The green, purple, and yellow dots at the bottom represent the users and the blue triangle at the top represents the air base station for each partition.

Acknowledgement. This work is supported by Science and Technology Project from State Grid Information and Telecommunication Branch of China: Research on Key Technologies of Power Emergency Communication System integrating LEO satellite, BeiDou Navigation Satellite System and 5G (SGXT0000XGJS2100236).

Energy-Efficient Networking for Emergency Communications

231

References 1. Li, B., Fei, Z., Zhang, Y.: UAV communications for 5G and beyond: recent advances and future trends. IEEE Internet Things J. 6(2), 2241–2263 (2019) 2. Kalantari, E., Shakir, M.Z., Yanikomeroglu, H., Yongacoglu, A.: Backhaul-aware robust 3D drone placement in 5G+ wireless networks. In: 2017 IEEE International Conference on Communications Workshops (ICC Workshops), pp. 109–114. IEEE, Japan (2017) 3. Liu, J., Sheng, M., Lyu, R., et al.: Performance analysis and optimization of UAV integrated terrestrial cellular network. IEEE Internet Things J. 6(2), 1841–1855 (2019) 4. Li, W.: Management Mechanism of Cell Outage Compensation in LTE Radio Access Network. Beijing University of Posts and Telecommunications, Beijing (2016) 5. Zhang, T.: A Cell Outage Compensation Method Based on Aerial Base Station. Beijing University of Posts and Telecommunications, Beijing (2018) 6. Yu, P., Zhou, F., Zhang, X., Qiu, X., Kadoch, M., Cheriet, M.: Deep learning-based resource allocation for 5G broadband TV service. IEEE Trans. Broadcast. 66(4), 800–813 (2020) 7. Zhong, X., Huo, Y., Dong, X., Liang, Z.: QoS-compliant 3-D deployment optimization strategy for UAV base stations. IEEE Syst. J. 15(2), 1795–1803 (2021) 8. Kimura, T., Ogura, M.: Distributed collaborative 3D-deployment of UAV base stations for ondemand coverage. In: IEEE Conference on Computer, IEEE INFOCOM 2020, pp. 1748–1757. IEEE Press, virtual (2020) 9. Yu, P., Yang, M., Xiong, A., Ding, Y., Li, W., et al.: Intelligent-driven green resource allocation for industrial internet of things in 5G heterogeneous networks. IEEE Trans. Industr. Inf. 18(1), 520–530 (2022) 10. Wang, Z., Liu, R., Liu, Q., Thompson, J.S., Kadoch, M.: Energy-efficient data collection and device positioning in UAV-assisted IoT. IEEE Internet Things J. 7(2), 1122–1139 (2020) 11. Chen, N., Rong, B., Zhang, X., Kadoch, M.: Scalable and flexible massive MIMO precoding for 5G H-CRAN. IEEE Wirel. Commun. Mag. 24(1), 46–52 (2017) 12. Li, X.: Research on IoT energy efficiency optimization strategies based on drones. Wuhan University of Technology, Wuhan (2020) 13. Wu, J.: Research and Implementation of Location Planning Method of Aerial Base Station for Capacity Enhancement. Beijing University of Posts and Telecommunications, Beijing (2020)

Space-Air-Ground Integrated Network Driven by 6G Technology Pingping Lin1(B) , Lei Liu1 , Xi Meng2 , and Michel Kadoch3 1 China Telecom Research Institute, Beijing, China

[email protected]

2 Beijing University of Posts and Telecommunications, Beijing, China 3 Ecole de Technologie Superieure, Montreal, QC, Canada

Abstract. Still, the Space-Air-Ground Integrated Network (SAGIN) is not mature yet, in its planning phases, making it unable to prevail in reality despite many important discoveries in improving ground, air, and satellite systems. SAGIN needs higher agility, flexibility, robustness, and scalability than 5G could provide, and thus it’s time to concentrate on the next generation of sophisticated 6G technologies to address the SAGIN ecosystem’s current issues. This paper gives a clear road map of how 6G might enhance the existing SAGIN infrastructure and boost certain value-added services. Keywords: SAGIN · 6G · Emerging space communication

1 Introduction As the number of smart devices [1] increases and network traffic climbs rapidly, a large and worldwide link has arisen as a critical technological need. Furthermore, in today’s world, a diverse variety of applications with varying needs must be handled. Smart transportation, long-distance connection, marine surveillance, interplanetary communication, smart cities, and disaster assistance are just a few of the potential applications for advanced networking services. Terrestrial networks alone will be unable to meet these enormous service-aware demands and provide optimum traffic solutions. As a result, it is expected that terrestrial networks, as well as space and aerial communication infrastructures, would collaborate to provide network services that reduce delays. With unmanned aerial vehicles (UAVs) or drones, we can see how low- and medium-range satellite constellations could be reinforced whenever three-dimensional (3D) network integration is used to create additional spectacular possibilities anywhere three-dimensional (3D) network integration is utilized. In this solution, to achieve high network data rates, low latency and high reliability, the Space-Air-Ground Integrated Network (SAGIN) is able to provide full-featured ubiquitous communication, computing and caching capabilities [2, 3]. The bulk of these apps is related to national security and catastrophe preparedness. To expand the scope of sensor and actuator operations at the micro-level, technology enablers such as the internet of things (IoT) may readily be considered. Edge and cloud aggregation have © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 S. Sun et al. (Eds.): ICSINC 2021, LNEE 895, pp. 232–242, 2022. https://doi.org/10.1007/978-981-19-4775-9_29

Space-Air-Ground Integrated Network Driven by 6G Technology

233

introduced a new level of delay mitigation that may be enhanced by unmanned aerial vehicles (UAVs). As IoT is incorporated into smart cities and manufacturing, UAVs are becoming more important. To further expand and improve network coverage and customer equipment maintenance, UAVs must be integrated with the rest of the network components. With better speed, capacity, and service-aware activity entitlement, we expect a new wave of ground-breaking and remarkable networking technologies. Rising demand for SAGIN-based services means 5G will be unable to keep up with demand despite 5G appears to be superior to existing cellular technology in many respects. We must consider a multi-dimensional SAGIN-aware design method to prevent network intrusion at the ultra-low edge and fully cognitive levels. 6G is being viewed as the ultimate solution to SAGIN-based user service orientation [4, 5]. In the purpose of tackling existing challenges and promoting ubiquitous and high-capacity global connectivity, at present, 6G research is focusing on the construction of non-terrestrial networks (NTNs) [6–8]. 6G envisions a three-dimensional (3D) heterogeneous architecture in which non-terrestrial stations such as UAVs, HAPs, and satellites complement terrestrial infrastructures, which is distinctive from previous generations of wireless networks aiming at providing connectivity for a quasi-bidimensional space [9–12]. These components can provide not only cost-effective on-demand coverage in congested and unserved areas, but also trunking, backhauling, high-speed mobility support, and high-throughput hybrid multi-play services. The importance of NTNs is realized in routine operations especially. In this paper, we show how 6G’s multi-dimensional capabilities may significantly improve SAGIN’s on-demand value addition infrastructure [13, 14]. Ubiquitous 3D coverage (space, air, ground, and underwater networks) will provide space for the construction of 6G technology. Cognitive radio and real-time intelligent edge are improved by the SAGIN framework’s intelligent connection due to its intrinsic distributed AI. Furthermore, better stratification would assist SAGIN’s integration with 6G while covering a range of network points of view. It is perfect for the SAGIN scenario to fit the new types of dynamic spectrum utilization and content-driven routing approaches. Viability of FL through particular implementations. Finally, the article outlines the difficulties encountered while applying federated learning to the practical scenario.

2 Emerging Space Communication Technologies in 6G-SAGIN 2.1 Space Access Network Satellite monitoring and relaying, as well as spatial data transmission devices, are all handled by the space access network (SAN). In 6G-SAGIN, the SAN may be utilized to establish communication amongst different communicative components. In most instances, geostationary satellites are used to provide network connection as well as Ku band coverage. 506 megabits per second are the data transmission rate. It has the potential to cover an area of 100–6000 km2 . The modem and dish antennas may be used both inside and outside., could be used in the same way. Focusing on using medium and low earth orbit satellites, recent initiatives improve the connectivity. Satellites in low and

234

P. Lin et al.

medium earth orbit will offer 64 Kbps and 1 Gbps separately. The words “satellite internet access” (SIA) and “satellite internet access” (SAN) refer to satellite communication that is exclusively two-way, are often interchanged. Portable satellite modems and internet connections through satellite phones have grown increasingly popular in recent years. In the broadcast model, terrestrial transmit solutions with one-way receivers, on the other hand, are used with minimum modifications. At any given time, these technologies have the ability to empower between 100 and 4000 end users. With the proper, multiple access mechanisms for software and time division, this range may be extended. All issues include signal latency, geostationary orbit placement, atmospheric interference, fresnel zone, and line of sight that this technology has to deal with. Using SAN and SIA together, however, is expected to benefit 6G-SAGIN. 2.2 Satellite Communication Satellite communication may easily be included in a 6G network. Satellite communication, with its vast range of uses and incredibly accurate communication capabilities, would become an indispensable component of 6G. In digital communications, low earth orbit (LEO), medium earth orbit (MEO), and geostationary earth orbit (GEO) satellites are used (GEO). LEO satellites orbit the earth at a height of 160 to 2000 km above the surface. The MEO constellation’s satellites orbit between 2000 and 36,000 km above the Earth’s surface. Geostationary orbiting satellites (GEO) orbit the Earth at a constant height of 22236 km above the surface. Other kinds of satellite communication systems are utilized for a wide range of purposes. Fixed satellite service (FSS) is used to send feeds to television stations, radio stations, and broadcast networks, and other kinds of satellite communication systems are utilized for several purposes. In most instances, the FSS requires Earth to space services. According to our results, both radio navigation and space research utilize the 14–14.25 GHz and 14.25–14.3 GHz frequencies. Using 12–18 GHz and a 0.5-m-radius dish antenna, TV channels may be transmitted. Digital video broadcasting (DVB) technologies, such as DVB-T, S, C, and H, play an important part in data transfer via set-top boxes. A variety of frequency bands are utilized for space-to-earth (S2E) communication, including 137–136.025 MHz, 137.025–137.175/137.825 MHz, 148–149.9/150.05 MHz, and 156.7625–156.8375 MHz. Besides, earth-to-earth (E2E) communication frequencies are 156.7625–156.8375 MHz. As a consequence, satellites may be used for mobile data services, television broadcasting, radio broadcasting, military applications, as well as broadband internet services in the 6G network. To directly receive radio service from satellites, the Satellite RAN (SARN) technique may be employed. Table 4 shows a variety of space communication scenarios. 2.3 Laser Space Communication A well-established method of networking is the employment of a laser to communicate in free space. Using optical telescopes as a bridge between the terrestrial and space components may help in long-range communication. Several research projects are ongoing to deploy laser space communication in a range of situations. The most common types of laser communication are from Low earth orbit, medium earth orbit, and geostationary

Space-Air-Ground Integrated Network Driven by 6G Technology

235

earth orbit are the three types of earth orbit. Techniques centered on the stratosphere and troposphere, on the other hand, are being explored. Typical situations include space-toground (S2G), space-to-space (S2S), air-to-air (A2A), and air-to-ground (A2G). When transmitting a signal, an average data rate of 1 Gbps is obtained. However, it may vary and dip as low as 0.15 Gbps, with a top speed of 10 Gbps. The following are the objectives of such implementations: (a) creation of a global telecommunications backbone, (b) megaconstellation of satellites, (c) data relay network design with mixed radio-frequency capabilities, (d) coverage of rural and remote regions, and (e) in-flight communication. 2.4 Deep Space Optical Communication Another type of space communication, deep space optical communication (DSOC), might be a big help for the 6G-SAGIN. Communication speeds are enhanced by a factor of a hundred due to the DSOC. For downlink operations, DSOC is expected to need a significant amount of bandwidth. The DSOC, created by NASA’s JPL lab, is expected to conduct optical communications in the near-infrared (NIR), with a wavelength of 1.55 mm. In 2022, NASA’s Psyche mission will utilize the DSOC technique, as scheduled. A flying laser transmitter and a base station will be the two components of the system. A telescope with a 22 cm aperture would be used to focus a 4 W laser in the satellite laser unit. The ground station’s uplink and downlink operations must be coordinated with a total transmitter module power of 100 W. In uplink mode, the ground station would utilize 1.064 mm. The average data rate will be 292 kilobits per second, according to projections. We think that 6G-SAGIN will improve in the next few days with the assistance of the DSOC. 2.5 Optical PAyload for Space Communication To keep an active optical connection between space and ground stations, Optical Payload for Space Communication (OPASC) was created. However, in the 6G-SAGIN situation, we believe it is appropriate for future use case development. In space communication, OPASC is able to interpreting distorted signals. It’s useful for creating procedural parts using the optical connections supplied. The typical data rate is about 50 megabits per second. The OPASC would be used by 6GSAGIN to help with mission operations and application use cases. Using tracking data and relay services technology, the OPASC would begin coordinating between the satellite and ground station (TDRSS). To give a superior downlink, 6GSAGIN may combine approaches based on the near-earth network (NEN) or the ground network (GN).

3 Main Pillars of 6G-SAGIN 3.1 Physical Layer Innovation To make the system feasible, the physical layer of the 6G-SAGIN to be built must meet certain requirements. To begin, to apply different frequency bands to different applications, the frequency band should be used optimally. Next, the Ku and Ka Q/V frequency

236

P. Lin et al.

bands will be effectively deployed. The second stage is to simulate the propagation channel: 6G-SAGIN requires a communication channel with a high degree of interruption mitigation and a low bit error rate. When building a channel that can enhance system capacity, propagation delay is another factor that needs to be considered. Finally, since 6G-SAGIN will operate simultaneously at the physical, data connection, and network levels, there will be no need to monitor latency, throughput, reliability, or energy efficiency. Multi-carrier modulation, antenna design, network coding, routing loss, and channel status evaluation are all essential considerations. Finally, spectrum allocation may be done in a manner that avoids issues like Bayesian equilibrium. In this case, the use of multi-layer drone deployment and sub-bands may be an important solution (Fig. 1).

Fig. 1. HOWF: a 6G-SAGIN optical network architecture.

3.2 Mobile Crowd Sensing Traditional drones and SAGIN-centric technologies still have some problems, such as network densification, cache capacity, and edge computing capabilities. As a consequence, an improved mobile crowdsensing method to prevent 6G-SAGIN from edge caching problems may be deployed. As a consequence, to address the above-stated

Space-Air-Ground Integrated Network Driven by 6G Technology

237

issues, a better mobile edge network (MEN) must be created. As an alternative, a multiaccess radio access network with UAV-based cellular architecture might be developed. In this manner, UAVs may serve as base stations for other UAVs or ground/satellite stations, delivering HetNet-based services. The mmWave and beamforming strategies outlined here may aid SAGIN’s and caching capabilities in 6G computing. Four interconnected pillars include a cross-layer interactive manager, crowdsensing module, publisher/subscriber module, and decision-making tool. Use convolutional deep neural networks as a decision-maker. To keep up with approved flow charts, satellites, UAVs, and ground stations may be compelled to use software-defined networks (SDN). Those flowcharts should be utilized anytime there is a management problem. The 6G-SAGIN cooperation gateway would service all kinds of UAV-based cells, backhaul interfaces, and ad-hoc interfaces after obtaining an accurate judgment from the underlying caching engine. Problems such as cognitive behavior, prediction-aware facilitation capacity, and communication connection optimization, on the other hand, should all be taken into account in this strategy. 6G-SAGIN routing difficulties may be alleviated by improved 6G-SAGIN centric scheduling and predictive modeling. 3.3 Intelligent Offloading A networking system’s agility, for example, allows it to operate effectively and share a vast amount of spectrum. 6GSAGIN infrastructure also aims for the timely fulfillment of such criteria. Task offloading in a fully bi-directional way may be envisioned with the joint participation of VNF and SFC. When it comes to UAVs, the SAGIN network can now accommodate high-altitude platforms thanks to this type of offloading (HAPs). The options for 6G-SAGIN provides mission offloading services are the followings. To figure out how services are offloaded from orbit to the ground station, a satellite cooperative sensing mission may be utilized. HAPs, for example, may transfer raw images acquired by satellites to mobile and static ground components. MEC-enabled 6G-SAGIN user equipment which could subsequently send the analytical results to the satellite can carry out the Image processing,. Supplementing a HAP-based relaying option improves the offloading service. Through virtualization and data flow embodiment, all three tiers of the 6G-SAGIN architecture may be combined by SDN controllers to share physical resources. Assign groundwork to the aerial network, in the meanwhile, which still having access to a 3D network architecture. In the 6G-SAGIN, we may expect improved capacity, coverage, mobility, wireless backhauling, and resilience. Space-air duties may also be delegated to the ground station with a more sustainable design and a more intelligent layer. The 6G-SAGIN shall examine reconfiguration of agility-aware resource scheduling to abstract and virtualize heterogeneous resource sharing. 3.4 Super IoT The Internet of Things (IoT) is a horizontal layer allowing disparate components to communicate with each other. These Internet of Things systems are extensively used in several areas. On the 6G-SAGIN, the super IoT version of implication is expected to be installed. The addition of asset tracking, monitoring, and control capabilities to the current IoT-based scenario is referred to as “super IoT.” The super IoT idea may be

238

P. Lin et al.

used with 6G-SAGIN to improve asset monitoring, both fixed and mobile. Super IoT would also address 6G-SAGIN problems like security, service platforms, and reliance on a centralized support infrastructure. Because objects are such an important part of IoT, super IoT would broaden its reach to include human scanning. Industry 4.0 would become highly important if super IoT and 6G-SAGIN platforms were included. It will be utilized for energy monitoring, sales data generation, logistic operation management, and service leveraging, among other things. The implementation of super IoT in the SAGIN vision might dramatically improve the security of people and their systems in an emergency. 3.5 Stringent Authentication Strict authentication will be a key enabler for success in the 6G-SAGIN era. Data manipulation, identity fraud, interception, and other threats must all be protected against. This problem may be solved using hash-based chains. In this architecture, security drones will connect 6G-SAGIN specific user components including smart cities, smartphones, smart industries, and people in a localized way. The hash-based authentication framework will oversee the whole drone-based communication infrastructure. The dynamic handshake communication method is extensively used in this authentication. It works like this: the drone-assisted security manager (SM) confirms border crossing operations,

Fig. 2. Hash-based chain authentication architecture.

Space-Air-Ground Integrated Network Driven by 6G Technology

239

allowing for the creation of handshaking. With the assistance of the SMs, all 6G-SAGINserved communicative devices begin to establish bi-directional signaling. In this case, the selected SM gathers transactions and creates hash chain blocks. To support the SMs and encrypt and decode the material IDs, the hash chain infrastructure is used. To support the SMs and encrypt and decode the material IDs, the hash chain infrastructure is used. The Markle tree in the inferred hash chain would not only offer authentication, but also non-repudiation and secrecy. Figure 2 depicts the hash-based authentication architecture. 3.6 Service Function Chaining In the 6Gage, it is problematic to coordinate with object, especially in the SAGIN architecture, which installs large-scale infrastructure. As a result, merging virtual network functions (VNF) with service data routing might be an attractive option. Normally, although service function chaining (SFC) is an NP-hard problem, a heuristic greedy strategy will solve it. The aggregation ratio (AR) is the approach of assessing the computationcommunication trade-off in addition to the SFC. The difference between total VNS and necessary VNFs divided by total VNFs is defined as AR. In the 6G-SAGIN, SFC can eliminate resource blockage, resulting in improved near-optimal performance. It is possible to integrate SFC with 6G-SAGIN in a hybrid architecture. With the aid of VNFs, all three levels of SAGIN may be virtualized, allowing for the implementation of virtual

Fig. 3. Service function chaining in 6G-SAGIN.

240

P. Lin et al.

link mapping and VNFs embedding ideas. SFC is capable of appropriately routing data flow and identifying where VNFs should be placed in the 6G-SAGIN physical network. By raising the AR ratio, SFC can handle multiple junction 6G-SAGIN traffic scheduling problems. The SFC architecture of 6G-SAGIN is shown in Fig. 3. 3.7 Ultra-Dense Cell-Free Massive MIMO The ultra-dense cell-free massive MIMO design is helpful to 6G-SAGIN considerably. Its goal is to combine large-scale MIMO and cloud-based RAN technology. To solve the current cellular-edge-focused problem, our cell-free method leverages macro-level diversity. It also means that channel hardening is kept to a minimum. The advantages of the cell-free network design are MEC readiness, very low latency, energy efficiency, ultra-reliability, extremely high capacity, and low-power high data rate. In 6G-SAGIN, UAVs may use cell-free networking to take advantage of sophisticated central processing units on-the-fly service mitigation (CPU). In the SAGIN era, a colossal array of antennas will encircle the user components. The CPU is connected to a large number of dispersed antennas to ensure uplink and downlink coherence because the network lacks cells. Using the cloud-RAN connection, over several access points or antenna nodes (AP), processing may be distributed.

4 Future Research Direction 4.1 Under Water Communication 6G will revolutionize all aspects of connection, particularly subsea or conversation when submerged. 6G-SAGIN may need real-time communication with the subsea network. As a result, very low frequency (VLF) signals in the 3–30 kHz band should be examined, as they have the potential to penetrate up to 20 m of saltwater. For VLF signaling, we should expand the transmission range of current submarines. The transmitting antenna should be big enough to take up underwater broadcast VLF frequencies. Ground-to-submarine (G2SM) channels should be one-way broadcast, but a bidirectional link can be formed by rising to periscopic depth. Only text messages and minimal music may be transmitted at a low data rate since VLF can only identify 300 bits per second. 6G would allow for a greater coverage area to be studied, and it is expected that in the right place, it may be utilized in conjunction with VLF. Extremely low frequency (ELF) of 3–300 Hz is used in military applications to penetrate hundreds of meters of saltwater. We can establish a viable link with the submarines utilizing the UAVs by employing a dipole antenna array. At frequencies as low as 60–900 Hz, the NATO-based Ramon god of gateways can connect with underwater modems. The 6G-SAGIN, on the other hand, maybe used with acoustics or a mix of radio and acoustics. 4.2 Dew Computing Dew computing is a new computing paradigm. The goal of this computing paradigm is to offer full user experience-focused edge services via new dew services. Since it

Space-Air-Ground Integrated Network Driven by 6G Technology

241

substantially enhances user experience dew computing may be considered a viable option for implementing 6G-SAGIN. Using a single super-hybrid, between the Dew device and the cloud, the Dew-cloud architecture spread peer-to-peer communication. With 6G, direct device-to-device communication is possible due to a complete technique of plugin services that includes computational domain and synchronization activities. A dew server may be installed within the onboard system of UAVs and satellites to act as a dew client. As a consequence, additional local computers and associated tools may be combined in the shape of a dew cluster with the dew client. Such clusters may give real-time data in a cached format, as requested by the dew client. This approach contains a privacy-preserving component, since it may utilize dew client apps to work on the top of dew database management systems. Dew clients may use such systems to execute customized dew scripts, such as network table storage and auto-update. Overall, a dew client may be able to deliver its customers with a full internet data services experience with little or no real-time backhaul network connection disturbance. As a consequence, there may be less dependence on 6G backhaul, and a more traffic- and privacy-conscious SAGIN ecosystem may emerge. 4.3 Cross-Layer Communication At various levels of abstraction, 6G-SAGIN is an expected function. As a consequence, the SAGIN ecosystem’s sustainability depends on a cross-layer communication mechanism. For addressing cross-layer communication problems, 6G-SAGIN should explore a variety of options. In the 6G, for example, the SAGIN infrastructure should be appropriately placed to monitor fast fading channels. Downlink synchronization is another aspect of cross-layer optimization that must be followed. To use cross-layer augmentation in the 6G-SAGIN, channel estimation and time advancement are required. The size of the HARQ buffer should be constantly adjusted depending on the channel representation need. New methods for updating satellite location and UAV, as well as user equipment handover rules, should be investigated when dealing with trepidation across layers. With improved QoS mode, cross-layer facilitation will be improved by more flexibility and energy efficiency. UAVs’ on-board protection and satellites may be investigated to enhance cross-layer interaction.

5 Conclusion This work paves the way for a SAGIN that can support 6G while remaining transparent. The requirements and criteria for transforming SAGIN’s current situation into a more future 6G-SAGIN were imagined. We discussed how UAVaaS and 6G spatial communications might help SAGIN’s infrastructure. After resolving many remaining research issues, we expect that 6G-SAGIN may be completely realized. The goal of this paper is to provide new perspectives on SAGIN so that future paths for next-generation network technologies may be proposed. With the assistance of 6G technology, SAGIN, we found, may become a real part of our society. We discussed the limitations and main applications of such potential use cases. SAGIN’s future technological advances will provide a solid basis for improving humanity’s complete social components.

242

P. Lin et al.

References 1. Mamdouh, M., Elrukhsi, M.A.I., Khattab, A.: Securing the internet of things and wireless sensor networks via machine learning: a survey. In: 2018 International Conference on Computer and Applications (ICCA), pp. 215–218 (2018) 2. Bhat, J.R., Alqahtani, S.A.: 6G ecosystem: current status and future perspective. IEEE Access 9, 43134–43167 (2021) 3. Khelifi, H., et al.: Named data networking in vehicular ad hoc networks: state-of-the-art and challenges. IEEE Commun. Surv. Tutor. 22(1), 320–351 (2020). First quarter 4. De Lima, C., et al.: Convergent communication, sensing and localization in 6G systems: an overview of technologies, opportunities, and challenges. IEEE Access 9, 26902–26925 (2021) 5. Chen, S., Sun, S., Kang, S.: System integration of terrestrial mobile communication and satellite communication—the trends, challenges and key technologies in B5G and 6G. China Commun. 17(12), 156–171 (2020) 6. Gao, H., Xiao, Y., Yan, H., Tian, Y., Wang, D., Wang, W.: A learning-based credible participant recruitment strategy for mobile crowd sensing. IEEE Internet Things J. 7(6), 5302–5314 (2020) 7. Zhang, J., et al.: A secure decentralized spatial crowdsourcing scheme for 6G-enabled network in box. IEEE Trans. Ind. Inform. 18, 6160–6170 (2021) 8. Emu, M., Choudhury, S.: Towards 6G networks: ensemble deep learning empowered VNF deployment for IoT services. In: 2021 IEEE 18th Annual Consumer Communications & Networking Conference (CCNC), pp. 1–4 (2021) 9. Fang, X., Feng, W., Wei, T., Chen, Y., Ge, N., Wang, C.-X.: 5G embraces satellites for 6G ubiquitous IoT: basic models for integrated satellite terrestrial networks. IEEE Internet Things J. 8(18), 14399–14417 (2021) 10. Wang, H., Zhang, P., Li, J., You, X.: Radio propagation and wireless coverage of LSAA-based 5G millimeter-wave mobile communication systems. China Commun. 16(5), 1–18 (2019) 11. Dong, W., Xu, Z., Li, X., Xiao, S.: Low-cost subarrayed sensor array design strategy for IoT and future 6G applications. IEEE Internet Things J. 7(6), 4816–4826 (2020) 12. Li, J., Yin, H., Zhang, X.: Satellite communication used in the non-geostationary Fixedsatellite service systems and Earth-exploration satellite service (passive) systems. In: 2019 28th Wireless and Optical Communications Conference (WOCC), pp. 1–5 (2019) 13. Chen, E., Tao, M.: User-centric base station clustering and sparse beamforming for cacheenabled cloud RAN. In: 2015 IEEE/CIC International Conference on Communications in China (ICCC), pp. 1–6 (2015) 14. Li, O., et al.: Integrated sensing and communication in 6G a prototype of high-resolution THz sensing on portable device. In: 2021 Joint European Conference on Networks and Communications & 6G Summit (EuCNC/6G Summit), pp. 544–549 (2021)

Federated Learning Based 6G NTN Dynamic Spectrum Access Pingping Lin1(B) , Xi Meng2 , Lei Liu1 , and Michel Kadoch3 1 China Telecom Research Institute, Beijing, China

[email protected]

2 Beijing University of Posts and Telecommunications, Beijing, China 3 Ecole de Technologie Superieure, Montreal, QC, Canada

Abstract. With the advent of 6G, dynamic spectrum sharing (DSS) has become a critical technology to Non-Terrestrial Networks (NTNs). In 6G DSS, machine learning (ML) will provide a better understanding of network dynamics, as conventional DSA techniques are ineffective in the scenarios where wireless network analysis is lacking. This work presents a distributed machine learning framework for the DSA problem by using federated learning (FL) methods. FL technology can protect user endpoint privacy throughout data dissemination in an open cooperative computing environment. We investigate the benefits, difficulties, as well as the feasibility of federated learning for future 6G applications. Keywords: 6G NTNs · Federated Learning (FL) · Dynamic Spectrum Access (DSA) · Dynamic Spectrum Sharing (DSS)

1 Introduction Space-Air-Ground Integrated Network (SAGIN) is projected to provide fully function of ubiquitous communication, computing, and caching capabilities, and enable high network data rates, low latency, and high dependability [1]. The bulk of these apps are related to national security and catastrophe preparedness. To expand the scope of sensor and actuator operations at the micro level, technology enablers such as the internet of things (IoT) may readily be considered [2]. In recent years, the range of cellular mobile networks has considerably increased as a result of the widespread deployment of 5G [3]. Studies showed that 5G mobile networks have brought about a significant reduction in mobile network latency, which allow for the use of amounts of wireless equipment, sensors, and other devices, supporting the development of nascent scenarios such as driverless cars, augmented reality, and other new technologies, and thus contributing to the widespread growth of the Internet of Things [3]. Although 5G seems to be better than other existing cellular technology in many ways, increasing expectations for SAGIN-based services make it not able to keep up with demand [4]. In order to prevent ultra-low edge and fully cognitive network intrusion, a multi-dimensional SAGIN-aware design method must be considered. 6G is being © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 S. Sun et al. (Eds.): ICSINC 2021, LNEE 895, pp. 243–250, 2022. https://doi.org/10.1007/978-981-19-4775-9_30

244

P. Lin et al.

viewed as an ultimate solution to SAGIN-based user service orientation. In the purpose of promoting ubiquitous and high-capacity global connection, and tackling existing problems well, 6G research is concentrating on the development of non-terrestrial networks (NTNs) presently. To further improve the spectrum efficiency, the Third Generation Partnership Project (3GPP) has been working on the standards that introduce dynamic spectrum sharing DSS to 5G New Radio (NR) services. In contrast to traditional spectrum management policies that allocate spectrum to operators’ exclusive possession, the existing DSS strategy is effective in terms of improving spectrum efficiency in 5G and beyond 5G networks, because hierarchical spectrums access structure with secondary users (SUs) is employed, or unlicensed users who access spectrum resources without causing significant interference [5–7]. Owing to the 3D feature of Space-Air-Ground Integrated Network, DSS will play an even more pivotal role in 6G NTN. Machine learning (ML) based methods will be introduced to solve the DSS problem and improve the performance. The benefits of centralized and distributed learning are combined in federated learning (FL) methods. Federated learning decentralizes model training and keeps data where it was generated, avoiding the need for unrefined data to be transmitted. FL was mentioned in the 3GPP technical report (TR 22.874) as a possible option for supporting ML model transfer and dissemination in 5G networks. FL is mentioned in 3GPP Technical Report (TR) 22.874 as a possible option for supporting ML model transmission and dissemination in 5G networks. Federated learning in 6G disseminates the model through the Broadcast Channel (BC) and collects model updates via Multiple Access Channels (MAC). Due to restricted uplink bandwidth and model updates that vary from users, the aggregation process is typically the bottleneck, making the process sluggish and costly. To solve this problem, quantized information transfer, partial device involvement, and periodic aggregation are frequently utilized to improve communication efficiency. 1) Model Compression: Even though FL simply gathers gradient parameters to reduce communication overhead, a wireless network with billions of devices accumulates communication overhead. As a result, rather of uploading full-accuracy model parameters, low-accuracy quantization operators may be used to reduce local model updates. Communication overhead is reduced overall by sending quantized updates. 2) Partial Device Participation: Because local devices and networks have heterogeneous data sizes, channel conditions may vary significantly depending on device location, transmit power, and noise level. It is not feasible to include all gadgets throughout the whole training procedure. Because of the large number of participating devices, communication is costly and time-consuming. Furthermore, due to the Straggler’s impact, devices may be unable to react (inactive). 3) Periodic Aggregation: Aggregation in each training iteration, like conventional distributed learning, results in substantial communication overhead. To minimize communication overhead, once the server distributes a model, participating devices execute a sequence of local iterations and periodically synchronize with the server in FL. The update direction may be different from the global gradient direction, particularly for non-i.i.d settings, because the ways of all devices doing local updates is unsynchronized. It is worth mentioning that the above-mentioned strategies for reducing communication are intricately linked. The payload of the model parameters determines the number

Federated Learning Based 6G NTN Dynamic Spectrum Access

245

of devices that may be active at a given MAC channel capacity. In the meanwhile, when the MAC capacity is constant, between the payload of the model parameters and the aggregation frequency, there is a clear trade-off. From this article, a federated learning is presented based method to deal with the spectrum allocation issue in DSS systems. We will go through the key characteristics of joint learning that set it apart from other machine learning frameworks and assess the viability of FL through particular implementations. Finally, the article outlines the difficulties encountered while applying federated learning to the practical scenario.

2 Basics of DSS and FL 2.1 Dynamic Spectrum Sharing Two kinds of spectrum management techniques are commonly utilized to effectively use the increasingly limited spectrum resources: static and dynamic [8–12]. Spectrum resources are dynamically distributed to licensed service provider SPs and unlicensed SPs in static spectrum management; this method makes use of radio resources while alleviating spectrum shortages without providing additional spectrum resources for unlicensed SPs. An opportunistic DSS approach enables unlicensed users to access radio resources while licensed users are not using them. Primary users (PUs) and secondary users (SUs) are the terms for licensed and unlicensed users, respectively. As a result, the tagged base stations may identify the distribution of the closest spectrum holes, which is suited to the appropriate access policy. A good example for DSS in practice is the Citizen’s Broadband Radio Service (CBRS) System.

Fig. 1. Dynamic spectrum sharing in 6G

246

P. Lin et al.

As shown in Fig. 1, in 6G there are new chances for spectrum sharing due to emerging technologies such as cognitive radio, geolocation databases and the usage of higher frequencies. New users must have equal access to spectrum while existing users are protected. It is necessary to evaluate coexistence using computer simulations and measurements based on the best available information. As a result, AI approaches like as machine learning, deep learning, and federated learning can enhance spectrum sharing efficiency dramatically [13]. 2.2 Federated Learning in Wireless Communications and Networking In wireless communications, machine learning algorithms are used to more effectively distribute spectrum resources. For example, deep reinforcement learning (DRL) can help rationalize the allocation of dynamic spectrum management. The DRL agent, in particular, adopts a behavior based on observations of the environment state, calculates a corresponding evaluation, and switches to a new state. The DRL agent’s objective is to develop a policy that maximizes the total reward. When there are many agents in a Multi-Agent Reinforcement Learning (MARL) system, this becomes an optimization issue for all of the agents’ policies [14]. This is due to the fact that each agent’s particular rewards and system states are unique.

Agent 1 Action

Observation

Environment

Local state 1

Local state 2 ĂĂ

Local Reward

Local Reward

Private Data Policy

Server

Agent 2

Action Private Data Policy

Local state N Action

ĂĂ

Policy Federated Learning

Agent N Local Reward

Private Data Policy

Fig. 2. Multi-agent reinforcement learning in federated learning framework

As shown in Fig. 2, federated learn can provide a perfect computing platform for MARL. The main reasons are as follows.

Federated Learning Based 6G NTN Dynamic Spectrum Access

247

Limitations of Traditional Techniques: In traditional machine learning methods, users submit raw data to a centralized server for training in order to build a model for future interference. The raw data includes private information because to the large amount of data that must be utilized in the training process. Distributed learning, on the other hand, simply needs the cloud to deliver a model to users and train the model using data from the users’ local area. There is no need for additional contact. Due to restricted data sharing and collaboration, however, in distributed learning, all users improve their models without considering the variance of the environment. Advantage of FL: FL is a potential option for addressing the aforementioned issues by moving from a centralized to a more realistic decentralized learning paradigm. FL creates a global representation by combining distributed training models from local devices. Each end user, in particular, effectively trains its local model from a small dataset and transmits model changes to a server administrator. The server creates a global model based on the data gathered and distributes it to the users. A sufficient number of communications turns can ensure the correctness of the global model. Table 1 lists the benefits and drawbacks of the conventional ML framework and FL. 3) Analysis of Optimization: FL optimization relies on Stochastic Gradient Descent (SGD), a method of training deep networks for statistical problems that has been proven to be successful. For an unbiased estimation of the full gradient, independent and identically distributed (i.i.d) training data sampling is required. Indeed, assuming that each device’s observation is i.i.d. is unrealistic. Model quantization, partial device involvement, and periodic aggregation must also be considered in order to obtain a high degree of communication reduction.

3 FL Based 6G Dynamic Spectrum Sharing Framework System Framework: Figure 3 shows how FL is deployed in a 6G DSS framework. The signal-to-Interference-plus-Noise Ratio (SINR) characterizes all aspects of the SUs network as a quality measurement, including background noise, transmission power, and interference between concurrent transmission pairs. Furthermore, all the FL agents work together to make the spectrum assignment decision. Spectrum Access Policy: To optimize the joint reward in a DSS system, we adopt a decentralized policy gradient method in each FL agent. A policy that has been initialized is distributed to all agents initially. Each agent’s neural network is updated based on its own gradient. The agent empties its buffer, observes the environment, takes an action based on its policy, and receives a reward from the environment at the end of each communication cycle. Each agent acquires an updated local policy network after adequately repeating the above-mentioned actions. The global policy network is then renewed by aggregating the updated local policy networks at a central server. The policy network in our system is implemented in the output layer by a Recurrent Neural Network (RNN) with a softmax activation function.

248

P. Lin et al.

Fig. 3. Dynamic spectrum sharing framework for 6G

User Determination: User selection is critical for achieving an efficient wire-free communication system by accelerating convergence and counteracting non-i.i.d data bias. To this purpose, we are considering using a separate deep Q-learning system to conduct user selection. The agent learns to select devices in each communication cycle to reduce communication overhead while minimizing system validation error. In the proposed system, only to communicate model weights. As a result, any data samples don’t need to be collected or authenticated by the agent from mobile devices, and the agent allows FL to maintain the same level of anonymity. There is an implicit link between the data distribution on a device and the local model weights obtained by executing SGD on those data. The server solely depends on model weight information to decide which device will enhance the global model the most. To make optimum use of the spectrum resources in the 6G system, we model an SUs spectrum accessing approach. Assume N SUs and M channels (N > M), with each SU having access to just one channel at a time. SUs are not allowed to transmit when PU is using the spectrum to avoid interference from unlicensed users. However, any SU may cause problems for the others. We use the previously stated spectrum access architecture to help SUs avoid interfering with one another.

4 Future Directions and Challenges 4.1 Joint Processing The communication system with an SAS in the CBRS is suitable for using the aforementioned dynamic spectrum access strategy. At the same time, the observation of RL is purely dependent on the instantaneous power of other SUs. In order to further enhance the spectrum efficiency, it is a good choice to increase the extended dimensions of observation, such as the temporal and spatial statistics of the spectra. In addition, based on their mobility and peak-time traffic, users’ access behaviors are predicted by the local agent

Federated Learning Based 6G NTN Dynamic Spectrum Access

249

through adding a supplementary policy network. However, implementing the subsidiary neural network needs additional memory and energy resources. Except for the CBRS systems, the SUs are configured to perceive the PUs’ activities before their access. For example, to supply a less false alarm probability and higher detection probability which can further improve our work. 4.2 Privacy Issues It is always a problem to keep the maximum participated devices during FL training, especially in an irresponsible network environment. What’s more, to save energy, batterypowered IoT devices adopt a special strategy of distributed learning and choose to quit a certain round of training. Promising the local user’s data privacy is also a significant feature of FL. It is possible for adversaries to collect critical information from model changes. Newer methods, such as secure multiparty computation (SMC), differential privacy, or secure aggregation, attempt to improve the privacy of federated learning by sacrificing inference performance for privacy. It is difficult to understand and balance these costs in implementing private federated learning systems, both theoretically and practically. 4.3 Asynchronous FL Optimization FL model is sensitive to the Straggler effect, which also provides better convergence guarantees. The asynchronous FL model suits practice. In terms of equipment, due to differences in hardware, network connections and battery power, system parameters in the entire network may be substantial heterogeneous. It is necessary to analyze and study the convergence boundary and expansion capability of popular algorithms such as SGD to provide convergence guarantees for the loss functions in Asynchronous FL. When learning a single global model, statistical heterogeneity creates significant challenges for measuring convergence behavior in a joint environment. Specifically, when data is non-i.i.d delivered among network devices, if the chosen devices execute too many local updates, algorithms like FedAvg might diverge in practice. FedProx recently developed is to better analyze FedAvg’s performance in heterogeneous contexts. The basic concept of FedProx is that system heterogeneity and statistical heterogeneity are inseparable. FedProx alters the FedAvg to some extent by enabling partial work to be allocated among devices based on the underlying system limits, and then safely includes the partial work using the proximal term. In addition, applying the FL framework in RL, the widely used Q-learning or DQN (discrete actions) in solving DSS problems, which requires a solid theoretic proof when making convergence analysis based on FedAVG or FedProx. Therefore, this article introduces the policy gradient as an alternative approach to update the shared model by averaging aggregated local policies.

5 Conclusion This article provides a high-level framework design to solve the dynamic spectrum access challenges using machine learning. The resulting framework gives a road map for the

250

P. Lin et al.

operation of 6G NTNs while simultaneously safeguarding user privacy and increasing spectrum efficiency. In comparison to other learning approaches, our proposed method can optimize system throughput via distributed policy gradient algorithms. Additionally, we also discussed the challenges and future directions of federated learning for DSA in 6G era.

References 1. He, L., Guo, Q., Zhong, J., Wang, X., Li, M., Teng, Y.: 5G network performance analysis and verification based on ubiquitous electricity internet of things. In: 2020 IEEE 4th Information Technology, Networking, Electronic and Automation Control Conference (ITNEC), pp. 2613–2617 (2020) 2. Koch, D.B.: An Internet of Things approach to electrical power monitoring and outage reporting. In: SoutheastCon 2017, pp. 1–3 (2017) 3. Jeon, J., Ford, R.D., Ratnam, V.V., Cho, J., Zhang, J.: Coordinated spectrum sharing framework for beyond 5G cellular networks. In: 2019 IEEE Global Communications Conference (GLOBECOM), pp. 1–6 (2019) 4. Liu, S., Lin, H., Huang, C.: Design and implement domain proxy based CBRS system for 5G. In: 2019 IEEE VTS Asia Pacific Wireless Communications Symposium (APWCS), pp. 1–6 (2019) 5. Aloqaily, M., Bouachir, O., Boukerche, A., Ridhawi, I.A.: Design guidelines for blockchainassisted 5G-UAV networks. IEEE Netw. 35(1), 64–71 (2021) 6. Wang, S., Shin, O., Shin, Y.: Social-aware routing for multi-hop D2D communication in relay cellular networks. In: 2019 Eleventh International Conference on Ubiquitous and Future Networks (ICUFN), pp. 169–172 (2019) 7. Shen, H., Deng, Y., Xu, W., Zhao, C.: Rate-maximized zero-forcing beamforming for VLC multiuser MISO downlinks. IEEE Photonics J. 8(1), 1–13 (2016) 8. Kim, S.-M., et al.: Opportunism in spectrum sharing for beyond 5G with sub-6 GHz: a concept and its application to duplexing. IEEE Access 8, 148877–148891 (2020) 9. Khalaf, Z., Nafkha, A., Palicot, J.: Enhanced hybrid spectrum sensing architecture for cognitive radio equipment. In: 2011 XXXth URSI General Assembly and Scientific Symposium, pp. 1–4 (2011) 10. Tang, Z., Shi, S., Chu, X.: Communication-efficient decentralized learning with sparsification and adaptive peer selection. In: 2020 IEEE 40th International Conference on Distributed Computing Systems (ICDCS), pp. 1207–1208 (2020) 11. Hu, S., Zhang, L.Y., Wang, Q., Qin, Z., Wang, C.: Towards private and scalable cross-media retrieval. IEEE Trans. Depend. Secure Comput. 18(3), 1354–1368 (2021) 12. Stine, J.A.: Model-based spectrum management: Loose coupling spectrum management and spectrum access. In: 2011 IEEE International Symposium on Dynamic Spectrum Access Networks (DySPAN), pp. 628–631 (2011) 13. Kim, Y.J., Hong, C.S.: Blockchain-based node-aware dynamic weighting methods for improving federated learning performance. In: 2019 20th Asia-Pacific Network Operations and Management Symposium (APNOMS), pp. 1–4 (2019) 14. Yeo, S., Lee, S., Choi, B., Oh, S.: Integrate multi-agent simulation environment and multiagent reinforcement learning (MARL) for real-world scenario. In: 2020 International Conference on Information and Communication Technology Convergence (ICTC), pp. 523–525 (2020)

Green Communication Architecture Based on Cloud Radio Access Network for Demand Response Resources of Virtual Power Plant Zhengyuan Liu1 , Chuan Liu2(B) , Xinyue Zhao1 , Fei Zhou2 , and Shidong Liu2 1 Beijing University of Posts and Telecommunications, Beijing, China 2 Global Energy Interconnection Research Institute Co., Ltd., Nanjing, China

[email protected]

Abstract. With the increasingly prominent role of virtual power plant in effectively obtaining demand response resources and power grid peak shaving and valley filling, how to build a low-carbon and efficient communication system in the scenario of virtual power plant will be a very valuable research point. This paper will pool the originally isolated BBU based on the advanced 5G communication technology and C-RAN network architecture to improve the energy utilization efficiency of the BBU, and configure green energy collection devices and energy storage devices for each BBU and RRH to further reduce the use of fossil energy. The simulation results show that the network structure proposed in this paper can reduce the use of fossil energy and realize the peak cutting and valley filling of power grid, which has important reference value for the construction of communication system of virtual power plant. Keywords: Virtual power plant · C-RAN · Green communication · Genetic algorithm

1 Introduction With the gradual rise of global temperature and the intensification of environmental problems, the concept of “carbon neutralization” has become the focus of current research. Connecting demand response resources to power grid may be an effective way to solve this problem. However, due to the fluctuation of demand response resource output, blind access to demand response resources will reduce the stability of power grid. Therefore, in the current power grid, demand response resources have not been accessed on a large scale [1]. The proposal of virtual power plant provides a new way to access demand response resources. The virtual power plant can combine the traditional thermal power plant, distributed generator, controllable load and energy storage device in a certain range through the communication system, and use the control center for coordination. Therefore, the research on the communication system of virtual power plant will greatly improve the stability of virtual power plant operation and has important research value [2]. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 S. Sun et al. (Eds.): ICSINC 2021, LNEE 895, pp. 251–258, 2022. https://doi.org/10.1007/978-981-19-4775-9_31

252

Z. Liu et al.

2 Concept and Key Technology of Virtual Power Plant 2.1 Definition of Virtual Power Plant Although the concept of the virtual power plant has already been submitted for more than 10 years, the framework of the virtual power plant is still undefined [3]. In a narrow sense, virtual power plant is a polymer composed of a group of traditional thermal power units, wind power, solar power units and energy storage devices, which is controlled by the control center; In a broad sense, virtual power plant can be combined with controllable load technology, and use power side load to form a virtual whole to participate in power grid regulation [4]. 2.2 Key Technologies of Virtual Power Plant Coordinated control technology: the control targets for virtual power plants include mainly DGs, energy storage systems, controllable loads, and electric vehicles. Since the concept of virtual power plant emphasizes external functions and effects, it is an important and difficult point of cooperative control of a virtual power plant to integrate various demand response resources in order to realize stable power output according to system demands [5]. Intelligent metrics: intelligent metrology is an important component of virtual power plants, and is an important basis for realizing demand response resource monitoring for virtual power plants. The most basic function of the intelligent metrology system is to automatically measure electricity, gas, heat, water consumption or production in a user’s residence, and provide real-time power supply and demand information to a virtual power plant [6]. Information and communication technology: the virtual power plant adopts twoway communication technology, which can not only receive the status information of the queue of each unit, but also send control information to the control object information [7]. The communication technologies applied to the virtual power plant mainly include virtual private network, power line transmission, GPRS, 3G, 4G, 5g, Wi Fi, Bluetooth, ZigBee and other communication technologies. Different service requirements can be applied according to different service requirements [8].

3 Green Communication Technology Based on C-RAN 3.1 C-RAN Communication Architecture As a flexible architecture scheme that can greatly improve resource utilization efficiency. C-RAN centrally manages computing processing resources through centralized baseband processing unit (BBU) and distributed radio frequency pull-off unit (RRH), and improves resource utilization efficiency by flexibly adjusting the pairing relationship of BBU-RRH [9]. C-RAN system resources can be divided into wireless resources and computing processing resources. Wireless resources mainly refer to the wireless transmission resources owned by the whole system, which is determined by the channel capacity between RRH

Green Communication Architecture Based on Cloud Radio Access Network

253

and UE. Computing processing resources mainly refer to the computing resources of the BBU, which are used to process the uplink and downlink data of each RRH [10]. The network architecture of regional virtual power plant is shown in Fig. 1 below:

Fig. 1. Network architecture diagram of regional virtual power plant.

3.2 Virtual Power Plant Collaborative Architecture The main task of virtual power plants is aggregation and control for demand response resources. Therefore, to a certain extent, regional virtual power plants can cooperate with each other through peer interconnection, so as to aggregate multiple regional virtual power plants into wide area virtual power plants, and the wide area virtual power plants coordinate and dispatch each regional virtual power plant, It can further strengthen the use scope of virtual power plant and the ability of overall planning of power resources. The structure of wide area virtual power plant is shown in Fig. 2 below:

Fig. 2. Wide area virtual power plant collaboration architecture.

Each regional virtual power plant has its own independent regional BBU pool with certain data processing capacity. Some data processing tasks can be processed in the regional BBU pool, and then the processed data can be sent to the wide area BBU pool, which can greatly reduce the amount of data transmission between the wide area BBU pool and the regional BBU pool. The network architecture of wide area virtual power plant is shown in Fig. 3 below:

254

Z. Liu et al.

Fig. 3. Wide area virtual power plant network architecture.

The BBU pool and RRH in virtual power plant have green energy collection devices and energy storage devices. The energy storage device can store the excess electric energy in the low power consumption period, release energy in the peak power consumption period, so as to achieve the purpose of peak cutting and valley filling [11]. When the energy storage device has surplus energy, part of the energy can be transmitted to other regional virtual power plants through the power grid. The energy flow architecture of regional virtual power plant communication system is shown in Fig. 4 below:

Fig. 4. Energy coordination architecture of regional virtual power plant.

3.3 Mathematical Model Suppose a regional virtual power plant has i = {1, … , N} BBUs, j = {1, … , M} RRHs, k = {1, … , W} UEs. Energy consumption of each BBU consists of two parts: dynamic energy consumption and static energy consumption. The dynamic energy consumption is shown in Eq. (1). The energy consumption expression of the ith BBU is shown by Eq. (2). The energy consumption expression of the j-th RRH is shown in Eq. (3). It is necessary to sum the power allocated by the RRH to each connected UE. The grid energy consumed by BBU and RRH in time slot can be obtained by subtracting the energy supply of green energy from the energy consumption of equipment at that time, as shown in Eqs. (4) and (5). The electricity stored in BBU and RRH is the sum of the remaining energy of the previous time slot and the energy consumed or supplemented by the current time slot,

Green Communication Architecture Based on Cloud Radio Access Network

255

as shown in Eqs. (6) and (7) respectively. Sum the grid energy consumed by all BBUs and RRHs in each time slot, that is, the total grid energy consumed by the system, as shown in Eq. (8). PiD (t) = αri (t)

(1)

PiBBU (t) = piS (t) + piD (t)

(2)

PjRRH (t) =

W 

sjk (t)pjk (t)

(3)

k=1

ECiBBU (t) = PiBBU (t) − PiGreen (t)

(4)

ECjRRH (t) = PjRRH (t) − PjGreen (t)

(5)

ch arg e

(t)

(6)

ch arg e

(t)

(7)

EiBBU (t) = EiBBU (t − 1) + Pi EjRRH (t) = EjRRH (t − 1) + Pj ECtotal =

N   i=1

t 0

ECiBBU (t)dt +

M   j=1

t 0

ECjRRH (t)dt

(8)

4 Example Analysis We take the plant area as an example scenario. On weekdays, the network traffic generated in the plant area will increase with the arrival of staff and the increase of the number of connected terminal equipment, reach a peak, and gradually decrease with the staff gradually leaving the plant area. During non working hours, the network traffic generated in the plant is regarded as a constant value. Thus, a possible network load sequence is obtained, and then the energy consumption of each time slot is calculated according to this sequence. The energy consumption rate of each time slot is shown in Fig. 5(a) below: Assuming that the green energy collection device can only collect solar energy, the device will not be able to effectively collect green energy before sunrise and after sunset. During this period, the energy supply of BBU and RRH mainly depends on the power grid and energy storage equipment. The energy collection rate of the green energy collection device in each time slot is shown in Fig. 5(b).

256

Z. Liu et al.

(a)

(b)

Fig. 5. Energy consumption and energy collection rate per slot of BBU and RRH.

The above energy consumption sequence and green energy generation sequence are input into the genetic algorithm, and the peak valley electricity price is used. The specific flow of the algorithm is shown in Fig. 6 below. The average fitness curve of genetic algorithm is obtained through simulation experiment, and the curve is processed to obtain the average power unit price curve, as shown in Fig. 7 below. It can be seen that after about 400 generations of iteration, the fitness of the population has tended to be stable, and the average unit price of electricity is about 0.65 yuan/kWh. When BBU and RRH do not use energy storage devices and green energy generation devices, the average unit price of electricity is 0.978 yuan/kWh. Therefore, the use of green energy collection device and energy storage device can effectively reduce the use of fossil energy, cut the peak and fill the valley and reduce the load pressure of power system.

Fig. 6. Flow chart of optimization algorithm.

Green Communication Architecture Based on Cloud Radio Access Network

257

Fig. 7. Optimization results of genetic algorithm.

5 Conclusion In the scenario of virtual power plant, 5G technology will be a technology with great research value and application value because of its characteristics of low delay, large connection and high speed. This paper mainly discusses the 5G green communication architecture based on C-RAN in the scenario of virtual power plant, and applies the idea of virtual power plant in this architecture, adding green energy collection device and energy storage device. Through simulation experiments, this communication architecture can effectively reduce the use of fossil energy, cut peak and fill valley, and reduce the load pressure of power system, It has great reference value for the research of communication system of virtual power plant. Acknowledgments. The authors acknowledge Science and Technology Project of State Grid Corporation of China: “Key technology Research and Communication Network Systems for Power Plant (No. 5700-202117190A-0-0-00)”.

References 1. Bhuiyan, E.A., Hossain, M.Z., Muyeen, S.M., et al.: Towards next generation virtual power plant: technology review and frameworks. Renew. Sustain. Energy Rev. 150, 111358 (2021) 2. Yu, S., Fang, F., Liu, Y., et al.: Uncertainties of virtual power plant: problems and countermeasures. Appl. Energy 239, 454–470 (2019) 3. Yavuz, L., Önen, A., Muyeen, S.M., et al.: Transformation of microgrid to virtual power plant–a comprehensive review. IET Gener. Transm. Distrib. 13(11), 1994–2005 (2019) 4. Rahimiyan, M., Baringo, L.: Real-time energy management of a smart virtual power plant. IET Gener. Transm. Distrib. 13(11), 2015–2023 (2019) 5. Oest, F., Radtke, M., Blank-Babazadeh, M., et al.: Evaluation of communication infrastructures for distributed optimization of virtual power plant schedules. Energies 14(5), 1226 (2021) 6. Wang, Z., Liu, R., Liu, Q., Thompson, J.S., Kadoch, M.: Energy-efficient data collection and device positioning in UAV-assisted IoT. IEEE Internet Things J. 7(2), 1122–1139 (2020) 7. Zajc, M., Kolenc, M., Suljanovi´c, N.: Virtual power plant communication system architecture. In: Smart Power Distribution Systems, pp. 231–250. Academic Press (2019)

258

Z. Liu et al.

8. Kolenc, M., Nemˇcek, P., Gutschi, C., et al.: Performance evaluation of a virtual power plant communication system providing ancillary services. Electric Power Syst. Res. 149, 46–54 (2017) 9. Yu, P., Yang, M., Xiong, A., Ding, Y., Li, W., et al.: Intelligent-driven green resource allocation for industrial internet of things in 5G heterogeneous networks. IEEE Trans. Ind. Inform. 18, 520–530 (2020) 10. Chen, N., Rong, B., Zhang, X., Kadoch, M.: Scalable and flexible massive MIMO precoding for 5G H-CRAN. IEEE Wirel. Commun. Mag. 24(1), 46–52 (2017) 11. Zeng, D., Zhang, J., Gu, L., et al.: Energy-efficient coordinated multipoint scheduling in green cloud radio access network. IEEE Trans. Veh. Technol. 67(10), 9922–9930 (2018)

Real-Time Bandwidth Prediction and Allocation Method for Smart Grid Communication Network Services Bozhong Li1 , Fang Chen1 , Zifan Li1(B) , Xingyu Zhao1 , Zhuojun Jin2 , and Zhengyuan Liu2 1 State Grid Information and Telecommunication Branch, Beijing, China

[email protected] 2 Beijing University of Posts and Telecommunications, Beijing, China

Abstract. The real-time and accuracy of business flow forecasting is essential to the normal and stable operation of the data network, and it is also a key technology to improve the efficiency of the power grid. Therefore, this paper proposes an AUGRU-based real-time prediction algorithm for power communication network traffic and a momentum-based traffic correction method. Subsequently, according to the real-time prediction results, we also proposed a dynamic bandwidth allocation algorithm based on DDQN to ensure that the allocated bandwidth can meet the predicted results. Finally, the simulation results show that our algorithm can effectively guarantee the service delay and bandwidth resource requirements. Keywords: Component · Formatting · Style · Styling · Insert (Key words)

1 Introduction With the rapid development of smart grids, the reliable, efficient, and stable operation of the power communication network is an important guarantee for the safety and stability of the power system [1]. In the context of large-scale interconnection of power grids, in order to meet the needs of intelligent and real-time power system operation and control, the power system needs to be able to predict in real time the service bandwidth requirements for a certain period of time in the future, and reasonably allocate service bandwidth resources. In addition, the power system’s forecast of service bandwidth should also add a compensation mechanism, based on the comparison of real-time bandwidth and predicted bandwidth, to make certain compensation for the future predicted bandwidth to prevent insufficient allocated bandwidth or excessive bandwidth causing waste of resources [2].

2 Real-Time Traffic Forecast Model Based on AUGRU 2.1 Real-Time Traffic Forecast Model Based on AUGRU In orde7r to solve the problem of periodic traffic prediction in power communication network, we proposed a prediction model based on GRU with attentional update gate © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 S. Sun et al. (Eds.): ICSINC 2021, LNEE 895, pp. 259–266, 2022. https://doi.org/10.1007/978-981-19-4775-9_32

260

B. Li et al.

(AUGRU) [3]. AUGRU uses the attention coefficient to directly control the update of the hidden state. The update formula of the hidden state of AUGRU is: u˜ t = at ∗ ut

(1)

◦   ht = 1 − u˜ t ht + u˜ t◦ h˜ t

(2)

The model we proposed in this paper is structured into two parts, namely the longterm flow information module and the real-time flow information module [4]. The longterm traffic information module models the business traffic information to be predicted in a long period [5]. The real-time traffic information module models the real-time module of business traffic. These two modules enable the model to learn the law of traffic changes in the business in the long-term [6]. The calculation method of the attention score in our model is as follows: attention score = sigmoid (Ht ∗ ht )

(3)

Ht is the output of the last unit after processing long-term traffic information, and ht is the input of the real-time traffic information module, that is, the traffic information of the business at time t. 2.2 Data Organization In this paper, we use the business traffic data of the past 24 h to predict the business traffic data of the current hour. According to the analysis of this data set and the test results using a certain time range, it is shown that such a division has a higher prediction accuracy. Using LSTM neural network is many-to-many, that is, for each input xt , we can get a corresponding output yt . The structure of input data and output data is shown as Fig. 1.

Fig. 1. Model input data and model prediction data structure diagram

Real-Time Bandwidth Prediction and Allocation Method

261

As shown in the figure above, the process of forecasting is to input the data of the past week into the long-term sequence processing module, and at the same time input the data of the past 30 min and the data to be predicted as a real-time sequence to the short-term sequence processing module. 2.3 Traffic Forecast Correction Mechanism Through a large number of experiments, we found that the flow data predicted by the realtime flow prediction model based on AUGRU has some differences with the real value, but the overall change law is consistent [7]. Based on this, we refer to the momentum algorithm and propose a correction method based on momentum [8]. The specific formula is as follows:   (4) aup = μup aup + 1 − μup θ adown = μdown adown + (1 − μdown )θ

(5)

Among them, aup represents the ascending rate, adown represents the descending rate, μup represents the retention in the ascending direction, which can be regarded as the resistance value, and μdown represents the retention in the descending direction. Next, we will introduce how to use the forecast correction mechanism to adjust the forecast results based on AUGRU. The correction method is as follows:  t−1 t t Bt = ypred + aup ifypred > yactual , update aup (6) t−1 t t t B = ypred + adown ifypred < yactual , update adown After the AUGRU-based model is used to predict, compare the current flow value in the prediction result with the flow value of the previous moment. If the value of the current moment is greater than the value of the previous moment, the current flow value is corrected by adding aup , Update the value of aup at the same time; in the same way, if the value of the current moment is less than the value of the previous moment, the flow value at the current moment is added to adown for correction, and the value of adown is updated at the same time. The result of direct prediction and the prediction result after momentum correction is shown as Fig. 2. After 900 times of training, the prediction results of the model are as Fig. 2. As can be seen from the above figure, compared with the original predicted flow, the corrected predicted flow is basically consistent with the actual flow data, indicating that the momentum based flow prediction correction algorithm proposed by us can effectively correct the flow prediction results.

262

B. Li et al.

(a)

(b)

(c) Fig. 2. (a) is the prediction result of using AUGRU-based model directly and (b) is the result of adding the forecast correction mechanism. (c) is the absolute value of the error between the two methods and the real value respectively

3 Real-Time Bandwidth Allocation Method Based on DDQN After establishing the AUGRU-based real-time forecasting model of business traffic, we must also consider how to allocate reasonable bandwidth and routing for the business based on the predicted value. In response to this problem, we proposed a real-time allocation method of service bandwidth based on DDQN [9]. DDQN needs to define the state space and reward function  according to our background. First, we need to define the state space. Let At = A1,t , . . . , A|V |t represent the number of services on each node in the network at time t, Bt = B1,t , . . . , B|W |t represents the number of services carried on each link of the network at time t. The network status can be defined by (At , Bt ). We Use Ct to represent the location information of the current service, which can be represented by a one-hot code, and the vector length is the number of nodes. If the current service is located at the i-th node, the i-th position in the vector is 1 and the rest are 0. The state space of DDQN can be expressed as St = (At , Bt , Ct ). Since our goal is to guarantee the bandwidth of the service while considering the service delay, the design of the reward function must consider the rationality of the service bandwidth allocation and ensure that the service can reach the destination address within an acceptable delay [10]. At the same time, since there are requirements for indicators in the actual power production process, it is necessary to set constraints on each indicator.

Real-Time Bandwidth Prediction and Allocation Method

To build a reward function for reinforcement learning: ⎧ ⎨ r1 Does not meet business needs   R s, a, s = r2 Invalid action ⎩ r3 Meet business needs

263

(7)

where, r1 is the reward when the bandwidth allocated to the service after the selected action is less than the bandwidth prediction value, r2 is the feedback reward when action a is an invalid action. r3 means that the network can meet the real-time bandwidth requirements of the service after the path is allocated to the service. The main factor considered at this time is the average service delay. Therefore, the reward value at this time is the average service delay [11]. In addition,a value function needs to be defined. In DDQN, firstly we use MainNet  to find maxa Q s , a ; θi and Action (θi is a parameter of MainNet), and then we use TargetNet to find the Q value of this Action to form the Target Q value. The final Loss function to learn is: 

(8) L(θ ) = E (TargetQ − Q(s, a; θi ))2     TargetQ = r + γ Q s , maxa Q s , a ; θi ; θi−

(9)

In the above formula, L(θ ) represents the loss function in DDQN, and γ is the discount factor. Q(s, a; θi ) represents the action state value function.

4 Simulation The simulation scenario is shown below. The power communication system has 17 nodes and 15 links. There are four services in the communication network, namely {[0, 10], [0, 15], [1, 15], [2, 12]}. Marked as s1 , s2 , s3 and s4 , the priority is from high to low. The scene graph of the model is shown as Fig. 3.

Fig. 3. Experimental scene graph

This paper adopts comparative experiments, respectively using Naive Bandwidth Allocation Algorithm (N-BA), and Dynamic Bandwidth Allocation Algorithm based on High-order Moving Average Model (HMAM-Bandwidth Allocation Algorithm). DBA),

264

B. Li et al.

AUGRU-based real-time traffic prediction model and DDQN-based real-time bandwidth allocation method (AUGRU-DDQN). As shown in Fig. 4(a), for s1 high-priority services, because the N-BA algorithm uses a fixed bandwidth allocation model, for high-priority services, it can always meet its bandwidth requirements first. The AUGRU-DDQN algorithm proposed in this paper uses the AUGRU algorithm for real-time bandwidth prediction, and after the prediction is revised, the accuracy of the prediction is somewhat improved compared with the prediction accuracy of the HMAM-DBA algorithm. When bandwidth allocation is performed, the real-time bandwidth of DDQN is used The allocation model looks for the optimal bandwidth allocation fraction, so the delay is greatly shortened compared with the HMAM-DBA algorithm, and the overall performance is similar to the N-BA algorithm [12].

(a)

(b)

(c)

(d)

Fig. 4. The time delay of S1 ,S2 ,S3 ,S4 service varies with load conditions

Figure 4(b) shows the change of the delay of the s2 service with the network load. As shown in the figure, compared to the N-BA algorithm, the HMAM-DBA algorithm and the AUGRU-DDQN algorithm have significantly improved delay performance. Among them, the delay performance of the AUGRU-DDQN algorithm compared to the N-BA algorithm is increased by about 37%-44% when the network load is high. This is because the N-BA algorithm allocates too much bandwidth to high-priority services when the network load is high.

Real-Time Bandwidth Prediction and Allocation Method

265

As shown in Fig. 4(c), for the s3 service, compared with the N-BA algorithm, the HMAM-DBA algorithm and the delay have been reduced by about 14%, and the LSTMDBA algorithm has been reduced by about 17%. Therefore, in general, the AUGRUDDQN algorithm proposed in this paper performs well when the communication network is under low load and high load. As can be seen from Fig. 4(d), for low-priority s4 services, the N-DBA algorithm is significantly better than the HMAM-BA algorithm and the AUGRU-DDQN algorithm in reducing latency. Especially when the communication network is under high load, the delay is reduced by about 65% compared with the latter two. This is because the fixed bandwidth allocation method of the N-DBA algorithm always reserves a certain bandwidth for the s4 bandwidth, which also results in the overall performance of the NDBA algorithm in the s1-s3 business is not as good as the latter two. The AUGRU-DDQN algorithm performs better than the HMAM-BA algorithm under high load conditions because the AUGRU has higher prediction accuracy and benefits from the threshold set for s1-s3.

5 Conclusion In this paper, we propose an AUGRU-based real-time forecasting algorithm for power communication network traffic, and then a momentum-based forecast correction algorithm, which can effectively correct the forecast results. The experimental results show that the modified prediction result is accurate. Then we proposed a real-time bandwidth allocation algorithm based on DDQN, which can effectively prevent insufficient bandwidth allocation and resource waste caused by excessive bandwidth allocation by setting thresholds and secondary allocation. However, this paper also has some shortcomings to be improved. First, the method of using AUGRU to predict and then use momentum to correct is not straightforward. Secondly, AUGRU also has a certain length limit in the prediction sequence. Although it has been greatly improved compared to RNN and LSTM, if the sequence is too long, there will be information omission problems, that is, the previous information cannot be remembered. Finally, for the real-time bandwidth allocation algorithm, the upper limit of the bandwidth of different priority services is artificially set based on experience, which may not be able to cope with the surge of certain service traffic.

References 1. He, Y., Mendis, G.J., Wei, J.: Real-time detection of false data injection attacks in smart grid: a deep learning-based intelligent mechanism. IEEE Transactions on Smart Grid 8(5), 2505–2516 (2017) 2. Yu, P., Zhou, F., Zhang, X., Qiu, X., Kadoch, M., Cheriet, M.: Deep learning-based resource allocation for 5G broadband TV service. IEEE Trans. Broadcast. 66(4), 800–813 (2020). Dec. 3. Zhou, G., et al.: Deep interest network for click-through rate prediction. In: KDD, pp. 1059– 1068 (2018) 4. Li, L., Ota, K., Dong, M.: When weather matters: IoT-based electrical load forecasting for smart grid. IEEE Commun. Mag. 55(10), 46–51 (2017)

266

B. Li et al.

5. Wang, D.: Bandwidth prediction for business requirement of electric power communication network with deep-learning. In: 2018 3rd International Workshop on Materials Engineering and Computer Sciences (IWMECS 2018). Atlantis Press (2018) 6. Zhiqiang, X., Jun, L., Zi, L., Yanping, Z.: Queue-theory-based service-section communication bandwidth calculation for power distribution and utilization of smart grid. In: 2015 8th International Conference on Intelligent Networks and Intelligent Systems (ICINIS), pp. 137–140. Tianjin (2015). https://doi.org/10.1109/ICINIS.2015.11 7. Zhao, Z., Xiong, S., Zhou, J., Chen, X.: Backbone network bandwidth prediction and plan technology research based on Smart Grid telecommunication service analysis. In: 2010 International Conference on Power System Technology, pp. 1–10. Hangzhou (2010). https://doi. org/10.1109/POWERCON.2010.5666411 8. Cenggoro, T.W., Siahaan, I.: Dynamic bandwidth management based on traffic prediction using deep long short term memory. In: 2016 2nd International Conference on Science in Information Technology (ICSITech), pp. 318–323. Balikpapan (2016). https://doi.org/10. 1109/ICSITech.2016.7852655 9. Ridwan, M.A., Radzi, N.A.M., Abdullah, F., Al-Mansoori, M.H.: Implementation of universal dynamic bandwidth allocation algorithm in smart grid environment. In: 2018 IEEE 7th International Conference on Photonics (ICP), pp. 1–3. Kuah (2018). https://doi.org/10.1109/ ICP.2018.8533186 10. Webster, R., Munasinghe, K., Jamalipour, A.: A population theory inspired solution to the optimal bandwidth allocation for Smart Grid applications. In: 2014 IEEE Wireless Communications and Networking Conference (WCNC), pp. 2958-2963. Istanbul (2014). https://doi. org/10.1109/WCNC.2014.6952927 11. Guo, J., Liu, F., Lui, J.C.S., Jin, H.: Fair network bandwidth allocation in iaas datacenters via a cooperative game approach. IEEE/ACM Trans. Networking 24(2), 873–886 (2016). https:// doi.org/10.1109/TNET.2015.2389270 12. Qi-yu, Z., Bin, L., Run-ze, W.: A dynamic bandwidth allocation scheme for GPON based on traffic prediction. In: 2012 9th International Conference on Fuzzy Systems and Knowledge Discovery, pp. 2043–2046. Sichuan (2012). https://doi.org/10.1109/FSKD.2012.6234355

Attack Portrait and Replay Based on Multi-spatial Data in Grid System Zhiyuan Pan1 , Lisong Shao2 , Xinxin Song1 , and Hui Guo3(B) 1 State Grid of China Technology College, Jinan 250000, China 2 State Grid Electric Power Research Institute, Beijing 100092, China 3 State Key Laboratory of Networking and Switching Technology, Beijing University of Posts

and Telecommunications, Beijing 100876, China [email protected]

Abstract. The power system ensures the stability of national economic development, so the security of power system is particularly important. The attack portrait can intuitively display the attack information, help the security personnel intuitively analyze the attack, and is of great significance for the security analysis inside the power system. In this paper, the k-means algorithm improved by LOF algorithm is used to cluster the attacks to obtain the attack types, and the five features with the highest matching degree with the feature database of a specific attack are used as the attack labels to construct the attack portrait and realize the visualization of the attack portrait. Through experiments, it is found that this method can well display the attack information of internal attacks in power system, which is of great significance for security personnel to analyze attacks. Keywords: Power system · Attack portrait · LOF · K-means

1 Introduction The power grid security accidents occur frequently, in order to effectively maintain the security of power grid, it is of great practical significance to study the characteristics of network attack behavior, design attack detection model and depict the attack portrait. The attack portrait refers to collecting and analysising a lot of information about the attack, and labeling the attack behavior. The attack portrait includes portrait outline and portrait content. The portrait outline is the most intuitive feature content of the portrait, and the portrait content is the description of the detailed information of the portrait. The significance of the attack portrait is to help the monitoring personnel understand the attack from a multi-dimensional perspective, This paper proposes an attack portrait framework based on attack behavior analysis of power grid system. And it provides theoretical support and scientific basis for network security managers to protect in advance. In this paper, our contributions are as follows: (1) Comprehensively consider the impact of the data collected in the power monitoring system, and conduct correlation and fusion analysis of these multi-source data; © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 S. Sun et al. (Eds.): ICSINC 2021, LNEE 895, pp. 267–274, 2022. https://doi.org/10.1007/978-981-19-4775-9_33

268

Z. Pan et al.

(2) We optimized K-means algorithm and combined with outlier detection technology LOF algorithm, which can separate outliers in data during clustering; (3) The extracted attack features are matched with the attack feature database by SVC algorithm, and the top five features with the highest matching degree are extracted as attack labels.

2 Related Work The analysis of attack portrait uses the attack element to draw the attack graph, form the attack portrait based on the attack element, and improve the recognition accuracy. Portrait analysis technology was used to describe and analyze characters in the early stage. Reference [1] applies user portrait technology to groups with auxiliary vision defects. They use the ontology to construct the portrait of each target user and use the portrait to improve the interaction mode of the program. Reference [2] analyzes the user consumption data of power grid companies, and proposes a multi-perspective fusion framework for constructing user portraits. Reference [3] clusters and analyze users’ behavior data in combination with the improved k-means algorithm to obtain users’ portraits. The research shows that the accuracy of the user portrait constructed by this method has been greatly improved. In the power grid system, most of the portraits focus on the user side, and only a few portraits focus on the equipment. Reference [4] analyzes the defect set of power equipment with the improved clustering algorithm, and builds power equipment defect fault portrait model. In order to improve the efficiency and accuracy of network attack identification, reference [5] proposed a threat intelligence portrait analysis method for attack identification and realized the threat intelligence portrait analysis. However, the reference needs expert knowledge and analysis experience, which may have an impact on the structure of portrait analysis. To sum up, the construction of attack portrait can accurately analyze the behavior differences of different attacks according to the attack portrait, but there are few researches on network attack portrait at present. Therefore, this paper proposes a construction method of attack portrait.

3 Methodology In this paper, the traditional clustering algorithm k-means is optimized for the sensitive problems of noise and initial point selection. This paper mainly combines the outlier detection technology LOF algorithm to separate the outliers in the data in advance during clustering, so as to improve the accuracy of clustering. 3.1 LOF Algorithm The LOF algorithm is suitable for the identification of outliers in high-dimensional data. The LOF algorithm judges the degree of outlier by giving each object a factor. The larger the factor, the greater the degree of outlier. The specific calculation method is as follows:

Attack Portrait and Replay Based on Multi-spatial Data

269

(1) Calculate the k- distance dk (p) of data object p: in a given data set D, dk (p) is the distance between the data object p and its k-th nearest neighbor. (2) Find k-distance neighborhood Nk (p) of object p: for data set D, all object sets within k-distance dk (p) of data object p: Nk (p) = {q ∈ D|d (p, q) ≤ dk (p)}

(1)

(3) Calculate the reachable distance reachdisk (p, o) from the data object p to the data object o: reachdisk (p, o) = max{dk (o), d (p, o)}

(2)

That is, the k-th reachable distance from object o to object p is at least the k-th distance of o, or the real distance between o and p. (4) Calculate the local reachable density lrdk (p) of data point p: lrdk (p) = 

|Nk (p)| o∈Nk (p) reachdisk (p, o)

(3)

where |Nk (p)| is the number of neighborhood points Nk (p) of p. (5) Calculate the local outlier factor LOFk (p) of the data object p:  lrdk (o) LOFk (p) =

o∈Nk (p) lrdk (p)

|Nk (p)|

(4)

If LOFk (p) is greater than 1, it means that the p is more likely to be an abnormal point. The LOF algorithm overcomes the problem of different density in the data set. The data processed in this paper is multidimensional„ and the attack has the problem of different density. Therefore, we select the LOF algorithm. 3.2 K-means Algorithm The k-means algorithm has high efficiency for clustering in large data sample space. K-means is an iterative algorithm, and its algorithm flow is as follows: (1) Select k points from data set D as the initial clustering centers. (2) Allocate each data point in D to the nearest class cluster. (3) Calculate the average value of vector coordinates of data points in each cluster to update the cluster center of this cluster. (4) Repeat steps (2) and (3) until the cluster center converges. The elbow rule (SSE) is used to determine the optimal k value used by k-means algorithm. The calculation formula of SSE is as follows: SSE =

k  

||d − ci ||2

(5)

i=1 d∈Di

where, k is the number of class clusters and ci is the cluster center of the i-th class cluster.

270

Z. Pan et al.

3.3 The Algorithm Used in This Paper This paper first uses LOF algorithm to calculate the anomaly coefficient of each data point, and then the data objects whose anomaly coefficient is greater than the threshold are marked as outliers, and these outliers are excluded and filtered. These outliers are ignored in the cluster center calculation of k-means algorithm to reduce the impact of noise data on clustering. Aiming at the problem that the k-means algorithm is sensitive to the initial cluster centers„ we first count the value range of the data set samples excluding the outliers in each dimension is calculated first, and then the initial cluster center is obtained by the average difference method according to the required number of cluster centers.

4 Workflow As shown in Fig. 1, the attack portrait framework includes three modules: data processing, portrait construction and portrait visualization.

Fig. 1. Attack portrait architecture

4.1 Data Processing The goal of data processing module is to fuse multi-source heterogeneous data and format the fused data. Network traffic processing: (1) Traffic selection: extract Syslog type traffic and discard irrelevant traffic. (2) Payload extraction: unpack the traffic from the physical layer to the application layer to extract the payload. (3) Decoding: decode the payload to prevent codes garbling. Log processing: (1) Log filtering: remove redundant information from logs. (2) Word segmentation: use the word segmentation tool to segment the logs. Format: after word segmentation in the log, the traffic features are extracted. The features are encoded by real numbers, and the heterogeneous data are transformed into feature vectors. Finally, they are transformed into feature matrix for storage.

Attack Portrait and Replay Based on Multi-spatial Data

271

4.2 Portrait Construction Firstly, we use the LOF algorithm to calculate the anomaly coefficient of each data sample, and complete the classification of outliers. Then, we use k-means algorithm to cluster the sample data, obtain the category of attack, and take it as the outline of attack portrait. The multi-dimensional attack features extracted by clustering method are matched with the attack feature database by SVC algorithm, and the features are obtained as attack labels. Finally, the first five labels with the highest adaptability are selected as the portrait content. Finally, the outline and content of the portrait are combined to form an attack portrait. 4.3 Portrait Visualization We combine the detailed information such as the source IP, destination IP and occurrence time of the attack with the tag of the attack portrait, and visually display the attack portrait in the form of result list and tree graph.

5 Experiments 5.1 Datasets In this experiment, the data set we used is the real data collected in provincial company of State Grid. It includes network traffic and system log, in which the system log contains information from multiple dimensions. The traffic information is in the form of PCAP package, and the log information is XLS file. 5.2 Method Evaluation Aiming at the problem that k-means algorithm is sensitive to abnormal data and initial cluster center, this paper uses LOF algorithm to improve k-means algorithm to optimize clustering results. The elbow method is used to select the optimal number of clusters. The SSE value changes as shown in Fig. 2. When k = 3, the slope of SSE change curve slows down. Therefore, we use k = 3 for clustering experiment comparison. This paper compares the proposed method with the traditional K-means algorithm, and compares the clustering effects of the two algorithms through SSE and contour coefficient. The contour coefficient of a point is defined as: s=

disMeanout − disMeanin max(disMeanout , disMeanin )

(6)

where, the disMeanin is the average distance between this point and other points in this category, and disMeanout is the average distance between this point and non points in this category. The closer s is to 1, the better the classification. The SSE calculation formula

272

Z. Pan et al.

Fig. 2. The change trend of SSE with cluster number

Table 1. Comparison of two groups of algorithm indexes Algorithm

Index SSE

Contour Coefficient

K-means

8.83 × 106

0.6689

The algorithm in this paper

2.26 × 106

0.8031

is shown in formula (5). The smaller the SSE value, the better the clustering effect of the model. The comparison results are shown in Table 1. As can be seen from Table 1, the contour coefficient of the algorithm in this paper is greatly improved than that of K-means, and the SSE is reduced by more than half compared the original algorithm. 5.3 Experimental Result Through the above experiments, we can well classify the attacks and get the types of attacks. Through feature extraction and label selection, we can get the attack portrait content more accurately. By testing the real data of the power grid system, we can see that the real data classification of the power grid system includes “ICMP service access exception”, “DHCP service access exception”, “Illegal port opening” and etc. Now we choose the real alarm classification "ICMP service access exception" as an example to show the effectiveness of our attack portrait method. Figure 3 shows the portrait of ICMP attack.

Fig. 3. Attack portrait

Attack Portrait and Replay Based on Multi-spatial Data

273

Firstly, we divide the attack data into 45 categories by using the improved k-means algorithm. For the “ICMP service access exception” attack, we extract many features, such as "ICMP”, "link detection”, and etc. After matching these features with the attack feature database, we get the top five features with the highest matching degree as the attack labels. The attack labels obtained by this attack are: “ICMP”, “link detection”, “Telecontrol front-end server”, “vertical encryption” and “important alarm”. Their matching degrees are 99.8%, 99.6%, 98.9%, 97.0% and 97.4% respectively.

6 Conclusion In this paper, we mainly use the correlation between power grid multi-source data, takes the attack categories as the portrait outline, and take the specific attack features as the portrait content to construct the attack portrait. In the process of obtaining the attack category, this paper improves the k-means algorithm by using LOF algorithm, and obtains the attack category by removing the discrete value clustering method. This paper matches the extracted features with the attack feature database, and selects the top five features with the highest matching degree as the attack labels. Through the experiment on the real data of the power grid system, it can be obtained that the attack feature matching degree used by the attack portrait can reach more than 97%, which verifies that our method has achieved good results. Acknowledgements. Thanks to State Grid for supporting the data and funds to this project. Thanks to Science and Technology Project Funding of State Grid Corporation of China(Research on Key Technologies of Comprehensive Management and Control of Network Security Threats and Process Deduction for Power Grid Monitoring System in 2020-2021,5108-202040036A-0-0-00).

References 1. Torres-Carazo, M.I., et al.: Ontology-based user profile modelling to facilitate inclusion of visual impairment people. In: International Conference on Model and Data Engineering, pp. 386–394. Springer, Cham (2017) 2. Fei, P., et al.: A multi-view fusion framework for constructing user portraits. Computer Science 204(1), 179–182 (2018) 3. Wang, Y., et al.: Design and implementation of user portrait system. Comp. Appl. Softw. 35(3), 8–14 (2018) 4. Zhang, P., Wang, W., Zhao, D.W., Si, X.F.: Power equipment defect user portrait construction based on text mining. Sci. Technol. Wind, 177–180 (2019) 5. Yang, P.A., Liu, B.X., Du, X.Y.: Threat intelligence portrait analysis for attack identification. Comput. Eng. 46(1), 136–143 (2020) 6. Geng, J.C., Guo, Z.M., Li, X.L., Su, J., Yuan, S.G., Niu, S.X.: Verification method of distribution network line transformer relationship data based on lof and SVM. China Testing 04, 49–54 (2021) 7. Zhang, J.: Research on Library Knowledge Discovery Service Based on user portrait. Library and Information 06, 60–63 (2017)

274

Z. Pan et al.

8. Yao, Y., Zhang, H., Hao, Q., Xu, S.D.: “Ontology based user portrait construction method”, Network application branch of China Computer Users Association. In: Proceedings of the 22nd Annual Conference on new network technology and application of network application branch of China Computer Users Association in 2018. Network application branch of China computer Users Association: Beijing Key Laboratory of information service engineering, Beijing Union University (2018)

Radar Signal Classification Based on Bispectrum Feature and Convolutional Neural Network Haoyun Liu1 , Zesheng Zhou2 , Bingyang Li3 , Jia Zhu1 , Xiaojun Jing1(B) , and Bohan Li4 1 Beijing University of Posts and Telecommunications, Beijing, China

{lhy0824,jxiaojun}@bupt.edu.cn

2 Shandong Institute of Aerospace Electronics Technology, Shandong, China 3 Jiangsu Automation Research Institute, Lianyungang, Jiangsu, China 4 University of Southampton, Southampton, UK

Abstract. Radar signal classification is the key link in electronic information warfare, but as radar modulation becomes more sophisticated and the electromagnetic environment of the battlefield becomes complex, it is increasingly difficult to classify the radar signal. Aiming at the problem of low accuracy of radar signal classification in a low signal-to-noise ratio environment, a classification method based on bispectrum feature and convolutional neural network is proposed, it increases the accuracy of signal classification by taking advantage of bispectrum, which suppresses the Gaussian noise and retains the phase information. Therefore, the images of the signal bispectrum after pre-processing and data enhancement can train convolutional neural networks to obtain deeper signal features. Experimental results show that radar signal classification based on bispectrum features and convolutional neural networks can effectively improve the effect of radar signal classification. Keywords: Radar signal classification · CNN · Bispectrum

1 Introduction With the development of information technology, electronic countermeasures are gradually replacing guns and tanks to become the leading role in modern warfare. In order not to be detected and decrypted, radar modulated mode is evolving towards low interception characteristics and complexity. With civilian signal and environmental noise aliasing, the electromagnetic environment of modern warfare is harsh and brings challenges to radar signal feature extraction and recognition [1]. As early as the 1970s, radar emitter identification technology has been studied, the initial radar modulation types were simple and could be distinguished by time or frequency domain information. On this basis, researchers built a feature database based on the modulation types and by comparing with the database, measured signals can be identified [2]. This method is simple and fast to calculate, but it depends on prior knowledge and can’t recognize a new form of the radar signal, so the researchers combined © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 S. Sun et al. (Eds.): ICSINC 2021, LNEE 895, pp. 275–281, 2022. https://doi.org/10.1007/978-981-19-4775-9_34

276

H. Liu et al.

the time domain and frequency domain information to classify the signals, called time– frequency analysis. Zhe Zhang proposed a compound modulation recognition method based on FRFT and phase jump detection [3]. The FRFT was applied to the signal square, and the phase characteristics of the signals were compared with the spectral envelope characteristics. After the popularity of deep learning methods, scholars began to apply deep learning to signal classification. Yutao Huang proposed a deep learning method by mining the abstract features hidden inside the signal ambiguity function and using a convolutional neural network to extract the contour features of the ambiguity function to improve the accuracy of radiation source signal classification [4]. With an appropriate initial signal feature and using deep learning algorithms, relatively good results can be gained in radar signal classification [5]. A classification method based on bispectrum feature and convolutional neural network is proposed in this paper. In this method, a convolutional neural network is trained on the self-made dataset which is built on a bispectrum image of signals. The bispectrum feature is gained by nonparametric estimation. And this model is used for classifying radar signals. Experimental results proved the effectiveness of this method.

2 Bispectrum Feature of Signals at Low SNR In this paper, the signal spectrum is selected as the feature in radar signal classification, which is obtained by the nonparametric estimation method. During the estimation, the signal bispectrum retains the original signal amplitude, phase, frequency, and other information, and has an inhibitory effect on Gaussian noise, therefore bispectrum is suitable for radar signal classification under the condition of a low signal-to-noise ratio. 2.1 Bispectrum Theory Bispectrum is a high-order statistic that describes the higher-order statistical properties of a stochastic process. When calculating the bispectrum, the triplet autocorrelation of the stochastic process is calculated firstly, which is also its three-order cumulants. Then, by applying the discrete Fourier transform (DFT) to the three-order cumulants, we get the bispectrum of the stochastic process [6]. Several important characteristics of bispectrum in practical applications are as follows: (1) The bispectrum after DFT of a Gaussian stationary process is zero. Therefore, the bispectrum has a good inhibitory effect on the gaussian noise, which is also the theoretical basis for signal feature extraction and classification by bispectrum under the condition of low SNR. (2) Bispectrum keeps most of the original information of the radar signal. It only loses the linear phase feature of the non-Gaussian signal. For a complex modulation type of signal, there is rich phase information. Therefore, bispectrum features apply to the pattern recognition field where phase information is more important.

Radar Signal Classification Based on Bispectrum Feature and Convolutional

277

2.2 Nonparametric Estimation Method The bispectrum of radar signal can be calculated by non-parametric estimation method: after dividing the signal sequence, the third-order cumulant of the signal is estimated first, and then the bispectrum is obtained by taking the Fourier transform of the cumulant sequence. The specific steps are as follows: (1) The input signal is divided into M segments, each containing K samples. Then Each of these subtracts its mean. (2) For data in segment l: x(l) (0), x(l) (1), · · · , x(l) (K − 1), estimates its third-order cumulant: 1 K2 x(l) (n)x(l) (n + i)x(l) (n + j)l = 1, ...M (1) c(M ) (i, j) = n=−K1 K In this formula, K1 = max(0, −i, −j); K2 = min(K − 1, K − 1 − i, K − 1 − j). (3) Take the average value of the third-order cumulant of all segments as the third-order cumulant estimate of the whole observed data. 1 K c(M ) (i, j) cˆ (i, j) = (2) M =1 M (4) Calculate bispectrum estimation. L L Bˆ IN (ω1 , ω2 ) = cˆ (i, j)ω(i, j)e−j(ω1 i+ω2 j) (3) i=−L

j=−L

In this formula, L < K − 1, ω(i, j) is a two-dimensional lag window function. The bispectrum estimation result of the signal is shown in Fig. 1.

Fig. 1. Three-dimensional and two-dimensional bispectrum of BPSK signal.

3 Signal Classification Based on Convolutional Neural Network 3.1 Convolutional Neural Network The images of the bispectrum of the signals contain rich phase features, and the amplitude values at each point are represented by RGB color depth. By applying a convolutional neural network, the bispectrum images classification problem can be solved effectively. The convolutional neural network has a deep network structure and can extract nonlinear features from complex data which is suitable for image recognition and classification. With the characteristics of local connection and weight sharing, the model parameters are reduced and the structure is simplified significantly.

278

H. Liu et al.

3.2 Experimental Setup To verify the effectiveness of bispectrum for signal classification and the performance of different network structures, experiments are designed in this section and the experimental process is shown in Fig. 2.

Fig. 2. Experimental process.

Data Set Seven types of modulation radar signals are used as the original data, and the amplitude, carrier frequency and SNR conditions of each signal vary within a range. The bispectrum of the signal is obtained by the nonparametric estimation method. By using data augmentation methods like rotation, mirror, random shear, adding noise, chromatic aberration, etc., and reshaping the images to a resolution of 128 × 128, a dataset including radar signal bispectrum images described by RGB is built. Quantities of different radar signal bispectrum images are shown in Table 1. Table 1. Quantity of different types of signals

Some of the images in the dataset are shown in Fig. 3.

Radar Signal Classification Based on Bispectrum Feature and Convolutional

279

Fig. 3. Signal bispectrum in the dataset.

3.3 Experimental Results Two features were used as inputs to the convolutional neural network for comparison: time-domain images and bispectrum images. The classification results of time-domain images and bispectrum images are shown in Table 2. Three different CNN structures were applied for training: simple structure for faster training, LeNet, and AlexNet. The classification result is shown in Table 3. Table 2. Accuracy of input features in different SNR Dataset

10 ~ 2 dB

2 ~ −2 dB

−2 ~ −6 dB

−6 ~ −10 dB

Time-domain

0.8316

0.7273

0.4328

0.3062

Bispectrum

0.9559

0.9309

0.8784

0.7356

Table 3. Accuracy of models in different SNR Model

10 ~ 2 dB

2 ~ −2 dB

−2 ~ −6 dB

−6 ~ −10 dB

Simple

0.9176

0.8931

0.7859

0.6488

LeNet

0.9437

0.9388

0.8528

0.7109

AlexNet

0.9559

0.9309

0.8784

0.7356

As shown in Table 1, while using signal time-domain waveform as the learning input of the CNN directly may have a fair accuracy when SNR is ideal, it results in poor performance of the model in the condition of low SNR. Bispectrum, however, has a better effect on the accuracy of all conditions, it suppressed most of the noise effectively. As shown in Table 2, the overall accuracy of the three models is more than 89% when the SNR condition is better than −2 dB, and with SNR lower than −6 dB, LeNet and AlexNet still have accuracy near 70%, indicating that CNN has a good effect in radar signal bispectrum classification. Among them, duel to its deeper network, AlexNet shows the best performance. When SNR is from −6 dB to −10 dB, the confusion matrix of the AlexNet model is shown in Fig. 4.

280

H. Liu et al.

Fig. 4. Confusion matrix of AlexNet.

As is shown in the above confusion matrix, the classification accuracy of LFM signals and QPSK signals reaches 100% under the condition of very low SNR. The bispectrum feature of LFM signals and QPSK signals are quite different from the other five modulation types, which can also be seen in Fig. 3.

4 Conclusion In this paper, a radar signal classification method based on bispectrum feature and convolutional neural network is proposed. Firstly, divide the received signal sequence, and for each segment calculate its three-order cumulant value. Then take the average of each segment as the whole estimate and apply DFT on it to get a signal bispectrum. Build dataset by bispectrum images after preprocessing and data augmentation and train AlexNet on it. With the application of bispectrum and CNN, the accuracy of signal classification in a low SNR environment is significantly improved. The application scenarios of this method under the condition of low SNR include realizing high altitude platform communications [7], UAV swarm signal reception and control [8], and CNN for Spectrum Sensors [9], etc. The radar signal classification method proposed has a high value in engineering applications.

References 1. Jialu, L., Huaidong, S., Bin, Z.: Radar signal classification based on bayesian optimized supportvector machine. Journal of Physics: Conference Series, 1952 (2021) 2. Jiahuang, S., Jianchong, H., Yongcheng, Z.: Summary of rapid recognition of radar emitter signal. Electro. Info. Warf. Technol. 32(05), 5–10 (2017) 3. Zhe, Z.: Radar Signal Recognition and Parameter Estimation Based on FRFT. Harbin Engineering University (2020) 4. Yutao, H.: Radar Emitter Signal Recognition Based on Deep Learning and ambiguity function. Kunming University of Science and Technology (2020) 5. Junsheng, M., Gong, Y., Zhang, F., Cui, Y., Zheng, F., Jing, X.: Integrated sensing and communication-enabled predictive beamforming with deep learning in vehicular networks. IEEE Commun. Lett. 25(10), 3301–3304 (2021) 6. Chandran, V., Elgar, S.L.: Pattern recognition using invariants defined from higher order spectra: one dimensional inputs. IEEE Trans. Signal Processing 41(1) (1993)

Radar Signal Classification Based on Bispectrum Feature and Convolutional

281

7. Aerial RIS-Assisted High Altitude Platform Communications. In: IEEE Wireless Communications Letters. https://doi.org/10.1109/LWC.2021.3091164 8. 3-D Deployment of UAV Swarm for Massive MIMO Communications. In: IEEE Journal on Selected Areas in Communications. https://doi.org/10.1109/JSAC.2021.3088668 9. Junsheng, M., Youheng, T., Dongliang, X., Fangpei, Z., Xiaojun, J.: CNN and DCGAN for spectrum sensors over rayleigh fading channel. Wirel. Commun. Mob. Comput. 9970600:1– 9970600:12 (2021)

Blockchain and Edge Computing

Blockchain-Based Power Internet of Things Data Access Control Mechanism Xinyan Wang1 , Zheng Jia1 , Qi Wang1 , Dong Li1 , Jing Zhang1 , Xin Chen1 , and Han Yan2(B) 1 State Grid Henan Information and Telecommunication Company, Zhengzhou 450052, China 2 Beijing JingAn YunXin Science and Technology Ltd., Beijing, China

[email protected]

Abstract. In order to solve the problem of low security in the data access of the Power Internet of Things (PIOT), we design a blockchain-based PIOT data access control model. In this model, cloud servers and blockchain nodes cooperate with each other to optimize the existing data access control mechanism, so as to realize the secure storage of power terminal data in the PIOT. To solve the low efficiency of PIOT data access problem, based on the access control model of this article, a blockchain-based PIOT data access control mechanism is proposed. In terms of search efficiency, we use power terminal data indexing mechanism and query mechanism to improve the efficiency of data search. In terms of security, index encryption and decryption algorithms are used to improve the security of data access. In terms of performance analysis indicators, the performance of the proposed algorithm is analyzed from the three dimensions of system solution throughput, index generation time, and search time, which verifies that the proposed algorithm in this paper improves the security and availability of data access control. Keywords: PIOT · Access control · Blockchain · Encryption · Decryption

1 Introduction With the rapid growth of PIOT technology, more and more power terminals process intelligent data collection capabilities [1]. Due to limited processing capacity and small data storage space of the power terminals, the data of the power terminals is generally stored on cloud computing platforms or edge computing nodes. In this context, improving the ability of cloud platforms or edge nodes data access control is a critical factor to ensure the data security of the PIOT. [2] integrates cloud computing technology and data security management system, proposing a data security management mechanism based on cloud computing technology. [3] uses the combination of blockchain technology and general theory to propose a data security multi-party management protocol. [4] adopts verifiable data theory and tree access architecture, proposing a data search and encryption mechanism, which effectively improves data security. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 S. Sun et al. (Eds.): ICSINC 2021, LNEE 895, pp. 285–291, 2022. https://doi.org/10.1007/978-981-19-4775-9_35

286

X. Wang et al.

Through the analysis of blockchain technology and applications, it can be seen that blockchain technology has achieved more application results in data access control [5]. We use blockchain technology to optimize the existing data access control mechanism to realize the secure storage of power terminal data in the PIOT. In addition, to solve the problem of slow data retrieval in the blockchain, a keyword-based data indexing mechanism was proposed, which effectively improves the efficiency of data retrieval.

2 Blockchain-Based PIOT Data Access Control Model Based on the decentralization and anti-tampering characteristics of the blockchain, we construct a blockchain-based PIOT data access control model (Fig. 1).

Fig. 1. Blockchain-based PIOT data access control model

The blockchain-based PIOT data access control model includes four roles: power terminal, cloud server, blockchain, and data user. (1) Power terminal: The power terminal can generate necessary data according to demand, and these data is useful for the reliable operation of power business. (2) Cloud server: The use of cloud server equipment as the convergence node of the power terminal can improve the reliability and availability of the model. (3) Blockchain: Considering that the power terminal belongs to the field of the power company industry, the alliance chain technology is used to construct the blockchain. (4) Data user: The data of the power terminal can be used by different departments of the power company.

3 Data Access Control Mechanism 3.1 Blockchain- Based PIOT Data Access Control Mechanism In order to realize the efficient and safe use of power terminal data, we propose a blockchain-based PIOT data access control mechanism. The mechanism includes the following steps:

Blockchain-Based Power Internet of Things

1. 2. 3. 4. 5. 6. 7. 8. 9.

287

Power terminals generate data and index array; Power terminals send the ciphertext and keyword index array to cloud server; The cloud server encrypts and stores data; Data users register at the consensus node; Consensus nodes generate a shared key for successfully registered data users and save information; Consensus nodes generate a public decryption key for data users who have successfully registered key; Data users apply to consensus nodes for the use of specific data; Ledger nodes use a keyword search mechanism to search for data; The ledger nodes return the data to the data users.

In the step of generating data and index array by the power terminal, the power terminal packs the data that needs to be uploaded to the cloud server. To ensure data security, a shared key is generated through negotiation with the cloud server. To improve the data use efficiency, the data is classified and a key index array is generated. In the step of the power terminal sending the ciphertext and keyword index array to the cloud server, considering the time delay and energy consumption of data transmission, the cloud server is generally constructed using edge computing technology. Considering security, the shared key information between the cloud server and the power terminal is unknown to the edge node. In step of encrypting and storing the data by the cloud server, to facilitate the use of data, save data storage space. The cloud server uses a new encryption mechanism and data index mechanism to encrypt and store data. During the data user’s registration step with the consensus node, after the data user requests the shared key from the consensus node, it generates a key pair with its user ID and sends the public key to the consensus node. If for some reasons, the current data user is not using the power terminal data, the current user information needs to be logged out according to the user logout mechanism. In the step that the consensus node generates a shared key for successfully registered data users and saves information, according to the data user’s request, the shared key is generated for the data user, and after the data user has successfully registered, its identity information and shared secret key information are saved. In the step of generating the decryption key of data for successfully registered data users by the consensus node, to ensure data security, the consensus node generates a public decryption key for each user based on user attribute information, user identity information, etc. And the key will be saved in the ledger node. In the step of data users applying to the consensus node to use specific data, the data users send their own identity information and key features of the data that need to be accessed to the consensus node. The consensus node judges user identity legitimacy through the consensus mechanism. After passing the verification, the consensus node sends the data request information to the ledger node. In the step of searching data for the ledger node using the keyword search mechanism, the ledger node uses the keyword search mechanism to obtain data from the cloud server, encrypt it and return it to the data user. According to the data characteristics proposed by users, the ledger node uses a keyword search mechanism (see the content of “power

288

X. Wang et al.

terminal security mechanisms for data query”) to search for details. The cloud server returns the data to the ledger node through search. In the step of the ledger node returning data to data users, data users use decryption key to decrypt. 3.2 Power Terminal Security Mechanisms for Data Query To ensure the security of power terminal data, security query mechanism of power terminal data proposed is shown in Table 1. In the data generating keyword dictionary step, the cloud server generates a keyword dictionary  based on the attributes and characteristics of the data. In the step of generating a security index for the data by the cloud server, an index phrase Fωτ = {id (Fω1 τ ),id (Fω2 τ ),...,id (Fωn τ )} is created for the keywords ωτ ∈ . In the step of encrypting the security index, cloud servers randomly select a number ϕτ ∈ ZP as a random number, and calculate the keyword index using formula (1). Among it, g is the generator of bilinear mapping e: G × G → GT , and a ∈ Zp is a random element. x indicates the number of the owner of the current data on the cloud server. H (∗) indicates the hash function that can be used. This article uses the hash function in MD5 for implementation. I1 = g ϕτ x , I2 = g a(ϕτ +π ) · g ϕj H (ωτ ) , I3 = g π

(1)

Based on this, the key index sequence shown in formula (2) can be generated for the data key dictionary. I = {Iωτ = I1 ,I2 ,I3 }

(2)

In the step of using the index field, the data user uses SKUID and keyword W to generate a query trapdoor TDW , selects a random number a ∈ Zp , and uses the formula (3) to calculate the query keyword. T1 = g a(a+h(w)) , T2 = g aaˆx , T3 = g aˆx

(3)

Based on the result of formula (3), a query trapdoor can be generated, as shown in formula (4). TDw = {T1 ,T2 ,T3 }

(4)

The cloud server searches for corresponding keywords in the database according to the query trapdoor. Use formula (5) to find data that matches the query trapdoor. If found, return the result to the user. e(I1 ,T1 )e(I3 ,T3 ) = e(I2 ,T2 )

(5)

Blockchain-Based Power Internet of Things

289

Table 1. Security query mechanism of power terminal data 1. The cloud server generates a keyword dictionary  from the data; 2. For keywords ωτ ∈ , the cloud server creates index phrases Fωτ = {id (Fω1 τ ), id (Fω2 τ ), ..., id (Fωn τ )}; 3. The cloud server encrypts the security index; 4. The use of index fields: (1) The data user uses SKUID and keyword W to generate a query trapdoor TDW , selects a random number a ∈ Zp , uses formula (3) to calculate the query keyword, and uses formula (4) to generate a query trapdoor; (2) The cloud server searches for the corresponding keyword in the database according to the query trapdoor.

4 Performance Analysis The efficiency of data search has an important influence on the access and use of data. To verify performance of data search under algorithm of this paper, the blockchain-based PIOT data access control mechanism (PIoTDACMoBC) is compared with a traditional data access mechanism (Attribute-based data access control mechanism, DACMoA). First, compare the throughput of the system schemes under the two access control mechanisms. The experimental results are shown in Fig. 2. When data volume of the data access request increases from 400 bytes to 1200 bytes, the value of throughput does not change much. This shows that the data access throughput and data capacity gap between the two algorithms is small. Compared with traditional algorithms, the data encryption mechanism under this algorithm has less impact on throughput.

Fig. 2. Comparison of system solution throughput

From the analysis of throughput experiments, we can see that the search speed of data is of great significance to data access. The following compares the algorithm in this paper with the data search mechanism MKSSoPA in the literature [6]. The following compares the two mechanisms from the two dimensions of index generation and search phase. The result of generating the index is shown in Fig. 3. Compared with the comparison mechanism, the time cost of the index generated by

290

X. Wang et al.

PIoTDACMoBC has a small increase. It shows that PIoTDACMoBC is more efficient in generating indexes.

Fig. 3. Comparison of index generation time

The result of the search phase is shown in Fig. 4. We find as number of data attributes increases, the search time under both mechanisms increases rapidly. We think this is caused by the increase in data attributes, which increases the difficulty of data search. In the comparison of the search time of PIoTDACMoBC and MKSSoPA, the search time under PIoTDACMoBC is shorter. So PIoTDACMoBC can improve the efficiency of data search.

Fig. 4. Comparison of search duration

5 Summary To solve the data access security problem of the PIOT, we design a blockchain-based PIOT data access control model. Based on this model, a blockchain-based PIOT data access control mechanism is designed. We verify that PIoTDACMoBC improves the security and availability of data access control from three dimensions of system solution throughput, index generation time, and search time. Efficient storage of power terminal data is a prerequisite for data application.

Blockchain-Based Power Internet of Things

291

Acknowledgment. This work was supported by the science and technology project of State Grid Corporation of Henan Province “Research on integrating trusted computing and blockchain in power information security protection technology” (5217Q0210003).

References 1. Wang, Z., Liu, R., Liu, Q., Thompson, J.S., Kadoch, M.: Energy-efficient data collection and device positioning in UAV-assisted IoT. IEEE Internet Things J. 7(2), 1122–1139 (2020). https://doi.org/10.1109/JIOT.2019.2952364 2. Huang, L., Chen, L., Ma, M.: Big data security and privacy protection based on cloud computing. Electron. Technol. Softw. Eng. 24, 249–250 (2020) 3. Liu, F., Yang, J., Li, Z.B., Qi, J.Y.: A secure multi-party computing protocol for general data privacy protection based on blockchain. Comput. Res. Dev. 58(02), 281–290 (2021) 4. Zheng, Q., Xu, S., Ateniese, G.: Vabks: verifiable attribute-based keyword search over outsourced encrypted data. In: IEEE INFOCOM 2014 - IEEE Conference on Computer Communications, pp. 522–530 (2014) 5. Wang, X.L., Jiang, X.Z., Li, Y.: Data access control and sharing model using blockchain. J. Softw. 30(06) (2019) 6. Miao, Y., Ma, J., Liu, X., Li, X., Liu, Z., Li, H.: Practical attribute-based multi-keyword search scheme in mobile crowdsourcing. IEEE Internet Things J. 5(4), 3008–3018 (2017)

Consortium Blockchain Based Anonymous and Trusted Authentication Mechanism for IoT Tianhong Su1(B) , Wenjing Li2 , Di Liu2 , Shaoyong Guo1 , and Linna Ruan1 1 Beijing University of Posts and Telecommunications, Beijing 100876, China

[email protected] 2 State Grid Information and Communication Industry Group Co., Ltd., Beijing 100052, China

Abstract. IoT devices always work in an open network environment, and their security has been widely concerned. Authentication is the most important part of IoT security. The traditional centralized authentication scheme has the defects of prolonging the authentication time and being difficult to be supervised. The blockchain technology with decentralized characteristics provides a new solution for the distributed IoT system. This paper proposes an anonymous two-way authentication mechanism for IoT terminals based on the Consortium blockchain. An anonymous communication authentication code is designed during the authentication process to ensure the security of the terminal’s real public key. The decentralization and non-tampering characteristics of the blockchain are used to ensure the security of anonymous keys. The randomness of the key is guaranteed through the negotiation of multiple sets of random numbers, and the confidentiality and integrity of the communication data are guaranteed through the 256-bit elliptic curve algorithm and digital signature algorithm. Compared with the existing mobile IoT authentication schemes through experiments, the result shows that this scheme has better security and communication performance. Keywords: Internet of Things · Blockchain · Anonymous authentication

1 Introduction With the rapid development of 5G, Bluetooth and other communication technologies and the Internet of Things, more and more devices have joined the network communication, which has brought a great impact to the existing network stability [1]. At the same time, due to the particularity of IoT services, IoT terminals need to send private information such as location information, device usage habits, etc. when providing services. Malicious terminals can cause privacy disclosure of users by eavesdropping, replaying and modifying the transmitted information in public channels [2]. In the Internet of Vehicles scene, an experienced attacker can even track the owner of the vehicle through the vehicle owner’s trajectory, which brings serious security risks and seriously affects the popularization of the entire Internet of Things [3]. Authentication technology is the most important part of the security mechanism of the Internet of Things. A reasonable identity authentication technology can not only © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 S. Sun et al. (Eds.): ICSINC 2021, LNEE 895, pp. 292–302, 2022. https://doi.org/10.1007/978-981-19-4775-9_36

Consortium Blockchain

293

check the legitimacy of the access terminal but also effectively ensure the security of the internal node communication in IoT System. In the traditional public key Infrastructure (PKI) schema, the certificate authority (CA) will issue a certificate to the node before it joins the system. When the terminal communicates, it will use its private key to generate a signature for each message and attach it to the message. The receiver can verify whether the signature is legal through the sender’s public key [4]. Vasudev designed a secure message authentication protocol for communication in smart cities, which can ensure secure identity verification, but this scheme has the hidden danger of key leakage [5]. Ding et al. proposed a method to realize access control by storing Internet of Things attributes through blockchain, which avoids illegal data tampering through attribute storage [6]. But this schema is difficult to adapt to high concurrency scenarios. Su designed the IoV authentication system based on blockchain, and solved the problem of single point failure of traditional IoV by managing keys with blockchain. However, because the blockchain uses an authentication mechanism based on proof of work, it leads to Higher latency [7]. ARS combined the blockchain and IoV to design a blockchain-based IoV identity authentication system, which solved the problem of vehicle node and roadside unit authentication, but this scheme lacks an efficient consensus mechanism [8]. To sum up, because of the defects of the existing schemes, this article combines the characteristics of the IoT system and proposes a two-way anonymous authentication scheme between IoT terminals based on the consortium blockchain. This scheme combines the blockchain with smart contracts and uses anonymous keys to design communication authentication codes during the authentication process, which protects the privacy of real public keys and realizes efficient cross-domain anonymous authentication between IoT terminals. At the same time, the scheme uses symmetric cryptography, public-key cryptography, and digital signature technology in cryptography theory to realize the confidentiality and integrity of communication information, guarantee the security of terminal privacy, and prevent malicious terminals from stealing private information and bring security risks.

2 System Architecture The terminal authentication architecture mainly composed of the following parts: Trusted Agency (TA): TA is responsible for system initialization, registration of trusted service stations, registration of terminals, and deployment of blockchain smart contracts. TA is set to have certain communication and computing capabilities and will not leak user privacy. Law enforcement Agency (LEA): LEA responsible for reviewing TA’s information. The transaction information reviewed includes such as terminal public key registration requests and public key revocation requests. When there is a terminal message conflict, LEA can trace the source based on the blockchain, and only the information that has passed the review can be recognized as valid information. Our scheme assumes that other institutions cannot forge LEA signatures. Trusted Service Station: A trusted service station is an infrastructure for managing IoT terminals within a certain range. It consists of a signal sending device and a

294

T. Su et al.

high-performance server. The trusted service station can communicate wirelessly with IoT terminals within a certain range. It can receive and process the terminal’s instant information and provide identity authentication and other personalized services. Terminal: The terminal in the Internet of Things is equipped with tamper-proof communication equipment. The terminal can communicate with other terminals and trusted service stations through communication equipment. Smart contract: Smart contract is a trusted distributed application and the business logic of blockchain application. he smart contract in this system mainly automatically feeds back the public key query request of the terminal and determines the authentication result. The smart contract is deployed in the blockchain to ensure the accuracy of the calculation results. The architecture of the IoT authentication system based on the consortium chain is shown in Fig. 1.

Fig. 1. System architecture.

2.1 System Initialization In the system initialization phase, TA will create blockchain accounts for each TSS, generate smart contracts and deploy them to the blockchain network. Smart contracts allow any participant in the Internet of Things to supervise the blockchain and TA’s operations on the public key and promptly pursue responsibility in the event of a dispute, weakening the rights of the central agency and improving the fairness of the system. The system realizes the authentication process based on the public key cryptosystem and realizes the secure communication through the symmetric key. In the initialization phase, TA needs to determine the system cryptographic parameter L, where the system determines the elliptic curve E in the finite field through the parameter L. The final definition of the elliptic curve is shown in Eq. (1). TA will generate TSS identity information for each TSS, including TSS unique identifier, TSS public key, TSS private key, and publish the TSS unique identifier and TSS public key to the blockchain for identity

Consortium Blockchain

295

authentication between the IoT terminal and TSS. {(x,y) ∈ R2 |y2 = x3 + ax + b, 4a3 + 27b2 = 0} ∪ {0}

(1)

2.2 Authentication Authentication Between Terminal and Trusted Service Station. The system sets the authentication message to be valid as t, and the sender needs to attach the time information Ts every time a message is sent. Only when the difference between the time Ts of receiving the message and the current time Tc is less thant, the receiver determines that the current message is valid, otherwise the current message is discarded. Set the current trusted service station as T1 and the authentication terminal as V1. 1. T1 → V1: { RID ,RP ,Merk ,Ts} . The trusted service station broadcasts a beacon message within its communication range, including its registration ID(RID ), public key(RP ), public key identification(Merk ), public key identification is the root value of part of the Merkle  tree in the blockchain, and timestamp information (Ts). 2. V1 → T1: EncRp Msg1 ➀ When a new terminal enters the jurisdiction of T1, it will receive a T1 broadcast message. After V1 receives the message, it first verifies whether the timestamp information Ts meets the system requirements. When Tc − Ts < t, the terminal determines that the message is valid, otherwise it discards the current message. ➁ The terminal V1 verifies whether the T1’s public key exists according to the T1’s key identification, and if the public key exists, it determines that the broadcast message is valid. ➂ The terminal V1 generates a random number A, and packs A, RID, terminal registration ID (VID ), timestamp (Ts) information to generate an anonymous key certificate. V1 generates an anonymous public key (Pk), based on the proof and the elliptic curve encryption algorithm, and generates the corresponding private key (Sk) based on this public key. ➃ V1 signs the anonymous public key Pk with its own private key, encrypts the VID , timestamp information, and signature information with T1’s public key, and sends the ciphertext to T1. The message sent is Msg1 = (VID ,Ts,Pk ,SignVs (Pk )). 3. T1 → V1: EncVp (Msg2 ) ➀ After T1 receives the message from V1, T1 first verifies whether the timestamp information Ts meets the system requirements, and uses its own private key to decrypt the message after successful verification. ➁ After decrypting the message, T1 extracts the terminal ID and calls the SearchKeysByID method in the smart contract to query the public key of V1. If the query is successful and the public key has not been revoked, T1 records the public key information, otherwise discards the current data packet and terminates the authentication process. ➂ T1 uses the queried public key to verify the authenticity of the anonymous public key signature(SignVs (Pk )). If the verification is successful, it extracts the anonymous public key of the terminal V1, combines the RID and Pk to generate a communication authentication code (CAC) and stores it in the blockchain. ➃ T1 generates random number B, uses B, Ts, and CAC to generate a session key (SeRV ), the session key is a symmetric key. ➄ T1 uses the generated session key to generate the digest value of the communication authentication code through the HMAC algorithm, and sends

296

T. Su et al.

the random number B and the digest value to V1 after being signed by the private key. The final message sent is Msg2 = { B,SignRs (HMACSeRV (B,CAC)),Ts} . The structure of the communication authentication code is shown in Fig. 2

Fig. 2. Communication authentication code.

4. V1 → T1:{MsgHello } ➀ After V1 receives the message from T1, V1 uses the private key to decrypt Msg2 and verify whether the time stamp information is valid. ➁V1 uses RID , VID and Pk to generate a communication authentication code, and uses a random number B, a timestamp Ts, and a session key (SeRV ’) according to the same algorithm. ➂ V1 uses the same algorithm to generate CAC and the digest value of the random number B. V1 uses T1’s public key, CAC and the digest value of B to verify that the signature is correct. If it is correct, it proves that the SeRV ’ is valid. he terminal discards the data packet and notifies the T1 negotiation failure. ➃ V1 sends a message (MsgHello ) to the trusted service station T1 to prove that the authentication is complete. MsgHello is responsible for proving the success of the negotiation process. Terminal and Terminal Authentication. The terminal in this scheme requires both parties to complete two-way authentication and negotiate the session key before communicating. Only the terminal with a legal communication authentication code after authentication with TSS is recognized as a legal terminal, and legal terminals can communicate securely with each other after two-way authentication. Set the authentication terminal as V1 and V2, and the TSS participating in authentication is T1. V1 requests authentication with V2. 1. V1 → V2 : {EncV2p (CACV 1 ,SignSkV 1 (CACV 1 ),Ts)}.V1 signs its own communication authentication code with its own anonymous private key, encrypts the signature result with V2’s public key, and sends it to V2. 2. V2 → T1:{EncRp (M 1)}. V2 generates authentication information in the same format as V1 and adds it to the request information of V1 to form both parties’ authentication request information M1. The authentication request information is encrypted using T1’s public key and sent the ciphertext to T1. The final message(M1) sent to T1 is {SignSkV 1 (CACV 1 ),CACV 2 ,SignSkV 2 (CACV 2 ),Ts}. 3. T1 → V1: {EncSeRV1 (M 2)}. ➀ After T1 receives the message from V2, T1 uses the private key to decrypt M1 and verify whether the time stamp information is valid. ➁ After successful decryption, the smart contract for key query is called to query whether Pkv1 and Pkv2 exist. ➂ T1 uses Pkv1 and Pkv2 to verify whether the signatures of both parties are legal. If the signatures are legal, it returns Success, otherwise it returns False. ➃ T1 extracts the communication authentication code of V2, uses the

Consortium Blockchain

4.

5.

6.

7.

297

private key to sign CACV2 and timestamp, and sends it to V1 using the previously negotiated session key of V1. M 2 = {CACV 2 ,SignRs (CACV 2 ,Ts),Ts}. T1 → V2 : {EncSeRV2 (M 3)}.T1 extracts the communication authentication code of V1(CACV 1 ), uses its private key to sign CACV 1 and timestamp, encrypt it with the session key and send it to V2. M 3 = {SignRs (CACV 1 ,Ts),CACV 1 ,Ts}. V1 → V2: {EncPkV 2 (M 4)}. ➀ After V1 receives the message from T1, it uses the session key to decrypt the message, and verifies whether the message has expired through the timestamp. ➁ V1 verifies whether the signature of T1 is valid, and discards the message if the signature verification fails. ➂ V1 extracts the anonymous public key of V2 and generates a random number A, V1 uses Pkv2 , RID , A and timestamp to generate the Initial Key (Ik). After the generation is successful, v1 uses its own anonymous private key to sign Ik.➃V1 encapsulates the key agreement information M4, encrypts it with the anonymous public key of V2, and sends the ciphertext to V2. The final message sent is M 4 = {SignskV 1 (Ik,Ts),Ik,Ts}. V2 → V1: {EncPkV 1 (M 5)}. Before verifying the message of V1, V2 verifies the correctness of the signature of the trusted service station T1 according to the same steps as V1. If the signature is valid, it proves that the identity of V1 is legal. ➀ After V2 verifies the validity of V1, it first verifies the time stamp information to prove the validity of the message. ➁ V2 extracts Ik and uses V1’s anonymous public key to verify the validity of V1’s signature. ➂ V2 generates a random number B, combines Ik, Pkv1 and timestamp (Ts’) to generate a session key for both parties. The session key is a symmetric key.V2 uses the SeV2V to encrypt the random number B, and signs the ciphertext with its own anonymous private key. ➃ V2 packs the random number B and the ciphertext of B. V2 encrypt the packed result with the anonymous public key of V1 and send it to V1.The final message sent is M 5 = {Signskv2 (B),Ts ,EncSeV 2V (B)}. V1 → V2: {EncSeV2V (Msg)}. ➀ After V1 verifies the time stamp information, it uses the anonymous private key to decrypt the message sent by V2. ➁ V1 extracts the random number B,Ts’, and uses V2’s public key to verify the signature. ➂ V1 uses random numbers B, Ik, Ts’ to generate the session key using the same method as V2, and uses it to decrypt the received ciphertext information. If it succeeds, it proves that the session key negotiation process is complete. ➃ V1 and V2 can use the session key to communicate with each other. The key agreement process is complete.

3 Security Analysis Anti-eavesdropping attacks: Eavesdropping attacks monitor the communication channels of both parties to steal unencrypted sensitive information from both parties. The anonymous authentication scheme designed in this paper needs to perform two-way authentication with TSS and the target terminal and negotiate the session key before the terminal communicates. All private information in the channel needs to be encrypted before transmission, so there is no risk of private information leakage. Anti-replay attack: The anonymous authentication scheme designed in this paper adds timestamp information to each communication. The communication parties first verify whether the difference between the synchronization clock information and the

298

T. Su et al.

message timestamp is within the allowable range of the system after receiving the message, and only the messages within the legitimate time range are considered as legitimate messages. So our schema can resist the replay attacks. Anti-witch attack: A witch attack refers to a malicious terminal disguising as a legitimate terminal to release distorted messages. In the IoV system, it may cause the recipient to make wrong driving decisions and cause major traffic accidents. The security of the ECDSA algorithm used in this paper is based on the discrete logarithm problem on the elliptic curve finite group and the private key cannot be derived from the public key. Therefore, no matter in the authentication process between the terminal and the trusted service station or the authentication process between the terminal and the terminal, the malicious terminal cannot forge the legal identity. Public key anonymity: This scheme uses the communication authentication code instead of the traditional public key for communication. The true correspondence between CAC and the terminal is only stored in the trusted TSS. At the same time, the TSS will periodically check the communication authentication code and delete the expired communication authentication code through the smart contract, so that the malicious terminal cannot determine the correspondence between the anonymous public key and the terminal identity. Data security: only trusted blockchain nodes have access to the information stored on the blockchain, and unauthorized nodes cannot access terminal information in the blockchain. The chain uses hash pointer to connect various transactions to ensure that the data on the chain cannot be maliciously tampered and guarantee the integrity of the terminal information on the chain.

4 Performance Analysis This article uses Hyperledger Fabric 1.4.3 to deploy blockchain nodes and uses the default PBFT algorithm as the system consensus algorithm. This article mainly uses ECDSA signature algorithm and 256-bit ECC algorithm based on the SECP256K1 curve. The symmetric encryption algorithm used in this scheme is the 256-bit AES algorithm, and the digest algorithm is the SHA256. The testing tools used in this paper are shown in Table 1. Table 1. Test tool list. Tool

Function

Ubuntu 18.04

Experimental basic system

HyperledgerFabric1.4.3

Blockchain test environment

Docker

Hyperledger operating environment

Golang Java

Programming language

Consortium Blockchain

299

The symbols used in this article and their meanings are shown in Table 2. Table 2. Symbol table. Symbol

Function

RID

TSS unique identifier

Rp

TSS public key

Rs

TSS private key

Vp

Terminal public key

Vs

Terminal private key

Pk

Anonymous public key

Sk

Anonymous private key

CAC

Communication authentication code

HMACx

Hash authentication code

Encx

Encryption algorithm with key x

Signx

Signature algorithm with key x

Verx

Signature verification algorithm

Sex

Session key

4.1 Storage Cost According to the IEEE1609.2 standard, the certificate size is defined as 126 bytes, the terminal communication message load is set to 67 bytes, and the ECDSA algorithm signature length is 64 bytes. The storage cost calculated in this article mainly calculates the cost at the terminal. In addition to storing the registered public key and private key information, the IoT terminal in our scheme also needs to store the communication authentication code and session key, and set and authenticate the terminal. The number of nodes is Nv, the cost is 64 bytes for the private key, 64 bytes for the public key, 64 + 3 + 4 = 71 bytes for its own communication authentication code, and 64 * Nv bytes for the session key of the authentication terminal, totaling 199 + 64 * Nv bytes. The storage cost of traditional PKI scheme is related to the number of certificates (Nc), The storage consumption of SA-KMP mainly refers to the number of roadside units (Nr) in the system [9]. The cost for He’s scheme is 96 + 64 * Nv [10]. The storage cost of each scheme is shown in Table 3:

300

T. Su et al. Table 3. Storage cost. Scheme

Storage cost (Byte)

PKI

126 * Nc

TAN

37 * Nr

HE

128 + 64 * Nv

This paper

199 + 64 * Nv

4.2 Communication Cost The traditional PKI scheme needs to add a certificate when transmitting a message. The communication load is 126 + 67 + 64 = 257 bytes. Zhang’s scheme also needs to transmit a certificate, and the communication load is 257 bytes [11]. SA-KMP does not require a certificate to be attached to the message, and its communication load is 67 + 64 = 131 bytes. Our scheme proposed is based on the blockchain to manage the public key of the IoT terminal. The communication parties do not need to pass certificate authentication, and the communication load is 131 bytes. 4.3 Authentication Cost This article separately introduces the authentication process between the terminal and the trusted service station and the terminal and the terminal. This article takes the two delays and the node authentication delays. In this paper, 5 blockchain nodes are deployed in the blockchain to test the operating environment of IoT nodes. The delay of the scheme is shown in Fig. 3.

Fig. 3. Time delay of this scheme.

Considering that this scheme can be used as a mobile terminal and a cross-domain authentication scheme, the experiment mainly compares the current authentication scheme for the IoV system. The traditional PKI scheme needs to transmit the certificate

Consortium Blockchain

301

and check CRL during authentication, which brings a high delay. This article uses smart contracts to automatically trigger the authentication process, which reduces the processing delay. In our scheme, the symmetric session key is used for encryption when the terminal communicates. Most of the existing schemes perform the key agreement process after authentication, while this scheme completes the session key negotiation during the authentication process, which reduces the time delay and improves the efficiency. SA-KMP uses the pre-authentication method to reduce the authentication delay, but the use of multiple signatures and asymmetric encryption algorithms in the key negotiation phase leads to a higher negotiation delay [10]. Zhang’s scheme requires the participation of CA during authentication. it is easy to cause a single point of failure when the amount of concurrency is too high [11]. The communication cost for each scheme and the time delay for each scheme to complete the authentication and key agreement process is shown in Fig. 4. After simulation experiments under the same conditions, the results show that compared with the existing solutions, this solution reduces the communication cost and authentication delay. So compared with the existing scheme, our scheme has higher certification efficiency.

Fig. 4. Communication cost and delay

5 Conclusion Identity authentication is a key technology for the privacy protection of the Internet of Things. This paper proposes a blockchain-based anonymous authentication scheme for IoT terminals, when the IoT terminal enters the TSS communication range, it generates an anonymous key to communicate with other nodes instead of the real public key associated with identity. The anonymous key is recorded in the blockchain in the form of a communication authentication code and updated regularly, which improves the security of the real public key and the efficiency of cross-domain authentication. Before IoT terminals communicate with each other, they need to complete two-way authentication and

302

T. Su et al.

temporary session key agreement process through a communication authentication code. The authentication process between IoT terminals is automatically triggered by smart contracts, and the authentication result is permanently recorded ensuring the security and traceability of the results. Malicious terminals cannot interfere with the authentication process. The scheme described in this paper solves the disadvantages of high pressure on the central server and the difficulty of monitoring the data of the central server in IoT system through the blockchain. The experimental results on the Fabric system shows that the scheme has good performance in terms of communication efficiency and security. However, the communication authentication code of this scheme will bring a small amount of storage pressure to the IoT terminal. In the future, we will continue to optimize the consensus algorithm used by the system to improve system efficiency. Acknowledgment. This work was supported by National Key R&D Program of China (2020YFB2104503).

References 1. Feng, Q., He, D., Zeadally, S., Liang, K.: BPAS: blockchain-assisted privacy-preserving authentication system for vehicular ad-hoc networks. IEEE Trans. Ind. Inf. 16, 4146–4155 (2019). https://doi.org/10.1109/TII.2019.2948053 2. Yang, H., Liang, Y., Yuan, J., Yao, Q., Ao, Y., Zhang, J.: Distributed blockchain-based trusted multi-domain collaboration for mobile edge computing in 5G and beyond. IEEE Trans. Industr. Inf. 16, 7094–7104 (2020). https://doi.org/10.1109/TII.2020.2964563 3. Tan, H., Chung, I.: Secure authentication and key management with blockchain in VANETs. IEEE Access 8, 2482–2498 (2020). https://doi.org/10.1109/ACCESS.2019.2962387 4. IEEE trial-use standard for wireless access in vehicular environments - security services for applications and management messages. In: IEEE Std 1609.2-2006, vol., no., pp. 0_1-105 (2006). https://doi.org/10.1109/IEEESTD.2006.243731 5. Vasudev, H., Das, D., Vasilakos, A.V.: Secure message propagation protocols for IoVs communication components. Comput. Electr. Eng. 82(1), 106555 (2020). https://doi.org/10.1016/ j.compeleceng.2020.106555 6. Ding, S., Cao, J., Li, C.: A novel attribute-based access control scheme using blockchain for IoT. IEEE Access 7, 38431–38441 (2019) 7. Tianhong, S., Shao, S., Guo, S., Lei, M.: Blockchain-based internet of vehicles privacy protection system. Wirel. Commun. Mobile Comput. 2020, 1–10 (2020). https://doi.org/10.1155/ 2020/8870438 8. Shrestha, R., Bajracharya, R., Shrestha, A.P., Nam, S.Y.: A new type of blockchain for secure message exchange in VANET. Digital Commun. Netw. 6(2), 177–186 (2020) 9. Tan, H., Ma, M., Labiod, H., Boudguiga, A., Zhang, J., Chong, P.H.J.: A secure and authenticated key management protocol (SA-KMP) for Vehicular networks. IEEE Trans. Veh. Technol. 65(12), 9570–9584 (2016). https://doi.org/10.1109/TVT.2016.2621354 10. He, W.: Research on Key Management Mechanism of Internet of Vehicles Based on Blockchain Technology. Xidian University (2019) 11. Zhang, J., Li, F., Li, R., Li, Y., Song, J., Zhou, Q.: Research on identity authentication based on elliptic curve encryption algorithm in V2X communication. Automot. Eng. 42(1), 27–32 (2020). https://doi.org/10.19562/j.chinasae.qcgc.2020.01.004

Domain Name Management Architecture Based on Blockchain Zhenjiang Ma(B) , Feng Qi, and Wenjing Li Beijing University of Posts and Telecommunications, Beijing, China [email protected]

Abstract. The centralization problem embodied in the domain name system (DNS) is getting more and more serious. It is not only vulnerable to network attacks but also suffered from abuse of authorities. Blockchain as an effective technical means which could solve decentralization problems brings light to the domain name system. We propose a domain name management architecture based on multi-blockchain. The architecture we proposed uses a relay chain to achieve the management of different top-level domain names. Compared with the singlechain system, the scalability and compatibility of decentralized DNS are improved. Besides, an off-chain storage method is designed for improving the throughput. Then we give the processes of domain name registration, modification, and resolution. Finally, the analysis and conclusions are given. Keywords: Domain name system · Blockchain · Relay chain · Domain name management

1 Introduction The domain name system maps domain names that are easy for a human to remember to IP addresses. The DNS is a vital infrastructure for all network applications. The centralized domain name system has been running well for decades, it has the advantages of high efficiency, good availability, and strong scalability. However, the centralization of the system is reflected in management and technical implementation, which leads to its weak robustness, poor security, and imbalanced power [1, 2]. The centralization of the DNS is embodied in three aspects. They are naming centralization, data-publishing centralization, and rolution centralization. The first two are involved in the management and the last one is involved in technology. Firstly, naming centralization means that a unified domain namespace is managed by The Internet Corporation for Assigned Names and Numbers (ICANN). Secondly, the data-publishing centralization means that domain name data is released by authoritative third-party organizations, so users do not have the right to dispose of digital assets (domain names). Thirdly, resolution centralization means that when the DNS is working, each domain name query begins from the root server which is vulnerable to a single failure. This work was supported by grants from the National Key Research Program Projects (No. 2018YFB1800704). © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 S. Sun et al. (Eds.): ICSINC 2021, LNEE 895, pp. 303–309, 2022. https://doi.org/10.1007/978-981-19-4775-9_37

304

Z. Ma et al.

In order to solve the problems of the invalid and untrustworthy resolution, Domain Name System Security Extensions (DNSSEC) is proposed [3], each domain uploads its own delegation signer (DS) record to the parent and the DS is signed by the parent, then a chain of trust is formed by signed DS record from root servers to other authority servers. There are drawbacks to DNSSEC: firstly, DNSSEC is based on asymmetric encryption which also introduces additional overhead during the encryption process. The computational overhead caused by DNSSEC message encryption will increase the workload of authority servers. In addition, the coverage rate of DNSSEC is low [4], unable to form the closure of the trust chain from the beginning to the end. Therefore, although DNSSEC is proposed, it cannot effectively solve the problem of untrustworthy resolution in the DNS. Combining blockchain technology, we propose a new domain name management architecture, which uses the characteristics of multi-party participation and joint management of the blockchain to evolve the domain name system from a single-rooted tree-like structure into a multi-rooted net-like structure to realize decentralization of domain name management.

2 Background The design of the DNS improves usability and efficiency. However, the system is vulnerable to DDoS attacks due to centralization, and there are also other security problems such as domain name pollution and hijacking. The composition and operation of the traditional DNS are as follows (Fig. 1).

Fig. 1. Structure of domain name system

The blockchain originated from “Bitcoin: A Peer-to-Peer Payment System” published by Satoshi Nakamoto in 2008. It uses a blockchain structure to connect the front and rear blocks in series. Based on the irreversibility of the hash algorithm, it ensures credibility and resists tampering with the ledger data. Then the equal rights of participants are realized through the consensus algorithm. Blockchain uses consensus algorithms and cryptography to establish a trust system between unfamiliar users without an authoritative third party.

Domain Name Management Architecture

305

At present, there has been some progress in the research on the decentralization of the DNS. The decentralized DNS running on public blockchain includes Namecoin, Blockstack, Emercoin, Bitforest, and Ethereum’s ENS. Among them, Namecoin is vulnerable to tampering due to its low computing power [5]. Blockstack proposes a four-layer architecture and is mounted on Bitcoin which guarantees the security of the system but also leads to slow read and write speed [6]. Emercoin and Bitforest face the same problem of low throughput. ENS is mounted on Ethereum, when the transaction is intensive, it will cause a lack of computing resources and then reduce the availability and usability of the system [7]. However, the systems mentioned above split the domain namespace and only support the resolution of specific top-level domains. Also, the resolution of specific top-level domain names requires new plug-ins or tools, and the split domain namespace reduces the universality and usability. Some scholars have also proposed solutions compatible with the DNS. Wang et al. [8] proposed Blockzone, which uses an improved Practical Byzantine Fault Tolerance (PBFT) consensus algorithm to provide domain name resolution service. Liu et al. [9] proposed FI-DNS which replaces the single root with root alliance and completed storage of the domain name below the root by blockchain, but the domain name resolution tree is forced transferring into two layers which decreases the efficiency. He et al. [10] proposed TD-Root which uses the consortium blockchain to construct a root zone management solution that ensures strong consistency and security of root zone data. The above research results did not consider the decentralization of domain name publishing, also the resolution is running on a single-chain with low throughput. Polkadot [11] is a scalable multi-chain system. Polkadot provides a relay chain, on which there can be a large number of verifiable and globally dependent dynamic data structures, which we call para-chains. This feature can be used that we put domain name resolution on para-chains to improve flexibility and efficiency.

3 Domain Name Management Architecture Based on Relay Chain First of all, a unified domain namespace is maintained in our architecture. The relay chain is used to interconnect between top level domains (TLD), providing cross-chain transaction means and ensuring the security of the blockchain. Para-chains provide second-level domain name management and resolution services. Finally, trusted offchain storage of domain names is designed. Para-chains could improve the capability of domain name services. Both the relay chain and the para-chain are deployed on the consortium blockchain. 3.1 Overall Architecture The overall architecture designed in this paper is shown in Fig. 2. Compared with the traditional domain name system, the blockchain-based domain name system has the advantages of higher security and equality of all parties, especially for country code top-level domains.

306

Z. Ma et al.

Fig. 2. Blockchain-based domain name management architecture diagram

The central relay chain maintains root domain zone files and para-chains maintain top-level domain zone files. Each para-chain that is directly connected to the relay chain manages a top-level domain ledger. The architecture supports multi-level access. The first-level para-chain can be used as the second-level relay chain. Four roles are involved, as shown in the right of Fig. 2. They are collectors, observers, nominators, and validators. The collectors are groups that help validators create effective para-chain blocks. They are responsible for packaging new blocks and executing transactions (similar to miners). Then they submit a new block to one or more validators. The observers are not directly related to the block packaging process. They exist at a supervisory level and punish validators who perform illegal actions. Observers can be rewarded as long as they report and prove that at least one secured party has illegal behaviors. Illegal actions include signing two different blocks with the same parent block, or approving an invalid block on a para-chain, or registering a malicious domain name. The nominator is a group with rights and interests, and they entrust the security deposit to the validator. They invest in specific validators to maintain the blockchain. The validators have the highest authority to help package new blocks. The validator needs to pledge enough deposit and must run a relay chain client on a highly available and high-bandwidth machine. On each block, the node must be prepared to receive a new block on the submitted para-chain. 3.2 Domain Name Storage Architecture The storage architecture of the domain name is shown in Fig. 3. In the domain name storage architecture, we abstract the architecture into 3 layers. The blockchain layer defines different operation logics and stores all meta-transactions. Then, we separate the task of data location from the actual storage of data through the routing layer. We use zone files to store routing information. The storage layer is responsible for the actual storage of domain name data. All stored domain name data is signed by the key of the domain name owner. The off-chain storage method ensures that domain name owners can customize storage content without worrying about storage costs. We can verify the integrity of the data off-chain through the hash value stored in the control plane. The zone file of the domain name owner contains a URL record pointing to the domain name data. The data about to be written needs to be signed. Reading the data involves obtaining

Domain Name Management Architecture

307

the zone file and the complete file. The public key is used to verify the signature of the data. So we deal with writing data at high speed due to the immutability of zone files in the blockchain. 3.3 Domain Name Management

Fig. 3. Domain name storage architecture

Domain Name Registration. The process of domain name registration is shown in Fig. 4(a). Firstly, the domain name applicant as a user initiates a registration application to collectors and fills in the corresponding form (domain name, domain name ownership information, and domain name NS record). Collectors will process incoming requests in each slot, and then package them into candidate block and submit them to the validator. After the GRANDPA consensus, the validator will confirm the block and distribute the confirmed block to the collector. Then, the collector broadcasts it to all network. Finally, the blockchain returns the registration completed information to the user. Domain Name Update. The process of domain name change is shown in Fig. 4(b). Similar to the domain name registration process, the type of transaction is domain name updating. Domain Name Resolution. The policy process of domain name resolution is as follows. The user initiates a request to the resolver, and the resolver initiates a request to the corresponding para-chain. The required off-chain data is addressed through the smart contract and returned, and then returned to the user through the resolver.

4 Consensus Algorithm The consensus algorithm is used to ensure the blockchain for producing blocks continuously and keeping data consistency of nodes. We use the mixed consensus of BABE and GRANDPA to ensure the stable operation of the blockchain. Hybrid consensus ensures that even if the network speed is fast, there will be no risk of delaying transactions, and there will be no stuck and rollback attacks.

308

Z. Ma et al.

Fig. 4. Domain name management processes

Blind Assignment for Blockchain Extension (BABE). In each Slot, the validator uses the current slot number (Slot Number), the number of cycles (Epoch number), and the randomness of the chain (randomness) as the output to get a VRF output. If the obtained VRF is lower than a certain value, if it is, the validator gets a chance to produce a block. Each Slot generates a new block by the validator. BABE divides the time into epochs. Each epoch is divided into several time intervals (slot). In each slot, some nodes that are eligible to produce blocks (slot leaders) are selected from a large number of nodes through VRF (verifiable random function). Here we call it a block producer. Each time interval (slot) selected block generation nodes may be different, there may be more than one, or none of them may be selected. Generally speaking, the probability of a node being selected is proportional to the number of tokens it pledges. As shown below: The GHOST-Based Recursive Ancestor-Derived Prefix Protocol (GRANDPA). The GHOST-based recursive ancestor-derived prefix protocol provides near-instantaneous, asynchronous, and responsible security finality in the hybrid consensus blockchain (Fig. 5).

Fig. 5. BABE+GRANDPA

Hybrid consensus separates the finality decision from the block production mechanism. This is an efficient way to obtain the benefits of probabilistic finality and provable finality in our system. It also avoids the shortcomings: the possibility of unknowingly following false forks in probabilistic finality, and the problem of pausing in proving finality.

Domain Name Management Architecture

309

5 Conclusion First of all, the architecture we proposed enables multiple parties involved in domain name maintenance to have equal management rights. The multiple parties involved in domain name maintenance could pledge assets to validators to maintain the top-level domain and realize the goal of co-governance by multi-parties. Then, we use the relay chain structure and off-chain storage method. On the one hand, the management of domain names is assigned to different para-chains according to the top-level domains. On the other hand, the method of off-chain storage reduces the number of blockchain transactions. The combination of the two means improves the throughput of the blockchain. Finally, due to the characteristics of the blockchain, the data hash stored on the blockchain could be used to verify the integrity of the data. Thereby the credible management of the domain name is ensured and domain name hijacking or domain name poisoning are avoided.

References 1. Mockapetris, P.V., Dunlap, K.J.: Development of the domain name system. Comput. Commun. Rev. 18(4), 123–133 (1988) 2. Zhang, Y., Xia, C., Fang, B., Zhang, H.: An autonomous open root resolution architecture for domain name system in the internet. J. Inform. Secur. 2(4), 57–69 (2017) 3. Gourley, S., Tewari, H.: Blockchain backed DNSSEC. In: Abramowicz, W., Paschke, A. (eds.) BIS 2018. LNBIP, vol. 339, pp. 173–184. Springer, Cham (2019). https://doi.org/10.1007/ 978-3-030-04849-5_15 4. Hao Yang, E., Osterweil, D.M., Songwu, L., Zhang, L.: Deploying cryptography in Internetscale systems: a case study on DNSSEC. IEEE Trans. Dependable Secure Comput. 8(5), 656–669 (2011) 5. “Namecoin”, http://namecoin.info 6. Ali, M., Nelson, J., Shea, R., Freedman, M.J.: Blockstack: A global naming and storage system secured by blockchains. In: 2016 {USENIX} annual technical conference ({USENIX}{ATC} 16), pp. 181–194 (2016) 7. Emercoin: Emercoin Links & Resources. https://emercoin.com/en/documentation/links-res ources (2019) 8. Wang, W., Ning, H., Liu, X.: Blockzone: a blockchain-based DNS storage and retrieval scheme. In: Sun, X., Pan, Z., Bertino, E. (eds.) Artificial Intelligence and Security: 5th International Conference, ICAIS 2019, New York, NY, USA, July 26–28, 2019, Proceedings, Part IV, pp. 155–166. Springer International Publishing, Cham (2019). https://doi.org/10.1007/ 978-3-030-24268-8_15 9. Liu, W., Zhang, Y., Liu, L., Liu, S., Zhang, H., Fang, B.: A secure domain name resolution and management architecture based on blockchain. IEEE Symp. Comput. Commun. (ISCC) 2020, 1–7 (2020). https://doi.org/10.1109/ISCC50000.2020.9219632 10. He, G., Su, W., Gao, S., Yue, J.: Td-root: a trustworthy decentralized dns root management architecture based on permissioned blockchain. Futur. Gener. Comput. Syst. 102, 912–924 (2020) 11. “Polkadot Blockchain”, https://polkadot.network/

A Novel Layered GSP Incentive Mechanism for Federated Learning Combined with Blockchain Jiangfeng Sun, Guangwei Zhan(B) , Jiaxi Liu, and Yu Feng Engineering Research Center of Information Networks, Ministry of Education, Beijing University of Posts and Telecommunications, Beijing, China [email protected]

Abstract. Federated Learning (FL) has shown great potential as a solution to the problem of data islands. It enables collaborative modeling while adhering to data privacy and security. But how to ensure participants remain active in FL and make rewards reasonable are issues. Although various game theory models, such as Stackelberg, Cournot, and Leader price, give a pricing method for participants, these models don’t consider the roles in FL and the security. We presented the FL incentive mechanism, B-LSP, based on the Generalized Second Price Auction (GSP). This mechanism can overcome the issue of unmanageable incentives while calculating the reward values. Furthermore, a magnitude stratification is introduced to ensure the participants remain active and the basic need for data volume in FL. The requester sets the basic volume requirement for participants, Initial, and volume comparison standard, Interval, to ensure the basic effectiveness of FL and categorize participants into different layers according to their data volume. In B-LSP, requester can control its costs and keep the federation’s stability. And the blockchain is a crucial part to audit everyone’s contribution and guarantee our mechanism safe. The analysis results show that the B-LSP is more reasonable and scientific, and it satisfies both security and traceability when compared to other game theory models. Keywords: Federated Learning · Incentive mechanism · Smart Contract · B-LSP

1 Introduction Enterprises’ decisions are largely reliant on the power of data. However, their data often contain considerable amount of private information, and according to data privacy and protection laws, these data cannot be centralized or exchanged directly, posing challenges to mining data value. Federated Learning (FL), which has been proposed in recent years, may provide a solution of allowing users to collaborate without exchanging genuine facts [1]. Each member in FL retains its data locally to ensure that it does not breach existing laws and regulations. During the collaborative model training, members merely transmit the intermediate parameters, and subsequently utilize the aggregation algorithm to construct a complete model. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 S. Sun et al. (Eds.): ICSINC 2021, LNEE 895, pp. 310–318, 2022. https://doi.org/10.1007/978-981-19-4775-9_38

A Novel Layered GSP Incentive Mechanism for Federated Learning

311

Though FL is a viable answer, the key to applying FL is to persuade participants to share their data and keep the federation active. The primary issue is the incentive mechanism. Members will be unmotivated to join the federation if there are no fair incentives [2]. Most mechanisms concern about how to allocate the benefits provided by the final model, and many studies are based on the theory of Output Competition, Price Leader, and Analytic Hierarchy Procedure methods, such as Cournot and Stackelberg models [3]. These models show that price is negatively correlated with the output, which means that higher output causes lower price. However, this may not be consistent with the FL condition. Additionally, these methods ignore the cost generated by data and don’t distinguish requester and participants. 1.1 Contribution B-LSP considers the energy costs of various roles and is designed to cover the complete procedure of the FL task. The contributions of this article are as follows: • Introduce GSP to encourage members to compete for bonus: we treat extra bonus as auction objects and the data volume provided by members as their quotations, so that members compete for bonus on the data volume they agreed to share [4]. Each participant will be ranked based on volume, and the first one will get the bonus. • Introduce Magnitude Stratification Mechanism to regulate incentives: the requester determines the volume requirement and comparison standard for distinct layers, and the Incentive Control Line (ICL). Members are classified into layers based on the quotations. The ICL and member’s location decide the auction winner’s bonus. • Introduce Smart Contracts to ensure open and effective rules: B-LSP uses blockchain to implement data auditing and automatic incentives. The data audit will run automatically before FL training to check member’s data. The Shapley value is used to calculate member’s contribution to the federation [5]. The data authentication, quotation, winner’s information, bonus, and basic cost information are merged by a blockchain based smart contract [6] (Fig. 1).

Fig. 1. A novel layered GSP incentive mechanism

312

J. Sun et al.

The follow-up arrangements for this article are Part II, related work; Part III, B-LSP in FL; Part IV, Theoretical Analysis; Part V, Conclusion and Outlook.

2 Related Work Yufeng Zhan, Peng Li, Zhihao Qu and others have proposed a deep reinforcement learning (DRL) based incentive mechanism combined with the Stackelberg model [2]; Latif U. Khan, Shashi Raj Pandey, Nguyen H. Tran and others designed an incentive mechanism based on the Stackelberg model for the application of FL in mobile scenarios [7]. The Stackelberg model is widely used in incentive mechanism design which can ensure profitability and steady decisions for advantageous enterprises. However, Stackelberg model is lack of concern for SMEs (Small and Medium Enterprises) which are the essential part of FL. Jiawen Kang, Zehui Xiong, Dusit Niyato and others think that it is necessary to consider the cost of the members, this paper uses the hardware parameters of each member to calculate the energy consumption [8]; Ismael Martinez, Sreya Francis, Abdel-hakim Senhaji Hafid and others combined blockchain with federated learning by using Enterprise Operation System (EOS) to record and reward members’ contributions of FL, and proposed Class-Sampled Validation-Error Scheme (CSVES) to verify valuable participants and only reward these valuable members [9]; In addition, Yuan Liu, Zhengpeng Ai, Shuai sun and others proposed to build a FedCoin combined blockchain with FL which is a payment system for federated learning [10]. The combination of energy costs, blockchain, and member contributions makes incentive mechanism more reasonable and scientific, but we still lack a comprehensive solution to organically link these technologies. To meet all our objectives for fairness, security, and profit, we presented the B-LSP mechanism, which comprises of GSP, Magnitude Stratification, and Blockchain.

3 B-LSP in FL To separate members into distinct layers in B-LSP, the requester will set Initial and Interval. The member layer and Incentive Control Line are determined by the Initial and Interval. B-LSP ensures data quantity for federated learning by Initial and controls the maximum amount of extra bonus via Interval. The ICL is denoted as: Separatek = Initial + k ∗ Interval, k = 0, 1, 2, . . . , N

(1)

In B-LSP, the requester can ensure its basic interests by Initial and Interval, while the requester will pay the basic cost of members during the FL and the benefits from collaborative model will be allocated based on the contributions of each participant. BLSP consists of three essential technologies: 1. Layered GSP with blockchain; 2. Basic Cost Calculation Method; 3. Shapley Value Method and Smart contract. 3.1 B-LSP Process This paper proposes a mechanism that combines GSP and Magnitude Stratification (BLSP) to motivate members to provide as much data as possible while also rewarding members fairly for their contributions [11]. The procedure details of B-LSP are:

A Novel Layered GSP Incentive Mechanism for Federated Learning

313

1. The requester establishes the fundamental requirement for data volume based on its revenue aim, and the payment for participants’ base costs can be controlled by volume comparison standard. Additionally, B-LSP will calculate the bonus using the requester’s hardware parameters. 2. Participants compete for the extra bonus by their quotation, and for each quotation, they must upload data authentication to the blockchain. If the first and second quotations are at the same layer, the extra bonus is calculated by the difference between them; If they are not at the same layer, the extra bonus is calculated by the difference between the first quotation and the nearest ICL. 3. The quotation, data authentication, data date, hardware parameters and winner are all uploaded to blockchain for subsequent verification and data auditing. 4. After passing the data audit, start federated training using the quotations (promised data from each member) and record the training iteration, training time and communicating times for next computation. 5. After completing the federated model training, compute each member’s basic cost in the training procedure using the hardware parameters and allow the requester of this FL job pay the basic cost for each member. Furthermore, the extra bonus is calculated by requester and allocated to the winner based on the winner’s identification information on the blockchain. 6. Shapley Value can be used to calculate member contributions and upload them to the blockchain. When the final model creates business profits, the income will be allocated to members in proportion to their contribution. The above process is carried out automatically as a smart contract, and the requester will pay the costs of operation and maintenance in the FL job. The B-LSP process is shown in the Fig. 2 and the algorithm is Algorithm 1:

Fig. 2. B-LSP workflow with blockchain

314

J. Sun et al.

Algorithm 1 B-LSP Parameter: Users’ information and quotations, Results: Contract, (Cost, V, Allocation ratio) 1: if Every user agrees the contract then 2: Each member executes: 3: Price for the bonus 4: Requester executes: 5: Set Initial and Interval 6: Ranking and select winner 7: Get H(r), winner’s δ and Separatek 8: if winner - Separatek ≤ winner – second then 9: Num=winner - Separatek 10: else 11: Num = winner - second 12: end 13: V = f Num + g δ , Upload V and quotation to blockchain 14: Starting Federated Learning 15: Calculate allocation ratio and Cost 16: else 17: Restart the auction 18: end

3.2 Basic Cost Calculation Method The basic cost of FL can be estimated using hardware parameters [8]. The hardware detection can be expressed as H (i), the cost of computing power of the participants is calculated via f (Num), the calculation of the participants’ communication costs is g(δ). Suppose the amount of data provided by member i is Numi , its CPU cycle frequency is fi and the number of CPU cycles required for training one single data is Ci ; The effective chip capacitance is ξ. Table 1. Basic cost parameters and functions Parameters and functions

Denotation

Amount of data provided by member i

Numi

CPU cycle frequency

fi

Number of CPU cycles required for training one single data

Ci

Effective chip capacitance

ξ

Member i’s intermediate parameter size

δi

Hardware detection function

H(i)

Computing power

f(Num) (continued)

A Novel Layered GSP Incentive Mechanism for Federated Learning

315

Table 1. (continued) Parameters and functions

Denotation

Communication cost

g(δ)

Transmission power

ρi

Transmission bandwidth

B

Channel gain

hi

Background noise

N0

i Then, the local training time of member i is ci ·Num , the energy consumption of fi computing in a local iteration is f (Numi ) = ξ · ci · Numi · fi2 . Suppose that local iterations before member i communicates with the coordinator is iterationi , transmission bandwidth is B, transmission power is ρi , channel gain between i and coordinator is hi , background noise is N0 , the size of intermediate parameter transmitted is δi . We calculate  δi  the time for one communication between i and coordinator is ρi hi , the energy

B ln 1+

N0

consumption of one communication between i and coordinator is g(δi ) =

δi ρi  ρh . B ln 1+ Ni i 0

If i communicates with the coordinator t a second, the basic cost in federated learning is Costi = t · (iterationi · f (Numi ) + g(δi )). The hardware detection obtains the hardware parameters of each participant: H (i) = ci , fi , ξ, B, ρi , hi , N0 . If i obtains benefits in FL as Ui , then only when Ui ≥ Costi , members will join the federation and keep active. Therefore, the requester must pay the basic cost of FL for each member Costi . The basic cost calculation algorithm is as followed:

Algorithm 2 Basic Cost Calculation Input: User i, Number of iterations iterationi, Size of data set Numi, Size of i’s intermediate parameter δi, Total times of communication in FL total Output: Basic Cost of user i Costi 1: Initialize ci, fi, ξ, B, ρi, hi, N0 = H(i), Costk = 0, Costi = 0 2: Each data provider i executes: 3: for round t = 0; t < total; t++ do 4: for round k = 0; k < iterationi; k++ do 5: Costk Costk + f(Numi) 6: end 7: end 8: Costi Costi + Costk + g(δi) 9: Upload Costi to the blockchain 10: return Costi to requester 11: Requester executes: 12: Aggregate all users’ cost: Cost = ΣCosti

316

J. Sun et al.

3.3 Shapley Value and Smart Contract Shapley Value is a way of calculating the data’s contribution to the model. If the data of participant i is Di ; Xj is the eigenvalue vector of the j-th feature; S is a subset of the entire feature space, |S| is the number of features in the subset; k is the number of features. The Shapley value calculation method is defined as: 

τj =

S⊆x1,x2,...,xk \xj

  |S|!(k − |S| − 1)!   val S ∪ xj − val(S) k!

The contribution of data to the model is τDi =

(2)

n

j=1 τj . The relative contribution of τ Di . If the model creates income M, participant i will get pi ∗ M . τD

each member is pi = The Smart Contract consists of the whole procedure of B-LSP and a business contract which contains the members’ quotations, winners, extra bonus, and basic costs. Table 2. Commercial contract content Participant type

Name

Winner

Quotation

Basic cost

Allocation ratio

Extra bonus

Requester

name

No

Volume

Costi

P

None

Participants

name

Yes

Volume

Cost i

Pi

V

4 Theoretical Analysis By combining blockchain and game theory, B-LSP can give more data security and traceability than many existing solutions. Table 3 compares the advantages of B-LSP to other game theory models. Now, let’s pay more attention on the profit analysis. First, we consider no cost for members in FL. If the extra bonus for the winner is V, the income of the collaborative model used once is M, then member i can get benefits pi ∗ M . The member’s revenue in FL can be expressed as:  V + M ∗ Pi , i wins bonus (3) Ui = M ∗ Pi , i loses bonus we can deduct from this formula that if M ∗ Pi ≥ 0 (the utility of each member is more than zero), it will satisfy the individual rational assumptions in economics, then members will be willing to join the federation [12]. The B-LSP determine the extra bonus for the winner using the requester’s own hardware parameters:  f (Num i − Separate   k ), i and others in different layer  i − Separatek ) + g(Num V= f Numi − Numj + g Numi − Numj , i and others in the same layer (4)

A Novel Layered GSP Incentive Mechanism for Federated Learning

317

from the equation, we can know that 0 ≤ V ≤ f (Interval) + g(Interval) which means requester can control the amount of V. Second, let’s consider about cost. The cost of requester can be expressed as: Costr = V + Costi . Assume the requester’s expected revenue from the collaborative model is W, the income for the model used once by requester is r, the income of the model used once by others is M, and the requester can get p ∗ M per time. If there are k members, the requester’s revenue can be expressed as: 

Ur = W − V − f (Num) + g(δ) (5) W =

 ∞ 1

train

 r − k1 (f (Num) + g(δ)), requester use ∞ ∞ 1 p∗M + 1 r, others use

(6)

Due to the individual rationality, the requester’s income must be more than zero. It’s  ∞ ∞ k obvious that ∞ 1 p∗M + 1 r ≥ 0 which means 1 r− 1 (f (Num) + g(δ)) ≥ 0 is the key for requester to initiate a FL job. Table 3. The advantage of B-LSP than other methods Methods

Controllability

Stability

Security

Traceability

B-LSP

✓

✓

✓

✓

Stackelberg

✓

✓

×

×

Cournot

✓

✓

×

×

Leader price

✓

✓

×

×

5 Conclusion and Outlook The B-LSP help members in getting benefits and requesters in controlling their costs and FL model quality. Through B-LSP, all members can gain greater benefits than the existing widely used Stackelberg model and all members can be continually rewarded. The B-LSP mechanism in this article is a continuous, multi-party, fair and effective incentive mechanism for FL. The basic cost of FL is defined in this article as computing and communication energy cost. As a result, the remainder of this article will concentrate on fundamental cost accounting and data value appraisal in FL. Furthermore, how to assess the model’s value should also be studied in the future. Acknowledgements. This work is supported in part by the National Key Research and Development Program of China (No.2018YFB1403000). Engineering Research Center of In-formation Networks, Ministry of Education.

318

J. Sun et al.

References 1. Yang, Q., et al.: Federated machine learning: concept and applications. ACM Trans. Intell. Syst. Technol. (TIST) 10(2), 1–19 (2019) 2. Yufeng, Z., Peng, L., Zhihao, Q., Deze, Z., Song, G.: A learning based incentive mechanism for federated learning. IEEE Internet Things J. 7(7), 6360–6368 (2020) 3. Jingfeng, Z., Cheng, L., Antonio, R.K., Mohan, K.: Hierarchically fair federated learning. ArXiv 2004.10386 (2020) 4. Dinh, C.N., et al.: Federated learning meets blockchain in edge computing: opportunities and challenges. IEEE Internet Things J. 8(16), 12806–12825 (2021). https://doi.org/10.1109/ JIOT.2021.3072611 5. Wang, G.: Interpret federated learning with shapley values. ArXiv 1905.04519 (2019) 6. Hyesung, K., Jihong, P., Mehdi, B., Seong, L.K.: Blockchained on-device federated learning. IEEE Commun. Lett. 24(6), 1279–1283 (2019) 7. Latif, U.K., et al.: Federated learning for edge networks: Resource optimization and incentive mechanism. IEEE Commun. Mag. 58(10), 88–93 (2020). https://doi.org/10.1109/MCOM. 001.1900649 8. Jiawen, K., Zehui, X., Dusit, N., Shengli, X., Junshan, Z.: Incentive mechanism for reliable federated learning: a joint optimization approach to combining reputation and contract theory. IEEE Internet Things J. 6(6), 10700–10714 (2019) 9. Martinez, I., Francis, S., Hafid, A.S.: Record and Reward Federated Learning Contributions with Blockchain. In: 2019 International Conference on Cyber-Enabled Distributed Computing and Knowledge Discovery (CyberC), vol. 1, pp. 50–57. IEEE, Guilin, China (2019) 10. Liu, Y., Ai, Z., Sun, S., Zhang, S., Liu, Z., Yu, H.: FedCoin: a peer-to-peer payment system for federated learning. In: Yang, Q., Fan, L., Yu, H. (eds.) Federated Learning. LNCS (LNAI), vol. 12500, pp. 125–138. Springer, Cham (2020). https://doi.org/10.1007/978-3-030-63076-8_9 11. Kang, L.N., Zichen, C., Liu, Z., Yu, H., Liu, Y., Yang, Q.: A multi-player game for studying federated learning incentive schemes. In: 29th International Joint Confer-ence on Artificial Intelligence (IJCAI 2020), pp. 5279–5281. IJCAI, Japan (2020) 12. Lyu, L., Xu, X., Wang, Q., Yu, H.: Collaborative fairness in federated learning. In: Yang, Q., Fan, L., Yu, H. (eds.) Federated Learning. LNCS (LNAI), vol. 12500, pp. 189–204. Springer, Cham (2020). https://doi.org/10.1007/978-3-030-63076-8_14

Vehicle Searching in Underground Parking Lots Based on Blockchain Haoming Zhang1 , Xiaojun Jing1(B) , Quan Zhou1 , Junsheng Mu1 , and Bohan Li2 1 Beijing University of Posts and Telecommunications, No 10, Xitucheng Road, Haidian

District, Beijing, People’s Republic of China {zhanghaoming0314,jxiaojun,quanzhou,mujs}@bupt.edu.cn 2 University of Southampton, University Road, Southampton SO17 1BJ, UK [email protected]

Abstract. Now the area of the underground parking lot is becoming larger and larger. In such a large or even multi-level area, people usually spend a lot of time finding their vehicles. This problem is called reverse vehicle search. The unique problem of indoor positioning is that the GPS signal is limited, but the location is relatively fixed. The Internet of vehicles for vehicle search will face security and privacy risks. This paper proposes a vehicle positioning system based on a blockchain network to avoid relevant risks. In this system, the vehicles in the parking lot share each other’s positioning as nodes and store the data encrypted. Users can access the blockchain network to search the cars. Keywords: Blockchain · Vehicle searching · Low-signal

1 Introduction The area of underground parking lots tends to bigger nowadays. In such big or even multilevel areas, the passenger’s People usually get typically take a lot of time to find their vehicle. This kind of problem is referred to as reverse vehicle-searching. In essence, indoor positioning and navigating are the most critical issues in the reverse vehiclesearching problem. Compared with outdoor positioning, it is impossible to use GPS technology to locate because of the very weak GPS signal. There are currently three mainstream indoor positioning systems: 1) wirelesscommunication-based positioning, 2) sensor-based positioning, and 3) camera-based positioning. Among them, wireless-based positioning technologies are the most attractive method. The field of wireless positioning technology can be divided into widearea positioning and short-distance wireless positioning. Wide-area positioning can be divided into satellite positioning and mobile positioning; indoor short-distance positioning technologies mainly include WLAN, RFID, UWB, Bluetooth, and ultrasonic. The principle of WIFI positioning is that the device turns on WIFI, scans the surrounding AP signals, obtains the MAC address of the corresponding AP, and then determines the unique object based on the MAC address, and judges the distance based on the © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 S. Sun et al. (Eds.): ICSINC 2021, LNEE 895, pp. 319–324, 2022. https://doi.org/10.1007/978-981-19-4775-9_39

320

H. Zhang et al.

strength of the signal. There are two kinds of algorithms based on WIFI positioning: Triangular Positioning Algorithm (RSSI) and Fingerprint Algorithm. The accuracy of the triangulation algorithm is low, and the locations of all APs need to be known in advance. The fingerprint algorithm usually traverses the terminal after the network deployment is completed, collects the signal strength of the terminal monitored by the AP, and builds a database of location and signal strength. The WIFI fingerprint algorithm will eventually put the location data inventory in a central server. Considering the situation of vehicle positioning in underground parking lots, it is reasonable to introduce the concept of the Internet of Vehicles into vehicle positioning. However, in a typical car networking system, in-vehicle messages are generally only shared among a few nearby vehicles. But if you want to start searching for a specific target vehicle at any location, the vehicle location information needs to be more widely distributed to vehicles in the entire underground parking lot. It can be challenging to specify a single trusted entity for storage and distribution in unfamiliar peer nodes, and from the perspective of energy consumption and revenue, individual vehicles are not willing to play this role. In addition, there are security and privacy issues in the centralized storage and distribution of information. This paper proposes a vehicle positioning system based on blockchain to solve the problem of storage and distribution of vehicle positioning information in an indoor background. In this system, vehicle location information is stored in blocks, and each node can broadcast, update and synchronize information based on economic incentives. The remainder of this paper is organized as follows. In Sect. 2, the system of blockchain in vehicle searching is proposed. Section 3 concludes the paper.

2 System Implementation We consider a heterogeneous network, in which nodes are vehicles and smartphones. Smartphone nodes would join and disjoin the network based on its distance to its nearest vehicle. Meanwhile, the vehicle node in our system has a stable position. The system can execute the blockchain implementation of these on a smartphone or even vehicle ECU (electronic control unit) (Fig. 1).

Fig. 1. The architecture of the system

Vehicle Searching in Underground Parking Lots Based on Blockchain

321

The system mainly consists of the following three parts: node joining, location information generation, information storage, and distribution. 2.1 Node Joining When the vehicle enters the underground parking lot, the vehicle trip computer interacts with the management node of the property at the entrance to complete the registration of the vehicle information, including the registration of the public key private key pair and the vehicle identity information sent to the management node through the secure channel ask. 2.2 Location Information Generation There are three types of messages generated on the blockchain as below: Register Message Generation. When the vehicle stops moving, the on-board computer running the blockchain node starts broadcasting its own information to join the blockchain network. Locate Message Generation. After successful registration, the vehicle records several nodes with the shortest communication delay and generates positioning information. It is generally believed that a shorter communication delay represents a shorter distancein space. This information will be used to indicate the vehicle’s location. Reward/Punishment Message Generation. Based on the different behaviors of the vehicle, corresponding reward information or punishment information will be generated. Corresponding to the indoor parking lot scenario, in the system mentioned in this article, the reward and punishment information is designed as a token called P- Coin to pay for parking. We define honestly generating and actively disseminating information as acts of good faith, and falsifying the information as malicious acts.

2.3 Storage The vehicle location messages store on the blockchain. The structure of the block is shown in Fig. 2. The block header contains the hash value of the previous block, the time information when the block was created, the specific identifier of the parking lot, the block number, the location information (e.g., the id of the nearest car), the difficulty (one used to increase The difficulty value of the mining cost) and a calculated random number. The vehicle nodes in the parking lot need to find a specific random number to make the new block information meet the difficulty. When the specific value is found, the new block is also determined, and the node that determines and broadcasts the block gets the relevant reward.

322

H. Zhang et al.

Fig. 2. The architecture of the blockchain

3 System Evaluation We conducted two experiments to prove the efficiency of our system. The simulation is carried out on a Windows system, using Intel Core i7-10700 with a frequency of 2.90 ghz. We use a less difficult nonce value to reduce power consumption. At the same time, we use the PoW consensus protocol. Because PoS protocol is related to currency age, this concept is not suitable for the scene of looking for vehicles, and vehicles continue to join and leave, which will make the system based on DPoS consensusconsume a lot of energy to hold elections to redefine miners. 3.1 Simulation We believe that the bottleneck of the system lies in the delay of communication surrounding vehicles and the verification of signatures. Therefore, subsequent experiments will simulate these two steps.

Fig. 3. Mean delay time (once mine and broadcast) with node’s number

As shown in Fig. 3, with the increase in the number of nodes, the average communication time between single mining and nodes does not change significantly. In order to ensure that the data is tampered with, there are minimum requirements for the difficulty of hash calculation in mining operations. Therefore, the communication time cost increases linearly with the increase of nodes, but in this system, the influencing factor of time mainly comes from mining time.

Vehicle Searching in Underground Parking Lots Based on Blockchain

323

Fig. 4. Mean delay time (once mine and broadcast) with verification node’s number

Figure 4 shows the change of message signature verification time with the verification node threshold. Although the cost increases linearly, it is not high. Even if the number of signers is 50, multiple signatures can be verified in 1 s. This has littleimpact on practical use. 3.2 Conclusion A new reverse vehicle search method is proposed. The method is divided into two stages: vehicle registration stage and smartphone query stage. In the vehicle registration phase, the vehicle registers and records its location in the blockchain network. In the smartphone query stage, we propose a location strategy based on block information matching. Simulation results show that this method can update vehicle data at least milliseconds and search within 1 second. The simulation results verify the effectiveness of the system. Future research includes further simulation, real node communication, and other problems in wireless transmission.

References 1. Qiu, J., Grace, D., Ding, G., Yao, J., Qihui, W.: Blockchain-based secure spectrum trading for unmanned-aerial-vehicle-assisted cellular networks: an operator’s perspective. IEEE Internet Things J. 7(1), 451–466 (2020). https://doi.org/10.1109/JIOT.2019.2944213 2. Azaria, A., et al.: Medrec: using blockchain for medical data access and permission management. In: 2016 2nd International Conference on Open and Big Data (OBD). IEEE (2016) 3. Zhang, L., et al.: Blockchain based secure data sharing system for Internet of vehicles: a position paper. Veh. Commun. 16, 85–93 (2019). https://doi.org/10.1016/j.vehcom.2019.03.003 4. Huang, G., Hu, Z., Cai, H.: Wi-fi and vision integrated localization for reverse vehiclesearching in underground parking lot. In: 2017 4th International Conference on Transportation Information and Safety (ICTIS). IEEE (2017) 5. Liu, Q., Qiu, J., Chen, Y.: Research and development of indoor positioning. China Commun. 13(Supplement2), 67–79 (2016)

324

H. Zhang et al.

6. Zeng, Y., Zhang, R., Lim, T.J.: Wireless communications with unmanned aerial vehicles: opportunities and challenges. IEEE Commun. Mag. 54(5), 36–42 (2016). https://doi.org/10. 1109/MCOM.2016.7470933 7. Liu, J., et al.: Secure intelligent traffic light control using fog computing. Future Gener. Comput. Syst. 78, 817–824 (2018). https://doi.org/10.1016/j.future.2017.02.017 8. Zhang, L., et al.: Privacy-preserving cloud establishment and data dissemination scheme for vehicular cloud. IEEE Trans. Dependable Secure Comput. 17(3), 634–647 (2018)

Blockchain and Knowledge Graph Fusion Network Architecture Model Mengqi Han, Xiaojun Jing(B) , Jia Zhu, and Junsheng Mu School of Information and Communication Engineering, Beijing University of Posts and Telecommunications, Beijing, China [email protected]

Abstract. The development and application of blockchain technology is a hot topic today. With the support of encryption algorithm, smart contract and other technologies, blockchain has many advantages. We often use the concept of the knowledge graph in our research. In the context of the era of artificial intelligence, the knowledge graph is not only a simple semantic network describing relationships, but also can deduce a lot of richer data. This paper combines the two, and stores the data of the knowledge graph on the blockchain, making the data more secure and efficient to be used. This paper discusses the construction method and process of the network architecture of the integration of blockchain and knowledge graph by summarizing the method, and prospects the potential application in the military field. Keywords: Blockchain · Knowledge graph · Network architecture fusion

1 Introduction The term “knowledge graph” has been used a lot in recent years. In fact, it’s not a new term. In the official wikipedia entry, “Knowledge Graph” is the knowledge base that Google uses to enhance its search engine. In essence, a knowledge graph is a semantic network that reveals the relationships between entities, which can formalize descriptions of things in the real world and their interrelationships. As now the progress and development of science and technology, many advanced technology arises at the historic moment, the machine is expected as people can understand vast amounts of network information, hope can quickly, accurately and intelligent access to the information you need, in order to meet this demand, intelligent knowledge map arises at the historic moment, the research significance is to convenient people to access to information more quickly and efficiently. Knowledge map is today’s hot word is a branch of artificial intelligence, with the development of science, people gradually from the second generation of “perception” to the third generation of artificial intelligence ai “cognitive” high-speed development, and knowledge map qiao can better help computer more logical reasoning ability, based on past form the database can make more rich data reasoning. In addition, with the popularity of “Bitcoin”, the technology behind it, blockchain, is gradually known to people. Blockchain has extremely high application value such as: improve © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 S. Sun et al. (Eds.): ICSINC 2021, LNEE 895, pp. 325–330, 2022. https://doi.org/10.1007/978-981-19-4775-9_40

326

M. Han et al.

efficiency, reduce cost, reduce risk and promote mutual trust. This paper combines the two, fuses the data extracted from the knowledge graph and stores it in the alliance chain, so that the data has the advantages of decentralization, openness, transparency and so on, and is more secure and effective reasoning ability, based on past form the database can make more rich data reasoning. In addition, with the popularity of “Bitcoin”, the technology behind it, blockchain, is gradually known to people. Blockchain has extremely high application value such as: improve efficiency, reduce cost, reduce risk and promote mutual trust. This paper combines the two, fuses the data extracted from the knowledge graph and stores it in the alliance chain, so that the data has the advantages of decentralization, openness, transparency and so on, and is more secure and effective.

2 Related Work 2.1 Blockchain Blockchain is one of the hottest technologies right now. Blockchain is the technology behind cryptocurrencies, which is quite different from the basic language or platform. It is not a new technology in itself. Similar to Ajax, it can be said that it is a technical architecture. Blockchain does not need a central server, and can realize the openness, transparency and traceability of all kinds of stored data. One of the unique ways cryptocurrencies like Bitcoin store data is a self-referent data structure that stores large volumes of transactions. Each record is linked back to front, transparent, untampered, and easily traceable. The blockchain architecture enables each participant in a business network to share the same ledger, and when transactions occur, all of the ledger can be changed in a peer-to-peer synchronous manner. Ledger modification requires consensus algorithms, based on a consensus mechanism among network participants to ensure that transactions are jointly validated. Business networks meet government oversight, compliance and audit requirements. The security, authorization, and validation of transactions can be ensured by using cryptographic algorithms to ensure that the transactions of participants on the network are tamper-free. In addition, the blockchain also has programmable and Turing-complete smart contracts, which embed the contract terms related to asset transfer transactions into the transaction database so that transactions can occur only when business conditions are met. 2.2 Knowledge Graph Today’s ARTIFICIAL intelligence technology, mainly through image recognition, speech recognition and other pattern recognition technology, to complete the “perception” level of work. But really want to reach the level of “cognition”, knowledge graph technology is widely regarded at present, it has the hope to become “big. The Brain”. The construction and application of knowledge graph need the support of various information processing technologies [1]. Through knowledge extraction technology, knowledge elements such as entities, relationships and attributes can be extracted

Blockchain and Knowledge Graph Fusion Network Architecture Model

327

from some open semi-structured and unstructured data. The ambiguity among entities, relations and attributes can be eliminated by knowledge fusion, and the knowledge reasoning is to further excavate the implicit knowledge based on the existing knowledge base, so as to expand the knowledge base. Knowledge Extraction Knowledge extraction mainly includes entity extraction, relation extraction and attribute extraction. Entity extraction refers to the automatic recognition of named entities from the original corpus. Relationship between extraction. Knowledge Fusion Information fusion or knowledge fusion is entity fusion, also known as entity alignment operation. When we combine these data, we face the problem of entity alignment, also known as syndication. Entity alignment, also known as entity normalization, is still commonly used now is the clustering method. Knowledge Storage The output after information extraction and fusion is triple. We mainly use relational database, graph database and NoSQL database to store the triple effectively, so as to facilitate retrieval and invocation. Intellectual Reasoning Reasoning is a very difficult part of natural language processing. Whether it is question answering system or reading comprehension, the current technology can only achieve good results on the query based on knowledge base, but once involved in reasoning, it will face great challenges. This paper briefly introduces one of the reasoning methods in knowledge graph: Reasoning based on RDF (Resource Description Framework), which is Data Model in nature. In simple terms, RDF is a method and means to represent things, which is formally expressed as triples. RDF consists of nodes and edges. Nodes represent entities and attributes, while edges represent the relationship between entities and values of entities and attributes. Inference based on RDF is inference based on symbols and concepts. Application of knowledge Graph Knowledge graph provides a more effective way for the expression, organization, management and utilization of large, heterogeneous and dynamic data on the Internet, which makes the intelligence level of the network higher and more close to human cognitive thinking [2]. Intelligent Search: After the user enters the keyword that wants to query, the search engine not only looks for the keyword, but first carries on the semantic understanding. For example, after the query is segmented, the description of the query is normalized to match the knowledge base. The returned result of the query is a complete knowledge system given by the search engine after retrieving the corresponding entities in the knowledge base. In-depth Question Answering: Question answering system is an advanced form of information retrieval system, which can provide users with answers to questions in accurate and concise natural language. Most of question answering system will tend to a given

328

M. Han et al.

problem is decomposed into several small problems, and then pump matching answers one by one, to the knowledge base, this process is to question the attribute extraction, with the knowledge map attributes in the process of one to one correspondence, and automatically detect it in time and space alignment, etc., finally will be the answer to merge, with intuitive way to show to the user. Social Networking: Facebook launched Graph Search in 2013. Its core technology is a knowledge Graph that connects people, places, things, and more, and supports accurate natural language queries in an intuitive way, making these services close to the lives of individuals. It meets the user’s need to discover knowledge and find the most relevant people. Vertical Industry Application: In many vertical fields such as finance, healthcare, e-commerce, etc., knowledge graph is bringing better domain knowledge, lower financial risk, and more perfect shopping experience. More, such as education and research industry, library, securities industry, biomedical industry and some industries that need to do big data analysis. These industries are in urgent need of integrated and relevant resources. Knowledge graph can provide more accurate and standardized industry data and rich expression for them, helping users to obtain industry knowledge more easily.

3 Methodology 3.1 Intelligence Data Alliance Chain Blockchain can be divided into three types: public chain, private chain and alliance chain according to the different ways of nodes joining and exiting the blockchain system. These three types have different scope of application according to different characteristics: public chain allows any node to participate in it, nodes can join or quit freely; A private chain allows only a small number of trusted nodes to join. Nodes in the chain are trusted to each other but not to the public.; Only some authorized nodes are allowed to join the alliance chain, and the number and function permissions of participating nodes are set. Since intelligence does not require all nodes to participate in the blockchain, the alliance chain meets the requirements of reliable storage, reliable transmission and controllable access of intelligence data. The alliance chain allows some nodes to join the chain conditionally, which is completely open to the nodes in the chain and conditionally open to the nodes outside the chain. The entry and exit of intelligence data providers and receivers need to be authenticated and authorized by trusted nodes. The contents of authentication include qualifications, confidentiality conditions, utilization purposes, etc., while the contents of authorization include the confidentiality level, audit object, sharing scope, sharing degree, and sharing level of intelligence data [3]. The architecture of combining intelligence data with knowledge graph based on blockchain technology is shown in the figure below (Fig. 1).

Blockchain and Knowledge Graph Fusion Network Architecture Model

329

Fig. 1. Blockchain and knowledge graph fusion

4 Potential Research Directions 4.1 Attributive Reasoning If a field is missing from a collection of obtained intelligence, the knowledge graph can be compared to a database stored on the blockchain [4]. If the field accuracy of the verified entity reaches 90%, the missing part of the data can be found through attribute comparison to restore the integrity of the entity. 4.2 Target Recognition By training the relationship between categories in the detector, the knowledge graph can indeed improve the accuracy of our detection algorithm and explain the relationship between targets to a large extent. The steps are analyzed through the combination of ternary (object, predicate, subject). There is a parallel semantic pool for separate training, and finally merged into the original algorithm output feature map (Enhanced feature) [5]. Based on previous knowledge, the target is identified by inferring from the feature data. 4.3 Target Tracking Target tracking is an important branch of computer vision research. Based on the context information of image sequence or video, it models the appearance features and motion information of the target, so as to predict the motion state of the target while calibrating the position of the target. Using the knowledge map and block chain technology, will be A point of access to the entity attribute information data and in point B access to the entity attribute data and features in the knowledge base compared properties of chain blocks, if highly consistent, can reason out the logic relation of the two entities for the same entity, so as to complete the entity tracking from point A to point B.

5 Conclusion The research on blockchain and knowledge graph is still the focus of current academic scholars. By combining the two modes, this paper can greatly improve the security of the data in knowledge graph, and it also has a great application prospect in the military field. At present, the research in this field is still in the development stage, and the third generation of ARTIFICIAL intelligence, “cognitive” intelligence, is still in its infancy, so the knowledge reasoning technology in the knowledge graph still has many difficulties to overcome. In the future, the reliability and effectiveness of research can be greatly improved after overcoming these difficulties.

330

M. Han et al.

References 1. Hua, M., Lu, H.: Research hotspot analysis of blockchain based on knowledge mapping domain in China and abroad. Inf. Sci. 38(11), 70–79 (2020) 2. Liu, X., Bai, X., Wang, L., Ren, B., Lu, S., Li, L.: Review and trend analysis of knowledge graphs for crop pest and diseases. IEEE Access 7, 62251–62264 (2019). https://doi.org/10. 1109/ACCESS.2019.2915987 3. Mu, J., Gong, Y., Zhang, F., Cui, Y., Zheng, F., Jing, X.: Integrated sensing and communicationenabled predictive beamforming with deep learning in vehicular networks. IEEE Commun. Lett. 25(10), 3301–3304 (2021). https://doi.org/10.1109/LCOMM.2021.3098748 4. Wang, Z., Liu, R., Liu, Q., Thompson, J.S., Kadoch, M.: Energy-efficient data collection and device positioning in UAV-assisted IoT. IEEE Internet Things J. 7(2), 1122–1139 (2020). https://doi.org/10.1109/JIOT.2019.2952364 5. Zhang, R., Jing, X., Wu, S., Jiang, C., Mu, J., Yu, F.R.: Device-free wireless sensing for human detection: the deep learning perspective. IEEE Internet of Things J. 8(4), 2517–2539 (2021). https://doi.org/10.1109/JIOT.2020.3024234

Blockchain Performance Optimization Mechanism Based on Caching Strategy Xinyan Wang1 , Jing Zhang1 , Jizhao Lu1 , Beibei Zhu1 , Xiao Feng1 , and Han Yan2(B) 1 State Grid Henan Information and Telecommunication Company, Zhengzhou 450052, China 2 Beijing JingAn YunXin Science and Technology Ltd., Beijing, China

[email protected]

Abstract. As a distributed system, the research on blockchain related technologies is gradually showing an explosive growth trend. As one of the cores of blockchain, consensus algorithm is mainly responsible for ensuring the security of blockchain system. It has the disadvantages of large overhead and long communication delay. In order to improve the performance of the blockchain system in storing and querying user data, this paper takes the consensus algorithm as the research entry point, and designs the cache optimization raft consensus algorithm (COR), which can effectively shorten the waiting confirmation time in the user data storage stage, so as to improve the efficiency of user data storage. In addition, the system adds a cache layer to its traditional structure, further improve the query efficiency of the system. Finally, the system realizes the goal of system load balancing and improving system query efficiency based on three types: master node, slave node and supervisory information cache node, as well as two data structures: data identifier cache table and resource utilization queue. Keywords: Blockchain · Consensus algorithm · Cache optimization · Resource utilization queue

1 Introduction With the development and popularization of Bitcoin and Ethereum technologies [1, 2], the research and application of blockchain technology has gradually shown an explosive growth trend. With the gradual increase of blockchain applications such as digital currency, data integrity services [3], energy transactions [4], etc., the research on improving the efficiency of blockchain systems has gradually become one of the main directions at present. However, the traditional consensus algorithm such as (POW, Raft) has a higher communication overhead and a long communication delay [5]. Take Raft for example [6], after the transaction is submitted, the leader node needs to wait for a sufficient amount of user data to be packaged, and needs to be submitted to all backup nodes and receive more than half of the replies before returning a confirmation to the client. In the above process, the client has been in a waiting state during this process. It can be seen that the long delay caused by the consensus process has seriously affected the performance of the blockchain. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 S. Sun et al. (Eds.): ICSINC 2021, LNEE 895, pp. 331–337, 2022. https://doi.org/10.1007/978-981-19-4775-9_41

332

X. Wang et al.

Therefore, this article plans to obtain the scheme to effectively improve the storage and query performance of the blockchain system, starting from the blockchain system infrastructure and the consensus algorithm.

2 Blockchain Architecture Based on Caching Strategy 2.1 Overall System Architecture The information authentication and query system based on blockchain mainly consists of five parts: the client node of the user layer, the master and the slaver information cache node of the cache layer, and the primary node and the backup node of the storage layer. The structure is shown in Fig. 1.

Fig. 1. System architecture diagram

2.2 Cache Communication Mechanism For the cache layer, the system adopts a master-slave replication caching mechanism. This mechanism can copy the data of one information cache node to other nodes. Data replication is one-way, and can only be replicated from the master node to the slave node. The master node is mainly writing, and the slave node is mainly reading. In order to improve the stability and high availability of the information cache node, the system introduces a supervisor node (Supervisor). The node can send the status request command to the other information cache node server to return to the running state. When the supervisor node detects that the Master is down, it will automatically switch Slaver to Master and modify the configuration file. Then notify other servers to switch hosts through publish-subscribe mode.

Blockchain Performance Optimization Mechanism Based on Caching Strategy

333

3 Blockchain Performance Optimization Strategy 3.1 Storage Optimization Strategy 1. The client submits user data through communication with the primary node and cache node. First, the client sends a data upload request to the primary node. If the primary node cannot provide services, the client needs to wait for the election to be completed and submit it to the new primary node. Before submitting data, the client needs to verify the legitimacy of the data. The verification contents mainly include the legitimacy of fields, signatures and contents. 2. Compared with the traditional raft algorithm, COR algorithm adds the information cache node. After confirming the legitimacy of the user data, both the primary node and the information cache node will return to the client a user data identifier T encrypted by the sha256 algorithm. The calculation formula of the data identifier T is as follows: T = sha256(userinfo)

(1)

In the formula, T represents the unique identification of user data, and userinfo is the actual content of user data. It should be noted that identifier T is also the leaf node of the Merkle tree, which can help users locate user data in the blockchain. In addition, when the information cache node receives the client’s user data, it will persist the client’s user data into memory. After the blockchain completes the consensus, the information cache node will respond to the request of the primary node, recycle the storage space according to the situation, and clear the persistent user data. If the primary node fails, the information cache node will also provide user data temporarily stored in itself and not received by the new primary consensus node when the primary data node switches and recovers user data. 3. In the consensus stage, if the primary node confirms the legitimacy of user data, it will immediately reply to the calculated data identifier T to the client. If the client receives the same reply T from the primary node and the information cache node, it represents that the request of the client has reached data consistency between the primary node and the information cache node; If the data identifier returned to the client by the primary node and the cache node are inconsistent, the client will choose to close the node or retransmit the data until the returned data identifier are completely consistent. Since the information cache node completes the persistent storage of data, there is no need to worry about the loss of data in the next consensus process. It can be directly regarded as a successful data submission, and this submission of the client has experienced one RTT(Round-Trip Time). If the cache layer crashes as shown in Fig. 2, the client needs to wait for a while to confirm the information. This is because if the information cache node fails or the primary node conflicts with the user data in the information cache node, the cache layer cannot repair the primary node data. In order to avoid the consistency problem of the user data that has passed the verification, the primary node needs to wait for the completion of the consensus, and half of the backup nodes will return the success information to

334

X. Wang et al.

the client after receiving the successful data. If the cache layer works normally but the primary node stops serving, the system will recast the node, and the new primary node will find the primary node user data of the previous consensus stage from the information cache node and the backup node.

Fig. 2. COR algorithm consensus process

3.2 Query Optimization Strategy Query Request Tuning In the query optimization process, considering that most of the nodes need to participate in the maintenance of the blockchain and other tasks, in order to better achieve load balancing, the system adds a Supervisor node. In the process of resource utilization monitoring, the Supervisor node needs to maintain a node queue. The queue stores the resource utilization rate and port data of different Slaver information cache nodes, and is sorted according to the resource utilization rate of different nodes. The closer the node to the head of the queue, the lower the system resource utilization rate. When the client sends a query request to the blockchain system, Supervisor node will send the query request to the Slaver node corresponding to the lowest resource utilization port. When the supervisor node forwards a certain number of requests, it is necessary to recount the resource utilization and adjust the queue. Query Process Design 1. In the query phase, the user needs to submit the data identifier T to the supervisor node, and the supervisor node selects and forwards the request to the slaver node with low resource utilization according to the resource utilization queue. The slaver node can find the corresponding data in the cache according to the data identifier T. If the search is successful, the data can be returned to the user, and the activity corresponding to the identifier in the user data identifier cache table is increased by 1.

Blockchain Performance Optimization Mechanism Based on Caching Strategy

335

2. When the slaver node does not find the data identifier T in the cache, it will query the feature information in the user data identifier cache table. Due to memory constraints, the cache node may eliminate the cache according to the data activity, so there is a possibility that the data identifier search in the cache table fails. 3. If the data identifier is in the cache table, the node needs to submit the data identifier and the timestamp corresponding to the identifier to the blockchain for secondary query. The blockchain node uses the timestamp to quickly locate the block where the data is located, and determines the final location of the data according to the data identifier T. Finally, the query results are returned to the user and the information cache node, and the activity of the node in the cache table is set to 1. 4. If the data identifier is not in the cache table, it will be submitted to the blockchain for data query directly. Finally, the query result is returned to the user and the information cache node, the data identifier and the timestamp in the query result are added to the cache table, and the activity corresponding to the ID is set to 1.

4 Simulation Test 4.1 Experimental Environment The COR algorithm is tested in a local virtual machine cluster, and the configuration of each consensus node is shown in Table 1. In Table 1, the distributed network is set up with 4 nodes, and each node has a dual identity. For the cache layer, the system sets up 4 information cache nodes, where the master node is responsible for recording cache data, the two slaver nodes are mainly used to read data, and the supervisor node is used for fault detection. For the storage layer, the system sets up 4 consensus nodes, where the master node is responsible for data recording, and 3 backup nodes are used for blockchain failure backup and master node legitimacy verification. Table 1. Configuration of consensus nodes Machine name

CPU

Memory

Operating system

Cache layer

Storage layer

Node01

2

4G

Ubuntu18.04

Master

Backup node

Node02

2

4G

Ubuntu18.04

Slaver

Primary node

Node03

2

4G

Ubuntu18.04

Slaver

Backup node

Node04

2

4G

Ubuntu18.04

Supervisor

Backup node

4.2 Result Analysis The system assumes that the size of each data request is set to 100 bytes, and each block contains 50 data requests. Communication Delay Test For the communication delay experiment, the common raft consensus algorithm scenario and COR consensus algorithm scenario are analyzed respectively. The experimental

336

X. Wang et al.

results are shown in Fig. 3. Experiments show that when the information cache node works normally, the COR communication delay, that is, the user waiting for confirmation time, is significantly lower than raft, while when the information cache node fails, there is little difference between the two communication delays. This is because the information cache node can ensure that the data submitted by the user will not be lost when the storage layer master node fails, so that the user can stop the data submission process without ensuring that the data is formally chained.

Fig. 3. Communication delay test diagram

Response Delay Test The user query response delay experiment conducted 10 experiments on different query repetition rates with cache layer, and compared with that without cache layer. The experimental results are shown in Fig. 4. Experiments show that in the same case, the cache layer can effectively reduce the data query delay, and with the increase of query repetition rate and the number of concurrent requests, the response performance optimization is more obvious.

Fig. 4. Test diagram of query response delay

Blockchain Performance Optimization Mechanism Based on Caching Strategy

337

5 Summary By improving the blockchain structure adopted by the basic raft algorithm, this paper designs the information cache layer, which greatly improves the user’s query efficiency. In addition, the supervisor node of the information cache layer can also maintain a resource utilization queue for load balancing. Based on this structure, a cache optimization consensus algorithm is proposed, which separates the persistence part and the linearization part in the consensus process, and reduces the time required for users to wait for confirmation. Finally, through the above improvements, we can obtain the scheme to effectively improve the storage and query performance of the blockchain system. Acknowledgment. This work was supported by the science and technology project of State Grid Corporation of Henan Province “Research on integrating trusted computing and blockchain in power information security protection technology” (5217Q0210003).

References 1. Efanov, D., Roschin, P.: The all-pervasiveness of the blockchain technology. Procedia Comput. Sci. 123, 116–121 (2018) 2. Kondo, M., Oliva, G.A., Jiang, Z.M., Hassan, A.E., Mizuno, O.: Code cloning in smart contracts: a case study on verified contracts from the Ethereum blockchain platform. Empir. Softw. Eng. 25(6), 4617–4675 (2020). https://doi.org/10.1007/s10664-020-09852-5 3. Liu, B., Yu, X.L., Chen, S., Xu, X., Zhu, L.: Blockchain based data integrity service framework for IoT data. In: Inter’l Conf. on Web Services, pp. 468–475. IEEE (2017) 4. He, T.: Research on Key Algorithms and Technologies of Blockchain for Distributed Energy Trading Scenarios. University of Electronic Science and Technology of China (2020) 5. Futao, L.: Consensus mechanism in blockchain. China New Commun. 21(21), 12 (2019) 6. Yi, W., Sheng, Z.: Research on block chain consensus algorithm raft. Inf. Netw. Secur. 21(06), 36–44 (2021)

Multi Energy Coordinated Dispatching of Virtual Power Plant Based on Blockchain Zuhao Wang1(B) , Bing Zhou2 , Xusheng Yang2 , Na Li2 , Bing Wang2 , Shaojie Yang1 , Yu Chen1 , and Ao Xiong1 1 Beijing University of Posts and Telecommunications, Beijing 100876, China

[email protected] 2 State Grid Integrated Energy Service Group CO., LTD., Beijing 100052,

People’s Republic of China

Abstract. As one of the important means to promote distributed energy consumption, virtual power plant has some problems in information security, user trust and energy scheduling. But the emergence of blockchain technology provides a practical solution to such problems. Starting from the technical characteristics of virtual power plant, this paper designs a set of virtual power plant energy scheduling blockchain system framework including user authentication, block consensus and energy scheduling, combined with the characteristics of blockchain, such as traceability, high availability security and immutability. On this basis, in order to further achieve the goal of economy and stability of system energy scheduling, the system introduces the scenario analysis method to solve the volatility problem in the process of distributed energy scheduling, and genetic algorithm is introduced to help the system quickly and efficiently determine the distributed energy scheduling scheme with the lowest cost, and finally maximize the benefits of virtual power plant. Keywords: Virtual power plant · Blockchain · Scene analysis · Genetic algorithm

1 Introduction In recent years, with the increase of power demand all over the world, the scale of power system is expanding day by day, which makes the implementation of centralized power supply mode more difficult. Therefore, gradually establish the distributed planning and dispatching mode of power system, it has become an important direction for the sustainable development of power industry in the future [1]. As one of the main ways to realize distributed energy trading in China, virtual power plant still adopts centralized dispatching [2]. This way is not conducive to the improvement of security and fairness of energy dispatching. If the security of the dispatching control center fails, such as attack or failure, it will cause huge losses to the energy transaction. In order to solve the above problems, this paper introduces blockchain technology [3]. As a decentralized distributed accounting technology, blockchain technology can © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 S. Sun et al. (Eds.): ICSINC 2021, LNEE 895, pp. 338–344, 2022. https://doi.org/10.1007/978-981-19-4775-9_42

Multi Energy Coordinated Dispatching of Virtual Power Plant

339

timely grasp the changes of power resource schedulability, which is consistent with the principles of information equivalence, openness and transparency to be followed by distributed power energy scheduling. Based on the characteristics of distributed transaction and blockchain technology, this paper designs a blockchain framework suitable for distributed multi energy scheduling scenario. On this basis, combined with scenario analysis [4] and genetic algorithm [5], a distributed energy scheduling scheme is proposed. The scheme ensures the economy of multi energy transaction and can adapt to more complex energy dispatching environment.

2 Blockchain Distributed Energy Scheduling Framework The blockchain based virtual power plant distributed energy scheduling scheme includes three modules: user authentication, block consensus and energy dispatching. 2.1 User Authentication In the energy dispatching in this paper, the system uses SHA256 algorithm to encrypt the user’s relevant information, and the obtained hash value is used as the user’s unique identification. The encryption algorithm is shown in formula (1). SHA256 (UserInfo) = UID

(1)

where Userinfo includes enterprise ID, smart meter number and other information. After verifying that the enterprise information is accurate, the certified enterprise will obtain the key application permission provided by the VPP dispatching center, automatically generate the key locally and obtain the account address. 2.2 Block Consensus Considering that the blockchain system takes the virtual power plant as the application background, this paper improves the traditional POS (Proof of Stake) consensus algorithm, so that the system can better meet the energy scheduling requirements. The specific process of the consensus algorithm is as follows: (1) The miner node participating in the current round of transactions needs to invest a certain amount of tokens as the “coin age”, in order to become a candidate miner. With the increase of the number of mining rounds, the “coin age” will also increase. If the mining is successful, the “coin age” of miners’ tokens will be eliminated. (2) In the process of selecting accounting nodes, PoS consensus mechanism needs to comprehensively consider the deposit and currency age invested by miners to select an appropriate bookkeeping miner node M and give it the ability to generate blocks. If the selected miner cannot generate the block, POS will continue to select the next accounting node according to the weight among the candidate miners until the block is successfully packaged. (3) If the miner operates the new block according to the regulations of the system, he will get the token reward of the system. If the packaged block fails to pass the verification at other nodes, the deposit will be deducted from the bookkeeping miner node.

340

Z. Wang et al.

2.3 Energy Dispatching Miner node M will save the energy dispatching information in the form of smart contract in the blockchain. This contract includes the expected power generation, execution time, deposit amount, etc., and the funds participating in this round of transaction are frozen. After the transaction, the blockchain automatically executes the smart contract to obtain the information in the smart meter of the power generation enterprise. If the power generated by the distributed generation station is greater than the declared power of the transaction, the VPP dispatching center will store the remaining power to the energy storage facilities in the virtual power plant according to the transaction price; When the power generated by the distributed generation station is equal to the declared power, the transaction ends normally and the deposit is returned in full. When the power generated by the distributed generation station is lower than the power of the transaction, it is necessary to supply power from the external power grid and deduct the expenses from the deposit.

3 Distributed Energy Dispatching Mechanism 3.1 Cost Objective Function The mathematical model of energy dispatching of virtual power plant can be expressed as follows. In order not to lose generality, it is assumed that the power system dispatched by VPP includes controllable distributed energy units, wind wind energy units, optical energy units and other energy sources. S 

W DG ρs (Cs,t + Cs,t + CtPV + CtBT + CtE )

(2)

s=1

where S is the number of scenes; ρs represents the probability of scene s occurrence; W , C PV respectively represent the power generation cost of wind power and photoCs,t s,t electric in scene s at time t; CtDG represents the generation cost of controllable distributed energy at time t; CtBT represents the charging cost of the battery at time t; CtE represents the cost of purchased electricity at time t. 3.2 Distributed Energy Optimal Scheduling Based on the blockchain distributed energy trading framework, this paper designs a multi energy scheduling hybrid model. In this model, the power generation subject first needs to collect corresponding meteorological data according to its own power generation type. Considering that in the distributed energy power system, the output of uncontrollable energy sources such as wind energy is uncertain, the scenario analysis method is introduced into the power system. This method is an effective method to deal with stochastic programming with uncertain factors. Through the data simulation method, a set of scene sets are generated, the optimization problem with uncertain factors is transformed into the optimization problem of multiple deterministic scenes.

Multi Energy Coordinated Dispatching of Virtual Power Plant

341

This paper takes wind power as the main research object. The output of wind power is closely related to wind speed. Limited by the technical parameters of wind turbine, the variation of wind power output with wind speed shows phased characteristics. When the wind speed is lower than the cut-in wind speed vin or higher than the cut-out wind speed vout , the wind turbine will stop working. If the wind speed is between the rated wind speed vrat and the cut-out wind speed vout , the wind turbine will output at the rated power srat . In other cases, the output of the wind turbine depends on the wind speed. The relationship between the wind speed and the output power is shown in formula (3). ⎧ 0  ⎪  v < vin ⎪ ⎪ ⎨s v−vin vin ≤ v < vrat rat vrat −vin (3) Gw = ⎪ vrat ≤ v < vout ⎪ srat ⎪ ⎩ 0 v ≥ vout where srat is the rated power of the wind turbine, and vin , vout , vrat respectively represent the cut-in wind speed, cut-out wind speed and rated wind speed. Through the wind power model, we can calculate the edge generation power of the power generation subject, and the information will be uploaded to the blockchain. The accounting node of the blockchain is responsible for formulating the energy scheduling scheme with the best overall cost. In order to adapt the scheme to the complex energy scheduling environment, this paper introduces genetic algorithm. The introduction of genetic algorithm can help the system to determine the actual generation power of different generation entities more efficiently under the constraint of edge generation power. The fitness function of genetic algorithm is as follows. fitness = max(pred ) − pred

(4)

In formula (4), pred is the set of cost expectations calculated for all individuals in the current population, and the cost expectations are negatively correlated with the population fitness. Finally, the genetic algorithm is used to find the energy scheduling scheme that makes the overall generation cost of the virtual power plant the lowest in the marginal generation power range.

4 Simulation Verification 4.1 Simulation Background This paper takes Hebei Province with virtual power plant demonstration project as the research object, and obtains the meteorological data of Shijiazhuang and Zhangjiakou in Hebei Province through “Huiju data”. In order to verify the energy optimal dispatching scheme of virtual power plant, it is assumed that the configuration of power plant a is 6 × 36 MW wind turbine. Among them, the output power of the wind turbine under the rated wind speed of 7 m/S is 3 MW, the cut-in wind speed of 1 m/s and the cut-out wind speed of 10 m/s. The quadratic term coefficient of the wind power generation cost function is 1.2, the primary term coefficient is 100 and the constant term is 50. The configuration of power plant B is the same as 6 × 36 MW wind turbine. Among them, the output power

342

Z. Wang et al.

of the wind turbine under the rated wind speed of 7 m/S is 3MW, the cut-in wind speed is 1 m/s, and the cut-out wind speed is 10 m/s. The quadratic term coefficient of wind power generation cost function is 1.2, the primary term coefficient is 70 and the constant term is 0. The configuration of Plant C is 6 × 10 MW controllable energy generator set. The quadratic term coefficient of the power generation cost function is 2.5, the primary term coefficient is 80, the constant term is 0, and the power purchase price of the virtual power plant from the external power grid is 0.55 yuan/KWh. 4.2 Scene Analysis Firstly, taking the meteorological data of Shijiazhuang and Zhangjiakou 30 days before the forecast day as the experimental data, we obtain the four most likely wind speed scenarios by scenario analysis. Combined with the four specific wind speed scenarios and the distributed wind power output characteristic function formula (3), the generator unit output under different scenarios can be calculated. In this paper, the maximum unit output and minimum unit output in the four scenarios are taken as the edge generation power of the generator set. Figure 1 is the edge generation power diagrams of Shijiazhuang power plant.

Fig. 1. Total unit marginal power of power plant

4.3 Cost Calculation Considering that with the increase of the user scale of the virtual power plant, the traditional sample space traversal method to calculate the generating power per plant has high time complexity and low efficiency, genetic algorithm is introduced to quickly and efficiently obtain the optimal power generation and the lowest expected cost of the wind turbine that meet the constraints. The relevant parameters of genetic algorithm are set as follows: the population number is 200, the crossover probability is 0.8, the mutation probability is 0.005, and the number of iterations is 200. This paper takes the energy dispatching at 2:00 a.m as an example for analysis. Figure 2 shows the expected cost corresponding to different generation power of the power generation main body of the virtual power plant at 2:00 a.m.

Multi Energy Coordinated Dispatching of Virtual Power Plant

343

Fig. 2. Expected cost diagram at 2:00 a.m.

In Table 1, the error between the minimum expected cost, unit power obtained by genetic algorithm and the system optimal solution obtained by sample space traversal is very small. It can be seen that genetic algorithm is a feasible solution to solve the minimum expected cost. Table 1. Comparison table of solution methods at 2:00 a.m. Solution method

Time (h)

Wind power output of group A (MW)

Wind power output of group B (MW)

Cost (yuan)

Ergodic solution

2

2.99

35.75

64538.2

Genetic algorithm

2

3.00

35.50

64537.9

5 Summary This paper mainly studies the problems and specific solutions of distributed energy scheduling in the virtual power plant scenario. Considering the inherent consistency between the blockchain and the virtual power plant, a distributed energy scheduling framework based on the blockchain is designed. Each power generation entity jointly maintains the blockchain and completes the formulation of energy scheduling plan, so as to ensure the safety, fairness and stability of the virtual power plant. Besides, the blockchain accounting node is responsible for collecting the output information of all power generation entities in the virtual power plant, and using genetic algorithm to quickly and efficiently determine the distributed energy scheduling scheme with the lowest overall cost, so as to improve the overall benefit of the virtual power plant. Acknowledgment. This work was supported by Construction of industrial Internet platform application innovation promotion center for energy industry (No. TC200802B).

344

Z. Wang et al.

References 1. Yu, T., Feng, B., Wei, D., et al.: Source-network-load-storage coordinated optimal scheduling for active distribution network with distributed generation. Water Resour. Hydropower Eng. 52(6), 215–222 (2021) 2. Zhang, G.: Bidding Strategy and Coordinated Dispatch of Virtual Power Plant with Multiple Distributed Energy Resources. Shanghai JiaoTong University (2019) 3. Lu, Y.: The blockchain. State-of-the-art and research challenges. J. Ind. Inform. Integr. 15, 80–90 (2019) 4. Huo, H.: Muliti-objective Optimization and Policy Simulation of the Power System Scheduling Considering Virtual Power Plants. North China Electric Power University (2018) 5. Jie, Z., Bao, L., Wang, L.: Basic genetic algorithm and power grid intelligent dispatching. Science and Technology and Innovation (07), 24–25+29 (2019) 6. Jin, J.: Modeling Analysis of Economic Emission Dispatch with Wind Power. Nanjing University of Aeronautics and Astronautics (2015)

Research on Distributed Energy Trading Mode and Mechanism Based on Blockchain Chang Liu1(B) , Xuwei Xia2 , Yong Yan3 , Junwei Ma4 , and Feng Qi1 1 Beijing University of Post and Telecommunications, Beijing 100083, China

[email protected]

2 State Grid Ningxia Electric Power Co., Ltd., Electric Power Research Institute,

Yinchuan 750000, Ningxia, China 3 State Grid Zhejiang Electric Power Co., Ltd., Electric Power Research Institute,

Hangzhou 310000, Zhejiang, China 4 State Grid Shanxi Electric Power Company Information and Communication Company,

Taiyuan 030000, China

Abstract. Traditional distributed energy transactions mainly provide trading contracts by using centralized trading platforms, and provide power supply and transmission by cooperating with power grid. There are problems such as high service cost, low transaction efficiency, and lack of user privacy security. This paper puts forward a kind of suitable for distributed energy trading considered Internet fee and risk appetite of bilateral auction trading rules, and consider the risk appetite can avoid the user power purchase price is too low or distributed sell electricity price is too high can’t auction success so as to reach a deal, to ensure efficient distribution of resources and maximize the interests of both sides, Considering that the network fee can make the distributed energy as near as possible to trade, reduce the loss of energy in the transmission process, reduce waste. At the same time, this paper designs a system software architecture suitable for distributed energy transaction, and gives the corresponding complete transaction flow. Keywords: Distributed energy · Blockchain · Bilateral auction · Risk strategy · Energy trading

1 Introduction At present, the research and application of blockchain technology in the field of electricity has not been mature, but there have been many achievements. Literature 1 discussed the innovative application of blockchain technology in the field of distributed energy trading, analyzed the feasibility, and summarized the development status at home and abroad. Literature 2 introduced the new development of improving throughput performance, but did not give the specific transaction mode and method. Literature 3 adopted This work was supported by Science and Technology Program of State Grid Corporation of China. The program name is “Research and Application of Key Technologies of Distributed Power Trading Based on Blockchain”, and the program number is 5700-202035373A-0-0-00. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 S. Sun et al. (Eds.): ICSINC 2021, LNEE 895, pp. 345–350, 2022. https://doi.org/10.1007/978-981-19-4775-9_43

346

C. Liu et al.

the photovoltaic microgrid transaction game model based on the alliance chain, which reduced the transaction cost and improved the transaction price and speed of “remaining access”, but only analyzed the user side without considering the influence of photovoltaic output uncertainty on the user side. Literature 4 proposes a distributed microgrid direct transaction system, and designs two security protocols with digital signatures to resist tampering attacks and ensure the security of transaction settlement. However, we only look at the security for a certain period of time, not the entire transaction cycle. Literature 5 active power distribution network based on model predictive control algorithm more time scale coordination scheduling policy, implementation of renewable energy given to the greatest extent, although the multi-stage coordination scheduling strategy can largely distribute energy and load of volatility, but as a result of optimal operation are implemented through ring opening state, unable to realize the actual working condition of feedback correction. Literature 6 proposed a multi-micro grid economic dispatching method based on distributed convex optimization. This method takes the minimization of power generation cost as the optimization goal, and does not take the micro grid as an independent interest subject, which is not in line with the market development trend. Literature 7 puts forward the quotation strategy based on risk attitude. Literature 8 proposed the grid resource allocation based on the improved risk strategy of multi-unit continuous two-way auction. The existing blockchain-based transaction mode has the problems of high transaction cost, complex transaction mode, resource waste and low utilization rate. Aiming at the problems in the study, based on the characteristics of distributed energy trading, this paper designed is suitable for the overall architecture of distributed energy trading block chain system and the corresponding trading patterns, trading patterns and methods of the concrete are given, and the benefit of the micro network as an independent subject, reduces the transaction cost of trading main body, improve the trading main body.

2 Consider the Risk Attitude and the Quotation Strategy of the Network Fee 2.1 A Pricing Strategy that Considers an Attitude to Risk The Bidding Layer. The bidding layer is mainly for each party to submit their own offer, and first of all, they should calculate their own limit price respectively. The distributed generation party takes the larger part of the cost price Ci and the feed-in price Pb,grid of their own power generation as their limit price si , and the user purchasing party takes the selling price Ps,grid of the large grid as its limit price bj . That is:   si = max Pb,grid , Ci

(1)

bj = Ps,grid

(2)

After setting its own limit, each party has no information about other members’ offers until the new trading round. For distributed generation i, its own quotation is determined according to its own limit price si and the current optimal power purchase

Research on Distributed Energy Trading Mode and Mechanism

347

price bidb,max (k) (the highest quotation of users in the current trading round). For the power buyer j, it determines its own quotation according to its own limit price bj and the current optimal electricity selling price bids,min (k) (the lowest quotation of the power producer in the current trading round), so as to maximize the benefits of all parties.     bid s,min (k) − bid s,min(k) − max si , bid b,max (k) /ηi , (k = 1)  (3) bid s,i (k) = bid s,min (k) − bid s,min (k) − τi (k) /ηi , (k ≥ 2)      (k) /ηj , (k = 1) bid b,max (k) + min bj, bid s,min (k) − bid b,max  bid b,j (k) = (4) bid b,max (k) + τj (k) − bid b,max (k) /ηj , (k ≥ 2) k represents the trading round, η ∈[1, + ∞] represents the profit decline rate, and the smaller η is, the faster quotation convergence will be. τi (k) represents the target price of seller i in round k. τj (k) represents the target price of the power buyer j in round k. The Adaptive Layer. In the adaptive layer, the trading members use the adaptive learning algorithm to adjust their risk factors and change the target price to adapt to the market environment when the trade occurs or there is a new offer. The risk factor γ ∈ (−1, 1), when γ ∈ (−1, 0), the trading party belongs to the risk aversion type (who wants to obtain a higher probability of transaction with a lower trading profit), when γ ∈ (0, 1), the trading party belongs to the risk preference type (who wants to obtain a higher trading profit, but will reduce the success rate of the transaction). The adaptive learning process of risk factors is as follows:   (5) γ (k + 1) = γ (k) + δ(k) − γ (k) × θ δ(k) = (σ (k) + 1)×γ shout

(6)

σ (k) = {−0.05, 0.05}

(7)

Among them, γshout is the risk factor of the last quotation of the trading subject, which is mainly used to generate the target price of the trading subject. δ(k) is the expected risk factor, θ ∈ (0,1) is the learning rate of the algorithm, which affects the moving speed of the target price, where 0.5 is taken. To reduce the risk factor, the seller sets the expected risk factor to a value slightly lower than γshout , where σ = 0.05. To increase the risk factor, the seller sets the expected risk factor to a value slightly higher than γshout , where σ = −0.05. Competitive equilibrium price can be used to measure the degree of competition in the market. Its value is close to the high frequency transaction price within the estimated range, that is, the final market transaction price gradually approaches the competitive equilibrium price. The competitive equilibrium price is calculated as follows: pe =

k

(αi × pi )

i=1

αi is the weight of the transaction i, pi is the price of the transaction i.

(8)

348

C. Liu et al.

Market members are classified into two types based on how they react to market conditions: internal and external counterparties. An insider is a user whose limit price is higher than the competitive equilibrium price, or a distributed generation whose cost price is lower than the competitive equilibrium price, with a higher chance of transaction. An external party is a user whose limit price is lower than the competitive equilibrium price, or a distributed generation whose cost price is higher than the competitive equilibrium price. Trading members can calculate the target price based on competitive equilibrium price and limit price. The calculation method of insider traders’ target price is as follows:  pe + (pe − si ) × γ (k), γ (k) ∈ (−1, 0) (9) τi,inner (k) = pe + (MAX − pe ) × γ (k), γ (k) ∈ (0, 1)    pe + pe − bj × γ (k), γ (k) ∈ (−1, 0) (10) τj,inner (k) = pe × (1 − γ (k)), γ (k) ∈ (0, 1) τi,inner and τj,inner are the target prices of the seller i and the buyer j among the internal transaction parties respectively, and MAX is the maximum value of the seller’s offer among the internal transaction parties. In the calculation method of the target price of the external trading parties, it is only necessary to replace pe in the above formula with their respective limit price si , bj .

2.2 The Clearing Strategy Considered the Net Fee Calculate the Network Fee. The network fee 9 refers to the charge charged by the power company to recover the investment, operation and maintenance costs of the power grid and obtain a reasonable return on assets. Check and ratify the net fee is an urgent problem in the distributed energy trading, reasonable quantitative service value of energy transmission process and service value of distributed generation to provide power supply can effectively solve the problem of grid cost of equity and reasonable allocation of resources to reduce energy waste in the transmission process. It also can maximize the parties’ rights and interests. Here, MW-KM (Megawatt kilometer method) 10 is used to calculate the network fee. It is assumed that the distributed energy transaction is the only transmission and distribution service in the linear network, and the calculation process of the network fee is as follows: (1) Calculate the transmission and distribution capacity cost of each line Cc,w,l in the transmission network. (2) Exclude the load and generator not used for power transmission and distribution, calculate the power flow, network loss and operation cost Co,w,l of all branches of the whole network. (3) Calculate the average cost per megawatt kilometer for each line γw,l =

Cc,w,l +Co,w,l Pc,l ×Ll

(11)

Research on Distributed Energy Trading Mode and Mechanism

349

(4) Calculate the network transmission fee of item K power transmission and distribution business Fw (k) = γw,l × Pw,k,l × Ll (12) l

where Pc,l is the safe transmission power of line l, Pw,k,l is the transmission power of item k transmission and distribution service on line l, and Ll is the length of line l. Net Fee is Considered Two-Way Auction Market Clearing. Continuous bilateral auction refers to the situation in which buyers and sellers exist in a many-to-many form. Market participants can submit bids at any time. After the deadline of quotation, the bids are submitted to the auction market for matching. The buyers and sellers are matched according to the principle of “price first, time first”. When the quotations are the same, they are sorted according to the sequence of quotation submission time. The resulting clearing curve is shown in Fig. 1, where the solid line is the quotation curve of the electricity buyer and the dotted line is the quotation curve of the electricity seller.

Fig. 1. Double auction market clearing curve

The revised clearing curve after deducting network transfer fee is shown in Fig. 2. It can be seen that the consideration of network fee will affect the clearing result of bilateral auction market, so that the transaction with lower network fee can be reached, and the distributed energy can be absorbed nearby to reduce energy waste.

Fig. 2. Market clearing curve after considering internet fees

350

C. Liu et al.

3 Conclusion In this paper, introducing risk factors on behalf of the members of the market risk attitude, combined with block chain technology is put forward considering risk attitude and a net charge of bilateral bidding strategies, and design the system architecture of distributed energy trading and transaction process, experimental results show that using the proposed distributed energy trading mode, distributed energy can get to make better use of purchase, reduce the waste of resources, At the same time, compared with the transaction with the large power grid, the electricity seller can obtain greater profits, and the electricity buyer can save more electricity purchase costs.

References 1. Zhao, S., Xu, X., Wu, Z.: Innovative application of blockchain technology in the field of distributed energy trading. Electr. Appliances Energ. Effi. Manage. Technol. 11, 1–10 (2020) 2. Dorri, A., Hill, A., Kanhere, S., Jurdak, R., Luo, F., Dong, Z.Y.: Peer-to-peer energytrade: A distributed private energy trading platform. In: Proceedings of the 2019 IEEE International Conference on Blockchain and Cryptocurrency (ICBC), Seoul, Korea (South), pp. 61–64 (2019). https://doi.org/10.1109/BLOC.2019.8751268 3. Lu, J., Wang, Z.G., Yang, Q., et al.: Research on game model of photovoltaic microgrid transaction based on blockchain. Electr. Meas. Instrum. 57(11), 80–86 (2020) 4. Qi, F., Chen, L., Zoo, L., Hou, X.Y.: Distributed microgrid direct transaction system and its security protocol. Comput. Appl. Softw. 37(5), 327–333 (2020) 5. Dong, L., Chen, H., Pu, T.J., et al.: Multi-time scale dynamic optimal dispatch of active distribution network based on model predictive control. Proc. CSEE 36(17), 4609–4617 (2016) 6. Gremolata, D., Matamoros, J.: Distributed energy trading: The multiple-microgrid case. IEEE Trans. Industr. Electron. 62(4), 2551–2559 (2015) 7. Khorasani, J., Mashhadi, H.R.: A risk-based bidding strategy in an electricity multimarket. In: Proceedings of the 2011 19th Iranian Conference on Electrical Engineering, p. 1. Tehran, Iran (2011) 8. Zhao, X., Wei, C.: Grid resource allocation based on multi-unit continuous two-way auction based on risk strategy. Comput. Appl. 29(02), 602–605+610 (2009) 9. Xiao, J., Liu, R., Jing, Z., He, Y.: Analysis of improved transmission overnetwork pricing mechanism in UK. China Electr. Power 52(02), 53–60 (2019) 10. Zhong, X., Lo, K.L., Sun, J.: Research on the pricing method of transmission price. Electr. Power Equipment (06), 46–50 (2008)

BlockChain-Based Power Communication Network Cross-Domain Service Function Chain Orchestration Algorithm Xinyan Wang1 , Zheng Jia1 , Wencui Li1 , Beibei Zhu1 , Feifei Zhang1 , Ying Zhu1 , and Peng Lin2(B) 1 Information and Telecommunication Company, State Grid Henan Electric Power Company,

Zhengzhou 450000, China 2 Beijing Vectinfo Technologies Co., Ltd., Beijing 100082, China

[email protected]

Abstract. The low resource allocation efficiency of the service function chain caused by malicious nodes is a problem, a cross-domain service function chain orchestration model is proposed. The model uses the blockchain ledger to store the resource information and trust degree of each network domain, and solves the problem that the trust value of nodes is easily tampered with. Based on this model, a blockchain-based power communication network cross-domain service function chain orchestration algorithm is proposed. The algorithm includes the service function chain sending resource allocation requests to the inter-domain controller, the inter-domain controller obtains the underlying resource information, the interdomain controller arranges the service function chain according to the resource information, the inter-domain controller issues resource allocation instructions to each domain controller, and the inter-domain controller requests blockchain nodes to update their trust levels for nodes participating in resource allocation in each domain. Keywords: Power communication network · Cross-domain service · Service function chain · Resource allocation · Blockchain

1 Introduction Within power communication network, the demand in power business is increasing rapidly, network virtualization technology has adopted by power companies and equipment manufacturers [1]. In the electric power business, this service function chain has the characteristics of peer-to-peer efficient service and communication. The resource allocation in the network virtualization is a research content, and research scholars pay more attention to this research content. Literature [2] uses link sharing theory for link resource allocation. Literature [3] uses a multi-path partition strategy for optical scene. Literature [4] adopts mathematical theory and adopts a strategy of relaxing constraints to obtain an optimization strategy for resource allocation. For low calculation efficiency of integer © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 S. Sun et al. (Eds.): ICSINC 2021, LNEE 895, pp. 351–358, 2022. https://doi.org/10.1007/978-981-19-4775-9_44

352

X. Wang et al.

programming, literature [5] designs a resource allocation algorithm using dynamic programming. For improving the reliability of business-related resources, literature [6] uses heuristic algorithms to use key resources backup technology to effectively reduce degree of virtual network’s impact on the quality of service of the infrastructure provider. Literature [7] analyzes the characteristics of the 5G network, and optimizes the performance of the algorithm based on this feature.

2 Problem Description When network equipment operators provide basic equipment resources, network slicing technology is a technology recognized by industry companies. After network slicing technology is applied to basic equipment resource management, network equipment operators can get a higher utilization rate of basic equipment resources. The new network environment brought about by network slicing technology includes basic equipment network and slicing equipment network. The basic equipment network is represented by GS = (NS , ES ). The basic equipment network includes basic equipment nodes and basic equipment links. The operable attributes of the basic equipment node nsi ∈ NS are represented by cpu(nsi ). The transportable attribute of the basic equipment link ejs ∈ ES is represented by bw(ejs ). The slicing device network is represented by GR = (NR , ER ). The slicing device network includes slicing device nodes and slicing device links. The operable attributes of the slicing device node nri ∈ NR are represented by cpu(nri ). The transmittable attribute of the slice device link ejr ∈ ER is represented by bw(ejr ). This paper analyzes the characteristics of network function virtualization technology, the scalability and on-demand service characteristics of network function virtualization are discovered and used in the research of this article. The article uses Network Function Virtualization (NFV) technology to virtualize each underlying node into multiple types of network function nodes to satisfy service quality requirements. The requirements of each virtual node on the network can be solved and satisfied. The virtual node that realizes a certain function that is virtualized by each underlying node is called a NFV instance (NFVI). For a node of a service function chain, to improve the performance of service quality, when the business needs request arrives, the Internet Service Provider selects an optimal underlying node from the set of underlying nodes through a strategy formulated in advance. The service function chain node generates multiple NFVIs. The service function chain orchestration algorithm can select an optimal NFVI from these NFVIs to allocate resources for it.

3 Models and Algorithms The low trust and low availability of resource allocation algorithms in resource allocation in a cross-domain environment is a very urgent problem。This article designs a crossdomain business function orchestration model (in Fig. 1). The model includes four modules: inter-domain controller, blockchain, intra-domain controller, and underlying network. The inter-domain controller is responsible for receiving service function chain resource allocation requests, service function chain arrangement, and bottom network

BlockChain-Based Power Communication Network Cross-Domain

353

node trust management. Blockchain nodes use blockchain ledgers to store resource information and trust levels of various network domains. The controller in the domain is responsible for the orchestration. The infrastructure service provider is responsible for providing resources related to basic network equipment for the service function chain. Inter-domain controller

Intra-domain controller

Intra-domain controller

Underlying network

Underlying network

Blockchain

Fig. 1. Orchestration model of cross-domain service function chain for power communication network

This algorithm effectively reduces the influence of malicious nodes on the arrangement of network resources by modeling and analyzing the trust of each underlying node. The blockchain-based power communication network cross-domain service function chain orchestration mechanism includes the service function chain sending resource allocation requests to the inter-domain controller, the inter-domain controller to obtain the underlying resource information, and the inter-domain controller according to the resource information. The service function chain is arranged, the inter-domain controller issues resource allocation instructions to each domain controller, and the blockchain node updates the node’s trust value.

4 Key Process 4.1 Credit Evaluation of Network Nodes When network service providers are performing resource management, the occurrence of viruses in network equipment is their most relevant problem. After a virus occurs in a network device, the harm that these network devices can bring to the business and the network is very serious. Moreover, it is difficult to give clear solutions to these serious consequences within a certain period of time. The malicious nodes report the false amount of CPU to their controller in this domain, causing the resource allocation algorithm to fail due to insufficient underlying resource capacity. For avoiding this occurrence failures caused by these false alarms of malicious nodes, it is necessary to analyze the resources authenticity. Using ki to represent the initial information of the trust level of the nsi ∈ NS . The initial information of each bottom node is related to its location and performance.

354

X. Wang et al. i Use fjSFC to represent the j-th service function chain, and use Tft,n to represent the j

credibility when the underlying network node nsi ∈ NS allocates resources to fjSFC at time t, and use formula (1) for calculation. i i Tft,n = F(Tft−1,n , fjt−1,SFC ) j j

(1)

i Among them, F(Tft−1,n , fjt−1,SFC ) represents the bottom-level node trust update j function, which is calculated by formula (2). Mapf t−1,SFC = yes indicates that at time t-1, j

the underlying network node nsi ∈ NS successfully allocates resources to the virtual nodes in fjSFC . Mapf t−1,SFC = no indicates that at time t-1, the underlying network node nsi ∈ NS j

cannot successfully allocate resources to the virtual nodes in fjSFC , resulting in fjSFC ’s resource allocation failure. λyes represents the reward value obtained by the underlying network node after successfully allocating infrastructure. λno represents the penalty value that the underlying network node needs to receive after failing to successfully allocate infrastructure once.  t−1,ni Tfj + λyes , Mapfjt−1,SFC = yes t−1,ni t−1,SFC (2) F(Tfj , fj )= t−1,ni Tf − λno , Mapf t−1,SFC = no j

j

After obtaining the current trustworthiness value of the underlying node, it can be compared with the trustworthiness threshold to determine whether the node is a dishonest node. Use k Th to represent the trustworthiness threshold of the underlying node. When the trust degree of a node is greater than k Th , so this node is a trusted resource and can be used as a candidate node to participate in resource allocation. Otherwise, this node is a malicious resource and cannot participate in the resource allocation work. 4.2 Node Trust Value Consensus Algorithm From the service function chain orchestration sub-algorithm process, so the trust value of the basic network equipment is very important. If the trustworthiness value of the node cannot be stored safely, it is easy to be tampered with, deleted, etc., which will bring about a failure. In this existing research, two strategies are usually adopted for each bottom node to save it by itself, and to create a trust center to save it. Both of these strategies can easily lead to tampering and deletion of the value of trust. To solve this problem, this article uses the decentralization advantages of blockchain technology and uses blockchain modules to preserve the trust of nodes. The resources allocation in a cross-domain environment is a problem. So, each domain participating in resource allocation is fixed before resource allocation. Through analysis, this article uses alliance chain technology to build blockchain nodes. Through the analysis of the alliance chain technology, it can be known that the consensus algorithm needs to be optimized according to the problems solved in this article. This article uses Practical Byzantine Fault Tolerance (PBFT) as the basic consensus algorithm, and applies it to the node trust consensus process in this article.

BlockChain-Based Power Communication Network Cross-Domain

355

In the process of generating the update request by the inter-domain controller, the t inter-domain controller generates the bottom-level node trust update request Tupdata according to the result of each SFCR resource allocation. The request information includes the current SFCR resource request information, the resource allocation result, and the successful resource allocation. The collection of nodes and the collection of malicious nodes. In the process that the inter-domain controller sends an update request to the master node of the alliance chain, the inter-domain controller requests the master node of the alliance chain to implement a consensus mechanism. In the process of verifying data by each blockchain node, the consortium chain master node sends a consensus request to all blockchain nodes. In the process of writing data after receiving the message confirmation, after receiving other 2f different node confirmation messages, the node trust in the storage area is trusted according to the value of the node trust.

5 Simulation Environment and Algorithm Comparison When comparing the simulation environment and the algorithm, use the GT-ITM [8] tool for experiments. Because the negative impact of malicious nodes on the network is very large, this article mainly analyzes how to avoid the negative impact of malicious nodes. In order to analyze the performance of the algorithm in different scenarios, it is necessary to analyze the performance of the network according to the number and location of malicious nodes. Because the greater the number of networks, the greater the number of potential malicious nodes, and the greater the negative impact on the network. Because the larger the network, the greater the number of nodes in the network. Therefore, the simulation environment needs to analyze the algorithm in combination with different network scenarios. In order to analyze the malicious nodes of the network, it is necessary to select some nodes from the network nodes as malicious nodes for simulation. In the simulation environment, according to different network scale environments, a random selection strategy is adopted to simulate different numbers of malicious nodes. When comparing method performance, choose the appropriate algorithm to compare with the algorithm PCNCSFCOAoBC in this paper. After analyzing and selecting from multiple dimensions, this article uses the RAAPCNSFCoGA algorithm as the comparison algorithm. Because the algorithm RAAPCNSFCoGA uses a greedy strategy to achieve infrastructure management. Therefore, the RAAPCNSFCoGA algorithm can represent the general idea of infrastructure management. In the management of infrastructure resources, the effective use of infrastructure resources is a very important issue for relevant network operators. Through years of work experience and customer demand analysis results, it can be known that the higher the utilization rate of infrastructure, the more benefits it can bring to infrastructure providers. Based on the above analysis, Fig. 2 verifies the feasibility of infrastructure resource allocation under the respective algorithms. It can be seen from the figure that the algorithm in this paper is more efficient in infrastructure management than the comparison algorithm. Therefore, the algorithm in this paper can bring better benefits to infrastructure service providers.

356

X. Wang et al.

Fig. 2. Feasibility analysis of infrastructure resource allocation

Because the ideal goal of infrastructure resource management is to make better use of infrastructure resources. In order to verify whether the infrastructure resources are being used well, the two dimensions of the node resources and link resources of the infrastructure are analyzed in Fig. 3 and Fig. 4. From the analysis results in Fig. 3, it can be seen that although the resource utilization rate of the infrastructure nodes of the two algorithms is decreasing. However, the utilization rate of infrastructure node resources under this algorithm is higher than that of the comparison algorithm. It shows that the node resource utilization result of the algorithm in this paper is better than the comparison algorithm. It can be seen from the analysis result in Fig. 4 that although the utilization of infrastructure link resources of the two algorithms is decreasing. However, the utilization rate of infrastructure link resources under this algorithm is higher than that of the comparison algorithm. It shows that the link resource utilization result of the algorithm in this paper is better than the comparison algorithm.

Fig. 3. Node resource utilization of infrastructure resources

BlockChain-Based Power Communication Network Cross-Domain

357

Fig. 4. Link resource utilization of infrastructure resources

6 Conclusion In the process of network operation, the efficient use of infrastructure is one of the issues that network operators are most concerned about. Regardless of the utilization rate of the infrastructure, if the network service quality provided by the network operator is low, the negative impact on the network operator will be very large. In order to improve the network operation service quality of network operators, the security management of network equipment is a very important task. This paper studies the mining mechanism of malicious nodes from the point of view of identifying malicious nodes. Based on the mining results of malicious nodes, normal infrastructure resources can be used in advance to provide users with highly reliable infrastructure resources. Through simulation experiments and algorithm comparison analysis, the advantages of malicious node identification for network operators are verified. Acknowledgement. This work was supported by the State Grid Henan Electric Power Compan Science and Technology Project “Research and application of energy big data directory chain technology based on blockchain” (B117Q021K003).

References 1. Aijaz, A.: Hap-SliceR: A radio resource slicing framework for 5G networks with haptic communications. IEEE Systems Journal PP(99), 1–12(2017) 2. Mijumbi, R., Serrat, J., Gorricho, J.L., Boutaba, R.: A path generation approach to embedding of virtual networks. IEEE Trans. Netw. Serv. Manage. 12(3), 334–348 (2015) 3. Soto, P., Botero, J.F.: Greedy randomized path-ranking virtual optical network embedding onto EON-based substrate networks. In: Proceedings of the 2017 IEEE Colombian Conference on Communications and Computing (COLCOM), pp. 1–6. IEEE, Colombia (2017) 4. Chowdhury, S.R., Ahmed, R., Shahriar, N., Khan, A., Boutaba, R., Mitra, J., Liu, L.: Revine: Reallocation of virtual network embedding to eliminate substrate bottlenecks. In: Proceedings of the 2017 IFIP/IEEE Symposium on Integrated Network and Service Management (IM), pp. 116–124. IEEE (2017)

358

X. Wang et al.

5. Dehury, C.K., Sahoo, P.K.: DYVINE: Fitness-based dynamic virtual network embedding in cloud computing. IEEE J. Sel. Areas Commun. 37(5), 1029–1045 (2019) 6. Zheng, X., Tian, J., Xiao, X., Cui, X., Yu, X.: A heuristic survivable virtual network mapping algorithm. Soft. Comput. 23(5), 1453–1463 (2018). https://doi.org/10.1007/s00500-0183152-7 7. Raza, M.R., Fiorani, M., Rostami, A., Öhlen, P., Wosinska, L., Monti, P.: Dynamic slicing approach for multi-tenant 5G transport networks. IEEE/OSA J. Opt. Commun. Networking 10(1), 77–90 (2018) 8. Zegura, E.W., Calvert, K.L., Bhattacharjee, S.: How to model an internetwork. In: Proceedings of IEEE INFOCOM (1996)

Research on Intelligent Intrusion Detection Method of Power Information Network Under Cloud Computing Chenyi Xia(B) , Feixiang Ao, Jianping Xu, and Baoming Yao State Grid Yingtan Electric Power Supply Company, Yingtan 335000, China [email protected]

Abstract. Aiming at the problems of large workload, wide control range, complexity and prone to management vulnerabilities in the power information platform, this paper proposes an intelligent intrusion detection method based on cloud computing, and a cyclic intrusion detection model is designed based on the full connected recurrent neural network, finally, the anomaly nodes in the network are labeled by the forward propagation algorithm. Experimental results demonstrate that the proposed detection method can effectively detect different types of network anomaly nodes, and the security of power information network can be improved. Keywords: Power information platform · Intrusion detection · Neural network

1 Introduction With the continuous in-depth integration and development of cloud computing technology and daily work, study and life, the power system is gradually moving towards informatization and digitalization, which has become an indispensable part of the development of various enterprises at the present stage [1]. For national economic and social development [2], the power grid is an important infrastructure, intrusion detection system and encrypted authentication device [3]. However, the security of power information network has been greatly threatened under the cloud computing, there are still some problems that the safety accident cannot be found timely and accurately. Each security incident is isolated and there is no complex relationship between them. Incidents often occur in places where the process of the entire security incident cannot be traced and documented in time [4]. In order to solve the above problems, it is necessary to achieve a centralized monitoring of various network and system security resources. Therefore, this paper proposes an intelligent intrusion detection method of power information network based on the cloud computing. The method can improve the level of information system security, and closely combined with the actual situation through beneficial exploration and practice so as to achieve the desired effect.

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 S. Sun et al. (Eds.): ICSINC 2021, LNEE 895, pp. 359–364, 2022. https://doi.org/10.1007/978-981-19-4775-9_45

360

C. Xia et al.

2 Intelligent Intrusion Detection Method for Power Information Network Under the cloud computing 2.1 Intelligent Detection Based on Message Queuing Network Message processing technology provides message processing and message queuing functions [5]. It is known that the message queue can send retrieval information to each other [6] to enhance the identification and scanning of vulnerabilities in the process of intelligent intrusion detection of power information networks. On this basis, the detection is divided into two aspects: the overall detection of the power information network space and the detection of the host. Vulnerability detection of power information networks includes network mapping and port detection. The port is a vulnerable communication channel so a TCP session is used to connect to the target host in the process of port detection, according to the feedback information of the host, it analyzes whether there is vulnerability in the platform timely. Among the above mentioned intelligent detection methods, the message queue is used to record the detection results. Therefore, it is ensured that the method can realize one-to-one superimposed power information detection without omitting any space information points. 2.2 Intelligent Detection Vulnerability Risk Situation Intelligent vulnerability risk situation estimates the risk value by considering risk transmission factors, so as to obtain the network security trend of the whole power information platform [7]. Among them, the overall vulnerability risk of power information platform belongs to the vulnerability risk of confidentiality, integrity and usability. Assume that the vulnerability risk is represented by L(a1 , b1 , c1 ), where a1 is the attack, b1 is a service, and c1 is the weakness of b1 . Therefore, there is the formula: (1) L(a1 , b1 , c1 ) = G        In formula (1): G ∈ x , y , z represents the risk impact category; x , y and z represent the riskvalues ofconfidentiality, integrity and availability, and there is a security    attribute set q = x , y , z for the power information platform s. At this time, the direct risk value faced by s is: 



Fx (s) = x + y + z



(2)

Through the intelligent detection of the vulnerability risk situation, the reliable procedure of the detection results is strengthened.

3 Optimization of Intelligent Intrusion Detection Methods for Power Information Networks 3.1 Calculate Node Exception Score and Set Intelligent Detection Threshold According to the power information obtained, the node anomaly score is calculated and a detection threshold is set. Assuming that the statistical features of the node s extracted

Research on Intelligent Intrusion Detection Method of Power Information

361

by the feature extraction function Hs are recorded as Rtz , there are: Rtz = Hs × Wi

(3)

In formula (3): Wi represents a multidimensional vector. For all nodes, the statistical feature set is represented by F = (f1 , f2 , . . . , fi , . . . , fn ). Figure 1 shows the feature structure of power information network node.

Consistent access interface

Comprehens ive inquiry

Resource management

Task scheduling

Security mechanism

Inquiry service

Notification service

Core layer Artificial intelligence technology Resource layer Power information resources and storage resources

Fig. 1. Diagram of feature structure of power information network nodes

According to the feature structure shown in Fig. 1, and the feature fi of i has certain rules or specific spatial distribution. The anomaly score yi is used to express the degree of deviation between abnormal node i and normal nodes, and the calculation formula of the difference between abnormal node i and normal node is as follows: yi = Ddistance (Fi , θ )

(4)

In formula (4): Ddistance (∗) is the calculation function of deviation degree; θ is linear fitting. According to the anomaly detection principle, combined with formula (5), the node anomaly score is calculated as follows:      max Rp , Mik  k R (5) − M × log Ydf =    p a i +1 k min Rp , Mi In formula (5): Rp represents the number of network edges; k is the fitting parameter of power-law distribution; Mik represents node data. Ydf reflects the feature deviation degree of abnormal nodes. Generally, if the score Ydf exceeds the threshold σ , the node is proved to be an abnormal node, the node df is considered to be an abnormal node.

362

C. Xia et al.

3.2 Design of Intrusion Detection Mode by Fully Connected Recurrent Neural Network According to the forward propagation algorithm, the nodes in the network are labeled; The difference between the node features and the threshold is calculated, and the backpropagation fine-tuning algorithm is used to adjust the shared detection parameters in the fully connected recurrent neural network. The power information data of each node obtained by intelligent detection is input into the fully connected recurrent neural network. The types of abnormal nodes are analyzed, and the intrusion categories of abnormal nodes are distinguished, the cyclic detection of node sequence information of the whole network is realized. The specific implementation steps of the forward propagation algorithm are as follows: Step 1: power information network data input: the node information is xi , (i = 1, 2, …, N ), the weight matrix is A1 , A2 , A3 , the offset is p, activation function J is sigmoid function [8], classification function L is softmax function. Step 2: Detection output: the output value corresponding to xi is si . The specific formula is as follows: Di = A1 xi + A2 Ci + p1

(6)

In formula (6): A1 represents the weight from the input layer to the hidden layer; A2 represents the weight from the previous temporal hiding layer to the current temporal hiding layer; A3 represents the weight from hidden layer to output layer; p1 is the offset of the abnormal node in the hidden layer; Ci is the hidden layer node; Di represents the information input quantity of i node at D time; It is known that the objective function is as follows: g(α) = T (yi , si )

(7)

In formula (7): T represents the adjustment function; α is the target parameter. Step 3: input data information (xi , si ), i = 1, 2, ..., n; The detection

target α is initialized; After fine tuning the detection target α = A1 , A2 , A3 , p1 , p2 is obtained; The output value si of xi is calculated by forward propagation algorithm; Then the cross entropy between the output value and the label value is calculated. For each network parameter αi in α, the partial derivative is calculated;

4 Experimental Analysis The KDD Cup2019 dataset is used as the experimental dataset. There are 10 intrusions in this dataset. The types of intrusion data are DOS, R2L, U2R, port scan and vulnerability scan (PROBE). The power grid data for January to July 2020 of the city shall be selected. The specific data information is shown in Table 1:

Research on Intelligent Intrusion Detection Method of Power Information

363

Table 1. Experimental data Month Example Heterogeneous data/piece

Data of different deparments/piece

1

11025

5358

1058

2

9845

7895

20145

3

5895

12304

12014

4

2257

2369

6951

5

4892

8654

8231

6

7589

8562

21045

7

6985

7541

10023

The 4 groups of 4000 network sample subsets are randomly extracted from the KDD Cup2019 dataset, and 3000 normal data, 50 intrusion data, and 4 subsets have the same intrusion data type distribution in each subset. The proposed method is applied to cluster intelligent detection of intrusive data in 4 subsets. The results are shown in Table 2: Table 2. Detection results by the proposed method Sub dataset coding

Difference unit

A

15

B

9

C

20

D

5

DOS detection value unit

R2L readings unit

Difference unit

U2R readings unit

Difference unit

The PROBE readings unit

Difference unit

0

5

0

10

0

10

0

1

10

0

20

0

10

0

0

5

0

5

0

20

0

0

15

0

15

0

15

0

After detecting the intrusion data on the four subsets, the testing index mainly reflects the loss of network data. The testing result is shown in Fig. 2: It can be seen from Fig. 2 that the proposed method has less loss, specifically, on the four subsets, there are only 5 data losses at most, and most of the data losses are less than 5, which indicates that the proposed method has better protection for network data integrity when detecting intrusion.

364

C. Xia et al.

Loss data number/Unit

40 35 30 25

DOS R2L

U2R PROBE

20 15 10 5 0

500

1500 2000 2500 3000 3500 4000 The amount of data/Unit

Fig. 2. Data loss test results

5 Conclusion This paper proposes an intelligent intrusion detection method of power information network based on cloud computing. The advantages of the method are as follows: (1) There is only one difference between the detection result of the proposed method and the set value of the DOS intrusion data; (2) After the method has detected the intrusion data on the four subsets, the data loss on the four subsets is small and the value of data loss is less than 6; However, it also needs to focuses on the evaluation criteria of power information system operation and maintenance data integration.

References 1. Peri, M., Ani, I., Vladimir, N.: Life-cycle cost assessments of different power system configurations to reduce the carbon footprint in the Croatian short-sea shipping sector. Renew. Sustain. Energy Rev. 131(10), 110028 (2020) 2. Roselli, C., Marrasso, E., Tariello, F., et al.: How different power grid efficiency scenarios affect the energy and environmental feasibility of a polygeneration system. Energy 201(7), 117576 (2020) 3. Nancy, P., Muthuraj, K.S., Ganapathy, S., et al.: Intelligent intrusion detection system using fuzzy and deep learning approach for wireless sensor networks. IET Commun. 14(5), 123–130 (2020) 4. Juma’H, A., Alnsour, Y.: How do investors perceive the materiality of data security incidents. J. Glob. Inf. Manag. 29(6), 59–63 (2021) 5. Buetas, E., Abad, I., Cerrada, J.A., et al.: Message queuing telemetry transport (MQTT) security: A cryptographic smart card approach. IEEE Access 22(6), 115051–115062 (2020) 6. Fu, G., Zhang, Y., Yu, G.: A fair comparison of message queuing systems. IEEE Access 12(99), 1–10 (2020) 7. Zhang, J.G.: Correlation simulation of malicious behavior characteristics in network situational awareness system. Comput. Simul. 036(011), 280–283, 351 (2019) 8. Jgla, B., Xjyab, C., Yyfb, C., et al.: New fractional derivative with sigmoid function as the kernel and its models. Chin. J. Phys. 68(12), 533–541 (2020)

Research on Cloud-Edge Collaboration Architecture for Intelligent Acquisition of Digital City Information Based on 5G Customized Network Peng Ding, Qiuhong Zheng(B) , Yun Shen, Dan Liu, and Shuntian Feng China Telecom Corporation Research Institute, Beijing, China {dingpeng6,zhengqh}@chinatelecom.cn

Abstract. The construction of digital city is based on the information of the city. The aggregation and application of these massive city data need more intelligent top-level design. This paper proposes the design concept of “1 + 1 + N” for the intelligent acquisition of digital city information. Based on the 5G SA architecture, it proposes a cloud-edge collaborative architecture for the intelligent acquisition of digital city information based on 5G customized network, and builds a set of city multiple information perception platform, and introduced its two application schemes in urban governance scenarios Taking data collection and standardization as the core, data open sharing and enabling application as the concept, this paper is a positive practice of comprehensively using 5G communication technology, big data, Internet of things, cloud computing, edge computing and other new generation of information technology to help digital city construction. Keywords: Digital city · 5G customized network · Cloud-edge collaboration · Edge computing

1 Introduction Digital city construction plays a pivotal role in national economic and social development. Promoting the construction of digital cities is an urgent need to improve the level of urban informatization. The use of information technology to collect, process, mine and display city data will strengthen the city’s informatization capabilities and provide guidance for city planning, construction and management. Among them, the intelligent acquisition of data is the foundation of building a digital city, which is particularly important for subsequent application development and business analysis services. The collection of urban data needs to rely on a large number of Internet of Things devices, while wired networks and Wi-Fi networks have problems such as limited line layout and unstable signals, which cannot meet the network needs of a large number of data collection tasks. 5G has the characteristics of low power consumption and large connection, supports the connection of more than 100 billion network devices, and

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 S. Sun et al. (Eds.): ICSINC 2021, LNEE 895, pp. 365–376, 2022. https://doi.org/10.1007/978-981-19-4775-9_46

366

P. Ding et al.

Fig. 1. China Telecom 5G customized network.

meets the connection density index of 1 million per square kilometer [1]. The largescale application of 5G technology has brought opportunities for the construction of digital cities. As shown in Fig. 1, China Telecom’s 5G customized network is based on the 5G SA architecture and has three deployment modes, providing a 5G network with large bandwidth, low latency, high security, full coverage, flexible customization, and one-stop cloud-network-Application overall service. At present, the implementation of urban information collection mostly adopts a cloud-end architecture, and a large amount of data collected by end-side devices are uploaded to the central cloud for storage and processing. In fact, the processing of a large amount of collected data does not require much computing power, and uploading the data to the cloud indiscriminately will waste a lot of network bandwidth resources and cloud computing power resources. At the same time, services with high real-time requirements cannot meet their low-latency requirements in this scenario. Therefore, it is considered to introduce an intelligent urban information acquisition architecture based on cloud-edge collaboration, and sink the MEC (Multi-access Edge Computing) and UPF (User Plane Function) to the edge to relieve the pressure of uploading data to the cloud and enhance the execution efficiency of edge services. This paper will propose a 1 + 1 + N construction concept for the intelligent acquisition of digital city information, that is, to build “1” cloud-side collaborative architecture for intelligent acquisition of digital city information based on 5G customized network, and build “1” multi-source perception platform (Hereinafter referred to as “perception platform” and “this platform”), empower and connect “N” industry applications.

2 Architecture 2.1 Cloud-Edge Collaborative Deployment Architecture As shown in Fig. 2, the cloud-edge collaborative deployment architecture mainly includes the cloud side, the edge side and the end-device side. Cloud Side: Centralized management and control. The central cloud is the core provincial-level data fusion center. Its powerful computing, storage, and security capabilities provide a powerful capability warehouse and security guarantee for MEC edge

Research on Cloud-Edge Collaboration Architecture

367

applications. The perception platform deployed therein provides IoT device management, data aggregation, application development and business analysis services, and realizes the orchestration and management of provincial edge services.

Fig. 2. Cloud-edge collaborative deployment architecture.

Edge Side. Multi-level edge. In order to realize the provincial-level IoT perception, it is necessary to deploy edge nodes at multiple levels to achieve efficient aggregation and control of terminal data. On-site exclusive-level edge nodes are access nodes deployed in various actual business sites and used to carry various exclusive business applications. Shared edge nodes in counties and cities are regional nodes deployed in counties and cities in the province to carry shared services in the region [2]. Edge nodes can take many forms, such as edge intelligent servers, edge intelligent controllers, and edge intelligent gateways. In addition, edge nodes can use 5G modules to communicate with the cloud using 5G networks. End-Device Side. Comprehensive IOT perception. End devices refer to massive IoT terminals distributed in cities, such as cameras, sensors, VR devices, etc. Devices is accessed through MQTT, Modbus, Bluetooth, HTTP, UDP, TCP and other protocols to realize real-time aggregation of IoT data and terminal control.

2.2 Cloud-Edge Collaborative Architecture for Intelligent Acquisition of Digital City Information Figure 3 shows the cloud-edge collaborative architecture for intelligent acquisition of digital city information. The cloud side has three parts: the city data analysis platform, the resource and application management module and the device management module. The city data analysis module obtains data from the device management module, performs data cleaning and analysis, and conducts the full life cycle management of the

368

P. Ding et al.

data, including back-end services such as model training, cloud storage, and data analysis. The resource and application management module uses cloud native technology to manage edge nodes and edge applications. Taking city devices as K8S resources, the device management module uses K8S’s Operator to provide data collection, transmission, reception, and storage capabilities, as well as the ability to synchronize the status of cloud-side equipment.

Fig. 3. Cloud-edge collaborative architecture for intelligent acquisition of digital city information.

Message transmission between the cloud side and the edge side is based on the high-reliability message transmission modules located at both ends. A long link channel based on TCP can be established to realize the communication between edge application services and cloud application services. It is a directional full-duplex communication channel that allows any party to send a request or message to the other party between the cloud and the edge. And it supports multi-tenancy.

3 Multi-source Perception Platform 3.1 Platform Architecture Based on the aforementioned cloud-edge collaborative architecture, a multi-source perception platform construction scheme is proposed. This platform adopts a microservice design and a layered implementation mode. This mode can reduce the degree of coupling between modules and has flexible plug-and-play capabilities. At the same time, the capabilities of each module can be dynamically deployed in the form of microservices according to business scenarios to achieve business access capabilities and system performance improvements. The architecture of this platform is shown in Fig. 4.

Research on Cloud-Edge Collaboration Architecture

369

Fig. 4. Multi-source perception platform architecture.

Perception Layer. Obtain urban operation data through various sensor devices deployed in the city, and form the collection of monitoring information in various fields such as urban lifeline, urban components, ecological environment, public safety and services, communities and households. Network Layer. The network layer is the link that connects the perception layer and upper-layer applications. On the one hand, when the perception layer collects data, it first performs edge micro-model calculations through the smart gateway, and then transmits the perception layer data to the upper layer through the transmission network for corresponding processing; on the other hand, the control instructions issued by the upper layer are notified to the device to achieve Control of equipment. Adaptation Layer. The platform supports cross-operator, multi-standard, and multiprotocol adaptation, including general protocol adaptation, private protocol adaptation, and inter-network data access adaptation, etc., to analyze perception data (hereinafter referred to as the southbound interface for docking with the perception layer). Data Layer. The data layer is divided into device data and business data. Device data is related to end devices, including device connection, device analysis, and device interaction, etc. through the acquisition, storage, fusion, and forwarding of collected data to realize the aggregation and processing of terminal data. Business data includes perception data, log data, management data, alarm data, etc. Based on massive IoT data, create industry templates, build rule engines, and realize the generation and management

370

P. Ding et al.

of industry templates. At the same time, the central model is used to conduct in-depth analysis and mining of industry data to support upper-level application services. Use the collected city data to train and update the cloud model, and deliver the trained model to the edge to realize the collaboration between the center and the edge model. Service Layer. Discover new scenarios and new applications with data interaction and flow. Provide capability opening services, including: providing data push-based API interface services; providing corresponding services to urban decision makers, data owners, and application developers through a unified service portal, thereby realizing various smart application systems in digital cities Support (hereinafter referred to as the northbound interface with the application layer); in addition, provide a data interface with the big data center to realize the reporting and exchange of IoT perception data. Application Layer. Use the capabilities provided by this platform for vertical application development.

3.2 Data Processing Flow The platform’s processing flow of IoT perception data is shown in Fig. 5.

Fig. 5. IoT perception data processing flow.

As shown in Fig. 5, the data input of this platform provides a unified business portal on the data collection interface, which can be a domain name or several service IPs. The terminal or platform of the reporter actively reports the IoT-related data to the business portal. This platform can use the load balancing of the cloud platform to achieve random distribution of services, and the platform itself also has load balancing capabilities and intelligent business separation capabilities. At the entry module, the platform distributes requests from different industries to different microservices for processing, and different microservices are independent of

Research on Cloud-Edge Collaboration Architecture

371

each other, and jointly carry all businesses to achieve cluster processing. In addition, the upgrade of certain microservices does not affect the normal operation of other microservices, which facilitates later functional upgrades on demand. Data preprocessing includes data analysis, cleaning, supplementation, correlation, calculation, conversion, etc. Data analysis is based on data templates, interpreting all the reported data, and then cleaning invalid data through rules, and further processing the valid data. Associate, supplement, and convert data according to requirements, and finally achieve data formatting and normalization. One part of the preprocessed data is stored in the message middleware as the data source for northbound real-time push. The other is stored in a hashed manner according to the data usage and business identification, and stored in a file, a relational database, a non-relational database, or a memory component. These data are used as the data source for historical data acquisition, or as a basis for model training and data mining. Units that need to use data can subscribe to data in a variety of ways, including real-time push, timed push, API call, etc. Real-Time Push. Real-time push is suitable for time-sensitive business scenarios, such as smoke alarms. The data service module of this platform reads the latest business data from the message middleware in real time, and then pushes the data to the subscription platform in real time according to the subscription demand. Through streaming technology, the total delay from acquiring data to pushing data on this platform can be controlled within one second. Timed Push. Timed push is suitable for business scenarios that are not sensitive to time, such as data statistics. The platform regularly pushes relevant data to the subscription platform based on subscription requirements. API Call. This platform also provides API capabilities for external services. The subscription platform can actively call APIs to obtain data, such as obtaining inventory device data, yesterday’s business data, and single device business data. This platform sets strict authority control on data subscriptions. The above three methods of obtaining data all require permission management for data subscribers. When subscribing, you must first submit a data subscription application, specify the required data type, acquisition method, etc., and then the operator will approve the application form. After the approval is passed, the subscriber will have the data acquisition authority and method. After obtaining the data, relevant units can develop upper-level applications based on the obtained data.

3.3 Platform Capabilities Establish a multi-source perception platform, promote the resource utilization of physical equipment, realize the unified sharing and openness of perception data, and provide unified access to urban basic components and perception facilities and centralized management services throughout the life cycle. Based on 5G customized networks and public networks, it can provide IoT device management, data aggregation, application

372

P. Ding et al.

development and business analysis services to the whole province, forming a centralized management portal and data collection hub for urban IoT devices. Through “unified device identification, unified device access, unified IoT data standards, and unified resource sharing”, a map of urban IoT resources is constructed to realize the comprehensive perception, overall management and maintenance of IoT devices, and achieve high efficiency of IoT data Converge and share to support the “synchronous planning, simultaneous design, simultaneous construction, and simultaneous development” of smart cities. In summary, the perception platform provides the following capabilities: Terminal Access and Management. It has the ability to access and manage multistandard and multi-protocol IoT terminals from different manufacturers, different models and multiple operators. Device access supports protocols such as Modbus, MQTT, HTTP, UDP, TCP. It has the unified access and data collection capabilities of millions of Internet of Things terminals. Unified Data Specification. Develop a trinity standard specification of “access, data, and sharing”, unify the IoT data resource catalog, standardize device access standards and data open standards, ensure the uniformity and openness of the platform, and have the ability of unified centralized storage, cleaning and processing of massive data of IOT terminals and industrial data standardization. Unified Data Push. With unified data push capabilities, through strict data authority definitions, it realizes the open sharing and management capabilities of IoT data for vertical applications, tenants, and industry executives. Unified Portal Interface. Build a unified open portal for IoT resources to realize visibility, control and full life cycle management of devices. VPN module can be added, and administrators or users can access the portal through the VPN private network.

4 Implementation 4.1 Comprehensive Management of Urban Environmente Environmental governance needs to strengthen multi-organization and multi-department collaboration to achieve comprehensive prevention and control. At present, the acquisition of environmental data is the main problem that restricts environmental governance. The monitoring and prevention and control systems of each department are independent, and the data collection standards are different. It is impossible to integrate various environmental data to realize the monitoring and management of the global environmental state. Based on a multi-sensing platform and using multi-dimensional environmental monitoring data, a comprehensive management plan for the urban environment as shown in Fig. 6 is proposed. The sensing platform provides a data base and provides data collection, storage, fusion, and standardized services for upper-level applications. With its advantages in global data perception, it can achieve wider data acquisition, more intensive infrastructure construction, and more comprehensive data sharing applications.

Research on Cloud-Edge Collaboration Architecture

373

Fig. 6. Comprehensive management system of urban environment.

Urban environmental governance involves many aspects of data, mainly including: basic support data, water regime monitoring data, air quality data and environmental sanitation operation data [3]. Basic support data is the basis for realizing a map of environmental supervision. Geographic data and remote sensing images provide support for the positioning of environmental monitoring equipment and the trajectory planning of dynamic monitoring equipment; meteorological data provide assistance for fusion perception analysis and early warning. In terms of water treatment, it is necessary to improve the sewage treatment system engineering, strengthen the management of water sources, and transform urban drainage facilities. Various water quality data such as PH value, temperature, dissolved oxygen, turbidity, and organic carbon content are collected by various special sensors. By setting fixed-point cameras or drones for picture and video collection, real-time observation and VR real-time perception are realized, and AI analysis is carried out through highdefinition pictures and videos to realize the improvement of the accuracy of pollutant identification and pollution source analysis. Using the 5G network, different monitoring equipment can be set up according to the needs of different monitoring points. The inspection boat equipped with a multi-functional water quality detector collects information such as swimming water quality and water level to realize water source inspections

374

P. Ding et al.

and waterlogging disaster warning; sensors placed in fixed points such as smart manhole covers or urban sewers monitor the underground water level to realize urban roads Timely warning and response to stagnant water [4]. For atmospheric monitoring, static monitoring and analysis are difficult to achieve precise monitoring of atmospheric data. 5G networks, edge computing, and big data technologies should be comprehensively used to collect, process, and excavate the pollutants themselves and their reaction mechanisms. Digging to achieve accurate control of atmospheric information [5]. Real-time video AI analysis of emissions and emission gas composition monitoring are carried out for factory emissions, so as to capture and fixed-point remediation of factories that exceed emission standards. Carry out allweather monitoring of road dust and exhaust gas to refine the construction of urban road pollution models, and focus on prevention and control of heavily polluted road sections and time periods. Monitor the content of atmospheric dust particles and the content of SO2 , NO2 , CO2 and other gases in the air in the city and its surrounding areas, perform big data analysis on the collected air quality data, and determine the source of pollution based on a certain space node to determine the atmosphere for a period of time in the future The quality change trend is predicted. Sanitation work is an important part of urban environmental governance. Tag the scattered trash cans, garbage stations, transfer stations, compression stations, sanitation trucks, etc. with electronic tags, digitally manage sanitation personnel and sanitation facilities, and establish a unified operation file for their location, status, and cleaning process to facilitate management Supervision and mobilization. Use drones, road monitoring and other return videos to intelligently analyze the road clearing situation, automatically generate work orders for road sections that do not meet the requirements, and quickly deal with the nearby coordinating personnel based on the distribution of environmental sanitation personnel. Record the daily classification of garbage inventory at each garbage classification site, count the total inventory of garbage in each district, and then use big data analysis to guide the adjustment and improvement of garbage treatment plans and facilities. This platform will provide a powerful digital base for the comprehensive management of the urban environment, comprehensively using 5G, Internet of Things, artificial intelligence, big data, edge computing and other technologies to maximize the value of data, efficient task execution, and precise regulatory decision-making. 4.2 Community Governance Based on the 5G and cloud-edge collaborative architecture, it enables two major application scenarios of community security and community sensing, and brings a more convenient experience for community managers and community residents [6] (Fig. 7). Community Safety. Using visual inspection, AI and other technologies, on the basis of traditional community security applications, the combination of 5G customized network and cloud-edge collaborative architecture can provide richer application scenarios and application experiences.

Research on Cloud-Edge Collaboration Architecture

375

Fig. 7. Smart community governance plan based on perception platform.

Based on the 5G customized network to achieve global coverage, 5G micro-stations can be used for network enhancement in the indoor environment. Using the cloud-edge collaborative architecture, model inference tasks are executed on the edge side, and model training tasks are executed on the cloud side. The MEC and the data center are sinking to the community, using the municipal/provincial shared UPF, the edge side is equipped with an image processing server, the data mining equipment with computing power can perform model inference, and the data mining equipment without computing power can upload video data to The 5G gateway performs model inference. Integrate AI into video surveillance, effective prevention and adjudication of certain blame issues such as high-altitude throwing objects, corridor accumulation, damage to public properties, littering, etc.; use face recognition and positioning to capture the tracks of blacklisted persons; capture illegal operations and give timely warnings and accountability [7]. The solution can provide 7 * 24 all-day monitoring, which can reduce community labor costs and security costs, and adopt artificial intelligence technology to improve detection accuracy and inspection efficiency. Community Perception. Community perception ability involves all aspects of community service, and its perception ability determines the quality of community service. Mainly divided into the following categories: Public Equipment Management. Real-time data collection of water meters, electricity meters, and gas meters in each household is convenient for charging reminders and equipment abnormality monitoring; real-time smoke and gas sensing monitors fire alarms and gas leak alarms; monitoring the status of community street lights, manhole covers and other public facilities Timely warranty.

376

P. Ding et al.

Vehicle Management. Monitor the capacity of carports and parking spaces to realize parking space reservations and vehicle positioning, promote community parking space sharing, improve parking efficiency and parking resource utilization; monitor and promptly alert roadside vehicles. Health Management. Personnel carry equipment with positioning and physical condition detection functions. The equipment can monitor heart rate, body temperature, blood pressure and other physical parameters in real time. The data is uploaded to the cloud platform to build a health file and returned to the device for real-time monitoring of the physical state. When an emergency occurs It can automatically initiate a call for help function at any time, and property personnel/medical personnel can quickly locate the location of the personnel.

5 Conclusion This paper proposes cloud-side collaborative architecture for intelligent acquisition of digital city information based on 5G customized network, constructs a multi-source perception platform, and applies it to two typical digital city governance scenarios. With data collection and standardization as the core and data open sharing as the concept, comprehensively using 5G, big data, Internet of Things, cloud computing, edge computing and other new-generation information technologies. Relying on multi-source perception platform, various departments and enterprises can deploy IoT monitoring systems and obtain data services, and strengthen the overall management of heterogeneous network equipment, reduce the threshold and cost of innovative application development, and shorten the research and development cycle.

References 1. Zhang, J.: Edge computing method and engineering practice. J. Autom. Expo 2. China Telecom: China Telecom 5g customized network product manual (2020) 3. Yu, M.: Smart environment governance: a theoretical analysis framework. Comparison Econ. Soc. Syst. 209(03), 91–99 (2020) 4. Ma, D., Wang, H.: Application of 5g in water environment monitoring. Harnessing Huaihe River 498(02), 31–32 (2020) 5. Wu, Y.: Application of big data technology in atmospheric monitoring. Satellite TV Broadband Multimedia (21) (2019) 6. Zhang, Y.: Research and practice of 5g smart community construction. China’s Constr. Inf. 108(05), 80–82 (2020) 7. Hao, J., Jin, J., Wang, S., et al.: Smart light pole – nerve endings of smart city based on 5g. In: 2019 China Lighting Forum: Semiconductor lighting innovation and application and smart lighting Development Forum

Identification Hierarchical Cooperative Caching Strategy Based on Edge Computing Yutong Wen1(B) , Wei Bai2 , Xin Xu3 , Yang Lu2 , and Shaoyong Guo1 1 Beijing University of Posts and Telecommunications, Beijing 100876, China

[email protected]

2 Global Energy Interconnection Research Institute Co., Ltd., Beijing 102209, China 3 State Grid Chongqing Electric Power Co. Electric Power Research Institute,

Chongqing 401123, China

Abstract. In the IoT environment, in order to solve the problem that cache nodes can’t be effectively utilized and respond efficiently, this paper proposes a hierarchical cooperative caching strategy based on edge computing. Firstly, based on the designed IoT three-tier architecture model, the data communication model is established in combination with equipment request process. Secondly, node collaborative cache is realized by optimizing particle swarm optimization algorithm. To prevent particles from falling into local optimal solutions, Metropolis criterion is introduced to disturb solution space. In the simulation experiment, it is verified that the algorithm in this paper can reduce the network delay. Keywords: Internet of Things(IoT) · Edge computing · Particle swarm optimization · Caching strategy

1 Introduction With the rapid development of the IoT technology, smart home, smart grid and other fields have been widely used. At the same time, massive equipment data is generated, and traditional cloud computing model is difficult to meet data transmission rate requirements [1]. Therefore, it is very necessary to study cache strategies, such as the classic cache strategy FIFO, LFU [2], LRU and so on. LRU updates data with time interval from the last visit as the popularity index, but this strategy is easily interfered by accidental access data. Reference [3] proposed a joint collaborative caching framework, which supports Adaptive Bit Rate video streams in MEC and can effectively reduce data delay. According to the limitation of cache space,reference [4] proposed a cache optimization scheme based on monkey algorithm. Based on previous studies, considering device mobility and cooperative caching, this paper proposes a distributed cloud architecture and caching strategy based on edge This work has been supported by State Grid Corporation of China science and technology project “Research on Reliable Transmission Technology in Electric Internet of Things Based on IPv6” (5700-202058178A-0-0-00). © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 S. Sun et al. (Eds.): ICSINC 2021, LNEE 895, pp. 377–385, 2022. https://doi.org/10.1007/978-981-19-4775-9_47

378

Y. Wen et al.

computing. The framework ensures network security with blockchain technology and avoid single point of failure.At the same time, computing and storage capacity of mobile network edge is improved by deploying high-performance servers close to device source.

2 System Model 2.1 IoT Architecture Model This paper proposes a three-tier architecture model to deal with edge caching problem of IoT identification. The IoT device layer consists of a large number of fixed and mobile devices, which quickly obtain data by submitting access requests to the upper layer. The edge node layer caches frequently accessed identification data, and has functions of responding device requests, caching data, and forwarding identification logs. The access efficiency of device and hit rate of data are improved by setting edge node.The blockchain network layer is responsible for storing all identification data, and the log information is stored by special servers. Through cryptographic mechanism, blockchain ensures that stored data can not be changed and achieves data security and traceability (Fig. 1).

Fig. 1. IoT system architecture

2.2 Problem Modeling It is assumed that the number of IoT devices in cache domain is Y, represented by U = {u1 , u2 , · · · , uY }. The number of edge node is X, represented by B = {b1 , b2 , · · · , bX }. The number of identity that can be requested by devices is R,and there are H identity groups, represented by A = {a1 , a2 , · · · , aH }. And each size is normalized to l bit. In the edge node layer, the network topology is relatively stable, and each edge node can cooperate to cache id. The caching strategy can be represented by a binary matrix ni,q ∈ NX ×R . ni,j = 1 indicates that the identification Aq is precached in node Bi . In addition, the maximum cache space of node is Lai . When device Uj is within coverage of node Bi , Uj can communicate with Bi . Let the coverage radius be S. The binary matrix

Identification Hierarchical Cooperative Caching Strategy

379

mi,j ∈ MX ×Y represents the relationship between device and node. mi,j = 1 indicates that Uj is within the coverage of Bi . The influence of the identity popularitity on node hit ratio needs to be considered. Since each identity is requested at a different frequency, assuming that the request probability of device is approximately described by the Zipf distribution [5], the popularity of identity aq is θaq =

γ

aq ×

1 R

i=1 i

−γ

∀θaq ≥ 0,

R 

θi = 1

(1)

i=1

If γ is larger, the device requests are concentrated in the most popular identity set. Hit rate is used to evaluate the efficiency of caching strategy. The hit rate is percentage of the number of edge node hits to total number of device request, expressed as: h=

Q Z

(2)

3 Communication Delay Analysis Devices download data by submitting requests to edge nodes or blockchain networks. The communication link between device and node is set to the flat Rayleigh fading channel.The spectrum bandwidth from the device Uj to the node Bi is set to Wi,j . Define Pi as transmission power of Bi , Ni as gaussian white noise power of Bi , k as channel loss −τ , τ is the channel coefficient. The channel gain of Uj and Bi is expressed as gi,j = k · di,j loss index. The download rate of blockchain network is set to C0 , which is much lower than that of edge node. According to Shannon theorem, the average rate is: Ci,j = Wi,j log2 (1 +

Pi · g i,j ) N0

(3)

When a device makes a request, there are three ways to get the identity. 3.1 Communication Delay to Edge Node If the local node Bi has cached the identity aq requested by device Uj , the Uj will obtain the identity aq from Bi . Transmission delay is tbi ,j . If the local node cannot satisfy the device’s request, the request will be forwarded to its cooperative edge node Bk through Bi . The transmission delay between nodes is set to tbi ,bk . So the transmission delay required for this process [6] is tbk ,j .    X     l  , tbk ,j = (1 − nj,q ) n (4) tbi ,j = i,q  (tbi ,j + tbi ,bk )  Ci,j i=1,i=j  0

380

Y. Wen et al.

3.2 Communication Delay to Blockchain Network If the device request is not satisfied at the edge node layer, the request will be forwarded to the blockchain network. The delay from device to blockchain network is tw . Therefore, we can get the delay tj,q required by Uj to request an identity aq . ⎧ X ⎪ ⎪ mi,j · ni,q · ti,j ∀bi ∈ B, ∃ni,q = 0 ⎨ i=1 l  + tbi ,j , tj,q =  tw = (5) X   ⎪ C0   ⎪ (1 − n ) t otherwise ⎩ i,q  w i=1 0

The purpose of this paper is to obtain the optimal cache strategy through PSO and SA, and minimize the transmission delay of device. The hierarchical cooperative optimization problem between devices and nodes can be expressed as Y R min T = Paq · tj,q j=1 q=1 ⎧ R ⎨ ni,q · l ≤ Lai , ∀bi ∈ B q=1 s.t ⎩ mi,j , ni,q ∈ {0, 1}, ∀bi ∈ B, uj ∈ U , aq ∈ A

(6)

4 BPSO-SA Caching Strategy PSO is an efficient parallel search algorithm [7], but it is easy to fall into local optimization in individual coding process. SA is a random optimization algorithm, which can avoid the optimal solution of problem in a certain region, and ensure that PSO has good convergence and robustness. In this paper, with the goal of minimizing fitness value, a binary particle swarm optimization based on SA(BPSO-SA) is designed by combining pros of the two algorithms. 4.1 Particle Evolution Model Individual particle encoding uses integer encoding method, and each particle represents a caching strategy. The relationship between nodes and identifiers is denoted by the one-dimensional array {n11, · · · , n1R , n21 , · · · , nXR }. In the solution space, the current position of the particle s is represented by Xs = {xs1 , xs2 , · · · , xs,X ·R }, Xs (i) ∈ {0, 1}, where Xs (i)=xsi represents the relationship between the [i/R] + 1 node and the i mod R identity. xsi = 0 means that the node has not cached the identity, otherwise it has been cached. The fitness value of particle s at t time is the total transmission delay under t is expressed as the optimal value of this strategy. The individual extreme value Pbest t a particle, and the global extreme value gbest is expressed as the optimal value of all particles. To ensure that the caching strategy is constantly updated and keep up with the optimal strategy, sets the velocity of particle from the crossover of the current velocity

Identification Hierarchical Cooperative Caching Strategy

381

and the extreme value. Each particle follows the extreme motion.The velocity of particle indicates its direction of travel. The formula is as follows: Vs (t + 1) = Vs (t) ⊗ Pbest (t) ⊗ gbest (t)

(7)

Xs (t + 1) = Xs (t) ⊗ Vs (t + 1)

(8)

In the above formula, ⊗ represents the crossover operation. In order to retain excellent particles, the particles are updated only when the fitness value of the new particle is less than the old particle. As the particles move forward, their evolution speed will gradually tend to 0, that is, the particles are easy to fall into the local optimal. At this point, it is difficult for particles to update further because of the balance between the current position and the current velocity of the particles. Therefore, SA is introduced to disturb the current particles, so that the particle can recover its evolutionary ability. The particle evolution ability value is used to describe the evolution ability of current particles. ⎛ ability(s) = ⎝0.6 ·

X ·R 

same(Xst (i), Vst (i)) + 0.4 ·

i=1

where same(a, b) =

0, if a = b 1,if a = b

X ·R 

⎞ t (j), g t (j))⎠/(X · R) same(Pbest best

(9)

j=1

. The value is used to describe search ability of

particles. With the continuous evolution of particles, the caching strategies of the three are gradually similar. Set the ability threshold to Z. When the ability value is less than Z, the particle achieves local optimal. SA (Sect. 4.2) is used to restore its evolutionary ability and expand its search range. 4.2 Neighborhood Search Based on Metropolis Criterion Using the SA mechanism, the fitness value is reduced by the disturbance of replacement rule. Based on Metropolis criterion, the new position is accepted with P(1 → 2) probability. Among them, e2, e1 is the total delay after replacement and the delay obtained by PSO when the temperature is T, respectively.  1,   e2 < e1 p(1 → 2) = (10) 1 , e2 ≥ e1 exp − e2 −e T

382

Y. Wen et al.

4.3 Process of BPSO-SA

Input: parƟcle iniƟalizaƟon, velocity, number of parƟcles, iteraƟons N,etc Output: Array of opƟmized caching data 1. IniƟalizaƟon parameters, Calculate the fitness funcƟon value of parƟcle. 2. Set the extreme value Pbest and gbest of the current parƟcle. 3. Update parƟcles according to the formula 11, and decide whether to retain the new parƟcles based on the fitness value. 4. Check the posiƟon of parƟcles, and judge whether its ability value is greater than Z, if it is greater than Z, go to step 6. 5. To ensure the minimum delay, compare the current posiƟon of parƟcles with the individual extreme value, and if it is lower than this value, replace it with Pbest . In the same way, update gbest . 6. Using SA, the posiƟon of parƟcles is disturbed, and the probability of the new soluƟon being accepted is determined according to formula 13. 7. Check,if the maximum tempering Ɵmes is reached or the opƟmal soluƟon is fixed, output the opƟmal caching strategy, otherwise, return to step 6.

5 Simulation Experiment Analysis In this paper, the identification adopts EPC coding. EPC is a logistics technology based on RFID and Internet. The goal of EPC is to assign global unique identifiers. 5.1 Experimental Parameter Setting This paper uses Matlab as simulation tool to simulate device identification requests of Zipf distributed, and verify the effect of BPSO-SA in reducing delay. In this experiment, the number of identification is set to 750. The coverage radius of node is 500 m. The equipment obeys the Poisson distribution with a density of 0.02. It is assumed that all nodes have the same channel loss coefficient and transmission power. The setting of other simulation parameters are shown in the following Table 1:

Identification Hierarchical Cooperative Caching Strategy

383

Table 1. Parameters setting Parameters

Value

Bandwidth, Wi,j

20 MHz

Transmission power, Pi

6W

Channel loss index, τ

4

Channel loss coefficient, k

10–2

Noise power, N0

10–10 W

Size of each identification, l

10 MB

Zipf parameter, γ

0.56

Particle population size, K

100

The popularity-based caching strategy [8] aims at popularity of data and caches hot data to each edge node. In the random caching strategy, the cache probability of each data is the same. The node randomly selects several data and caches them. In order to verify the effectiveness of BPSO-SA caching strategy, this paper compares the caching performance of BPSO-SA with the above algorithms. 5.2 Result Analysis The left subfigure of Fig. 2 shows the comparison of cache performance between BPSOSA and PSO. As we can see, the delay of them decreases gradually. Compared with PSO, this algorithm achieves the optimal delay faster and the transmission delay is lower. The right subfigure of Fig. 2 shows the comparison of algorithm performance under different zipf parameters. As we can see, BPSO-SA has lower transmission delay and is better than other strategies. In the figure, only the delay of RC does not decrease with increase of parameters. This is because RC does not consider popularity of identities, but randomly selects cache items. With the increase of parameters, a large number of popular identities will be stored in edge nodes, and the delay consumption of device will be reduced.

Fig. 2. The delay changes with iterations comparison of delay with different Zipf parameter

384

Y. Wen et al.

Figure 3 shows the variation of transmission delay and hit ratio with the number of identifiers. Compared with popular cache strategy, this algorithm can get lower latency and higher hit ratio. As the number of identifiers increases, the delay of all caching strategies increases. Because the excessive number of identities will affect the caching ability of edge nodes. Edge nodes can only cache more popular content to meet the requests of most devices.

Fig. 3. Comparison of delay and hit ratio with different identifiers

6 Conclusion Based on the three-tier blockchain network architecture, we study a hierarchical cooperative cache scheme to minimize device transmission delay. This strategy uses BPSO-SA to realize the cooperative scheduling of local nodes, cooperative edge nodes and blockchain network. The simulation results show that the strategy proposed in this paper has better performance. In the future, we will consider the impact of device preferences on identification popularity, as well as further optimize the performance of the algorithm.

References 1. Ud Din, I., et al.: The internet of things: A review of enabled technologies and future challenges. IEEE Access 7, 7606–7640 (2019) 2. Sokolinsky, L.B.: LFU-K: An effective buffer management replacement algorithm. In: Lee, YoonJoon, Li, J., Whang, K.-Y., Lee, D. (eds.) DASFAA 2004. LNCS, vol. 2973, pp. 670–681. Springer, Heidelberg (2004). https://doi.org/10.1007/978-3-540-24571-1_60 3. Tran, T.X., Pompili, D.: Adaptive bitrate video caching and processing in mobile-edge computing networks. IEEE Trans. Mobile Comput. 18, 1965–1978 (2019) 4. Cui, B., Li, Z., Xue, J.: Research on cache replacement strategy for edge nodes. Softw. Guide 19(04), 131–134 (2020) 5. Rangan, S., Madan, R.: Belief propagation methods for intercell interference coordination in femtocell networks. IEEE J. 30, 631–640 (2012) 6. Ma, J., Wang, J., Liu, G.: Low latency caching placement policy for cloud-based VANET with both vehicle caches and RSU caches. In: Proceedings of the 2017 IEEE Globecom Workshops, pp. 1–6 (2017)

Identification Hierarchical Cooperative Caching Strategy

385

7. Bao, Z., Yu, J.: Intelligent Optimization Algorithm and its MATLAB Example, 2nd edn. Electronic Industry Press, Beijing (2018) 8. Zhao, X., Yuan, P., Ii, H.: Collaborative edge caching in context-aware device-to-device networks. IEEE Trans. Veh. Technol. 67, 9583–9596 (2018)

Intelligent Park Organism Based on 5G Edge-Cloud Collaboration Yuying Xue(B) , Peng Ding, Yun Shen, Huibin Duan, Xuezhi Zhang, and Yaqi Song Department of Service and Application Innovation, China Telecom Corporation Limited Research Institute, Beijing, China {xueyy,dingpeng6,shenyun6,duanhb,zhangxzh, songyq11}@chinatelecom.cn

Abstract. The emergence of new generation ICT technologies such as 5G, edge computing and AI promotes the rapid development of parks towards digitization, informatization and intelligence. Currently, there are some problems among the parks, such as information isolation, complex network requirements, high basic cost and insufficient optimization capabilities. To solve these problems, this paper proposes an intelligent park organism based on China Telecom’s 5G customized network and edge-cloud collaboration architecture to ensure resource collaboration, information exchange, and capability sharing in a single park or among multiple parks. The proposed intelligent park organism supports capacity training, capacity deployment and continuous optimization, in which, the capabilities can also be transferred and trained among multi-park through technologies such as federated learning. The intelligent park organism can provide functions such as three-dimensional video fusion, digital road, intelligent security, intelligent energy, etc. Through the closed-loop optimization of the intelligent park and the collaborative optimization among multiple intelligent parks, the service functions and management abilities of the intelligent park can continuously evolving. Keywords: Intelligent park · 5G · Custom network · Edge-cloud collaboration

1 Introduction Industrial park is an important carrier for industries, logistics, transportation and other industries, which has a huge market space. With the continuous improvement of national economy, the demand for intelligent parks in various industries is increasing, and there are more and more requirements for park functions, coupled with the guidance of national policies, which lead to explosive growth for the intelligent parks. The park is a characteristic functional block of the city, a regional convergence of all things interconnected, an important carrier for work and learning, and an important battlefield for epidemic prevention. The outbreak of COVID-19 has exposed some deficiencies of the management thinking and governance methods in the park. Based on the ability of information integration, innovation and management [1, 2], the intelligent park can improve the quality of work and life, promote industrial upgrading and economic development. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 S. Sun et al. (Eds.): ICSINC 2021, LNEE 895, pp. 386–397, 2022. https://doi.org/10.1007/978-981-19-4775-9_48

Intelligent Park Organism Based on 5G Edge-Cloud Collaboration

387

There are many problems in the construction of intelligent parks, including unsmooth data interaction, insufficient supporting related industries, uncoordinated management and operation. The construction of intelligent parks has been carried out in various parts, but most of which are still in the pilot stage, and the development direction needs to be explored [3]. Collaborative optimization of multiple parks is one of the main evolution trends of intelligent parks, which can effectively solve the above-mentioned problems of a single park. However, at present, the industry has little exploration of collaboration mechanism among multiple parks, and the direction is still unclear. In recent years, the development of new technologies has provided support for building intelligent parks and improving park management capabilities. Compared with 4G, the bandwidth, transmission rate, transmission stability and air interface delay of the 5G network have been greatly improved, and it has the ability to build an end-toend system, which can meet the needs of smart park scenarios [4]. Besides, the IoT (Internet of Things) technology provides global perception capabilities for intelligent parks. Therefore, 5G, artificial intelligence, cloud computing, edge computing and big data can provide intelligent parks with data analysis, edge intelligence, autonomous decision-making, autonomous optimization and other intelligent capabilities [5]. This paper proposes an MIPOs (Multiple intelligent park organisms) solution based on China Telecom’s 5G customized network and edge-cloud collaborative architecture to realize park resource sharing, capacity sharing, intelligent decision-making and comprehensive management [6]. Along with the large-scale deployment of 5G network, the MEC (Multi-Access Edge Computing) has been widely used, which can realize local data processing and edge cloud deployment of capabilities. The edge-cloud collaborative architecture integrates the cameras, sensors and basic equipment into management, technologies such as artificial intelligence, big data and knowledge graph empower the park to improve the intelligence level. Therefore, the park can provide intelligent services, such as scene roaming, target detection, vehicle recognition, parking guidance, and energy management. Single park and multiple parks have the capabilities of selflearning, self-adaptation, self-optimization and sustainable development. Through the self-optimization of IPO (Intelligent park organism) and the collaborative optimization of MIPOs, the intelligent collaboration and evolution of the park can be realized.

2 The Overall Architecture of the Intelligent Park Organism 2.1 Existing Problems in Building a Intelligent Park From the perspective of park network, intelligent parks have more complex network requirements. The existing network construction plan is single and the network deployment is messy, which is unable to meet the requirements of different industries and different types of parks, and the cost of network construction and post-maintenance is high, resulting in many enterprises unable to bear. From the perspective of park data, obsolete equipment in the park makes it difficult to connect to the network, which leads to difficulties in obtaining data in the park. A single park contains fewer scenes, and the quantity and quality of data obtained from the scene can’t be guaranteed, which seriously hinders ability training.

388

Y. Xue et al.

From the perspective of park intelligence, the phenomenon of information islands is serious. The park lacks the capability of data analysis and data integration. Most information collected by the equipment is only achieved for data conversion and cloud upload, without real information integration. From the perspective of the park system, the commercial software development capabilities of the park are relatively weak, and some parks lack top-level design plans. The various systems in the park are independent of each other, making it difficult to achieve effective cross-system operations, resulting in high operating costs for the entire park. From the perspective of multiple park, the lack of reliable and effective information collaboration and information linkage mechanisms among multiple smart parks has led to the problems of data sharing and non-interoperability among various parks. On the other hand, different industry parks have different requirements for data privacy protection, which restricts the self-learning and self-optimization capabilities of intelligent park organism, and it is difficult to realize the ecological construction and coordinated development of multiple parks. 2.2 Overall Scheme

Fig. 1. MIPOs solution diagram

Based on 5G customized network and edge-cloud collaborative architecture, IPO is constructed to ensure information exchange, resource coordination and capability sharing within the park and among multiple parks. The proposed scheme, as shown in Fig. 1, sinks the self-developed UPF (User plane function) to the edge data center in the park to offload the local service traffic to the MEC server, which is also deployed in the edge data center of the park. Besides, the proposed scheme also realizes intelligent collaboration and evolution between IPO through e-Cloud, which managed and operated by China Telecom. By adopting ICT (Information and communications technology) technologies such as artificial intelligence and edge computing, federal learning, and the proposed scheme can provide intelligent services such as 3D (Three dimensional) video

Intelligent Park Organism Based on 5G Edge-Cloud Collaboration

389

fusion, digital road, intelligent security and intelligent energy. The above capabilities can be coordinated and optimized among multiple smart parks. Intelligent Access and Intelligent Cloud Network: A data center is established in the park to support the access of multiple IoT devices. Data collected by devices such as sensors and cameras are directly offloaded to the MEC server in park data center via UFP for processing, so as to ensure data privacy and security. For information exchange among multiple smart parks, data is encrypted by federated learning technology and then transferred to e-Cloud for unified processing. 3D Visualization: In the proposed scheme, 3D model of the park is constructed, which can be integrated with multiple video streams of the cameras. The 3D model supports roaming and building dismantling. the client can directly display the camera video and the relevant data of the park, including camera positions, statistical data, etc. and realizes a map to display the scenes inside and outside the park. Intelligent Security: Based on AI technology, the proposed scheme can also provide functions such as sequential action detection, face recognition, vehicle detection, manhole cover offset detection, work clothes detection, park fireworks detection, important asset detection, etc. Based on the target trajectory tracking technology, it the real-time cross-camera trajectory tracking and target historical trajectory tracking services are also provided in the scheme. The above detection results can be visually displayed on the 3D model of the park to achieve safety management. Digital Road: The park is equipped with RSUs (Road side unit), lidars, cameras, millimeter-wave radars and other sensor equipments on the road side, and the cars are equipped with OBUs (On board unit). Based on AI technology and big data analysis, the allocation of traffic resources can be optimized greatly [7]. Digital road can provide parking guidance, blind spot detection, collision warning, vehicle path planning and other capabilities. Intelligent Energy: In order to realize the “carbon peak, carbon neutral” of the park, based on the collection, monitoring and analysis of the park energy data, the scheme can provide energy efficiency management and energy efficiency optimization functions, such as smart micro-grid. Coordinated Development of MIPO: The proposed scheme adopts federated learning, collaborative reasoning and other methods to carry out collaborative training and continuous optimization among multiple parks. Through the interaction of a single park internal cycle and the multiple parks cycle, the function and performance of MIPO can be continuously learned and optimized.

3 Intelligent Park Networking Solution 3.1 5G Customized Network Features Based on the China Telecom 5G SA network framework, 5G customized network provides network with flexible customization. The 5G customized network takes advantage

390

Y. Xue et al.

of cloud-network integration and provides “one-stop cloud-network-application overall services”. 5G customized network provides 5G + NICES multi-dimensional mode, the park is networked according to scene requirements (N-Network customization, IIntelligence edge, C-Cloud edge collaboration, E-application selection, S-Service guarantee). There is no need to establish 5G network infrastructure in the park, which reduces the cost of basic network investment. The park provides key capabilities for application development and reduce the complexity of informatization application development. China Telecom’s 5G customized network provides three modes, this project adopts the “neighbor mode” for network deployment. The mode has the characteristics of local processing and cloud-edge collaboration, and is suitable for scenarios with high requirements for exclusive wireless enhancement, low latency, local business processing and data local processing. This mode adopts the wireless side enhanced coverage and UPF/MEC on-demand sinking method, combined with super uplink, QoS enhancement, DNN, slicing and other technologies for flexible customization [8]. It has the following characteristics: • Private network or public network business collaboration, wireless side enhanced coverage. • UPF sinks to the computer room at the edge of the city or park, and you can choose to exclusive or share the edge UPF according to business needs. • Localized deployment of computing resources, MEC is deployed in the park computer room to provide local ICT services and support application sinking. • All data is transmitted to the intranet server or data center to ensure that the data local processing and achieve business isolation. 3.2 Intelligent Park Networking Deploy the network in two parks, set up indoor base station according to the needs of the scene to achieve full network coverage in the park. Based on edge-cloud collaboration, the park enjoys MEC exclusively to ensure data processing in the park, localized training and deployment of models. Information and resources between multiple parks are processed uniformly through e-cloud (Fig. 2). In this deployment scheme, edge side adopt an exclusive method. Sinking UPF and MEC to the edge side to provide localized deployment of computing and storage network resources. In order to ensure the privacy of the data in the two parks, all intelligent services or applications in the park are processed in the park MEC, and the devices such as cameras and RSU use 5G network and to convert structured and unstructured data shunt to MEC for AI perception processing. The intelligent service or application system in the park carried on MEC includes panoramic video 3D fusion system, multi-source vision fusion service, target trajectory tracking service, vehicle-road collaboration service, intelligent energy.

Intelligent Park Organism Based on 5G Edge-Cloud Collaboration

391

Fig. 2. Networking scheme of intelligent park

4 The Capabilities of Intelligent Park Organism The park adopts collaborative reasoning method to reduce model reasoning delay, this method supports continuous self-learning and self-optimization. At the same time, the park adopts incremental learning methods to balance the old and new data and reduce the impact of catastrophic forgetting on the model. The cloud-side performs full lifecycle management of applications and models, and capabilities can be simultaneously migrated to other parks. E-Cloud is responsible for the collaborative processing of data in multiple parks. The encrypted information of the park is uploaded to e-Cloud through the edgecloud collaboration mechanism. E-Cloud uses federated learning to conduct joint training of the model, the model is updated and distributed to the park MEC to realize the self-optimizing closed loop of intelligent services and applications. 4.1 Panorama of the Park-3D Video Fusion Subsystem The 3D GIS (Geographic Information System) model of the park is constructed. The 3D GIS model carries the video and multi-source perception data of the park, and the real-time captured images of the camera are projected to the 3D GIS model. The adjacent pictures collected by the camera are spliced and fused to construct an ultra-high resolution panoramic picture of the park. The 3D video fusion system can provide 3D models of indoor and outdoor scenes in the park, the model supports multi-view rotation and scene roaming. The detection results and related data of applications such as smart security and digital roads are fed back to the 3D video fusion system for display. The architecture of the 3D video fusion system is shown in Fig. 3. The system is connected to cameras, sensors and other equipment in the park. The information collected by the equipment is fused with the park 3D model for rendering, and the rendering results are transmitted to the Web front end for visual display.

392

Y. Xue et al.

Fig. 3. Architecture of 3D video fusion system

4.2 Intelligent Vision a. Vision AI Subsystem The visual AI subsystem can provide a variety of visual AI services, including fall detection, dangerous action detection, anti-static clothing detection, face mask detection, abnormal behavior detection, etc. When an abnormal behavior is detected and an alarm is issued, personnel can obtain the video and location information of the event in time. The subsystem includes data preprocessing module, image reconstruction module, visual AI inference module, and analysis module. The arrows in Fig. 4 are the data flow processing process and the data interaction relationship between systems.

Fig. 4. Architecture of visual AI subsystem

The video stream is transmitted to the data preprocessing module for key frame extraction, the image enhancement, and the image reconstruction module determines whether the image needs to be reconstructed and repaired. The key frames are transmitted to the visual AI reasoning module for visual AI reasoning and generate a video stream with added annotation information. The AI reasoning data output by the reasoning model

Intelligent Park Organism Based on 5G Edge-Cloud Collaboration

393

is sent to the analysis module. The module performs fusion analysis on the AI reasoning results in the same frame, module judges the event trigger state according to the business logic of the presentation layer and issues decision instructions. b. Target Trajectory Tracking The target trajectory tracking service can realize rapid positioning and intelligent trajectory tracking of pedestrians or vehicles in the park. The target trajectory service is deployed in MEC to realize localized processing, the service can track the target trajectory in both real-time and historical business scenarios. Target trajectory tracking realizes cross-camera and cross-scene recognition and retrieval through the process of target detection, feature extraction, function retrieval and re-recognition. The system architecture includes a data preprocessing module, image reconstruction and target trajectory tracking modules. b-1. Real-Time Trajectory Tracking The front end sends the target picture to the target trajectory tracking module, and obtains the structured features of the target through detection and feature extraction. The multichannel surveillance video stream is input to the data preprocessing module for key frame extraction and image enhancement. The target trajectory tracking module is input to the target structural feature for re-identification comparison. The video generation module generates a video with the re-identification result annotation information (Fig. 5).

Fig. 5. Architecture of real-time trajectory tracking subsystem

b-2. Historical Trajectory Tracking The front end sends the target image to the target trajectory tracking module for feature extraction. The data preprocessing module processes the historical video stream regularly, and inputs the preprocessed key frames into the target trajectory tracking module for structured data extraction and saves it to the database. Obtain the target historical structured data through the efficient retrieval module, and re-identify the target structured features to generate the target historical trajectory video (Fig. 6).

394

Y. Xue et al.

Fig. 6. Architecture of historical trajectory tracking subsystem

4.3 Digital Road Deploy RSU, lidar, millimeter-wave radar, cameras and other sensing equipment in the park, OBU equipment is installed on the vehicle. The business system is deployed in the MEC and the roadside edge computing unit in a distributed manner to realize multilevel intelligent collaboration. The park provides services such as blind spot detection, parking guidance and vehicle rear-end collision prevention. Services with extremely high latency requirements are deployed in the roadside edge computing unit, and services with relatively low latency requirements are deployed in the park’s exclusive MEC. c-1. Garage Access Service The garage entry and exit service adopts multi-level intelligent collaboration, and deploys the vehicle detection and recognition model on the roadside computing unit to perceive and recognize vehicles. The garage access service system is deployed in the exclusive MEC of the park, the system includes application platform, vehicle road collaborative business platform and RSU system management platform (Fig. 7). Vehicle approach: RSU uploads the vehicle information to the vehicle-road collaborative business platform, and pushes it to the third-party application platform for billing, automatic lever release. The platform sends the route guidance information to the RSU system management platform for parking guidance, so as to achieve non-parking access the garage. Vehicle appearance: RSU detects the vehicle information and uploads information to the vehicle-road collaborative business platform, RSU synchronously pushes the vehicle information to the third-party application platform for billing. The billing information is transmitted to the RSU system management platform through the 5G network for charged and released. c-2. Blind Spot Obstacle Detection The blind spot detection model is deployed in the RSU, and the model can quickly determine the type and location of obstacles (Fig. 8). The obstacle information is transmitted and stored to the MEC platform, and sent to the RSU system management platform to broadcast the blind area obstacle information, which is used for scenes with blind areas such as underground garage entrances and exits and turns. This function can be used in scenes where there are blind areas of vision, such as underground garage entrances and exits, turning, and so on.

Intelligent Park Organism Based on 5G Edge-Cloud Collaboration

395

Fig. 7. Garage access service scene

Fig. 8. Blind spot obstacle detection scene

4.4 Intelligent Microgrid The intelligent microgrid provides power generation, energy storage and optimized dispatch functions, which can improve the flexibility of the park’s electricity consumption and realize energy saving and electricity saving. The system collects and summarizes the data related to power generation and consumption in the park, and monitors the power consumption of different periods, regions and equipment in real time.

396

Y. Xue et al.

According to the above data, the power consumption in the park is reasonably distributed to improve the energy utilization efficiency (peak cutting and valley filling). The system collects data such as power generation, energy storage, switching status and power quality in real time. A management model is deployed on the edge to realize real-time monitoring and rapid processing of park data and improve the reliability of park power supply (Fig. 9).

Fig. 9. Intelligent microgrid scene

4.5 Intelligent Collaboration and Evolution of Park Organism a. Horizontal Federal Learning: Collaboration of Intelligent Vision Capabilities in Multiple Parks AI vision detection model training requires a large amount of high-quality data. The data that can be collected in a single park is relatively single and the amount of data is small, which cannot meet the accuracy requirements of AI models such as dangerous motion detection. In addition, the parks have high requirements for data privacy protection, so federated learning is used to implement data encryption sharing and joint training in multiple parks to improve the accuracy of each park model. On the other hand, the data extraction features of the intelligent vision function model are basically the same, the overlap of samples collected in each park is insufficient. Therefore, horizontal federated learning is used for model training and the number of training data is increased, all models are saved to e-Cloud. The model is updated by local training and parameter aggregation on the cloud. The park downloads the latest pre-training model from e-cloud, and trains the model according to the local data set. After the training, the model is encrypted and uploaded to e-Cloud. E-Cloud aggregates the gradients of each park to update the model parameters, and synchronously sends the model to the park for update (Fig. 10). b. Transfer Learning: Blind Area Obstacle Detection The types of blind area detection obstacles include vehicles, pedestrians, mounds, pits, etc. Because the amount of data that can be collected for obstacles in the park is different, and there are some overlapping features of these obstacles, training multiple obstacle

Intelligent Park Organism Based on 5G Edge-Cloud Collaboration

397

Fig. 10. Horizontal federal learning

models separately will take up a lot of bandwidth and edge computing power, so the federated transfer learning method is used for model training. The park downloads the pre-training model from e-Cloud, and realizes the personalized training of the pretraining model based on feature migration, sample migration, and model migration.

5 Conclusion This project deeply integrates 5G customized network, AI and edge computing technology to provide the vitality capabilities of intelligent parks, a large picture realizes a panoramic display of the park. The park provides intelligent management and intelligent service systems to reduce manpower consumption and improve the efficiency of park operation and maintenance management. The park adopts technologies such as incremental learning and collaborative reasoning to realize the continuous optimization of the life form of the smart park. Federated learning is adopted between multiple parks to improve model training efficiency, reduce resource consumption and promote the coordinated development of life bodies in multiple parks. This project provides a development direction for smart parks.

References 1. Akhunzada, A., Islam, S.U., Zeadally, S.: Securing cyberspace of future smart cities with 5G technologies. IEEE Netw. 34(4), 336–342 (2020) 2. Musa, S.: Smart cities-a road map for development. IEEE Potentials 37(2), 19–23 (2018) 3. Liang, F., Sun, L., Guo, Z.M.: Exploration and research on construction of new 5G smart park. Des. Tech. Posts Telecommun. 528(02), 56–59 (2020) 4. Zhang, R.: Research on key technologies of 5G mobile communication network. China Comput. Commun. 2006, 012028 (2018) 5. Batty, M., Axhausen, K.W., Giannotti, F., et al.: Smart cities of the future. Eur. Phys. J. Spec. Topics 214(1), 481–518 (2012) 6. Gharaibeh, A., Salahuddin, M.A., Hussini, S.J., et al.: Smart cities: a survey on data management, security and enabling technologies. IEEE Commun. Surv. Tutor. PP(4), 1 (2017) 7. Azgomi, H.F., Jamshidi, M.: A brief survey on smart community and smart transportation. IEEE (2018) 8. Dzung Van, D., Byeong-Nam, Y., Hung Ngoc, L., et al.: ICT enabling technologies for smart cities. ICACT Trans. Adv. Commun. Technol. (TACT) 8(1), 1190–1192 (2019)

Portable Citrus Detection System Combining UAV and Edge Equipment Heqing Huang1(B) and Michel Kadoch2 1 College of Electronic Engineering (College of Artificial Intelligence),

South China Agricultural University, Guangzhou, China [email protected] 2 Department of Electrical Engineering, Ecole de Technologie Superieure, Montreal, QC, Canada

Abstract. Deep learning computer vision research has been widely used agriculture field to reduce labor costs and provide technical support for fruit farmers. At present, most object detection algorithms cannot effectively extend to the mountain orchard scene. This paper designs a fast and efficient citrus detection system, which combines unmanned aerial photography and edge computing device. The system uses low altitude UAV to take an omnidirectional image of citrus at different angles. Then, we optimize the state-of-the-art object detection model, use the attention mechanism to enhance the detection effect of small citrus objects. We use UAV to collect citrus fruit images and use the improved model for detection. The experimental results show that the detection accuracy of the model for citrus fruit data set is 93.42%. The detection time of single citrus fruit image in the edge device is 310 ms. Keywords: Citrus detection · UAV · Edge equipment

1 Introduce Citrus trees are an essential category of fruit trees in China, the areas suitable for citrus planting are expanding every year. The object detection of citrus provides technical support for practical tasks [1–3]. With the rapid development of communication technology, this means that detection networks need richer conditions for identifying different kinds of targets and can be carried out in more meaningful devices and environments. At present, advanced detection algorithms have high accuracy, but many algorithms complete on high-performance computers. Their complex network structure and huge parameters make them unsuitable for the operation of mobile terminals; on the other hand, citrus data collection is difficult and dangerous in the mountainous orchard environment. The normal data cannot be covered around the fruit trees, which significantly impacts fruit farmers’ yield estimation and income increase. Therefore, the research on the rapid and all-around acquisition technology of citrus object detection is of great significance. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 S. Sun et al. (Eds.): ICSINC 2021, LNEE 895, pp. 398–403, 2022. https://doi.org/10.1007/978-981-19-4775-9_49

Portable Citrus Detection System Combining UAV and Edge Equipment

399

With a series of modern information technologies represented by 5g and artificial intelligence, they are changing rapidly. The object detection algorithm should adapt to the more rigorous recognition environment. To solve such problems, it is often necessary to use a variety of sensors or edge computing devices. Specifically, we used UAV aerial photography to collect and mark many citrus images in an all-around way; These images were sent to the Jetson nano edge computing device deployed with the object detection algorithm for real-time citrus detection, as shown in Fig. 1. At the same time, to improve model performance, the YOLOv5 object detection algorithm is improved by adding a convolution block attention mechanism (CBAM) [4].

Fig. 1. Portable edge computing device

2 Related Work 2.1 UAV Aerial Photography Detection With the proposal of intelligent agriculture, many researchers used UAV sensors for agricultural condition detection. Fan et al. [5] used UAV to collect data and hyperspectral remote sensing technology to detect crop growth. Liu et al. [6] used corn UAV remote sensing images to separate crops and soil, proposed a morphological method to extract the morphology of corn seedlings, and detected the number of plants in corn seedling images by harris corner detection algorithm. 2.2 Edge Computer Device Identification In the field of agriculture, Mazzia et al. [7] deployed a real-time object detection algorithm on the embedded platform to detect apples in orchards, which can fully help fruit farmers detect, count, and estimate yield. The mobile terminal achieved 83.64% detection accuracy and 30 fps/frame, meeting the accuracy and real-time performance of detection. SA et al. [8] used a micro air vehicle to collect multi-spectral images and

400

H. Huang and M. Kadoch

inferred the classification of dense weeds by using a cascaded convolutional neural network based on encoder-decoder architecture to help fruit farmers monitor and analyze the scene.

3 Method 3.1 YOLOv5 YOLOv5 has four different versions, including YOLOv5s, YOLO5l, YOLO5m, and YOLO5x. In these models, YOLOv5s has high detection speed and relatively few parameters. In this paper, we improve on YOLOv5s,. The model uses FOCS to slice the input image and adjust the size of the feature image. The model uses CSP cross level network structure to improve the learning ability of convolutional neural network. In the neck part, combined with the advantages of FPN and PAN, the features extracted from different convolution layers are fused. The structure diagram of the model proposed in this paper is shown in Fig. 2.

Fig. 2. When outputting the feature map, three YOLO layers are used to generate feature maps with three different resolutions, and the loss is calculated compared with the label.

3.2 CBAM When UAVs shoot citrus fruit images at low altitude, human operation of the aircraft may lead to problems such as different flight heights, lens jitter, different resolution of the captured fruits, etc. at the same time, in citrus fruit trees, mutual occlusion between oranges and occlusion of citrus fruits by leaves will lead to difficulties in model detection. Therefore, we add a lightweight convolutional attention module (CBAM) to the deep neural network to make the model pay more attention to the occluded or small resolution citrus fruit. This structure is an attention mechanism module combining space and time, which can be effectively inserted into the model to make the structure of the model more reasonable and effective. Figure 3 shows the process of input feature extraction by the attention mechanism module.

Portable Citrus Detection System Combining UAV and Edge Equipment

401

Fig. 3. The structure of attention mechanism consists of two parts: channel based attention and space-based attention.

4 Experiment and Results 4.1 Datasets The UAV equipment used in this paper is Mavic air 2, which has strong endurance effect and collects high-resolution citrus fruit images at the same time. We collected data by UAV shooting at low altitude around citrus fruit trees. This paper was collected 1800 citrus fruit images and manually labeled by Labelmg labeling software. We divide the data set. The training set contains 1300 images, and other citrus fruit images are equally divided into test set and verification set. In order to standardize the training data, we use the format of coco data set. The UAV equipment is shown in Fig. 4.

Fig. 4. Our device for taking citrus images.

4.2 Experimental Training Setting We use the transfer learning method to speed up the training speed of the model. The model first trains an effective weight from the coco data set. By fine tuning this weight, the amount of data required for model fitting is reduced. The original input resolution of the model is 608 × 608 × 3. The model trained 50 epochs on citrus fruit data set. We use a more effective loss function, ciou loss function and adam-w optimization algorithm β1 = 0.89, β2 = 0.99, ε = 109. During the test, we transplanted the improved citrus fruit detection algorithm into Jetson nano to test the detection accuracy and speed of the algorithm.

402

H. Huang and M. Kadoch

4.3 Experiment In order to analyze the contribution of each part of the model, we compared each improved module. Table 1 shows the ablation experiments on the citrus dataset. Table 1. Model ablation experiment Model

AP/% Speed/ms Speed/ms

YOLOv5s

91.03

270

270

YOLOv5s + CBAM 93.42

310

310

This paper evaluates the index selection of the model, average accuracy and image detection speed. Among them, average accuracy is a general judgment method in the field of deep learning object detection, which can effectively evaluate whether the model can accurately detect objects in the image. Formula 1–3 shows the calculation method of accuracy. TP × 100% TP + FP TP R= × 100% TP + FN  1 AP = P(R)dR P=

(1) (2) (3)

0

where p is the accuracy rate, which can judge whether the detection effect of the model is correct (%). R is the recall rate (%), which can judge whether all targets in the model detection image are completely detected. In Fig. 5, we show the detection effect of the model on citrus fruit.

Fig. 5. Detection effect of model on citrus image

We have carried out comparative experiments with the same advanced object detection models. As shown in Table 2.

Portable Citrus Detection System Combining UAV and Edge Equipment

403

Table 2. Model comparison experiment Model

AP/%

FPS(In 2080ti)/s

FCOS

90.76

47

YOLOv3

91.21

69

YOLOv4

91.97

73

Ours

93.42

76

5 Summary To solve the problems of complex citrus data acquisition and effectively deploy the detection model to the mobile terminal in mountain orchards. We designed a new portable citrus fruit detection method. Firstly, we use small UAV equipment to collect citrus fruit images in an all-round way. Then, we improve the YOLOv5 object detection model to pay more attention to the occluded fruit. This method is used to improve the detection effect and reasoning speed under different resolutions.

References 1. Lan, Y., Deng, X., Zeng, G.: Advances in diagnosis of crop diseases, pests and weeds by UAV remote sensing. Smart Agric. 1(2), 1–19 (2019). (in Chinese) 2. Harrell, R.C., Adsit, P.D., Munilla, R.D., et al.: Robotic picking of citrus. Robotica 8(04), 269–278 (1990) 3. Dorj, U.O., Lee, M., Yun, S.S.: A yield estimation in citrus orchards via fruit detection and counting using image processing. Comput. Electron. Agric. 140, 103–112 (2017) 4. Woo, S., Park, J., Lee, J.-Y., Kweon, I.S.: CBAM: Convolutional block attention module. In: Ferrari, V., Hebert, M., Sminchisescu, C., Weiss, Y. (eds.) ECCV 2018. LNCS, vol. 11211, pp. 3–19. Springer, Cham (2018). https://doi.org/10.1007/978-3-030-01234-2_1 5. Fan, X., Zhou, J., Xu, Y.: Research progress of UAV low altitude remote sensing monitoring agricultural information. J. Xinjiang Univ. (Natural Science Edition) (Chinese and English) 38(05), 623–631 (2021). (in Chinese) 6. Liu, S.B., et al.: Information extraction of maize seedling number based on UAV remote sensing image. J. Agric. Eng. 34(22), 69–77 (2018). (in Chinese) 7. Mazzia, V., Salvetti, F., Khaliq, A., et al.: Real-time apple detection system using embedded systems with hardware accelerators: An edge AI application. IEEE Access 8, 9102–9114 (2020) 8. Sa, I., Chen, Z., Popovic, M., et al.: : WeedNet: Dense semantic weed classification using multispectral images and MAV for smart farming. IEEE Robot. Autom. Lett. 3, 588–595 (2017)

Q-learning Based Computation Offloading Algorithm in Mobile Edge Computing Cheng Zhong1 , Shaoyong Guo2 , Pengcheng Lu1(B) , and Sujie Shao2 1 State Grid Xiongan New Area Power Supply Company, Heibei, China

[email protected] 2 Beijing University of Posts and Telecommunications, Beijing, China

Abstract. In order to improve the computation capability and reduce the energy consumption of the smart mobile devices (SMDs), mobile edge computing (MEC) is proposed. In this paper, a total cost minimization problem with task delay requirement and total computation resources constraints is formulated. Due to the original problem is non-convex, we propose a Q-learning based computation offloading algorithm. In order to verify the effective of the proposed algorithm, we compared with other two algorithms. The results shows the proposed algorithm can reduce the total cost. Keywords: Mobile edge computing · Q-learning · Computation offloading algorithm

1 Introduction The advancements in communication techniques and Internet of things (IoT) have paved the way toward realizing the computation-hungry and delay-sensitive mobile applications [1, 2]. However, because the computation capability and the energy of the smart mobile devices (SMDs) are limited, many related applications and services that cannot be processed by the SMDs. Considering these problems, people propose the mobile edge computing (MEC). MEC is considered as a novel technology, which has more powerful computation resources than the SMDs. The SMDs could offload the tasks to the MEC server, due to the powerful computation resources, the task offloading delay could be reduced effectively [3, 4]. Recently, there are a lot of literatures about the task offloading in MEC networks, and these works are focus on two different objectives. One objective is to minimize the task offloading delay. Li et al. proposed a delay-aware task congestion control and resource allocation method to minimize the total task offloading delay when considering the tasks random arrival [5]. Fang et al. investigated the task delay minimization problem. With considering the partial offloading, a bisection search iterative algorithm was proposed [6]. The second objective function is to minimize the task offloading energy consumption. In order to minimize the power consumption in a multiuser MEC system, a power consumption minimization problem was formulated. In order to solve this problem, an online algorithm method was utilized [7]. Yang et al. formulated a problem to minimize © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 S. Sun et al. (Eds.): ICSINC 2021, LNEE 895, pp. 404–410, 2022. https://doi.org/10.1007/978-981-19-4775-9_50

Q-learning Based Computation Offloading Algorithm in Mobile Edge Computing

405

the total energy consumption of the MEC server and sensors. Because the formulate problem is a non-convex problem, the authors transformed it into convex problems and utilized convex optimization method to solve it [8]. However, all the aforementioned literatures only considered one objective. In the MEC system, both task offloading delay and energy consumption are equally important. Therefore, these two metrics should be considered at the same time.

2 System Model The system model of MEC is introduced in this section, and the system cost minimization problem is formulated. 2.1 Network Model In this paper, there are one eNodeB equipped with a MEC server and I single antenna SMDs in the MEC system. Each SMD has one task needs to be processed. We define I = {1, ..., I} as the set of SMD, and we utilize {ci , si , timax } to describe the task i, where ci is the computation resources required to process task i. si is the file size of task i. timax is the maximal delay requirement of task i [9]. We denote a binary variable αi , if the task i is processed by the MEC server, αi =1. Otherwise, when the task i is processed by the SMD locally, αi = 0. Meanwhile, the task offloading vector can be defined as α = {α1 , ..., αi , ..., αI }. The bandwidth in this system is B, and we define bi as the bandwidth allocated to task i. In this system, we ignore the interference. According to the above analysis, the transmission rate from the SMD i to eNodeB can be expressed as   hi pi , (1) ri = bi log2 1 + n0 where hi is the channel gain between SMD i and the eNodeB, n0 is the noise power, and pi is the transmission power of SMD i to eNodeB.. 2.2 SMDs Computation Model fimax is the maximal computation resources of SMD i, and fil is the computation resources gave to task i by the SMD. When the task is computed by SMD, the task processing time can be expressed as T0i =

ci filoc

.

Based on [10], the local computation energy consumption of SMD i is  2 Eiloc = κiloc filoc ci ,

(2)

(3)

where κiloc is the conversion coefficient of SMD i. Based on the practical measurement, we set κiloc = 10−27 .

406

C. Zhong et al.

The task offloading delay and energy consumption is considered, and which are denoted as the system cost. Therefore, if the task i is processed by the SMDs, the cost of task can be expressed as t,loc T0i + βie,loc Eiloc , loc i = βi

(4)

where βit,loc and a βie,loc are the weights of task offloading delay and energy consumption if the task i is processed by the SMDs, respectively. 2.3 MEC Computation Model max as the maximal computation resources of MEC server, and f mec as the We denote Fmec i computation resources of MEC gave to the task i. Therefore, when task i is processed by the MEC, the processing time is

timec =

ci . mec fi

(5)

Similar to Eq. (3), the MEC server computation energy consumption of task i is Eimec = κ mec (fimec )2 ci ,

(6)

where κ mec is the conversion coefficient of MEC server. Based on the practical measurement, we set κ mec = 10−28 [10]. If task i is computed by the MEC server, we can utilize the following equation to express the task offloading delay timec =

si ci + mec . ri fi

And the task offloading energy consumption can be expressed   si ci mec + mec + κ mec (fimec )2 ci . Ei = pi ri fi

(7)

(8)

The same as Eq. (4), the cost of task offloading when the task i is processed by the MEC can be expressed as = βit,mec Timec + βie,mec Eimec , mec i

(9)

where βit,mec and βie,mec are the weights of task offloading delay and energy consumption for the task if the task i is processed by the MEC and SMD, respectively. 2.4 Problem Formulation The total cost in MEC system can be formulated as =

I    αi mec + (1 − αi )loc i i i=1

(10)

Q-learning Based Computation Offloading Algorithm in Mobile Edge Computing

407

According to the above analysis, we can formulate the cost minimization problem (P1) min  α,f

s.t.

(1−αi )T0i +αi Timec ≤ timax , mec 0 ≤ fimec ≤ Fmax ,

0≤

I 

mec fimec ≤ Fmax ,

(11a) (11b) (11c)

(11d)

i=1

αi ∈ {0, 1},

(11e)

mec and F mec are the maximal computation resources of MEC server and SMD i, where Fmax i,loc

respectively. f = f1mec , ..., fImec is computation resources allocation in the MEC server. Constraint (11b) is the task offloading delay requirement. Constraint (11c) and (11d) are the MEC server computation resource constraints. Constraint (11e) is the task offloading variable constraints. Because αi and fimec are product, and αi is a binary variable, we cannot utilize the traditional method to solve problem (P1).

3 Problem Solution Q-learning is a values-based algorithm in reinforcement learning, which will eventually learn a table Q-table. For example, if there are five states and four actions in a game, we can get the table [11, 12]: In Fig. 1 each row represents each state, and each column represents each action. The value of the table is the maximum expected reward in the future when taking each action under each state. Through the Q-table, we can find the optimal behavior in each state, and then get the maximum expected reward by finding all the optimal actions. In order to obtain the values in Q-table, the method is described as follows: (1) First, we initial all the data in Q-table to 0. (2) According to the current data in Q-table select an action and execute it. (3) The implementation process is not completed until the end of this round of training. Because all the data in Q-table are all 0, we have to adopt epsilon grey strategy to select. Epsilon grey strategy means that at the beginning, by setting a large epsilon, the agent can explore the environment and randomly select actions. As the agent understands the environment, reduce epsilon so that the agent begins to use the environment to perform operations. After selecting an action in the current state, we can use the Behrman equation to calculate the Q value. (4) After taking action and getting reward, we can update Q(s, a) with Q function. (5) The optimal Q-table can be obtained by repeating this process until the training stops.

408

C. Zhong et al.

Fig. 1. Q-table representation

The task offloading strategy of SMDs is only decided by the current state and action. Therefore, the task offloading strategy can be considered as Markov Decision Process, the reinforcement learning (RL) based Q-learning method can be utilized to solve this problem. In this system, we first define the state as s, action as a and reward for the RL as Q, Q = (s, a). The system status consists of two parts, s = (t , ac), where t is the system total cost . ac is the computation ability of MEC server. The action vectors are α and f. Considering the constraints in problem (P1), the optimal task offloading and computation resource allocation strategy τ ∗ is τ ∗ = arg min (s, τ ),

(12)

τ

where τ is the feasible task offloasding and computation resource allocation strategy. And then, by utilizing free Q-learning method, (s, a), and the corresponding value Q(s, a) is    k+1 + γ Qτ sk+1 , ak+1 |sk = s, ak = a , (13) Q(s, a) = Eτ t k+1 where γ is the discount factor, t is the system total cost of iteration k + 1. First, Q(s, a) can be obtained by s and a. And then, we can store Q(s, a) is in the Q-table. When the new data is smaller than the old one, we need to update the value.We denote k , Q(s, a) can be updated by the current cost function as t     k + γ min Q(s , a)), (14) Q sk , ak ← (1−υ)Q sk , ak + α(t α

where υ means the learning rate and min Q(s , a) means the benefits in memory. If the α

maximum benefit can be obtained in state s , it will also enter the state by selecting the action s in the next selection. Based on the above analysis, the data in the Q-table can be obtained.

Q-learning Based Computation Offloading Algorithm in Mobile Edge Computing

409

4 Results and Discussion Firstly, simulation parameters are made to confirm our analysis. The number of eNodeB is 1, and SMDs are uniformly distributed in eNodeB. The maximal transmission power of each SMD is 0.5 W. The total system bandwidth as 10 MHz.The maximal transmission power of each SMD is 0.5 W. The maximal computation resources of the SMDs and MEC server are 1 GHz and 5 GHz, respectively. The conversion coefficient of SMD and MEC server are 10−27 and 10–28 , respectively. The data size of task needs satisfy Gaussian distributions, si ∼N(400, 100), and the unit is KB. The computation resources of task needs satisfy Gaussian distributions, ci ∼N(1, 0.1), and the unit is Gb. The delay requirement of the task is 1 s. Two other algorithms are compared with the proposed algorithm. • Local computing (LC) algorithm: The tasks are only executed by SMDs locally. • MEC server computing (MC) algorithm: In this algorithm, all the tasks are computed by the MEC server.

Fig. 2. Total cost versus number of users

Figure 2 is the total cost versus number of user. The total cost increases with respect to the number of users. Because the resources are limited, when the number of users increases, each user will allocate fewer resources, the total cost increases. Compared with the LC algorithm and MC algorithm, we can see that, when the number of users grows from 1 to 7, the total cost of Q-learning algorithm can achieve the system performance gain about 59.33% and 25.73%, respectively.

5 Conclusion We studied a total cost minimization problem with task delay requirement and total computation resources constraints in MEC system. With the help of Q-learning method, a Q-learning based computation offloading algorithm was proposed. Simulation results indicate that the effectiveness of the proposed algorithm.

410

C. Zhong et al.

Acknowledgement. This work was supported by the State Grid Science and Technology Project “Research and application of 5G communication device for distributed energy monitoring service” (kj2021–025).

References 1. Li, S., et al.: Joint admission control and resource allocation in edge computing for Internet of things. IEEE Netw. 32(1), 72–79 (2018) 2. Li, S., Lin, S., Cai, L., Li, W., Zhu, G.: Joint resource allocation and computation offloading with time-varying fading channel in vehicular edge computing. IEEE Trans. Veh. Technol. 69(3), 3384–3398 (2020) 3. Chen, Y., Zhang, N., Zhang, Y., Chen, X., Wu, W., Shen, X.S.: TOFFEE: Task offloading and frequency scaling for energy efficiency of mobile devices in mobile edge computing. IEEE Trans. Cloud Computing 9, 1634–1644 (2019) 4. Mao, Y., You, C., Zhang, J., Huang, K., Letaief, K.B.: A survey on mobile edge computing: The communication perspective. IEEE Commun. Surv. Tutorials 19(4), 2322–2358 (2017) 5. Li, S., et al.: Joint congestion control and resource allocation for delay-aware tasks in mobile edge computing. Wirel. Commun. Mob. Comput. 2021(1), 1–16 (2021) 6. Fang, F., Xu, Y., Ding, Z., Shen, C., Peng, M., Karagiannidis, G.K.: Optimal resource allocation for delay minimization in NOMA-MEC networks. IEEE Trans. Commun. 68(12), 7867–7881 (2020) 7. Sheng, M., Dai, Y., Liu, J., Cheng, N., Shen, X., Yang, Q.: Delay-aware computation offloading in NOMA MEC under differentiated uploading delay. IEEE Trans. Wirel. Commun. 19(4), 2813–2826 (2020) 8. Zhang, G., Chen, Y., Shen, Z., Wang, L.: Distributed energy management for multiuser mobileedge computing systems with energy harvesting devices and QoS constraints. IEEE Internet Things J. 6(3), 4035–4048 (2019) 9. Yang, Z., et al.: Energy-efficient joint resource allocation algorithms for MEC-enabled emotional computing in urban communities. IEEE Access 7, 137410–137419 (2019) 10. Li, S., Wang, Q., Wang, Y., Tan, D., Li, W.: Delay-aware task congestion control and resource allocation in mobile edge computing. In: Proceedings of the 2019 IEEE 30th Annual International Symposium on Personal, Indoor and Mobile Radio Communications (PIMRC), pp. 1–6. Istanbul, Turkey (2019) 11. Christopher, W., Dayan, P.: Technical note: Q-learning. Mach. Learn. 8(3–4), 279–292 (1992) 12. Nash, T.J.: Asynchronous stochastic approximation and Q-learning. Mach. Learn. 16(3), 185–202 (1994)

Data Analysis I

Research on Smart City Platform Based on 3D Video Fusion Lai Wei(B) Beijing Research Institute, China Telecom Co., Ltd., Beijing, China [email protected]

Abstract. At present, It is a pain point that there are much isolated and scattered video data of urban cameras, which cannot provide strong technical support for command and decision-making of managers. This paper studies the smart city scheme with high precision and high real-time display of video data, using video fusion technology to build a digital smart city. The platform architecture includes access layer, fusion rendering layer and presentation layer. The access layer provides access of sensors, cameras as well as models for data processing. The fusion rendering layer realizes the deep integration of multi-channel video into 3D models, which mainly involves camera calibration, image stitching, image projection and rendering optimization. Presentation layer is responsible for interface between a capability set and Web applications. The platform provides abilities of comprehensive and full-space awareness of security and response to emergency, and brings great economic and social benefits. It can be widely used in civil aviation, customs, ports, smart cities and other fields. Keywords: Image stitching · 3D rendering · Smart city

1 Introduction At present, It is a pain point that there are much isolated and scattered video data of urban cameras. In the face of urban security monitoring and operation management, on the one hand, video data need monitoring personnel to switch monitoring screen in real time and actively [1–3]. On the other hand, it is often only a passive tool for post-inspection, lacking global dynamic control and rapid linkage of massive video data, which cannot provide strong technical support for command and decision-making of managers. By using 3D video fusion technology, this paper studies the smart city scheme with high precision and high real-time display of video data, and builds a digital smart city based on 3D panorama. The platform architecture includes three layers: access layer, fusion rendering layer and presentation layer, which is shown in Fig. 1. The access layer provides access of sensors, cameras and models to the fusion rendering layer according to certain standards. The fusion rendering layer provides video fusion and rendering optimization, to realize deep integration of multi-channel video in different positions and different perspectives into 3D models. Presentation layer is mainly responsible for interface between Web page and fusion rendering layer to provide the display of video data and model data. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 S. Sun et al. (Eds.): ICSINC 2021, LNEE 895, pp. 413–424, 2022. https://doi.org/10.1007/978-981-19-4775-9_51

414

L. Wei

Fig. 1. Platform architecture.

2 Access Layer Access layer provides access of sensors, cameras as well as 3D GIS/BIM/CAD model data for data processing in fusion rendering layer. 3D video fusion is based on 3D model. According to different characteristics of monitoring area, different modeling methods can be selected. Open laser scanning technology is used to measure scene with high precision and obtain 3D point cloud data. A large number of discrete sampling points are automatically reconstructed by 3D reconstruction technology. For areas without scanning conditions, 3D models can be generated using modeling tools based on CAD drawings or all available images, such as UAV oblique photography images, aerial images, satellite images, etc. [5, 6] Through the 3D modeling, a full-factor 3D model scene is generated, which create a realistic, geographical and accurate 3D earth model for users. Data sources of models usually include GIS, BIM, CAD, DEM, HD Map, OSM, etc. [4]. Due to different sources of data, attributes of data may be inconsistent, so the data needs to be cleaned according to standard before use. Cleaned data will include building floor information, road lane number and other attributes. Data cleaning makes model data lightweight and unified, and solves problems of operation laggings and the hybridity of visual effect. 3D video fusion involves deep integration of cameras. The camera access supports GB/T28181 standard, and supports integration of various video resources to form a unified video access platform. In presentation layer, camera objects can be marked on 2D or 3D situation map and associated with their video source, as is shown in Fig. 2. The corresponding monitoring video can be fused into the scene models and obtained by clicking on the map and other interactive ways. Furthermore, in order to obtain a comprehensive perception of urban information and facilitate command, the platform needs to integrate access control, smoke sensor, temperature sensor, lighting control, infrared, RFID, monitoring system and other data, and orderly organize all kinds of resource data in 3D scenes. In this way, rapid retrieval

Research on Smart City Platform Based on 3D Video Fusion

415

Fig. 2. Camera point is marked in 3D situation and associated with video source.

and positioning can be realized, and overall situation of key areas can be easily mastered. Moreover, attribute information of the key areas can be connected, and state view, prediction analysis and preventive maintenance of infrastructure in various fields of the city can be carried out, so as to realize the efficient control of resources.

3 Fusion Rendering Layer As the core level of the platform, fusion rendering layer realizes the deep integration of multi-channel video in different positions and different perspectives into 3D models, which can project images into the corresponding models. It breaks the limitation of traditional video surveillance system to split screen display and realizes panoramic real-time monitoring of the whole scene, which is convenient for timely command and disposal of various emergencies, and greatly improves practical efficiency of video monitoring. Video fusion technology is a full-space video enhancement technology based on computer vision, which mainly involves camera calibration, image stitching, image projection and rendering optimization. In order to meet the real-time and accuracy of monitoring, the fusion rendering layer is divided into initialization and normal monitoring. Initialization is to determine spatial position relationship of cameras through camera calibration, image stitching, and introduce optimization idea to determine stitching gap position of the two images. According to parameters determined in the initialization phase, normal monitoring is responsible for image projection and rendering optimization, that is, fuses video streams and 3D models and inputs them into 3D model rendering engine. The 3D model rendering engine is responsible for enhanced rendering and physical effects of 3D models and other graphical data. 3.1 Camera Calibration The process of video fusion is essentially the process of projection from image plane to model space point. The mathematical form of projection process can be expressed

416

L. Wei

as the conversion between several coordinate systems. The coordinate systems include image coordinate system, camera coordinate system, world coordinate system and model coordinate system. The image coordinate system represents the imaging plane coordinate system. The camera coordinate system represents the coordinate system with the camera as the origin. The world coordinate system is the spatial coordinate system with any point in the real space as the origin. The model coordinate system is the coordinate system with any point in the 3D scene as the origin. In order to reconstruct 3D information of objects from images, the correspondence between the object point in world coordinate system and its image point on image plane must be known, and this correspondence is determined by camera position and attribute parameters. Camera calibration is to calculate the corresponding relationship between the object point in world coordinate system and its image point on image plane, that is, to determine the camera parameters. Camera parameters include internal parameters and external parameters. Internal parameters include focal length, optical center, and non-vertical factor, which represents the relationship between camera coordinate system and ideal coordinate system. External parameters include translation matrix and rotation matrix between two cameras, which represents the position and direction of two cameras in world coordinate system. Therefore, the essence of calibration is to determine the transformation matrix from world coordinate system to image coordinate system. Camera calibration methods can be divided into two categories. The first one is direct estimation of camera position, optical axis direction, focal length and other parameters. The second method is to determine the transformation matrix of 3D space points mapped to two-dimensional image points by least square fitting. After obtaining camera parameters by camera calibration, a virtual camera matching the position and angle of the real camera can be added to 3D scene [7], so that viewports can be observed from perspective of the virtual camera in the 3D scene. 3.2 Image Stitching The internal and external parameters of the real camera can be used to set the parameters of view matrix and projection matrix of the virtual camera. Then vertex coordinates of models and texture coordinates of video images can be one-to-one correspondent. The texture information on video images can be mapped to corresponding position on the surface of 3D models, so as to realize the real-time fusion of multi-channel video images and 3D scene models. Ideally, when there are overlapping areas between video images collected by different cameras, the overlapping regions are seamlessly mapped onto the 3D model surface. However, the physical errors such as camera parameters, model accuracy and color differences of video images will lead to obvious seams in fused images, including geometric seams and color seams. Therefore, it is necessary to further deal with the texture of mapped images [8]. The adjustment correction algorithm such as feature point matching and texture triangulation topology can correct the texture coordinate of video images after texture

Research on Smart City Platform Based on 3D Video Fusion

417

mapping, to solve the geometric seams. On the basis of geometric correction and alignment, the texture color of overlapping areas is balanced to solve the color seams. Finally, the seamless fusion of real-time videos and 3D scene models is achieved. Moreover, the cameras for fusion cannot be installed in face, and once the camera installation position is determined, the camera perspective cannot be changed at will, so as to avoid dislocation at the stitching position in the later. In order to achieve better fusion effect, it is recommended that the overlap between two adjacent images is about 20% [9]. 3.3 Image Projection 3D video fusion is mainly divided into stereo plane and projection fusion [10, 11]. Stereo plane technology is to map multi-video images to a stereo plane in 3D space. Video content and scene models are still relatively separated, which has good results for the case of simple scene structure. Projection fusion technology is to map multi-video images to 3D models, so that the multi-video images and 3D scene are combined into a whole. The real-time video image is directly mapped to the surface of the 3D scene model as a texture by projection texture mapping technology. This method needs to deform images before texture mapping, and a single video processing requires more CPU computing. Given the parameters of image plane coordinate system and camera coordinate system, the position of an object in 3D world coordinate system can be calculated. According to the projection relationship between world coordinate system and model coordinate system, the coordinate of the object in model coordinate system can be calculated. This series of projection relations are calculated through camera calibration and image stitching, so as to realize projection of two-dimensional panoramic video images into 3D model world, and realize registration of video image information and 3D scene, so as to achieve the effect of augmented reality and a realistic world is obtained in 3D scene, which is shown in Fig. 3.

Fig. 3. Projection of 2D video images into 3D model world.

418

L. Wei

3.4 Rendering Optimization The key technology of rendering optimization involves block technology and LOD technology. Generally, urban-level large-scale scene involves a 3D scene model with an area of hundreds of square kilometers and even thousands of square kilometers. Due to the limitation of rendering function, a rendering instance cannot render an infinite scene. In view of this feature, urban dynamic loading strategy is adopted, that is, the scene is divided by block technology and presented on demand. Through the scene rasterization mechanism, the scene is divided into a rectangular area of 2 square kilometers. Each block has 4 square kilometers of a tile, which can support any scene display of seamless splicing. The scene in the field of vision is loaded asynchronously, and the scene outside the field of vision is unloaded asynchronously. In this way, users will not feel any sense of scene switching and can experience a very large world. LOD technology refers to multi-level detail. LOD technology determines resource allocation of object rendering according to position and importance of the model nodes in display environment, reduces numbers of faces and details of non-important objects, and thus obtains efficient rendering calculation. In addition, some other technologies such as real-time halo, dynamic shadow, volume cloud, pipeline clipping and culling based on GPU acceleration can also be introduced to realize real-time rendering of 3D scenes and 3D simulation of virtual reality. For example, global illumination will greatly increase the sense of reality. Through graphics research, spherical harmonic illumination information of many sampling points in 3D scene can be rendered dynamically by combining spherical harmonic GI and RTX technology, which provides global illumination information of dynamic and static objects around the scene, which is shown in Fig. 4.

Fig. 4. Shadow effect using global illumination.

Research on Smart City Platform Based on 3D Video Fusion

419

4 Presentation Layer Presentation layer is mainly responsible for interface between Web page and fusion rendering layer. Based on open capability API provided by fusion rendering layer, it provides users with 3D fusion model interaction, panoramic video surveillance and background management. By calling scene action of the cloud rendering service in Web application, and pushing rendered scene back to front end through WebRTC [12], 3D scene interaction under B/S architecture is completed. Cloud rendering features include video fusion, automatic cruise, trajectory drawing, floor thermal map, building disassembly, POI marking, region contour, lighting and weather rendering, which is shown in Fig. 5.

Fig. 5. Interaction between web page and rendering service.

For example, the automatic cruises can be carried out according to the angle and the speed of inspector’s daily cruises. Display of resource information associated with the route setting does not require any person-to-person switching. As long as the camera is completely covered relative to the scene, it will not cause omissions. The combination of system automatic cruise and personnel patrol can not only ensure 24-hour continuous monitoring of key areas, but also save manpower. At the same time, a large scene picture displayed by system is obviously superior to small-scale patrol of personnel, which is convenient for timely detection and rapid disposal of problems, and improves the efficiency of management and command. Building disassembly supports that when multiple floors alarm, you can click on a layer to view the alarm information, and switch to view the alarm of different floors, which is shown in Fig. 6.

420

L. Wei

Fig. 6. Building disassembly.

5 Application The 3D video fusion platform based on 3D video fusion technology in this paper can be applied to the research of smart city solutions. On the basis of constructing 3D models of cities, accessing city cameras and sensors data in the 3D scene, some core technologies can be added to the platform, such as trajectory tracking technology, AI recognition technology, and data storage technology. In this way, the driving of camera and sensor data on virtual models and the visualization in virtual scene are realized, providing an efficient and powerful means for simulation, prediction and governance of the smart city. 5.1 Target Trajectory Tracking In original monitoring system, tracking is very time-consuming and inaccurate, and if the area involved is too large, it is impossible to rely solely on manual to find the trajectory of suspicious persons. Target trajectory tracking technology carries on reidentify processing to video streams, realizes real-time tracking of the target person and fast retrieval of the historical trajectory. On the premise of good coverage of full-scene cameras, through correlation and comparison of time and space, historical events can be replayed and searched across frames through the retrospective of historical video, which can visually and panoramically facilitate rapid analysis, timely command and disposal. The process of cross-camera trajectory tracking is that the front end sends target pedestrian image to this module through API gateway, and the module obtains structural features of the target through pedestrian detection and feature extraction. At the same time, data preprocessing module input multiple monitoring video streams for key frame extraction and image enhancement. Reconstruction threshold decision module judges whether to reconstruct the image. Multiple key frames enter the target tracking module to obtain pedestrian features and re-identify and compare with the structural features of the target pedestrians. If the features are matched, the structural data are given to the front end, and the path is plotted, which is shown in Fig. 7. At the same time, a video is generated by video generation module to add annotation information of re-identification results, and forwarded to video storage.

Research on Smart City Platform Based on 3D Video Fusion

421

Fig. 7. Target trajectory tracking.

5.2 AI Recognition and Analysis AI recognition and analysis service is a multi-task target detection and target recognition processing for monitoring video streams, so that the platform can perform multivariate fusion analysis based on structural data processed by AI according to business logic. Firstly, monitoring video streams are connected to data preprocessing module through RTMP protocol. The preprocessed frame is sent to image reconstruction module and AI inference module to complete image enhancement and recognition tasks. The annotated frame and the original unprocessed frame are encoded to generate a real-time video stream. AI analysis module is used to output multi-source structural data of the frame and send them to structured data storage. At the same time, according to presentation layer business logic, based on multi-source structural data of the frame, event trigger is judged and decision instruction is issued to message service. POI point tags support hierarchical classification of labels. AI service can be docked to alarm and mark recognition results, which is shown in Fig. 8. A real-time dynamic information layer is added in the scene to realize ‘what you see is what you get’. You

Fig. 8. POI point tags with AI recognition results.

422

L. Wei

can view the corresponding videos and real-time information of targets and events by clicking on the labels. Floor thermal diagram supports that according to different density of each floor, building contour can be displayed by the way of temperature field diagram, and different density of people can be distinguished by different colors, which is shown in Fig. 9.

Fig. 9. Building contour with thermal diagram.

Based on recognition results of AI recognition and analysis service, the platform can realize personnel statistics, personnel traffic management, streaming monitoring, visitor system, vehicle flow statistics, vehicle traffic management, parking statistics, etc., so as to realize perception of vehicle distribution and personnel activities, issue tasks immediately for security personnel, and make timely response to potential hazards. 5.3 Data Storage and Management Data storage and management service provides storage, distribution, modification and other operations to video data and business data. During the whole operation of the platform, each service needs to interact with data storage and management service to realize addition, deletion, and modification of model data, video data and equipment data, and returns call results to Web client through response. Data storage and management module includes REST server, media server, database and video storage (NAS + disk array), which is shown in Fig. 10. The REST server provides authentication and management interfaces, includes user management, model management and streaming management, so as to ensures application security access and facilitate users to manage in all aspects. User management completes user creation and modification operation, updates key-value database and returns processing results. Model management returns model information imported by platform. Streaming management includes streaming registration, streaming access and streaming editing. The streaming registration assigns channels to the video stream of

Research on Smart City Platform Based on 3D Video Fusion

423

Fig. 10. Data management and storage service.

specified camera and configures channel information. According to the channel information, the video stream is accessed from specified camera to media server. Background management of Web portal initiates requests such as streaming editing, viewing, freezing, deleting to the streaming management module through API gateway. The streaming management module modifies channel information of the streams according to request message, updates key-value database, and returns processing results to the front end. The media server distributes video streams to storage platform, and supports RTMP, RTSP, GB28181 protocols [13]. According to specified recording strategy and encapsulation format to store the videos, it can configure storage space. Also, the media server supports video retrieval, download, HLS playback, real-time browsing and on-line merging and cropping of video files, supporting file format MP4/FLV/TS. The database is divided into key-value database, SQL database, and model database according to data type. The function is to store, add, delete, and check system model data, video data, and business data.

6 Summary This smart city platform scheme supports good scalability and has a flexible architecture to support new system access, which fully considers upgrading, expansion and interface capabilities with other application systems to meet further needs of users. In addition, the system directly constructs some regions in existing camera network, which has no restrictions on camera point position, angle, connection mode, pixel accuracy and other parameters, and protects existing investment. The platform helps cities to carry out video security management in key public areas, as well as global monitoring and intelligent analysis of the overall large scene. It provides a comprehensive and full-space awareness of security and response to emergency, which greatly improves application ability and actual effectiveness of video surveillance system, and saves a lot of manpower input under traditional video surveillance application

424

L. Wei

mode. It not only realizes timely treatment of sudden crisis events, reduces casualties and property losses, but also improves safety inspection and operation management level of cities, and brings great economic and social benefits. It can be widely used in civil aviation, customs, ports, smart parks, smart cities and other fields.

References 1. Sun, J., Tian, Z.: Image monitoring system and key technology in underground mine. Coal Sci. Technol. 42(1), 65–68 (2014) 2. Zhang, X.: Digital mine video monitoring system based on industrial ethernet. Safety in Coal Mines 49(12), 112–114 (2018) 3. Jin, Y., Zuo, B., Han, J.: Design of mobile video monitoring system for underground mine. Safety in Coal Mines 46(6), 104–106 (2015) 4. CAICT: Research report on digital twin cities (2019) 5. Xu, H.: Research on dynamic simulation of coordinated development of coalbed methane and coal in typical mining areas. Ind. Mine Autom. 46(3), 95–99 (2020) 6. Xu, H., Yin, D.: Design and realization of mine 3D visualization monitoring system. Safety in Coal Mine 47(7), 136–139 (2016) 7. Li, X., Zhang, J.: Research on key technologies of 3D simulation model and real image fusion. Urban Construction 18 (2019) 8. Yang, L., Du, Z.: 3D multi-channel video fusion based on texture coordinate correction. Science Paper Online (2019) 9. Hualong, X.U.: Mine real-time monitoring technology based on 3D video fusion. Safety in Coal Mines 52(1), 136–139 (2021) 10. Liu, Z., Dai, Z., Li, C., Liu, X.: A fast fusion object determination method for multi-path video and three-dimensional GIS scene. Acta Geodaetica et Cartographica Sinica 49(5), 632–643 (2020) 11. Nie, D., Liu, W., Zhang, J.: Fusion of video stitching in three-dimensional space. Computer Programming Skills and maintenance 12 (2017) 12. UNREAL ENGINE Homepage. https://www.unrealengine.com/zh-CN/. Accessed 05 Aug 2021 13. SRS Homepage. http://www.ossrs.net/releases/. Accessed 11 Aug 2021

Research on Intelligent Finance in the Era of Big Data Xuemei Wu1(B) , Quan Zhou2 , Ronghui Zhang2 , and Bohan Li3 1 China Electronics Standardization Institute, Beijing, China

[email protected]

2 School of Information and Communication Engineering, Beijing University

of Post and Telecommunications, Beijing 100876, China 3 University of Southampton, Southampton, UK

Abstract. With the development of the times, the socialist market economy system of China is becoming more and more perfect. For enterprises, with the situation where oligarchs control the market has gradually disappeared, the market is no longer controlled by the enterprise. The production materials have become open and transparent, and it is difficult for products to obtain huge profits. Therefore, enterprises must do a good job in financial management, emphasizing the improvement of capital use efficiency, and gaining vitality from the meager profits. At present, the existing financial management system is basically unable to effectively meet the development needs of enterprises. Therefore, it is necessary to construct intelligent finance and strengthen the effect of financial management through technologies such as big data. This paper starts with the problems existing in the construction of smart finance in the era of big data and comprehensively launches the research on smart finance in the era of big data. Keywords: Big data era · Intelligent finance · Blockchain · Financial management

1 Introduction In the current era, big data technology has become a necessary technology in the development of enterprises. The main reason is that big data technology can meet the needs of enterprises for data information, so as to better understand the market situation and effectively solve the market. At present, the big data technology of China has limited effects in the actual application process, because the application of big data technology in enterprises still requires supporting construction [1], such as the integration of smart finance in finance. Emphasizing intelligent management and intelligent analysis can maximize the effect of big data technology to meet actual work needs. Therefore, in order to better improve the effectiveness of financial management and make full use of big data technology, Chinese enterprises are actively constructing intelligent finance, which mainly integrates existing financial and intelligent financial work. The formation of the talent team hopes that the work will be fully developed through talents, and the © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 S. Sun et al. (Eds.): ICSINC 2021, LNEE 895, pp. 425–430, 2022. https://doi.org/10.1007/978-981-19-4775-9_52

426

X. Wu et al.

value of smart financial work will be fully utilized. It is advocated that smart finance replaces manual work, so as to improve the actual effect of financial work. Generally speaking, it lies in how to reduce human resources more effectively. It is the focus of current enterprise research in our country, and the existing staff is mainly for better building intelligent financial systems and supplementing some human logical thinking and sensory information.

2 Overview of AI Techniques in Intelligent Finance The latest generation of financial cloud will integrate technologies such as big data, artificial intelligence, blockchain, mobile Internet, and cloud computing [2]. It is based on the accounting theory of the matter method, with real-time accounting and intelligent finance as the core concepts, including financial accounting services, management accounting services, tax services, accounting services, treasury services, etc. The transformation of financial and digital intelligence will enhance the business innovation capabilities of enterprises. The practice has proved that intelligent financial services based on new technologies such as artificial intelligence can comprehensively improve the efficiency of financial operations. For example, robotic process automation (RPA) can replace the standardized, repetitive, and workload-intensive financial work, freeing financial staff from repetitive, low-value-added work and devoting themselves to high-value-added work such as management accounting.

3 Blockchain Technique in Intelligence Finance Blockchain finance is actually the application of blockchain technology in the financial field. Blockchain is an underlying technology based on Bitcoin, and its essence is actually a decentralized trust mechanism [3]. Collectively maintain a sustainable growth database by sharing in distributed nodes to achieve information security and accuracy. The application of this technology, as shown in Fig. 1, can solve the trust and security issues in transactions. Blockchain technology has become an optional direction for the future upgrade of the financial industry. Through the blockchain, both parties to the transaction can carry out economic development without the aid of a third-party credit intermediary. Activities, thereby reducing the cost of assets that can be transferred globally. Since 2016, major financial giants have also been following the trend and have launched blockchain innovation projects to explore the possibility of applying blockchain technology in various financial scenarios. In particular, Puyin Group took the lead in pioneering the “blockchain +” standard digital currency. Standard digital currency is the process of identification, evaluation, confirmation and insurance of assets through a third-party institution, and then written into the blockchain through rigorous digital algorithms to form the standard correspondence between assets and digital currency, which is called standard system Digital currency [4]. In order to realize the great leap forward and great development of blockchain finance, to promote the new development of China’s economy, accelerate the circulation of global

Research on Intelligent Finance in the Era of Big Data

427

Fig. 1. Credit verification system based on blockchain technology.

assets, and realize the dream of revival that has been struggling for generations to come, the Puyin Group will hold the Puyin Group in Guizhou on December 9, 2016. At the launch ceremony of the Guiyang Strategy for Blockchain Finance, the meeting will discuss the realization of the digital circulation of assets by the blockchain, the blockchain financial transaction model, and the application of blockchain services and public social industries. This conference will mark the beginning of the application of blockchain finance and the transformation and development of new financial ecology. On June 1, 2020, Xinhua News Agency was authorized to broadcast the “Overall Plan for the Construction of Hainan Free Trade Port” issued by the Central Committee of the Communist Party of China and the State Council. Standards and rules for security and blockchain finance” are one of the key tasks of Hainan Free Trade Port before 2035.

4 Knowledge Graph Technique in Intelligence Finance As one of the important areas of artificial intelligence, the application of knowledge graph-related technologies in the field of financial anti-fraud is increasingly widespread, which has played a significant role in promoting financial risk prevention. Jiang et al. [5] believe that the key to anti-fraud in commercial banks is to establish a risk control system supported by big data, using knowledge graphs, social network analysis, and other technologies to more effectively prevent fraud risks under complex models. Chai et al. [6] pointed out that relying on the powerful knowledge reasoning and logical judgment capabilities of the knowledge map can improve not only the accuracy of risk supervision and decision-making but also enhance its interpretability. Tao et al. [7] built a knowledge map of listed companies to intelligently monitor corporate risks, improve regulatory penetration, and ease regulatory time lag (Fig. 2).

428

X. Wu et al.

Fig. 2. Financial analysis system based on knowledge graph.

However, for the accounting case prevention business, this method not only has a low proportion of fraud samples but also affects the effect of model training. The analysis of the logic of the occurrence of the case is demanding, and the full use of machine learning models to predict fraud lacks corresponding interpretability.

5 Deep Learning Technique in Intelligence Finance Since the reform and opening up, traditional financial enterprises and Internet technologies have gradually cross-integrated. For example, in the original physical banks and online banks, new Internet financial models have been expanded, such as online financial management, electronic bank accounts, mobile banking, P2P models, Third-party payment platforms, crowdfunding, etc. In addition, our country has proposed the concept of a green, innovative and powerful Internet country, which has brought Internet finance into a period of rapid development. However, during the rapid development of Internet finance, many problems have emerged, such as policy lag and insufficient supervision. These problems may infringe the rights of all parties to the transaction. Moreover, Internet finance is highly subjective and prone to potential risks, causing huge losses to enterprises, individuals, and the entire industry, thus hindering the continued and healthy development of Internet finance. Deep learning, as shown in Fig. 3, is different from traditional artificial intelligence methods. It adopts the “reverse deduction thinking” method, based on a huge amount of data, with the help of the characteristics of autonomous learning of neural networks (the process of neuron building connection relations), and continuous learning through historical data. Improve and optimize to fit an optimal model, in which RNN (Circular Neural Network) is suitable for the learning of data features on time series. Due to the sequential nature of financial data, RNN can be used to fit historical data characteristics

Research on Intelligent Finance in the Era of Big Data

429

Fig. 3. The pipeline of using deep learning models for financial predictions.

as a prediction model, and the model can be continuously optimized during the prediction process. As the amount of data accumulates, the model can theoretically be infinitely optimal. Reasonable tracking of Internet financial data and realizing dynamic data prediction under certain algorithms, draws out risk factors that have a greater impact on Internet finance and proposes corresponding improvement suggestions, which is very important for Internet finance risk warning and platform management. Traditional sequence prediction methods are mainly regression models, such as autoregressive model ARIMA, etc. However, this method has poor fitting ability for complex financial data. At the same time, methods based on deep learning are gradually being used in sequence prediction. The most famous are the long short-term memory network (LSTM) and gated GRU methods. These methods are based on time series models. However, the learning of sequences is still inadequate and cannot be effectively obtained.

6 Conclusion In summary, in the era of big data in my country, intelligent finance has become the core of enterprise development, mainly to solve the financial problems of the enterprise better and make the financial work of the enterprise smoother, especially to realize intelligent finance. With the support of big data, data analysis can basically be separated from manual work. The existing staff mainly supplement more human intelligent sensory information and operate the intelligent financial system. Therefore, it is necessary to form a talent team to realize the effective application of intelligent finance better and improve the work of intelligent finance to ensure the smooth realization of the modernization of the enterprise.

430

X. Wu et al.

References 1. Polak, P., Nelischer, C., Guo, H., Robertson, D.C.: “Intelligent” finance and treasury management: What we can expect. AI Soc. 35(3), 715–726 (2019). https://doi.org/10.1007/s00146019-00919-6 2. Jun, H.E., et al.: A preliminary study on intelligent finance of CAS. Front. Data Comput. 3(2), 50–59 (2021) 3. Chen, D., Cheng, W.: Selection of intelligent financial management models for chinese enterprises towards an operational framework. In: Proceedings of the 2020 IEEE 18th International Conference on Industrial Informatics (INDIN), vol. 1. IEEE (2020) 4. Zheng, X.-L., et al.: FinBrain: When finance meets AI 2.0. Front. Inf. Technol. Electron. Eng. 20(7), 914–924 (2019) 5. Jiang, Z.M., Chen, J.F., Zhang, C.: The empowerment of financial technology on the risk management transformation of commercial banks. Contemp. Econ. Manage. 41(1), 85–90 (2019) 6. Chai, H.F., Wang, S., Tu, X.J., et al.: Intelligent innovative regulatory tools on financial technology: Concept, platform framework, and prospects. Chin. J. Intell. Sci. Technol. 2(3), 214–226 (2020) 7. Tao, R., Wu, J.C., Xie, S.Q., et al.: Applied research on deep learning and knowledge graph in intelligent regulation. Financ. Perspect. J. 8, 56–66 (2019)

Data Mining and Reasoning of Radar Radiation Sources Based on Knowledge Graph Lingxiao Li1,2,3 , Junsheng Mu1,2,3 , Fangpei Zhang1,2,3 , Xiaojun Jing1,2,3(B) , and Bohan Li1,2,3 1 School of Information and Communication Engineering, Beijing University of Posts

and Telecommunications, Beijing, China [email protected] 2 Information Science Research Institute of China Electronics Technology Group Corporation, Beijing, China 3 University of Southampton, Southampton, UK

Abstract. This paper mainly considers the data inadaptability issue of radar radiation sources, and proposes a corresponding similarity analysis method based on the knowledge graph, which helps to improve the reasoning performance of the knowledge graph module. In addition, an event-based knowledge graph reasoning framework is established for massive radar radiation source data and some effective information increment is obtained. Finally, simulation experiments verify the effectiveness of the proposed scheme. Keywords: Knowledge graph · Data mining and reasoning · Radar radiation source · Similarity analysis

1 Introduction 1.1 The Identification of Radar Signal and Radiation Source As radar equipment is used in more and more fields, such as drones, self-driving cars, space detection, etc. multiple types of radar waveforms are considered, which is a big challenge for current recognition systems. Previously used algorithms such as wavelet transform, correlation spectrum analysis, etc. have been difficult to meet the requirements and cannot identify the corresponding signals. Therefore, the research on radar signal recognition is of great significance. The Identification of Radar Signal: There are many signal recognition methods at this stage, which can be divided into four categories according to the extracted features: namely recognition based on fuzzy function, recognition based on time-domain features and frequency-domain features, recognition based on time-frequency domain features and other recognition methods [1–3].

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 S. Sun et al. (Eds.): ICSINC 2021, LNEE 895, pp. 431–438, 2022. https://doi.org/10.1007/978-981-19-4775-9_53

432

L. Li et al.

The Identification of Radar Radiation Source: Radar radiation source identification technology was born in the 1970s and was originally called radar fingerprint identification. With the development of artificial intelligence technology, researchers have gradually integrated the K Nearest Neighbor algorithm (KNN), Neural Network (NN), Support Vector Machine (SVM), Autoencoder (AE), Deep Belief Network (DBN), Convolutional Neural Network (CNN) and other algorithms, and have achieved good results in machine learning in the field of radar radiation source recognition.

1.2 Association and Mining of Signal Data As one of the core technologies of data fusion, data association has undergone nearly 50 years of development since the 1970s and has achieved fruitful theoretical results with the joint efforts of many scholars [4]. At present, the data association field is mainly based on the maximum likelihood algorithm based on the likelihood ratio of the observation sequence and the Bayesian algorithm based on the Bayes criterion. In addition to these two categories, many new methods have been proposed such as interaction multi-model-probabilistic data association algorithm, neural networks, fuzzy mathematics and other methods, these have further enriched the theoretical system of data association algorithms. Data mining, also known as Knowledge Discovery in Database (KDD). Data mining refers to the non-trivial process of extracting hidden and potentially valuable information from a vast amount of data. At present, data mining has become the current research hotspot issue in the field of database and artificial intelligence. The commonly used methods for data analysis using data mining mainly include classification, regression analysis, clustering, association rules, feature, change and deviation analysis, Web page mining, etc. They mine data from different perspectives [5].

2 Data Mining and Reasoning Technology Based on Knowledge Graph The knowledge graph is essentially a semantic network, which can also be called a multirelational graph. It consists of multiple types of nodes and multiple types of edges. It mainly involves data acquisition, data preprocessing, importing data into the knowledge graph, application layer construction, etc. The Logical Structure of the Knowledge Graph. Logically, the knowledge graph can be divided into two levels: the data layer and the model layer. The data layer is mainly composed of a series of knowledge, and the knowledge will be stored in units of facts. The model layer is built on top of the data layer, mainly to standardize the factual expression of the data layer. The Architecture of the Knowledge Graph. As it is shown in Fig. 1, the part in the dashed frame is the process of constructing the knowledge graph, which is continuously updated and iterated with the cognitive ability of users. Top-down and bottom-up are

Data Mining and Reasoning of Radar Radiation Sources

433

commonly used construction methods. Top-down refers to defining the ontology and data model first, and then adding the entity to the knowledge base. Bottom-up refers to extracting entity data from some open links first, adding data with higher confidence to the knowledge base, finally, the ontology model of the top layer is constructed [6].

Fig. 1. The architecture of knowledge graph

3 The Construction of Knowledge Base and Reasoning Method Based on Knowledge Graph In the process of reasoning and recognition, a knowledge map is needed to query and distinguish radar intelligence, as it is shown in Fig. 2.

Fig. 2. Radar entity recognition network structure

3.1 The External Database of Radar Signal Data The external database of radar radiation sources is mainly collected on the Internet, including Baidu Encyclopedia, Interactive Encyclopedia, UAV website, self-driving car website, etc. According to the collected results, we manually screen the crawled radar entities and their text data (including brief and text parts), and then use these radar entities as an external data entity set and perform entity segmentation. For the text data of each entity, we first perform the sentence segmentation work, which is based on the segmentation of punctuation such as (.!?). Then use the word segmentation function of the Stanford parser model to perform word segmentation for each

434

L. Li et al.

clause. Based on the Penn Treebank part-of-speech tagging table, the text segmentation results are tagged. After extracting the subject, the object and their relations, and doing a second expansion, finally output the relational triples according to the composition mode of the triples [7]. The main types of reference resolution: pronoun reference (personal pronouns and demonstrative pronouns); alias reference (for example Chuanpu = Trump); apposition reference; noun phrase reference. Because the radar signal mainly needs to complete the extraction of the default value (mainly pronouns) from the above-mentioned triple relationship, we mainly focus on the problem of pronoun reference. The model we adopted realizes the completion of pronouns in a single sentence by building a neural network with variable input length. Since the length of the word entities of different sentences is inconsistent, and the original neural network cannot process the input of variable length, we implement DTW (Dynamic Warping Algorithm) when inputting variables to make all input variables have the same length. 3.2 The Construction of Knowledge Graph for Radar Signal This section mainly explains how to build a knowledge graph of radar emitters from the bottom-up. The open-source knowledge graph we used is neo4j graph database, open KE, and the programming language used is Python, Ruby. The system architecture of radar radiation source inference technology is based on the knowledge graph. From the bottom to the top, it can be divided into three levels: radar knowledge system, knowledge base construction and knowledge application. The Construction of Radar Knowledge System. Abstracting and constraining knowledge is the basis for establishing a knowledge graph, which mainly includes ontology database and knowledge classification. An ontology library refers to a collection of ontology organized in an orderly manner, in which an ontology describes the attributes and relationships of concepts. Knowledge classification describes the classification of different concepts and entities, as well as the subordinate relationship. Build a Radar Signal Event Database. The construction of a radar signal event database mainly uses a graph knowledge base, that is, the neo4j graph database used in this article. The graph database is used to store the data of the knowledge graph because the graph database naturally satisfies the node-relation-node storage format. Knowledge Extraction of Radar Radiation Source. Under the constraints and guidance of the knowledge system, automatically extracting the names, attributes, attribute values and relationships between entities of entities from heterogeneous data. Knowledge Fusion of Radar Emitter. Integrating and optimizing knowledge from different sources, including entity alignment, entity attribute value determination, entity disambiguation, and entity-relationship completion, etc. Constructing Knowledge Graph Based on Event Database. If table TC stores the one-to-many relationship of Tables TA to TB , the relationship between EAi (the entity constructed in table TA ) and EBj (the entity constructed in table TB ) can be established by

Data Mining and Reasoning of Radar Radiation Sources

435

the mapping relationship recorded in table TC . The name of the relationship is determined by table TC , which is recorded as: EAi

R(C)

→

  EBj , when R TAi (pk), TBj(pk) ∈ TC .

(1)

3.3 Similarity Analysis New Similarity Analysis Method. According to the similarity analysis carried out in this paper, combined with the actual situation of radar radiation source data, use different similarity analysis methods different data types. Finally, the analysis results are summarized and fused to obtain the inference results. The methods of recognition: It is proposed to adopt neighbor-based metrics (Adamic-Adar index). Calculate the number of relative values of a common feature z of node x and node y, defined as follows: Similarity(x, y) =



z ∈ (x) ∩ (y)

1 . log|(z)|

(2)

Radiation Source Identification Model and System. First extract the event information, get the node data, analyze the number of shared neighbors in the node data, and assign weights to them according to the Adamic-Adar algorithm. Then find out the common neighbors of two nodes, calculate the degree of their common neighbors, take the inverse of the logarithm and add them. Compared with the direct accumulation of common neighbors in the common neighbor algorithm, taking the reciprocal of the logarithm is equivalent to calculating a weight for each common neighbor. The greater the degree of a common neighbor, the lower the weight. The method of measuring the similarity of the radiation source system is the same as that of the radiation source type, that is, the neighbor-based measurement (Adamic-Adar indicator). Signal Feature Vector. For data such as the frequency band of the carrier frequency and the maximum value of the carrier frequency, due to their different units, different character lengths, and the same signal characteristics of the radar radiation source, the Mahalanobis distance is used as the similarity calculation criterion. For a multivariate T T   vector =  x1 , x2 , x3 , ..., xp with a mean of μ = μ1 , μ2 , ..., μp and a covariance matrix of , its Mahalanobis distance is:  −1 DM (x) = (x − μ)T (3) (x − μ) Mahalanobis distance actually uses Cholesky transformation to eliminate the correlation between different dimensions and the nature of different scales. According to the information and data in the project, we take the number of carrier frequencies and pulse width as examples to explain and demonstrate binary data. As it is shown in Fig. 3, It is assumed that the horizontal axis represents the number of carrier frequencies and the vertical axis represents the pulse width, Remove the coordinate axis, get the following Fig. 4.

436

L. Li et al.

Fig. 3. Binary signal discrete data

Fig. 4. Discrete data (without coordinate axis)

Introduce a new coordinate axis according to the prompt information of the data itself: the origin of the coordinate is in the center of these points (calculated based on the average value of the points). The first coordinate axis (the blue line in the figure below) runs along the “spine” of the data point and extends to both ends, which is defined as the direction that maximizes the variance of the data. The second coordinate axis (the red line in the figure below) will be perpendicular to the first coordinate axis and extend to both ends. If the dimension of the data exceeds two dimensions, choose the direction that makes the data variance the second largest, and so on. As it is shown in Fig. 5. Then we need a scale of scale. The standard deviation along each coordinate axis defines a unit length. It is easier to find a reasonable unit by using the “68–95-99.7 rule”. To better present the chart, we rotate the picture. At the same time, let the unit length in each axis direction be the same. As it is shown in Fig. 6.

Fig. 5. Binary data under the new coordinate axis

Fig. 6. Binary data under the new coordinate system

In particular, the unit vector along the new coordinate axis is the eigenvector of the covariance matrix. Note that the undistorted ellipse will turn into a circle and divide the distance length by the standard deviation along the feature vector. The amount of coordinate axis expansion is the eigenvalue of the inverse of the covariance matrix. Similarly, the amount of coordinate axis shrinkage is the eigenvalue of the covariance matrix. Therefore, the more scattered the points are, the more reduction is needed to turn the ellipse into a circle. After the scatter diagram is constructed as a coordinate system, the Mahalanobis distance can be measured to determine the degree of similarity between the data. Simulation Analysis. The database sample information includes reconnaissance time, longitude, latitude, antenna, country name, place name, model, carrier frequency, carrier

Data Mining and Reasoning of Radar Radiation Sources

437

frequency value, bandwidth, PRI, pulse width, and pulse width value. In the experiment, the number of nodes in the network is set at n = 5000, the average degree of nodes is d = 10, and the maximum degree of nodes is m = 50. According to the L value, 10 values such as 0.05, 0.1, 0.15, 0.2…0.5 are set (the more the L value is Larger, the more difficult it is to find neighbor nodes), and 10 simulation networks are generated. Take each node in the network as the starting node once, repeat 5000 local neighbor discovery experiments, and consider the average of Precision, Recall and F-score of 5000 experimental results as the final result of the algorithm. The experimental results are as follows:

Fig. 7. Neighbor-based similarity analysis radiation

Fig. 8. Mahalanobis distance similarity analysis

Fig. 9. Visualization of time-similar source event relationships

It is shown in Fig. 7 that the similarity reaches the maximum at the parameter L = 0.2, which proves that the weight ratio can be selected as L = 0.2. Figure 8 shows the Mahalanobis distance similarity analysis, through the difference in the proximity between the two points, it can be judged whether the two are similar. Figure 9 shows the visual relationship of time-similar radiation source events after analysis and reasoning. It can be seen from Fig. 9 that the proposed similarity analysis method can effectively realize the reasoning of radar emitter event attributes which were based on time similarity, and form the radar emitter information increment.

438

L. Li et al.

4 Conclusion and Discussion In this paper, the research on the data mining and reasoning of radar radiation sources based on knowledge graph was carried out, the external database of the radar radiation source and the recognition map of radar radiation source were constructed. In addition, the knowledge map similarity analysis method of radar radiation source data was proposed, and established an event-based knowledge graph reasoning framework for massive radar radiation source data, realized the effective mining and intelligent reasoning of radar radiation source data. Related simulation experiments verify the effectiveness of the proposed scheme.

References 1. Guo, Q., Nan, P., Wan, J.: Radar signal recognition based on ambiguity function eaturesand cloud model similarity. In: Proceedings of the International Conference on Ultrawideband & Ultrashort Impulse Signals, pp. 128–134. IEEE (2016) 2. Vanhoy, G., Schucker, T., Bose, T.: Classification of LPI radar signals using spectral correlation and support vector machines. Analog Integr. Circ. Sig. Process 91(2), 305–313 (2017) 3. Zhang, R., Jing, X., Wu, S., Jiang, C., Mu, J., Yu, F.R.: Device-free wireless sensing for human detection: The deep learning perspective. IEEE Internet Things J. 8(4), 2517–2539 (2021) 4. Hou, X.: Research on Multi-sensor Multi-target Track Correlation Algorithm. Northwestern Polytechnical University (2006) 5. Lin, S.: Comparison and Analysis of ID3 Algorithm, Naive Bayes Algorithm and BP Neural Network Algorithm. Inner Mongolia University 6. Xihong, Y., Xiaodong, Q., Yunliang, Z.: Relevant word recognition based on citation coupling analysis method. J. Inf. 000(007), 161–164 (2014) 7. Qianqian, L., Keliang, Z.: Research on english entity relationship extraction in military field based on dependency analysis. Inf. Eng. 005(001), 98–112 (2019)

Exploration of English Learning in Cloud Classroom APP Based on Information Technology Platform Liping Zhang1(B) and Kaitlyn Huseyin2 1 Sichuan Vocational and Technical College, Suining 629000, Sichuan, China

[email protected] 2 Ecole Superieure De Mines Paris, Paris, France

Abstract. The continuous development of information technology and the deep integration with education have greatly promoted the process of education informatization. Traditional higher vocational English classroom teaching can no longer meet the needs of people in the information society. The development of the times urgently needs teaching innovation. At this time, project-based higher vocational English teaching based on cloud class teaching gradually emerged and attracted many attentions. This research aims to explore the project-based higher vocational English teaching model based on cloud class teaching under the information platform. Based on the current teaching status of English classrooms in higher vocational colleges, this article proposes how to construct a project-based higher vocational English teaching model based on cloud class teaching. On this basis, students in higher vocational colleges are the teaching objects, and the higher vocational English textbooks are specific the unit is the teaching content. Through the practice of project-based higher vocational English classroom teaching based on cloud class teaching, the effect of this teaching mode is explored, problems and deficiencies are found, and conclusions are drawn. The results of the survey on the reasons for the use of the cloud class show that the top three items are the completion of homework and tests and other learning tasks, the teacher’s classroom requirements to use the cloud class, and the viewing of notifications and learning resources, accounting for 73.95%, 65.73%, and 55.91%, respectively. This shows that students need to strengthen their self-application of cloud classes, and higher vocational English teachers did not thoroughly explore the project-based English learning model based on cloud classes when teaching. Keywords: Higher vocational English · Information platform · Teaching mode · Cloud class course APP

1 Introduction With the progress of the times, human society has entered an information society [1, 2]. The development of information technology has a revolutionary impact on the development of education, and innovation has become the development direction of the education © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 S. Sun et al. (Eds.): ICSINC 2021, LNEE 895, pp. 439–447, 2022. https://doi.org/10.1007/978-981-19-4775-9_54

440

L. Zhang and K. Huseyin

era [3, 4]. As a form of class management and classroom teaching organization and management, cloud class courses not only integrate the advantages of many platforms, but also make up for the deficiencies of many platforms [5, 6]. The project-based higher vocational English teaching model based on cloud class teaching plays a supporting role in activating the classroom atmosphere and mobilizing learning enthusiasm [7, 8]. Many scholars have discussed the cloud class and higher vocational English teaching mode. For example, Hunstig M believes that the cloud class has its own advantages not only to avoid the interference of external information brought by social media, but also to avoid the interference of external information brought by social media. Other network teaching platforms have insufficient installation, timely feedback, monitoring, etc. [9]; Marselis SM conducts front-end analysis from the three aspects of learner characteristics, learning fields, and learning goals, and conducts the design and design of WebQuest teaching mode supported by cloud classes [10]; Tobin JJ believes that the teaching mode based on cloud class is helpful for cultivating exploratory talents, and it has the effect of cultivating students’ sense of independence, cooperation and innovation [11]. Based on the current teaching status of English classrooms in higher vocational colleges, this article proposes how to construct a project-based higher vocational English teaching model based on cloud class teaching. On this basis, students in higher vocational colleges are the teaching objects, and the higher vocational English textbooks are specific the unit is the teaching content. Through the practice of project-based higher vocational English classroom teaching based on cloud class teaching, the effect of this teaching mode is explored, problems and deficiencies are found, and conclusions are drawn.

2 Project-Based Higher Vocational English Teaching Mode Based on Cloud Class Teaching Under the Information Platform 2.1 Construction of a Project-Based Higher Vocational English Teaching Model Based on Cloud Class Teaching Under an Information Platform (1) Selection of learning resources Modern higher vocational English teaching is no longer a simple textbook plus PPT that can meet the needs of students, and the project-based teaching model based on cloud class teaching focuses on the push of micro-class videos for student self-learning resources [12]. Among them, the micro-classes are mainly designed by teachers combining teaching content, incorporating materials related to the majors or interests of students in higher vocational colleges, and selecting pertinent knowledge points for design. Taking into account the different teaching goals in higher vocational English textbooks, the listening and speaking parts are mainly guided learning plans, focusing on vocabulary and basic sentence patterns; while the reading materials can be presented in the form of “guided learning plans + video”, except for correctness. In addition to clear explanations of vocabulary and sentence patterns, videos should be used to introduce topics to guide students to think. At the same time, micro-class resources are used to improve students’ extensive reading, reasoning and other skills and reading ability; while writing teaching can adopt the “PPT + video” method to standardize students’ writing format.

Exploration of English Learning in Cloud Classroom APP

441

(2) Teaching link 1) Absorption and internalization The first 15 min of classroom teaching can be designed as the stage of knowledge absorption and internalization. At this stage, teachers will incorporate knowledge points that need to be absorbed and internalized into task teaching. Through task design, cultivate English communicative ability, communication ability, teamwork ability and problem analysis ability. In task-driven teaching activities, teachers should design purpose-oriented tasks based on curriculum knowledge points, such as combining the skills required by vocational students in future work, guiding students to solve problems that may be encountered in future work through the absorption and internalization of knowledge points. (2) Problem solving link Through the internalization of knowledge in the first 15 min of classroom teaching, the complicated problems encountered by students in self-study and completion of tasks can be answered in the next 30 min. Based on the results reported by the group representatives, guide students to ask questions, ask questions among students, and ask questions from teachers to students, comprehensively answer students’ doubts, encourage students to actively participate in problem discussions, dare to express their ideas, and then cultivate students’ language expression skills. 2.2 Project-Based Higher Vocational English Teaching Practice Based on Cloud Class Teaching (1) Design and construction of learning resources Take the first unit of Vocational English Book 1 as an example, list all the knowledge points in the course and sort them into categories. According to different content, they can be transformed into online teaching resources, or conduct classroom demonstrations, or group discussions. And so on, present the knowledge points that need to be learned to students in different ways. Ask students to work in groups to find answers from the text based on the prompt words for the questions. Collaborate among group members and between groups to cultivate the ability of students to learn actively. (2) Pre-class knowledge learning Students learn knowledge points by logging into the cloud class teaching platform on the mobile terminal to complete their pre-class knowledge. Students can also use the course platform to give feedback to teachers or other classmates that they can’t solve by themselves in order to solve them quickly. Students can also use the cloud class platform to report their learning to teachers and provide a basis for teachers to teach in class. In order to ensure the learning effect of all students, the student’s learning progress should be checked at any time in the backstage of the cloud class. If the progress is different from

442

L. Zhang and K. Huseyin

the expectations, a reminder message can be sent to the students online, and the students’ performance can also be verified by publishing pre-class self-study test questions. (3) Internalization of classroom knowledge 1) Set up a multimedia platform and integrate a variety of teaching resources. Realize teaching on the multimedia platform, including scene materials, process display, diagrams, PPT, animation materials, etc., to make teaching alive. You can also use the cloud class resource platform to send teaching courseware to students’ mobile phones. When students listen to the teacher’s lectures, they miss hearing due to one or other reasons. With this platform, students can learn again, and some students are slow to think. Can’t keep up with the pace of teachers, students can use this platform to learn at their own pace. 2) Question feedback: The cloud class speech function can send messages at any time or participate in group discussion speeches. Teachers can adjust the focus of the lecture based on the students’ question feedback, adjust the focus of the lecture, or respond individually to the questions that are generally questioned. In the course of teaching, teachers’ questions should be targeted. One is to check the preclass learning tasks and check the completion of students; the other is to study students’ feedback on the mobile platform based on the teaching content to determine their own demonstration content, teachers’ demonstrations must be standard, and operations must conform to teaching standards and enterprise production standards. Only in this way can students’ impressions be deepened and professional knowledge can be turned into practical guidance. In this process, active-thinking students have a high degree of participation, and students with poor learning ability through “pre-class learning resources” will also be improved. (4) Design of after-school teaching activities 1) After-school homework push: After-school homework is an investigation of students’ learning situation and an important part of education and teaching. Teachers can use the cloud class to send homework, specific homework requirements, topic completion time, etc., can be edited online, and sent to students after the lecture. Compared with the traditional verbal layout or blackboard layout, this layout method has obvious advantages. The first is that the homework information is retained for a long time, and students can complete it according to their own schedule; the second is to prevent students from forgetting to record the homework and cannot be remedied; the third is to have time to think about the inner relationship between the homework and the content taught by the teacher. A counter-effect of homework level, forcing teachers to consider the relationship between homework and course content when setting homework questions.

Exploration of English Learning in Cloud Classroom APP

443

2) Difficulty feedback: For difficult points in the course content, or knowledge points missed in the classroom due to various reasons, doubts, etc., you can ask questions or discuss them by initiating a private chat. The object of private chat can be the teacher or your classmates. For example, when a student initiates a private chat with the teacher, the student does not know much about the structure of the aperture and asks his own question. At this time, the teacher can directly answer the aperture structure through private chat. If there are more people asking questions about the knowledge point of aperture, the teacher can also directly send a message to all class members to explain the aperture structure again. 3) After-class test: After-class test is an important means to check the learning effect of the classroom platform. The test method can be discussion, question and answer or selection. Test questions can be pushed through the cloud class, and the system can collect test papers according to the deadline, score and count the distribution of scores. Students can also check the test results on their own, and reflect on their knowledge points that they need to continue learning based on their scores. You can also divide the students into several groups, arrange group tasks for the grouped students, promote teamwork ability, and deepen the understanding and memory of the course content through mutual discussion and thinking. 2.3 Teaching Optimization Algorithm-Teacher “Teaching” Stage In the teaching phase, the teacher’s duty is to improve the average grade of the entire class. At this stage, all learners update their positions based on the difference between the current teacher’s position and the average position of the class. In the “teaching” phase, learners update according to the following equation: Xmean

Np 1  = Xi Np

(1)

i=1

Differencei = ri (Xteacher − TF ∗ Xmean )

(2)

newXi = Xi + Differencei

(3)

Among them, Xi represents the knowledge level of the i-th learner before learning, newXi represents the knowledge level of the i-th learner after learning; learning step ri = rand(0, 1) is a random number between [0, 1]; TF = round[1 + rand(0, 1)] is the teaching 1 Np Xi factor that determines the average value, and its value can be 1 or 2; Xmean = Np i=1 is the average of all learners; Xteacher is the best solution of the current generation. Teachers are the best learners among all learners.

444

L. Zhang and K. Huseyin

3 Investigation and Research on the Project-Based Higher Vocational English Teaching Model Based on Cloud Class Teaching Under the Information Platform 3.1 Research Methods (1) Questionnaire survey method Through the design of project-based higher vocational English teaching related questions based on cloud class teaching, they are presented in the form of questionnaires. The subjects of a series of questionnaires in this research are mainly students from higher vocational colleges. The main purpose is to explore efficient learning methods and provide new ideas and directions for the teaching of higher vocational English courses 3.2 Design and Distribution of Questionnaires (1) Design of the questionnaire The questionnaire includes two parts: the current situation of cloud class use and the student satisfaction scale. It involves investigating questions such as the reasons for using the cloud class, the evaluation of the cloud class, and the attitude. (2) Issuing the questionnaire This survey selected 15 majors of school H to conduct a questionnaire survey. There are two forms of questionnaire distribution. One is to contact the instructors and head teachers of each major class and send the questionnaire link to the students through them; The other is to use the time of students in self-study classes and evening self-study to conduct questionnaire surveys in the classrooms of various professional classes, allowing the survey subjects to scan the code through WeChat to enter the questionnaire star questionnaire page, and at the same time, a simple questionnaire explanation was carried out on the spot to help the survey subjects understand the research intention of this research and help them understand the meaning of each option. A total of 1032 online questionnaires were collected, of which the number of valid questionnaires was 998, and the questionnaire effective rate was 96.71%.

4 Data Analysis of the Project-Based Higher Vocational English Teaching Model Based on Cloud Class Teaching Under the Information Platform 4.1 Reasons for Using Cloud Class The results of the survey on the reasons for using cloud classes are shown in Table 1. The top three items are completion of homework and tests and other learning tasks, teachers’

Exploration of English Learning in Cloud Classroom APP

445

classroom requirements to use cloud classes, check notifications and learning resources, accounting for 73.95%, 65.73%, 55.91%, respectively. Table 1. Reasons for using cloud classes Reasons

Numbering

Number of people

Proportion

Teacher classroom requires the use of cloud class

1

656

65.73%

Finish homework and

2

738

73.95%

Learning tasks such as testing

3

558

55.91%

View notifications and learning resources

4

409

40.98%

Join a study group for group discussion

5

389

38.98%

Seek help from teachers and classmates when encountering problems

6

79

7.92%

proportion

Number of people

7.92

6

Numbering

%

5

79 38.98

4

40.98

389 409

55.91

3

558

73.95

2

738

65.73

1 0

100

656 200

300

400

500

600

700

800

Number of people Fig. 1. Reasons for using cloud classes

It can be seen from Fig. 1 that the main reason for students to use cloud class is to view learning resources and complete learning tasks. The promoters of cloud class are teachers, and students need to strengthen their self-application of cloud class. Higher vocational students are in a passive position in the process of using cloud classes to learn. The main reason why students use cloud classes is that teachers require students to use cloud classes for learning during the teaching process. And after investigation, it is found that students’ common cloud class functions are sign-in, in-class learning courseware, and online testing, which are all basic functions, which indicate that teachers have not explored the project-based English learning model based on cloud class in depth.

446

L. Zhang and K. Huseyin

4.2 Current Situation of Students’ Satisfaction with the Cloud Class Platform Higher vocational students are relatively familiar with the Internet. Cloud Class is an interactive teaching tool based on the Internet, and its platform image will affect students’ satisfaction in using Cloud Class to a certain extent. The satisfaction of students with the cloud class platform is shown in Table 2. The proportion of the number of people who choose each topic is very consistent and consistent with more than 70%. Table 2. The current status of student satisfaction with the cloud class platform Satisfaction

Numbering

Number of people

Proportion

Good overall impression

1

740

74.15%

Functional design satisfaction

2

708

70.94%

Satisfactory presentation of course content

3

733

73.45%

Satisfactory system operation speed

4

706

70.74%

Number of people

800

740

proportion

%

733

708

706

Number of people

700 600 500 400 300 200 100

74.15

70.94

73.45

70.74

0 1

2

3

4

Numbering

Fig. 2. The current status of student satisfaction with the cloud class platform

As shown in Fig. 2, more than 70% of students are satisfied with the overall impression and functional characteristics of the cloud class. In the process of students using the cloud class, most of the students are satisfied with the cloud class, and some students have higher requirements for the cloud class, indicating that the cloud class still needs improvement and needs to be improved continuously.

Exploration of English Learning in Cloud Classroom APP

447

5 Conclusion The project-based higher vocational English teaching based on cloud class teaching under the information platform combines modern information technology, conforms to the development background of the “Internet+” era, changes the traditional teaching mode, realizes “student”-centered teaching, and promotes learners self-learning. This research studies the project-based classroom teaching based on cloud class teaching under the information platform of higher vocational English courses in theory and practice. Practice has shown that this teaching mode can stimulate students’ interest in learning, enhance students’ learning initiative, improve students’ learning behavior, and significantly promote the improvement of classroom teaching effects, basically achieving the expected teaching effects, and the students’ satisfaction with the project-based English classroom teaching based on cloud class teaching under the information platform is relatively high. Acknowledgments. A Research on Construction of Public English Flipped Classroom Based on Project Teaching model (16SB0298).

References 1. Wu, X.: Application of artificial intelligence in higher vocational English teaching mode. J. Phys. Conf. Ser. 1852(2), 022089 (2021) 2. Liao, Y.: Discussion on the reform of mixed-English teaching mode of higher vocational English based on “Internet +”. J. Contemp. Educ. Res. 4(12) (2021) 3. Yang, C.: Research on the mixed teaching mode of TPACK in software testing course. Recent Pat. Comput. Sci. 11(4), 302–311 (2018) 4. Xia, D.: Research on the mode of ESP course under the background of the application of technological transformation. Int. J. Technol. Manage. 000(004), 27–29 (2016) 5. Wang, L., Liu, D.: Research on the professional development mode of primary and secondary school English teachers. Int. J. Technol. Manage. 000(005), 47–49 (2016) 6. Yan, M.: Research on the proper countermeasures of constructing the student centered English classroom atmosphere. Int. J. Technol. Manage. 000(004), 54–56 (2016) 7. Zhao, Y.: Research on the college English teaching innovation methods based on the theory of multiple intelligence and language cognitive. Int. J. Technol. Manage. 000(005), 25–27 (2016) 8. Susilo, Y.O., Liu, C.: The influence of parents’ travel patterns, perceptions and residential self-selectivity to their children travel mode shares. Transportation 43(2), 357–378 (2016) 9. Hunstig, M., Hemsel, T., Sextro, W.: High-velocity operation of piezoelectric inertia motors: experimental validation. Arch. Appl. Mech. 86(10), 1733–1741 (2014). https://doi.org/10. 1007/s00419-014-0940-0 10. Marselis, S.M., Yebra, M., Jovanovic, T., et al.: Deriving comprehensive forest structure information from mobile laser scanning observations using automated point cloud classification. Environ. Model. Softw. 82, 142–151 (2016) 11. Tobin, J.J., Looney, L.W., Li, Z.Y., et al.: The VLA nascent disk and multiplicity survey of perseus protostars (VANDAM). II. Multiplicity of protostars in the perseus molecular cloud. Astrophys. J. 818(1), 73 (2016) 12. Logothetis, T.A., Flowers, C.M.: Squaring the circle by attempting to teach a lab class in the cloud: reflections after a term in lockdown. J. Chem. Educ. 97(9), 3018–3022 (2020)

Data Analysis of University Innovation Fusion Based on Big Data Technology Yan Zhang1,2(B) and John Madurai3 1 Shenyang Institute of Technology, Fushun, Liaoning, China

[email protected]

2 Nueva Ecija University of Science and Technology, Cabanatuan City, Philippines 3 George Mason University, Virginia, USA

Abstract. In the new generation, the requirements for higher education training become much higher. As an important part of Chinese higher education, private colleges and universities develop many persons with kinds of abilities for local economic and social development. At present, in order to improve training quality, our Chinese private colleges and universities must effectively integrate the two situation, change the traditional educational concept and innovate the talent training mode, aim to better deliver much more qualified persons to the society. Keywords: Private colleges · Big data · Mass entrepreneurship and innovation education · Professional education · Analysis of integration path

1 Current Situation of Mass Innovation Education in Private Universities There are more than 2,000 private colleges and universities in China, which not only solve the practical needs of some social individuals to take higher education, but also provide many qualified persons for economic and social development. The mass innovation education has been launched for long time in Chinese private colleges and universities, and we had got some achievements. Some colleges and universities carry out the mass innovation education courses and include the career planning into the content of the courses. Many private colleges students began to actively participate in all kinds of entrepreneurship competition activities and improve their innovation and entrepreneurial ability. As a target to train high quality application-oriented person, Shenyang Institute of Technology, under the guidance of the school, improve the “1344” innovation and entrepreneurship education and working system [1], strengthen the school two-level mode, for our school to deepen the reform of innovation and entrepreneurship education top-level planning, collaborative implementation, cultivation service and innovation demonstration work. The specific “1344” innovation and entrepreneurship education system content is: 1 is centered on cultivating students’ innovation spirit; 3 is the threeinnovation platform for creativity, innovation and entrepreneurship. “4” is to bring theoretical teaching, practical teaching, extracurricular activities, and school and enterprise © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 S. Sun et al. (Eds.): ICSINC 2021, LNEE 895, pp. 448–453, 2022. https://doi.org/10.1007/978-981-19-4775-9_55

Data Analysis of University Innovation Fusion

449

linkage into the whole process of professional talent training; “4” is the four key indicators of college students’ innovation and entrepreneurship competition, student research and entrepreneurship driving employment, to improve the sustainable development of innovation and entrepreneurship education [2]. However, in the meanwhile, there are still some pending problems need to be solved. Especially, mass innovation education has not formed a complete theoretical system and training system, and we cannot produce advantage in the mass innovation education training. As far as our school is concerned, the corresponding teachers are still relatively weak, and the guidance ability of private college students in mass innovation education and innovation is still relatively poor. These problems not only impact the quality of training, but also impact the continually development of private universities. For private colleges and universities, whether the graduates trained by private universities can quickly meet the requirements of economic development has become one of the important evaluation standards of the public to the running level of private colleges and universities. The research on innovation and entrepreneurship education in private universities can enrich the concept, modes, methods and training mechanism of higher education, promote more scientific and standardization, improve the teaching and innovation level of higher schools, and comprehensively improve the innovation ability and comprehensive quality to cope with the increasingly severe employment situation of the current society [3]. Therefore, we should take mass innovation education as an important training content and strengthen the integration of the concept of professional education.

2 Innovation and Entrepreneurship Education and Professional Education Help Mass Innovation Education Organize teachers to carry out mass entrepreneurship and innovation learning. Teachers who have the consciousness of mass innovation is the premise of consciously infiltrating the components in daily teaching activities. Therefore, each teacher must keep pace with the times, and update the traditional teaching mode and concept. At the same time, we need to establish the awareness of mass entrepreneurship and innovation, we must clarify the purpose and requirements of mass innovation education, improve the consciousness of infiltrating mass innovation education in the activities, provide convenient conditions for promoting its education, and implement in practice. Take Shenyang Institute of Technology as an example, according to the guiding ideology of major training, the school has set up four curriculum platforms, public education platform + general education platform + professional education platform and mass innovation education platform. In order to indicate students to establish the awareness of massive innovation in the process of talent training, get the basic methods of innovative thinking and entrepreneurial practice, and improve the ability to apply professional knowledge to creatively solve practical problems [4]. According to their own professional characteristics, each professional talent training program is suggested to set up about 10% of the total credits of innovation and entrepreneurship education system credits, including in-class credits and after-curricular credits.

450

Y. Zhang and J. Madurai

At the same time, Shenyang Institute of Technology implements the education mode of “collaborative education, alternating work and study, integration of science and practice, and integrating learning and application”. The school aims at training applicationoriented talents, and pays attention to the improvement of students’ practical ability, to strengthen the “competition”. Promote teaching, to promote learning, to promote practice “traction action, requires the same course of the theory teaching and practice link taught by the same teacher, if conditions, in the experimental training courses, introduce big competition cases in professional curriculum, enhance the integration of innovation consciousness and innovative thinking and professional courses in the first classroom [5]. Improve the integration education system, entrepreneurship service system in the process of talent training, the school gradually improve the professional education-competition training-business temper integration education system chain, classroom teaching-project guidance-entrepreneurship training-entrepreneurship training-business incubation of business service system chain, parallel cross, under the concept of promoting X”, the innovation and entrepreneurship education into teaching in each link of undergraduate course, student activities and social practice. In 2018–2020, students won 262 national innovation awards and 2,373 provincial awards, among the top among the universities in the province, 7 from 2014–2018 to 4 in 2016 to 2020 to 260 in 2021. College students can enhance their awareness of innovation, and exercise and improve their insight, thinking, imagination and hands-on operation ability (Table 1). Table 1. Statistics on the awards of college students’ E&I competition (2018–2020) Project name

2018

2019

2020

National first prize

9

9

3

State second class

27

23

21

National award

36

69

57

National excellence award

8

8

5

52

71

159

Provincial second prize

124

119

321

Provincial third prize

190

189

558

75

149

417

1

8

39

522

645

1580

Provincial first prize

Provincial awards Provincial individual awards Total

Data Analysis of University Innovation Fusion

451

3 Integration Method of Mass Innovation and Professional Education in Private Colleges in China For China’s private colleges and universities in China, in order to promote the effective integration of the two, we must change the traditional education concept, further clarify the goals of talent training, strengthen the construction of curriculum system and teachers, and build a practice platform to innovate teaching means. Only can we effectively promote the effective combination of the two through multiple measures [6]. 3.1 Change the Educational Concept and Cover All the Innovation and Entrepreneurship Courses For Chinese private universities, In the concept of talent training, we should pay more attention to students’ professional skills and innovative and entrepreneurial thinking training, Guide students to better understand the significance of mass entrepreneurship and innovation education. In Shenyang colleges as an example, the school opened 16 innovation entrepreneurship, career planning and employment guidance courses (college students employment and entrepreneurship guidance, innovation entrepreneurship training, innovation credit, professional innovation activities, innovation credit, innovation practice, innovation practice, innovation practice, innovative thinking, innovative thinking training, business management, entrepreneurial business model design and innovation, career planning, job search OMG-college students employment guidance and skills development, personal development and employment guidance, college students entrepreneurship summary and practice and career planning-experiential learning). Innovation and entrepreneurship courses can be selected by professional colleges and modular [7]. It is a required course covering the whole school with a total credit of not less than 2 credits. The provincial textbook “13th Five-Year Plan” Entrepreneurship: “Theory and Practice” is compiled. 3.2 Integrate the Innovation and Entrepreneurship Education into the Whole Process of Talent Training Take Shenyang Institute of Technology as an example, the school requires all professional talent training plan to create favorable conditions for the cultivation of students’ innovation and entrepreneurship ability, and build an all-round innovation and entrepreneurship platform closely combined with its own professional background. To help students to establish the awareness of mass innovation training, get the basic methods of innovative thinking and entrepreneurial practice, and improve the ability to apply professional knowledge to creatively solve practical problems. All the professionals, according to their own professional characteristics, it is recommended to set up about 10% of the total credits of innovation and entrepreneurship education system credits, including in-class credits and extra-curricular credits [8]. At the same time, our school has established a cross-school study mechanism and signed cooperation agreements with Shenyang University of Aerospace, Shenyang Institute of Engineering and Liaoning University of Petroleum and other universities, realizing the sharing of educational resources.

452

Y. Zhang and J. Madurai

3.3 Promote the Integration of Big Innovation Competition and Big Training Projects Shenyang Institute of Technology around “to promote education, to race, promote learning, to practice” work policy, strive to carry out big competition and big training project activities, encourage professional teachers will big competition project and professional teaching organic integration, consolidate the big training projects and competition fruitful results, timely promote the competition results, make the competition concept of teachers and students, form positive feedback [9]. Establish and improve the school two levels supporting encouraging policies and measures, for teachers and students in the training and competition to provide people, financial, material, perfect competition training, competition selection mechanism, form big competition and big training project integration pattern, allow students to use competition results replace course credit, encourage cross-professional, college cooperation, solve the problem of hardware resources and software resources integration, stimulate big innovation competition and big training project vitality [10]. From 2018 to 2020, students from our school participated in 71 national projects, 178 provincial projects and 50 university-level projects, among which more than 70% of the projects included the provincial innovation competition, and students had great participation and high enthusiasm.

4 Conclusion In the context of the new era, Private colleges and universities want to better adapt to the development of The Times, We must innovate the talent training mode, We need to promote the mutual integration, In the process of talent training, Should actively adapt to economic and social development, Further enhance students’ innovation and entrepreneurial ability, Encourage and support students in conducting innovative and entrepreneurial practices, Further enrich the students’ innovative knowledge, Improve students’ entrepreneurial skills, This requires private universities to change traditional educational concepts and methods, Reestablish the talent training objectives, Adhere to the concept of people-oriented education, and leverage and leverage the government, social and universities themselves, Explore a talent training model suitable for economic and social development, Send more qualified talents for China’s economic and social development.

References 1. Opinions of the General Office of the State Council on Deepening the Reform of Innovation and Entrepreneurship Education in Colleges and Universities, p. 36. State Office (2015) 2. Circular of the General Office of the Liaoning Provincial People’s Government on Printing and Printing the Implementation Plan for Deepening the Reform of Innovation and Entrepreneurship Education in Colleges and Universities in Liaoning Province, p. 70. Liaozheng Office (2015) 3. Report on the Achievements of Shen Gong’s Exploration and Practice in Promoting Teaching Reform with Innovation and Entrepreneurship Education Reform System

Data Analysis of University Innovation Fusion

453

4. Anxian, B.: Present situation of innovation and entrepreneurship among applied college students—based on the questionnaire survey of landscape architecture specialty of Sanming college. J. Jilin Inst. Agric. Sci. Technol. (27), 20–23 (2018) 5. Huang, L.: Research on innovation and entrepreneurship education in applied undergraduate colleges in China—taking F college as an example Fujian demonstration university (2019) 6. Jiao, L., Wang, Y.: Construction of practice teaching system of innovation and entrepreneurship education in colleges and universities in China. J. Liaoning Inst. Educ. Adm. (2), 45–46 (2015) 7. Gong, L.: Reflections on the reform of undergraduate education in applied colleges—based on the combination of innovative entrepreneurship education and professional education. J. Natl. Inst. Educ. Adm. (9) (2016) 8. Zhao, Y.: A study on the model, problems and countermeasures of entrepreneurship practical education for college students in China. Heilongjiang High. Educ. Res. (10), p.17–20 (2018) 9. Wang, Z.: History of Innovation and Entrepreneurship Education in China, pp. 23–25. Social Science Literature Press, Beijing (2016) 10. Chen, X.: Innovation and entrepreneurship education throughout the whole process of talent training in Colleges and universities. China High. Educ. Assoc. Educ. (12), 4–6 (2010)

Investigation and Analysis of Online Teaching Documents Based on Data Mining Technology Yuan Liu(B) and Baoquan Men Henan Vocational College of Agriculture, Zhengzhou 451450, Henan, China [email protected]

Abstract. The development and education of big data (BD) and new technologies have provided technical support for education and teaching reform and promoted the progress of online teaching. In order to study the innovative research and development of the online teaching management model of higher vocational education under the background of BD, this article is based on online teaching management, using case analysis, document analysis and other methods to collect data from the database and build it based on online teaching. Model, and read and analyze a large number of relevant documents through the literature survey method. The experimental results proved that more than 50% of the population were dissatisfied with the current teaching system and believed that it did not play its due role, and more than 70% agreed with the reform of the teaching model. The experimental results also show that with the development of BD, online teaching is becoming more and more perfect. The current online teaching data is about 80% better than five years ago, and students’ acceptance of online education is getting higher and higher. This shows that, as a trend, online teaching has become more and more important. Keywords: Big data · Higher vocational education · Online teaching · Teaching mode

1 Introduction The growth of large industry data and BD technology and its applications are closely related. Academic Research “BD industry,” although originated in industry practices, but far behind the growth of the practice. Domestically, the current research is mainly based on large data industry and government industrial policy planning, industrial growth proposals, relatively large data industry at home and abroad, and other factors affecting industrial growth. Large internal data industry, the lack of suitable theoretical perspective. Components and governance mechanisms constitute [1]; View from abroad, although not many studies, but researchers have begun from a business point of view to explore the ecology of large data industry. BD ecosystem is divided into three levels of data distribution channels, namely the core value chain, value chain extending middle and macro levels [2]. Extend the value chain to the core value chain as the core technology providers, market data vendors, suppliers data suppliers, data products and complementary service © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 S. Sun et al. (Eds.): ICSINC 2021, LNEE 895, pp. 454–460, 2022. https://doi.org/10.1007/978-981-19-4775-9_56

Investigation and Analysis of Online Teaching Documents

455

providers, and other components of direct end-user data; and a macro-economic level BD important ecosystems such as government agencies, regulators, investors, industry associations, academic and research institutions, standardization bodies, start-ups and business organizations. And various other competitors, stakeholders and members of other regional [3]. With the continuous deepening of online learning, there are more and more collections of short message texts generated through online learning between students and teachers, so how to use these data to analyze students’ online learning the characteristics of interaction are very important [4]. It is precisely because of this demand that BD came into being. BD is formed by the intersection of multiple disciplines and has been widely used in various fields such as: medical care, e-commerce, finance, education, etc. Mining meaningful or potential information in data is not the ultimate goal of data mining, but to effectively summarize and organize the mined data to form effective data that can help people solve practical problems and provide data support for decision-making services. With the promulgations of the opinions of the Ministry of Education, it is clearly proposed to take “normal monitoring of basic teaching status data” as an important work content of colleges and universities. Colleges and universities have started to establish information platforms for monitoring and evaluating education quality to realize the normal monitoring of higher education teaching status. However, most of the monitoring content is a basic statistical analysis of the teaching staff, student-to-teacher ratio, teaching funds, professional courses, teacher evaluation, student performance, student employment and other related data, and lack of in-depth data mining. Some universities through the establishment of internal teaching quality monitoring system, using the feedback information to quality improvement decisions, but for the collection, summary and analysis of monitoring data, mostly adopts the traditional manual method, makes the workload big, complicated and error-prone, longer data processing time also make the quality evaluation can’t timely feedback effect [4]. Some studies believe that the operation of quality assurance mechanism is to find existing problems and gaps in time, and to help teachers, students and teaching managers improve teaching and learning, rather than simply ranking, selecting and eliminating [5]. However, the current data analysis means cannot meet the requirements of further guidance [6]. Began to put forward some colleges should enhance data collection, analysis and mining application level and ability, with the aid of specialized analysis software to carry out the depth of the data mining, to do a good job of teaching quality information service, thus to put forward the suggestion, make a diagnosis, realize the tracking in the whole process of the internal teaching quality monitoring and real-time feedback, effective quality control and improvement. In addition, it is proposed to use big data technology to drive the establishment of teaching quality monitoring mechanism, strengthen the early warning function of quality monitoring, continuously and dynamically collect relevant data of teachers’ teaching and students’ learning, and use data mining, statistical modeling and other technologies to carry out in-depth and multi-dimensional mining of teaching operation state data. Monitor and give early warning on whether the running status is close to the preset target, so as to form a teaching monitoring mechanism with timely feedback [7].

456

Y. Liu and B. Men

The definition of BD ecosystem, ecological structure made of large data model of the system, a detailed analysis of its components, has opened up a new theoretical perspective for large data chemist depth industry knowledge and understanding. Secondly, from the perspective of network management to explore BD ecosystem governance mechanisms proposed restraint mechanisms, incentives, coordination and integration of three harnessing mechanism, to further expand the governance mechanisms of large data ecosystem. BD ecosystem governance. Third, the model structure and governance model based on BD ecosystem, the status of vocational e-learning in-depth analysis, and make relevant suggestions based on the analysis concluded, with some reference value, significance.

2 Online Teaching Methods of Higher Vocational Education Under the Background of Big Data A large number of data are produced in the normal operation of higher vocational colleges, and various problems are often encountered in the process of analyzing these data. Under normal circumstances, most institutions do not feel a lack of data, but rather that it is too large and complex to take effective measures to extract useful information and use it. Especially in the face of the need to deal with data from multiple databases and different specifications of the data, its complexity can be imagined, which is also the main problem in the analysis of data engineering. To this end, in the process of human exploration, Data Warehouse (Data Warehouse, referred to as DW or DWH) concept should be produced. Researchers found that after get more data in the database, for a specific point in time or period after the analysis of target data processing storage together to real-time data warehouse, to realize the target data format and standardized, practical solution from multiple target data in the database is not consistent and difficult to deal with [8]. According to the above reasons, data mining has great potential in the teaching evaluation of secondary vocational colleges, and can play an unimaginable role, and will certainly have a significant impact on the relationship management between teachers and students. But data mining is well used in other areas as well. Such as familiar with secondary vocational school educational administration management system, employment statistics, and so on can use data mining, because it contains huge amounts of information, based on the information and data processing of mining, can know the factors affecting employment has in part because of education quality of teaching, it can better provide effective decision-making direction for leaders. The teaching method of using multimedia for display has not yet played a subversive change. Students just passively accept knowledge blindly. However, it is impossible to learn things other than knowledge points in the classroom, ignoring the improvement of its various aspects of ability [9]. With the continuous development of educational technology and information technology, and the continuous deepening of educational reforms, many schools have realized teaching informatization and chose the online teaching mode [10]. t=

1 1 fnm ∗ ln fnm n=1 ln x

(1)

Investigation and Analysis of Online Teaching Documents

 d=

x m=1

wn ∗ (rnm − uqm )2

457

(2)

Q collection of computing resources corresponding to resources requested by the user similarity: cosm(r, uq) = α ∗ cos(r, uq) + (1 − α) ∗

1 m δij n=1 m

Simulation parameters are calculated as follows:  (qrp,rpn − qrpm )(quq,uqn − quq )  F= n 2  (quq,uqm − quq )2 n=1 (qr − quq )

(3)

(4)

3 Innovative Experiment of Online Teaching Management Model for Higher Vocational Education 3.1 Subject This paper conducts a field survey of a school in this city, classifies online and offline teaching, and aims to optimize the online teaching management mode of higher vocational education. In-depth research has built a system for the optimization of online teaching in colleges and universities, and propose possible scientific opinions based on BD in university reform promoting online. 3.2 Establish a Model Evaluation Index System Definitive conclusions can be drawn from the actual observed. In general, the rating index system includes three levels of rating: a gradual breakdown of the relationship and an improvement. Wherein the evaluation and evaluation of two relatively abstract, not as a direct basis for the evaluation. 3.3 Statistics Data mining is naturally inseparable from the corresponding data mining software, which mainly includes commercial and open source data mining software. The commonly used mining tools include: SPSS statistical analysis software, EnterpriseMiner data analysis software, Intelligent Miller data mining tool, Weka data mining software, etc. This paper mainly uses SPSS 22 statistical analysis software for data analysis.

4 Innovative Experimental Analysis of Online Teaching Management Model for Higher Vocational Education We have made statistics on traditional teaching effects in recent years, and digitized them through templates, so that the data can intuitively show online teaching results, as shown in Table 1:

458

Y. Liu and B. Men Table 1. Online teaching effect 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020

Teaching efficiency 2.17

2.31

1.92

Teacher effect

2.08

2.13

2.39

Student’s result

1.97

2.23

2.24

Student interest

2.03

2.45

2.2

Teaching cost

1.85

2.23

1.86

1.98

2.21

2.42

2.36

2.31

2.3

2.24

1.94

2.02

1.92

2.32

2.02

2.46

1.91

2.19

2.19

2.24

1.86

2.48

Teaching efficiency

Teacher effect

Student interest

Teaching cost

2.32

2.12

2.41

2.91

2.28

1.98

2.54

2.35

2.2

2.45

1.83

1.8

2.48

1.95

2.37

2.66

student's result

3.1

2.91

2.9

Paremeter value

2.7 2.5 2.3 2.1 1.9

2.42 2.46

2.41

2.48 2.35 2.31 2.45 2.322.32 2.37 2.24 2.23 2.21 2.2 2.17 2.24 2.23 2.2 2.12 2.19 1.98 2.19 2.02 1.972.03 1.94 1.92 1.92 2.02 1.95 1.91 1.86 1.85 1.83 1.8 1.86

2.66 2.45 2.48

1.7 1.5 1.3 2011

2012

2013

2014

2015

2016

2017

2018

2019

2020

Years Fig. 1. Changes in online teaching

From Fig. 1, we can see that under traditional teaching in recent years, students’ grades and interest in learning have not improved much. This is because traditional teaching mostly uses a single teaching model of blackboard + chalk, which is not attractive to students, and the cost of teaching is increasing. We also make statistics on online teaching. Due to the short time of online teaching, we only count online teaching in the past 6 years, as shown in Table 2:

Investigation and Analysis of Online Teaching Documents

459

Table 2. Various parameters of online teaching 2015

2016

2017

2018

2019

2020

4.47

3.82

3.85

4.19

4.12

4.09

Teacher effect

4.64

5

4.84

4.53

4.88

4.2

Student’s result

5.33

4.78

4.79

5.3

5.02

4.86

Student interest

5.74

5.24

5.91

5.29

5.58

6

Teaching cost

2.39

2.42

2.34

2.23

2.21

2.28

Teaching efficiency

From Fig. 2, we can see that the efficiency of online teaching and the degree of attraction to students are higher than the traditional teaching mode, and the teaching cost is similar to the traditional teaching mode. This shows that online teaching is a new trend in teaching mode improvement. It is enough to solve the problem of concentration of students and professional problems of teachers in online teaching.

Teaching efficiency

student's result

Teacher effect

Student interest

Teaching cost

7 6

Value

5

5.33 5.74 4.47

4.64

4 3

2.39

4.78 5.24 5 3.82 2.42

5.91 4.79 3.85

4.84

5.3 5.29 4.19 4.53

2.34

2.23

5.02

5.58

4.12 4.88

2.21

6 4.86 4.09

4.2

2.28

2 1 0 2015

2016

2017

2018

2019

2020

Years Fig. 2. Online teaching efficiency

5 Conclusions With the growth of online learning and the exchange of people’s online learning deepening various learning platform has accumulated a big number of semi-structured data. Wide range of data, it will organize its hidden value behind them, so as to improve our teaching. Education Data mining is mining education system exclusive data, the confusion of education data into useful information to better understand the students to make educational decisions for us to optimize our teaching. The education system’s

460

Y. Liu and B. Men

unique data includes both teachers and students interactive chat data generated, including education management data generated online management school students. Acknowledgements. Supported by Higher Education Teaching Reform and Practice Foundation of Henan Province: Study and Practice of Constructing High-quality Skilled Talents Cultivation System in Vocational Colleges from the Perspective of “Three-wide Education Model” (No. 2019SJGLX642).

References 1. Guo, Y.: Research on the hierarchical teaching of higher vocational mathematics. Farm Staff (576(05)), 180 2. Lu, W., Sheng, Y.: Research “Suspension of Classes and Non-stop Schools”. Int. Public Relat. (102(06)), 122–123 (2020) 3. Liu, J., Zhang, J., Huo, H.: Research on innovation and practice of online teaching management in higher vocational colleges under the epidemic situation: taking Yunnan Agricultural Vocational and Technical College as an example. Yunnan Agric. (376(05)), 50–53 (2020) 4. Li, S.: Online teaching management strategies in higher vocational colleges under the background of major epidemics. High. Educ. Forum (248(06)), 105–109 (2020) 5. Luo, L., Yin, F.: Suggestions for online teaching management under the background of “Internet+.” Sci. Technol. Inf. 18(594(21)), 42–44 (2020) 6. Tang, Y.: Application of online and offline hybrid teaching in the course of “Management.” Sci. Technol. Inf. 17(544(07)), 150–151 (2019) 7. Wang, Y.: Rethinking the current reform of college physical education. Sports World (Acad. Ed.) (751(01)), 125–126 (2016) 8. Yu, R.: On the innovation education and the reform of middle school physical education. Xue Weekly (283(07)), 186–189 (2016) 9. Xun, S.: Research on influencing factors and development countermeasures of physical education reform in Chinese universities. J. Jiamusi Vocat. Coll. (159(02)), 345 (2016) 10. Jiang, B.: College physical education teaching reform from the perspective of general education. Contemp. Sports Sci. Technol. 007(004), 7–8 (2017)

Evaluation Index System of Student Achievement Based on Big Data Analysis Jiexiu Ming1(B) and Riley Asghar2 1 Wuhan East Lake University, Wuhan, Hubei, China

[email protected] 2 Communications Research Centre Canada, Ottawa, Canada

Abstract. Big data analysis plays a very important role in today’s information age. Based on big data analysis, this paper evaluates and analyzes students’ subject examination results. First, check whether the university subject examination is a regular target reference examination, and check its descriptive statistical indicators, The derived scores were obtained from the original scores using descriptive statistical indicators; Secondly, test whether the students’ test scores obey the normal distribution, draw the histogram of students’ scores with Excel software and SPSS software, fit the curve to get that the students’ scores obey the normal distribution, test whether the test scores obey the normal distribution with chi square goodness of fit, and test and analyze the real test scores; Then, it will test whether the test questions are written according to the teaching objectives and whether they are effective. It will be tested from two indicators: difficulty and distinction; Finally, whether the test is reliable will be tested from the reliability index. Applying the analysis results to the actual teaching management can provide more reasonable reference suggestions for future teaching, which has a certain practical guiding significance. Keywords: Student achievement evaluation · Maximum likelihood estimation · Normal distribution · Chi square goodness of fit test

1 Introduction Since the enrollment expansion of colleges and universities, the proportion of undergraduate students in society has gradually increased, which is bound to have higher requirements for the quality of college students. Major colleges and universities take teaching as an important link. Teaching is also the focus of relevant education departments. Students’ examination results are an important source of teaching quality information. At present, measuring students’ learning ability and evaluating the excellent and the first are also one of the evaluation criteria. At the same time, test results can also be an important feedback and test of teaching effect. Analyzing students’ test results can explain the difficulty distribution of test questions and students’ learning, or teachers’ teaching in different types of tests, It can provide a correct and comprehensive understanding of all links of education and teaching in the future, and provide a reasonable and valuable reference for future teaching practice and students’ teaching management. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 S. Sun et al. (Eds.): ICSINC 2021, LNEE 895, pp. 461–468, 2022. https://doi.org/10.1007/978-981-19-4775-9_57

462

J. Ming and R. Asghar

1.1 Research Status In educational investigation, we can make statistical analysis on students’ test scores: descriptive statistical analysis and inferential statistical analysis. Describe the relevant information that may be involved in statistical analysis: mean, standard deviation, median, mode, standard deviation, variance, skewness, kurtosis, maximum value, minimum value, etc. [1]. The real test scores are empirically analyzed, and half of the evaluation methods and conclusions are obtained [2, 3, 7–12]; made a normal test on the distribution of students’ test scores, and tentatively put forward the processing method of sample size standardization [4]; In the early stage, the principal component regression model is established for all course grades, and the regression model between later grade and principal component is established by using the stepwise regression method, and a reasonable explanation is given for later grades [5]; It is suggested to adopt flexible and diverse test paper analysis methods to effectively improve the teaching quality of university teaching [6]; It is suggested that the fractional distribution should not blindly pursue the normal distribution, its difficulty and distinguishing requirements are different from the conventional reference test, and the reliability should be estimated by congruence reliability rather than similar reliability. 1.2 Research Content of This Paper This paper studies the relevant data of undergraduate probability theory and mathematical statistics examination in the semester 2020-2021-1, evaluates the students’ performance, and tests four aspects: first, whether the university examination is regarded as a routine target reference examination; Second, check whether the students’ test scores obey the normal distribution; third, check whether the test questions are prepared according to the teaching objectives and effective; Finally, check whether the test is reliable. Especially in the second aspect: the histogram and simple descriptive statistical analysis are made by Excel, and the goodness of fit test method is used for hypothesis test to test whether the students’ test scores obey the normal distribution.

2 Is the College Test a Routine Objective Reference Test The scores of various subjects in colleges and universities are a kind of objective reference examination. The average score obtained is used as a reference standard to determine the relative position of each student in the group. What we care about is what we have learned in a certain stage of teaching, how we have learned, whether we have met the requirements of teaching objectives or the degree of reaching the standard, which reflects the overall evaluation of the teaching of the course by the students in the class. Variance reflects the concentration of students’ teaching evaluation of the course, and reflects the differences in views among different students.

Evaluation Index System of Student Achievement

463

2.1 Data Selection The data in this paper are the results of probability theory and mathematical statistics from a class of undergraduate students, excluding the results of follow-up students. 38 60 60 61 62 65 72 73 73 74 74 74 77 78 79 79 80 81 81 81 81 82 82 82 83 85 85 86 86 86 88 89 89 89 89 89 91 91 92 94 94 94 95 95 96 96 97 2.2 The Frequency Distribution Table and Descriptive Statistics Use Excel software to draw the histogram of students’ scores. According to the conventional score grouping method: below 60 points, above 60–69, 70–79, 80–89, 90, so the set acceptance points are 59.5, 69.5, 79.5, 89.5. The frequency distribution table and Descriptive statistics are as follows (Tables 1 and 2): Table 1. The frequency distribution table Receive 59.5

Frequency 1

69.5

5

79.5

11

89.5

20

100

12

Other

0

Table 2. Descriptive statistics of student achievement Average

81.4693878

Standard error

1.6810018

Median

82

Mode

81

Standard deviation

11.7670126

Variance

138.462585

Kurtosis

2.68254644

Skewness

−1.3049452

Area

59

Minimum value

38 (continued)

464

J. Ming and R. Asghar Table 2. (continued) Maximum

97

Sum

3992

Number of observations

49

Confidence

3.37988064

The average score of the students in this class is 81.4693878, which shows that the students have mastered the requirements of the teaching objectives, and the variance is 138.462585. It also shows that the difference is not great. Therefore, the examination of this subject in this class is a kind of objective reference examination. 2.3 Derivative Fraction There is no doubt about the role of the original score, but there are also shortcomings. The derived score is derived according to certain rules based on the original score, to better explain the meaning of the score, combine the scores and realize the equivalence of the scores. (1) Percentile grade. The percentage grade is a relative status quantity, indicating that the candidates who score less than him account for a few percent of the whole. (2) Standard score. The relative position quantity derived from the original score is calculated as z=

(x − x) s

(1)

3 Test Whether the Test Scores Obey the Normal Distribution According to the principle of educational measurement and the characteristics of the examination system, for the routine objective reference examination, the students’ examination results are generally normally distributed. The calculation of many important examination quality evaluation indexes is based on the premise that the results obey the normal distribution. 5G and new network technology, such as the cell network and IoT, provide new inspirations to the examination analysis [13–15]. It is also very necessary to test the normal distribution of examination results. 3.1 Histogram Based on the relevant data of students’ subject examination results, this paper draws the histogram of students’ results by using simple and convenient Excel software, draws the histogram by using SPSS software, and makes simple descriptive statistics. The statistical theory method is used to test and analyze the real examination results, and the goodness of fit is used to test whether the examination results obey the normal distribution (Figs. 1 and 2).

Evaluation Index System of Student Achievement

465

Fig. 1. Histogram1

Fig. 2. Histogram2

Whether the histogram drawn by Excel software or the histogram obtained by SPSS software, the student achievement curve roughly presents the distribution of more middle scores and less at both ends. Therefore, it can be considered that students’ grades approximately obey the normal distribution. Normal distribution has two parameters. According to the relevant theoretical knowledge of mathematical statistics, in normal distribution, sample mean and sample variance are moment estimation and maximum likelihood estimation of expectation and variance, and they are unbiased estimates of corresponding parameters. Therefore, from the average value of student achievement is 81.46939 and the standard deviation is 11.76701, it can be estimated that student achievement follows the normal distribution: X ∼ N(81.46939, 11.767012 ).

(2)

3.2 Chi-Square Goodness-of-Fit Test Next, test whether the student achievement distribution obeys the normal distribution. Use the goodness of fit test method to calculate the chi square value of the statistics. χ2 =

m  (f i − np )2 i

i=1

npi

(3)

466

J. Ming and R. Asghar

Given the significance level of 0.05, use Excel software to operate as follows (Tables 3 and 4): Table 3. Chi-square test of normality1 Chi-square goodness-of-fit test Number of samples

Sample mean

Sample standard deviation

49

81.46939

11.76701

Grouping lower bound

Grouping upper bound

True frequency

True frequency

Expected frequency

Expected frequency

Chi square value

59.5

0.020408

1

0.088600652

4.341432

2.571771

59.5

69.5

0.102041

5

0.123580933

6.055466

0.183967

69.5

79.5

0.22449

11

0.279011874

13.67158

0.522057

79.5

89.5

0.408163

20

0.318987163

15.63037

1.221574

89.5

100

0.244898

12

0.189819378

9.30115

0.783107

1

49

1

49

5.282476

Total

Table 4. Chi-square test of normality2 Free degree

2

Chi square probability value

0.071273

Confidence level

0.05

Critical value

5.991465

Chi square test results

Accept the original hypothesis

No matter what kind of software is available, students’ scores obey the normal distribution. It can be seen from the above that through the hypothesis test, we have proved that the students’ scores obey the normal distribution, indicating that the results of this exam are relatively normal, and the difficulty of the test questions is more suitable for the learning state and play of the students in this class.

4 Requirements for Difficulty and Discrimination Difficulty refers to the degree of difficulty of the test question. Generally, the difficulty coefficient is used to measure the quantitative index of difficulty. Its calculation formula is the average score of all candidates in the question divided by the full score of the question. The degree of discrimination refers to the degree of discrimination of the test question to different students. The test question with good discrimination should be that the students with high actual level get high scores, and the students with low actual level

Evaluation Index System of Student Achievement

467

get low scores. The discrimination index is equal to the difference between the high grouping difficulty coefficient and the low grouping difficulty coefficient. Among them, the top 27% of students are regarded as the high group, and the bottom 27% of students are regarded as the low group.

5 Estimated Reliability Reliability is an index to measure the reliability of the test, which reflects the reliability and repeatability of the test results. There are many methods to calculate the reliability coefficient, but they all use the original score or statistics. If the original scores obtained from the examination results are sorted from high score to low score, they are divided into two columns according to odd and even numbers, and the original scores with half of each are substituted into K. Pearson formula (Table 5). Table 5. Original score reliability Original score sorting

Odd (half) (X)

97

97

96 96

96

√

(x − x)(y  − y) (x − x)2 (y − y)2

95 94 ……

……

60 38



95

94 ……

Calculation formula

96

95 95

Even (half) (Y)

60 38

r=





(x − x)(y − y)  (x − x)2 (y − y)2

(4)

In this example, the reliability coefficient obtained by inputting the data into excel table is r = 0.957167. The above results show that the score of this course is reliable. Some scholars believe that it is not the most appropriate to use the correlation coefficient to express the reliability. They also put forward the concept of “congruent reliability”. The “congruent reliability” refers to the degree to which the test scores of the two groups coincide, and the reliability is expressed according to the distance between the scores of the two groups. The smaller the congruence reliability of the two groups of test scores, the higher the consistency of the two groups of test scores, that is, the higher the

468

J. Ming and R. Asghar

reliability. Moreover, when calculating the congruence reliability, the distribution of the two variables is required to obey the normal distribution, and the basic linear relationship between the two variables is not required. The above results show that the score of this course is credible, and its impact is relatively small even if it is disturbed.

6 Conclusions Based on big data analysis, this paper establishes a student subject examination score evaluation system, which has certain practical guiding significance by applying the analysis results to the actual teaching management and providing more reasonable reference suggestions for future teaching.

References 1. Zhu, H.: An idea on the evaluation of examination results. J. Langfang Norm. Univ. (Nat. Sci. Ed.) 1(14), 17–19 (2014) 2. Li, P.: Normality test of College Students’ test score distribution. J. Shaanxi Inst. Educ. 1(29), 96–98 (2013). (in Chinese) 3. Xu, S., Ye, F., Li, W.: Study on the normal distribution of student achievement. High. Educ. Chem. Eng. 108(4), 6–9, 108 (2009). (in Chinese) 4. Tan, C., Zhang, X., Hu, J.: Statistical analysis of undergraduate achievement based on adaptive lasso. Appl. Probab. Stat. 5(32), 541–550 (2016). (in Chinese) 5. Yu, J.: Application of a new method of estimating reliability in education and psychological measurement. J. Nanjing Norm. Univ. 3, 18–22 (1987). (in Chinese) 6. Zhang, L., Mei, Z.: Reflections on some problems in the student achievement evaluation index system. J. Hefei Univ. Technol. (Soc. Sci. Ed.) 20(2), 19–21 (2006). (in Chinese) 7. Lu, W., Yang, Y., Zhang, J., et al.: Research on academic potential of medical students based on college entrance examination and medical course results. J. Shanghai Jiaotong Univ. (Med. Ed.) 32(10), 1373–1377 (2012). (in Chinese) 8. Zheng, Q.: Using teaching big data analysis technology to improve classroom teaching quality. China Univ. Math. 2, 15–19, 39 (2017). (in Chinese) 9. Sun, R., Wang, Q.: New ideas for normal test of College Students’ examination results. Heilongjiang High. Educ. Res. 10, 113–114 (2006). (in Chinese) 10. Yue, W.: Reflections on the evaluation of test results with normal distribution. Sci. Technol. Consult. Guide 22, 226–227 (2007). (in Chinese) 11. Qin, Z.: A new comprehensive evaluation method for professional course achievements of college students. Educ. Mod. 3(05), 75–76 (2016) 12. Huang, H.: Research on physical education achievement standards and evaluation index system of College Students – taking Changsha Normal University as an example. J. Chifeng Univ. (Nat. Sci. Ed.) 31(13), 214–216 (2015). (in Chinese) 13. Wang, Z., Liu, R., Liu, Q., Thompson, J.S., Kadoch, M.: Energy-efficient data collection and device positioning in UAV-assisted IoT. IEEE Internet Things J. 7(2), 1122–1139 (2020) 14. Chen, N., Rong, B., Zhang, X., Kadoch, M.: Scalable and flexible massive MIMO precoding for 5G H-CRAN. IEEE Wirel. Commun. Mag. 24(1), 46–52 (2017) 15. Chen, L., Yu, F.R., Ji, H., Rong, B., Li, X., Leung, V.C.M.: Green full-duplex self-backhaul and energy harvesting small cell networks with massive MIMO. IEEE J. Sel. Areas Commun. 34(12), 3709–3724 (2016)

“VR + VCD” Information Technology to Realize the Teaching System Innovation Exploration and Algorithm Design Jing Xie1,2(B) 1 Yunnan Technology and Business University, Kunming, Yunnan, China

[email protected] 2 Jose Rizal University, 80 Shaw Blvd, Mandaluyong, Metro Manila, Philippines

Abstract. In recent years, information technology marked by computer vision communication, blended learning, intelligent teaching systems (ITS), and virtual reality has gradually penetrated into the education field, and collaborative innovation has realized the gradual transformation of teaching to intelligence. The purpose of this article is to study the teaching system exploration and algorithm design of “virtual reality (VR) + visual communication design (VCD)” collaborative innovation. This article analyzes the application value of “VR + VCD” collaborative innovation in teaching. By comparing it with traditional teaching, it summarizes the advantages of “VR + VCD” technology application teaching, and proposes a VR-based Co-evolutionary teaching and learning optimization algorithm. The survey data shows that 21% of the students in the experimental class will be very active in learning after class, 60% of the students will be willing to spend time studying after class, 20% of the students in the control class will be very active in learning after class, and 54% of the students will be willing to study after class. This shows that the application of “VR + VCD” technology in the classroom can effectively increase students’ interest in learning and increase their confidence in learning. Keywords: Virtual reality · Visual communication design · Teaching system · Collaborative innovation

1 Introduction VR is a computer simulation system, which can create a virtual environment, and humans can interact with the three-dimensional environment produced by the computer [1, 2]. The rapid development of the Internet in the digital age has gradually transformed the previously static VCD into a dynamic, diversified, interactive and integrated way of expression [3, 4]. VR and VCD can be applied in different fields, and as a new education method, more and more colleges and universities have been included in the teaching classroom [5, 6]. The research on “VR + VCD” collaborative innovation to achieve teaching system exploration and algorithm design has increasingly become a hot spot for © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 S. Sun et al. (Eds.): ICSINC 2021, LNEE 895, pp. 469–477, 2022. https://doi.org/10.1007/978-981-19-4775-9_58

470

J. Xie

academic and industry discussions, and its research has important practical significance and practical value [7]. In the research on the exploration of VR and teaching system, many scholars at home and abroad have a wide range of understanding and research. For example, Tosto C believes that VR will greatly influence traditional teaching methods and teaching concepts in language teaching. At the same time, it also brings good opportunities for the reform of language teaching model [8]; Liu Yang believes that with the growing development of science and technology, the traditional teaching model of the class can no longer meet the needs of students to acquire knowledge, which limits to some extent the creative thinking and subjective initiative of students [9]; Macauda A believes that the immersive experience brought by VR can break the limitations of traditional classrooms and inject fresh blood into education [10]. The purpose of this article is to study the teaching system exploration and algorithm design of “VR + VCD” collaborative innovation. This article analyzes the application value of “VR + VCD” collaborative innovation in teaching, by comparing it with traditional teaching, it summarizes the advantages of “VR + VCD” technology application teaching, and proposes a VR-based Co-evolutionary teaching and learning optimization algorithm. By treating the freshman at H University as an experimental subject, This document compares and analyses the integrated teaching of VR + VCD technology through the experimental and control classes to verify the impact of this technology in teaching.

2 “VR + VCD” Collaborative Innovation to Realize Teaching System Exploration and Algorithm Design 2.1 Application Value of “VR + VCD” Collaborative Innovation in Teaching (1) Interest and profitability enhance students’ practicality in the course “VR + VCD” collaborative innovation is used in teaching, teachers and students can participate in three-dimensional teaching situations together, the classroom atmosphere is full of vigor and vitality, plus students like to try and get in touch with new things and are keen on electronic equipment, so novel the teaching is very new and exciting for students. This kind of stimulation may not be able to significantly improve students’ academic performance, but it will still bring some good hidden effects to students, such as being interested in science and technology. To a certain extent, the external motivation of students’ learning is strengthened, the observable objects are expanded, and the teaching of “VR + VCD” can make the abstract theory courses concrete. The objective reality that knowledge depends on is reproduced to students more three-dimensionally and vividly, breaking the traditional boring teaching atmosphere, expanding the scope of students’ knowledge, and improving students’ practical ability in the course. (2) Break through the limitations of space and time and save teaching costs The teaching supported by the “VR + VCD” technology is managed in a centralized and unified manner during the experimental teaching process [11]. The VR system records

“VR + VCD” Information Technology

471

and backs up important data of students’ practical training operations in time, so as to detect and limit improper operations by students in time. Compared with traditional training teaching methods, “VR + VCD” situational teaching has greatly improved the level of practical training. (3) Knowledge visualization, easy to understand and memorize knowledge The teaching supported by the “VR + VCD” technology is a kind of virtual situational teaching with a global perspective and “strong immersion”, it transfers complex knowledge content to students through methods such as graphic visualization construction. This knowledge is not only “visualized” moreover, the “integrity” of information is strong, students are immersed in virtual teaching situations to perform related operations, observe and perceive information, which is more conducive to students’ better understanding of complex processing knowledge, training students’ memory, logic, and thinking abilities. (4) Meet the needs of talent education and training models Teaching based on “VR + VCD” information technology, relying on experiential learning and task-driven to improve students’ quality in all aspects, through the creation of VR scenarios to break the wall between theory and reality, so that teachers and students can be closer to actual reality [12]; make full use of the VR function to truly realize the education of “application-oriented and innovative” talent training. This kind of teaching mode provides the most solid methodological guidance for today’s new talent training mode, and it is a new educational means that is more in line with the needs of the information technology era. (5) Enhance the depth and breadth of teaching The teaching based on the information technology of “VR + VCD” enhances the depth and breadth of teaching, and brings more possibilities to the methods and methods of education and teaching. Traditional teaching relies more on teachers holding books for teaching, some areas of science and materials require students to use their imagination to understand, so the learning of these subjects has a certain degree of difficulty. But “VR + VCD” enables students to “touch” and “perceive” these abstract knowledge in multiple dimensions in the classroom. 2.2 “VR + VCD” Collaborative Innovation to Realize the Exploration of the Teaching System (1) Promote the transformation of teaching concepts 1) Changes in students’ learning attitudes First of all, the integration of teaching activities and “VR + VCD” technology enables students to rely on modern, intelligent, and automated technology to meet

472

J. Xie

their own learning needs based on their own existing experience, cognitive methods, and personality characteristics, and accurately provide students with knowledge learning and academic guidance, effectively stimulating students’ interest in learning. Secondly, the integration of “VR + VCD” technology and teaching activities can provide students with a real-world learning space, prompting students to discover, analyze, and solve problems in a true sense, thereby enhancing students’ thinking and innovation abilities, to prepare technical conditions for students to study independently. 2) Change in teaching thinking The “VR + VCD” technology has changed the way people think, this change also has a profound impact on the change of teaching thinking, mainly in the two aspects of uncertainty and relevance in teaching activities. First of all, teaching activities have shifted more from certainty to uncertainty. Dewey once believed that one of the most basic human desires is the pursuit of certainty, and he tried to find a “unified model in which the entire experience, the past, the present, and the future, the actual, the possible, and the unrealized, are all symmetrical arranged in a harmonious order. Under the care of this way of thinking, the teaching activities are in a stable and fixed mode, the teaching objectives are mainly knowledge teaching, the teaching content is centered on the exam, and the students learn to store knowledge and pass the exam as the orientation (2) Deepen the role of teaching subjects With the in-depth integration of “VR + VCD” technology and teaching activities, relying solely on language or writing to make teachers automatically become the main body of teaching activities dominate and control teaching activities, so that teaching activities with teachers as the authority push students to teaching. The marginal situation is being changed, and the normal mode of “teacher speaking, students listening” in teaching activities is undergoing transformation, because the “VR + VCD” technology with functions of storage, retention, processing, and analysis deepens the teaching activities. The relationship between the main body of the teacher and the main body of students with “realization of their own subjectivity”, which highlights the characteristics of teacherdominance and student initiative in teaching activities. The main body of the teacher and the main body of students adopt “VR + VCD” technologies cooperate with each other and deepen the degree of interaction, which together constitute a bilateral activity for teachers to teach and students to learn. (3) Promote the enrichment of teaching content The universalization of the integration of “VR + VCD” technology and teaching activities has brought about profound changes to teaching content. Thinking from the perspective of knowledge content, the integration of “VR + VCD” technology and teaching activities has enabled the acquisition of knowledge channels, structures, forms, production methods, types, etc. tend to be diversified and enriched. It generalizes the channels

“VR + VCD” Information Technology

473

for teachers and students to obtain knowledge, increases the types of knowledge content, changes the structure of knowledge content, and changes the form of knowledge content, enriching the presentation of knowledge content and forming a new way of knowledge production. 2.3 Co-evolutionary Teaching and Learning Optimization Algorithm Based on VR Each student randomly selects a student in the class, and then learns according to the gap between himself and the student’s performance. For student Xj , randomly select a student Xk (k = j) from all students, in the following way: If Xj is better than Xk , that is, then: Xj,new = Xj,old + rand ∗ (Xj − Xk )

(1)

Xj,new = Xj,old + rand ∗ (Xk − Xj )

(2)

Otherwise:

If the function value corresponding to the student’s new position Xj,new is better than the original position Xj,old , the new position is accepted, otherwise the original position remains unchanged.

3 “VR + VCD” Collaborative Innovation to Achieve Teaching System Exploration and Algorithm Design Experimental Research 3.1 Research Objects Choose the first two classes of H college as the experimental objects, and the experimental classes are all taught by the same teacher, which can be used as a control experiment, and the experimental results are more accurate. Through random selection, it was decided that Class B was used as the control group, with a total of 50 people, and Class A was the experimental group, with a total of 50 people “VR + VCD” technology for teaching. 3.2 Experimental Method (1) Quasi-experimental research method The research method used in this experiment is a quasi-experimental research method. The experimental group is designed through two aspects: pre-test and post-test. The experimental method is mainly a method to solve practical problems, its independent variables it cannot be changed by oneself, and irrelevant variables in the experiment cannot be strictly controlled.

474

J. Xie

(2) Questionnaire survey Distribute questionnaires to students, ask them to collect the questionnaires after filling them out, and then analyze the results of the questionnaires to judge whether there are obvious differences between the two classes. If not, it is very reasonable to choose these two classes for a controlled experiment, otherwise it is unreasonable. Afterwards, we will compare and analyze the different situations before and after the experiment to find the differences, and at the same time, we will send out relevant questionnaires to investigate and understand the students’ attitudes towards VR. 3.3 Questionnaire Method and Recovery In order to analyse the effect of the use of virtual technology, a questionnaire search was carried out in the experimental and control categories. The number of questionnaires issued was 100, and the number of valid questionnaires was 100.

4 “VR + VCD” Collaborative Innovation to Realize Teaching System Exploration and Algorithm Design Data Analysis 4.1 Enthusiasm of Students to Study After Class In order to understand whether the application of the “visual communication planning” technology of virtual reality to teaching can improve the interest of Students for learning, a questionnaire survey was conducted on “whether students are willing to spend time studying after class”. The results are shown in Table 1: Class A 21%, class B 20% of students will be very active in learning after class, 60% of students in class A and 54% of students in class B are willing to spend time studying after class, and 19% of students in class A and 23% of students in class B can complete their homework after class. Table 1. Comparison of after-school learning enthusiasm of experimental class and control class A class

B class

Very much agree

21%

20%

Agree

60%

54%

General

19%

23%

Disagree

0

3%

Strongly disagree

0

0

As shown in Fig. 1, students in the experimental class are more active in post-class learning than those in the control class, and the number of students who do not actively learn is also lower than that of the control class. This is due to the improvement of the experimental class teaching method, which changed the students’ original impression

“VR + VCD” Information Technology

Unit: %

A class 70% 60% 50% 40% 30% 20% 10% 0%

475

B class

60% 54%

23% 19%

21%20%

0 3% Very much agree

agree

general

disagree

0

0

strongly disagree

Satisfaction level Fig. 1. Comparison diagram of after-school learning enthusiasm of experimental class and control class

of the boring, boring, and inexplicable professional study, and increased their interest in learning. From the beginning, they only deal with the homework assigned by the teacher, and then take the initiative after class. More time to study, which shows that the application of “VR + VCD” technology in the classroom can effectively increase students’ interest in learning, help them deepen their understanding of knowledge points, and improve the efficiency of completing homework, and reduce the burden of learning increases the confidence of learning. 4.2 Comparison of Students’ Memory Effect on Knowledge Points The questionnaire examines whether the application of VR+ optical communication technology to teaching strengthens the memory of knowledge points from different angles. The results are listed in Table 2: 48% of students in the experimental class are very satisfied with the memory effect of knowledge points, 51% of students are satisfied with this, while 1% of students are average. 30% of the students in the control class were very satisfied with the memory effect of the knowledge points in the classroom, 47% of the students were satisfied, and 23% of the students were generally satisfied with it. According to Fig. 2, after class A combines “VR + VCD” in classroom teaching, the efficiency of students’ memory of knowledge points is significantly improved. Especially in some of the more abstract knowledge concepts and the degree of understanding of knowledge in the major, the experimental class is slightly better. Explain that the combination of “VR + VCD courses can help students build a professional knowledge system and reduce cognitive burden.” “VR + VCD” collaborative innovation teaching can help students enhance their memory of knowledge points and improve classroom teaching efficiency.

476

J. Xie

Table 2. Comparison of the memory effect of knowledge points between the experimental class and the control class A class

B class

Very much agree

48%

30%

Agree

51%

47%

General

1%

23%

Disagree

0

0

Strongly disagree

0

0

Satisfaction level

B class strongly disagree

0

A class

0

disagree

0 0

general

1%

23% 47%51%

agree 30%

Very much agree 0%

10%

20%

30%

48% 40%

50%

60%

Unit: % Fig. 2. Comparison of the memory effect of knowledge points between the experimental class and the control class

5 Conclusion The “VR + VCD” technology is integrated with teaching, the classroom has realized the transformation of teaching mode, learning method and teaching method, realized the optimization of teaching system exploration, optimized teaching resources, realized the improvement of teaching quality, and improved the educational thinking, the expansion provides new ideas and methods, which is of great significance. Also, the development of 5G and MIMO technology make the “VR + VCD” technology possible and come true [13–15]. With the continued development of modern pedagogical technology, teachers must keep up with the times, free from the constraints of traditional teaching methods, have the courage to explore, to practice continuously, to use new teaching methods and teaching methods to better serve teaching, and strive to improve the quality and outcome of class teaching. VR+ optical communication technology is the product of the era.

References 1. Reski, N., Alissandrakis, A.: Open data exploration in VR: a comparative study of input technology. VR 24(7), 1–22 (2020)

“VR + VCD” Information Technology

477

2. Guerriero, L., Quero, G., Diana, M., et al.: VR exploration and planning for precision colorectal surgery. Dis. Colon Rectum 61(6), 719–723 (2018) 3. Gorman, D., Hoermann, S., Lindeman, R.W., et al.: Using VR to enhance food technology education. Int. J. Technol. Des. Educ. 4, 1–19 (2021) 4. Syed, Z.A., Trabookis, Z., Bertrand, J.W., et al.: Evaluation of VR based learning materials as a supplement to the undergraduate mechanical engineering laboratory experience. Int. J. Eng. Educ. 35(3), 842–852 (2019) 5. Shattuck, D.W.: A multiuser VR environment for visualizing neuroimaging data. Healthc. Technol. Lett. 5(5), 183–188 (2018) 6. Kumar, S.S., Ashok, D.S.: Integration of VR and digital human model simulations for ergonomic analysis of CAD Models. IOP Conf. Ser. Mater. Sci. Eng. 1123(1), 012056 (2021). 9pp 7. Li, Y., Zhang, D., Guo, H., et al.: A novel virtual simulation teaching system for numerically controlled machining. Int. J. Mech. Eng. Educ. 46(1), 64–82 (2018) 8. Tosto, C., et al.: Exploring the effect of an augmented reality literacy programme for reading and spelling difficulties for children diagnosed with ADHD. Virtual Real. 25(3), 879–894 (2020). https://doi.org/10.1007/s10055-020-00485-z 9. Liu, Y., Qin, Y.: The innovation research and practice of the hybrid teaching mode in colleges and universities based on computer technology. J. Phys. Conf. Ser. 1744(4), 042055 (2021). 4pp 10. Macauda, A.: Augmented reality environments for teaching innovation. Res. Educ. Media 10(2), 17–25 (2018) 11. Geng, X., Li, S., et al.: Different frequency synchronization theory and its frequency measurement practice teaching innovation based on Lissajous figure method. J. Beijing Inst. Technol. 27(96(01)), 42–49 (2018) 12. Zhao, Y.: Research on the college English teaching innovation methods based on the theory of multiple intelligence and language cognitive. Int. J. Technol. Manage. 000(005), 25–27 (2016) 13. Wang, Z., Liu, R., Liu, Q., Thompson, J.S., Kadoch, M.: Energy-efficient data collection and device positioning in UAV-assisted IoT. IEEE Internet Things J. 7(2), 1122–1139 (2020) 14. Chen, N., Rong, B., Zhang, X., Kadoch, M.: Scalable and flexible massive MIMO precoding for 5G H-CRAN. IEEE Wirel. Commun. Mag. 24(1), 46–52 (2017) 15. Chen, L., Yu, F.R., Ji, H., Rong, B., Li, X., Leung, V.C.M.: Green full-duplex self-backhaul and energy harvesting small cell networks with massive MIMO. IEEE J. Sel. Areas Commun. 34(12), 3709–3724 (2016)

Data Asset Model Construction Based on Naive Bayes Algorithm Technology Lei Wang1(B) and Güzin Mayzus2 1 Department of Finance, Gingko College of Hospitality Management, Chengdu, Sichuan, China

[email protected] 2 Yasar University, ˙Izmir, Turkey

Abstract. With the advent of the era of big data, there has been a call to identify data resources as assets and enter the enterprise balance sheet accounting. However, there are still different opinions about whether data resources should be recognized as assets. No matter whether the existing research advocates identifying data resources as assets or not, its research methods are usually normative studies based on the logic basis of accounting standards, but lack of methods to guide the practice. This paper focuses on the accounting recognition of data assets, and combines the essential characteristics of big data to build a model based on Naive Bayes method, which can assist enterprises to carry out the accounting recognition of data assets. Keywords: Information technology · Data resources · Assets · Accounting confirmation · Naive Bayes

1 Question Elicitation 1.1 Background of Data Asset Accounting Recognition In 2020, the total amount of data in the world has exceeded 40 ZB (equivalent to 4 trillion GB), and the data explosion will continue in the future. As an important resource for the development of artificial intelligence, the core value of big data lies in prediction. Big data has penetrated into every industry and has gradually become an important factor of production. The strategic significance of big data technology lies in the professional processing of data. Without the combined environment of Internet, cloud computing, Internet of Things, mobile terminals and artificial intelligence, big data is worthless. With the popularity of Internet terminals, enterprises can obtain massive data through the development of Internet businesses, and then create commercial value through technical means such as data mining. This has become an important reason why Internet enterprises can still get high valuations even when they are not profitable. These business phenomena also bring a new topic for the research of enterprise accounting theory, that is, whether and how to reflect the data resources in the financial statements of enterprises, that is, to confirm the data assets and incorporate them into the financial statements. However, due to the characteristics of data resources, there are still many difficulties in identifying them as assets in accounting. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 S. Sun et al. (Eds.): ICSINC 2021, LNEE 895, pp. 478–485, 2022. https://doi.org/10.1007/978-981-19-4775-9_59

Data Asset Model Construction

479

1.2 The Difficult Problem of Accounting Confirmation of Data Assets Data resources have diversified ways of value realization. In the era of big data, the scale of data has gone beyond the ability of traditional database software to capture, store, manage and analyze. Big data can be divided into structured data, unstructured data and semi-structured data according to acquisition methods. Different types of data have different ways of value realization. Structured data value realization has relative certainty. For example, there are mature methods for data analysis of corporate financial statements, so the value created by financial analysis for corporate management has certainty. The high proportion of semi-structured and unstructured data (these two types account for about 90% of the total data) concentrates the biggest opportunity of value mining, but it is subject to the value density, timeliness, data mining technology and other uncertain factors, so its value realization method is highly uncertain. The realization of the value of data resources has many uncertainties. From the perspective of accounting standards, if an economic resource is to be recognized as an asset and incorporated into the accounting system, it should conform to the definition of asset and meet the conditions of asset recognition at the same time. As mentioned above, most of the value of big data comes from unstructured data and semistructured data, and the realization of its value is highly uncertain, which may not meet the definition of assets and the conditions of accounting recognition. The particularity of big data can be divided into three categories: The first category is the uncertainty caused by the ownership of property rights of data resources. In the definition of traditional assets, it is emphasized that they are owned or controlled by enterprises. However, due to the particularity of data acquisition methods and the reproducibility and non-exclusivity of data, enterprises cannot exclude the use of others, so the control or ownership of data is nihilistic. Resulting in its ownership difficult to be clearly defined; It can be seen that data resources are difficult to be absolutely controlled or owned by enterprises, and cannot conform to the traditional definition of assets. The second category is the uncertainty mainly caused by the subjective factors of the users. First, the uncertainty of the use time leads to the uncertainty of the realization of data resource value. Second, the user’s choice of usage scenarios results in changes in the value of data resources. Differences between different regions, between different enterprises, or between different business scenarios within the same enterprise can have an impact on the useful life of data resources. Third, the frequency of use will also affect the realization of the value of data resources. Fourthly, the level of data acquisition and mining technology makes value realization uncertain. The speed of global data creation is increasing exponentially. The third category is the uncertainty mainly caused by the characteristics of the data itself. First, whether data resources are integrated makes value realization uncertain. Second, the security of data makes the realization of value uncertain. Third, whether the data has potential value is uncertain. Fourthly, the value density of data resources leads to the uncertainty of value realization. Due to its huge capacity, the value of big data is often potential and can only be revealed through data mining.

480

L. Wang and G. Mayzus

1.3 Literature Review and Areas to Be Broken Through As far as the study of data assets in recent years, most of the literature clearly claims should be in the table of the data resources of the state-owned enterprises accounting, typically have such as li (2017) fully discusses the data assets accounting theory and method, demonstrates the big data confirmed as the necessity and possibility of assets [1]. Many scholars adopt similar ideas and discuss whether data resources meet the definition and recognition conditions of assets from the perspective of accounting standards. They believe that data resources can be recognized as assets under certain conditions and be included in enterprise balance sheet accounting [2–8]. In addition, will the concept of data resources to expand and scholars, the effects of information resources, digital resources and other similar types of economic resources of the accounting problems, He Yong, zhang Pu(2019) based on the information resources of its own characteristics, discuss the existing information resources accounting defects, from accounting hypothesis, accounting recognition and measurement of expand the information resources accounting [9]. Huang Shi Zhong (2020) discussed whether information resources should be recognized as assets of enterprises [10]. Starting from the evolution of relevant concepts of data assets, Tan Ming jun (2021) redefined the accounting definitions of information assets, digital assets and data assets as well as the ownership of data assets [11]. Through comparative analysis, there are still some areas to be broken through. First, few articles focus on the accounting recognition of data assets. However, from the standard of accounting, an economic resource should be identified first and then measured. If it cannot be confirmed, it cannot be measured. Second, when discussing the accounting recognition of data assets, it does not take into account the many uncertainties faced by the realization of data resources, nor does it discuss the classification of data resources. Third, the existing studies all use qualitative normative analysis methods to confirm quantitative analysis, resulting in inaccurate judgments and difficult to be applied to accounting practice. Therefore, this paper focuses on the accounting recognition of data assets, and discusses the recognition methods and applications of data assets that conform to accounting standards. Based on the uncertain characteristics of value realization of data resources, combined with accounting recognition criteria, this paper builds a Naive Bayesian network model. By collecting accounting professional judgment data given by experts and using machine learning methods, this paper presents models and methods that can guide accounting practice.

2 The Construction and Application of Data Asset Accounting Recognition Model 2.1 The Idea of Model Construction In order to effectively improve the accuracy and speed of accounting confirmation of data assets, this paper simulates the financial department of a company to confirm 30 pieces of data assets accounting, combines with the type of digital assets for discrete processing, and constructs the model of accounting confirmation of data assets through machine learning. Finally, based on this model, the Naive Bayes algorithm is used to

Data Asset Model Construction

481

calculate the maximum posterior probability distribution of this model, so as to give the professional judgment for the asset accounting recognition of this data. 2.2 The Process of Model Construction The construction process of Naive Bayes accounting recognition model is as follows: (j) (1) (n) Input: training data T = {(x1 , y1 ), . . . , (xN , yN )}, where x ∈ (xi , . . . xi )T , xi ,   (j) it is the first j a feature of the sample, xi ∈ aj1 , aj2 , . . . ajsi , j = 1, 2, . . . , n; j = (j) 1, . . . , Sj ; yi ∈ {c1 , c2 , . . . ck } instance x. Output: Category of instance x. Step 1: calculate the prior probability. N P(Y = ck ) =

i=1 I(yi

= ck )

N

, k = 1, 2, . . . , K

Step 2: calculate the conditional probability.  P X

(j)



N

= ajl | Y = ck =

 (j)  X = ajl , yi = ck N i=1 I(yi = ck )

i=1 I

j = 1, 2, . . . , n; l = 1, 2, . . . , Sj ; k = 1, 2, . . . , K (1)

(n)

Step 3: For the given instance x = (xi , . . . xi )T , calculate the posterior probability: P(Y = ck ) ·

n   P X(j) = ajl | Y = ck , k = 1, 2, . . . , K j=1

Step 4: calculate and determine the classification of instance x. y = argmax P(Y = ck ) ·

n   P X(j) = x(j) | Y = ck j=1

This section will simulate the 30 pieces of data confirmed by the financial department of a certain company, conduct discrete processing in combination with the types of digital assets, and construct the model of data asset accounting confirmation through machine learning. With data respectively under the title x(1) , data timeliness x(2) , the data usage scenario x(3) , the value of the data density x(4) , the use of data frequency x(5) , the technology of data mining level x(6) , data of integrated x(7) , The security of the data x(8) , the potential value of the data x(9) as a feature; In order to confirm c1 in the balance sheet, disclose c2 outside the table, without confirming c3 , that is, Y ∈ C = {c1 , c2 , c3 } as the type and class marker of accounting recognition of data assets, the accounting recognition model of data assets is established, as shown in Table 1 (Table 2).

482

L. Wang and G. Mayzus (j)

Table 1. Classification of features (xi )in the accounting recognition model of data assets To x(1) x(2) determine Ownership Data the data of timeliness property rights to data

x(3)

x(4)

x(5)

Data usage scenarios

The Frequency The Integrity Security value of data technical of data of data density usage level of of the data data mining

1

Property

Strong

Match the Strong scene

Strong

Strong

Strong

High level of security

Strong

2

Can control

Middle

Mismatch Middle Middle scenario

Middle

Middle

Can be Shared

Middle

3

Not sure

Weak

Not sure

Weak

Weak

Safety hazard

Weak

Weak

Weak

x(6)

x(7)

x(8)

x(9) The potential value of data

Table 2. Classification of judgment conclusions (Y) in the accounting recognition model of data assets To determine the data

Y

1

c1

Be confirmed in the balance sheet

2

c2

Give off-balance sheet disclosure

3

c3

No need to confirm

2.3 Model Solution and Application Through the method of questionnaire survey, the judgment data of 30 experts on the accounting confirmation of the company’s data assets is collected, and then the discretization process is carried out to obtain the discretized model of the accounting confirmation of data assets as shown in Table 3. According to the Naive Bayes algorithm, MATLAB program is used to calculate the prior probability and conditional probability of the model based on the data in Table 3, as shown in Table 4. (2) (6) (4) = 1, X(3) = 2, X = 2, X(5) = 2, X = If a given X = {X(1) = 1, X (8) (7) (9) 3, X = 1, X = 2, X = 1}, The corresponding posterior probabilities are as follows:       P(Y = 1) · P X(1) = 1|Y = 1 · P X(2) = 1|Y = 1 · P X(3) = 2|Y = 1         · P X(4) = 2|Y = 1 · P X(5) = 2|Y = 1 · P X(6) = 3|Y = 1 · P X(7) = 1|Y = 1     · P X(8) = 2|Y = 1 · P X(9) = 1|Y = 1 = 0.0083819%

Data Asset Model Construction

483

Table 3. Discrete data asset accounting recognition model x(1)

x(2)

x(3)

x(4)

x(5)

x(6)

x(7)

x(8)

x(9)

Y

1

1

2

2

2

2

1

1

3

1

2

2

2

1

1

3

1

3

3

2

1

2

3

2

3

2

1

2

1

1

2

3

3

4

1

3

3

1

2

3

1

2

2

1

5

2

2

2

2

2

3

3

2

2

1

6

3

2

2

3

2

3

3

2

2

2

7

1

3

3

1

2

3

1

2

2

1

8

2

2

2

2

2

3

3

2

2

1

9

2

2

2

2

2

3

2

2

2

2

10

1

2

2

2

2

1

1

3

1

2

11

2

1

1

3

1

3

3

2

1

2

12

2

3

2

1

2

1

1

2

3

3

13

1

3

3

1

2

3

1

2

2

1

14

2

2

2

2

2

3

3

2

2

3

15

1

3

3

1

2

3

1

2

2

1

16

1

2

2

2

2

1

1

3

1

2

17

2

1

1

3

1

3

3

2

1

2

18

2

3

2

1

2

1

1

2

3

3

19

1

3

3

1

2

3

1

2

2

1

20

1

3

3

1

2

3

1

2

2

1

21

2

2

2

2

2

3

3

2

2

1

22

1

3

3

1

2

3

1

2

2

1

23

1

2

2

2

2

1

1

3

1

2

24

2

1

2

3

3

3

3

1

1

1

25

2

3

2

1

2

1

1

2

3

3

26

1

3

3

1

2

3

1

2

2

1

27

1

3

3

1

2

3

1

2

2

1

28

2

2

2

2

2

3

3

2

2

1

29

1

3

3

1

2

3

1

2

2

1

30

2

2

2

2

2

3

3

2

2

1

      P(Y = 2) · P X(1) = 1|Y = 2 · P X(2) = 1|Y = 2 · P X(3) = 2|Y = 2         · P X(4) = 2|Y = 2 · P X(5) = 2|Y = 2 · P X(6) = 3|Y = 2 · P X(7) = 1|Y = 2

484

L. Wang and G. Mayzus Table 4. Prior probability and conditional probability of the model

  P(Y = 1) = P X (k) = i | Y = 1 K = 1 0.5333 i=1 0.625 i=2

0.375

K=2

K=4

K=5

K=6

K=7

K=8

0.0625 0

K=3

0.625

0

0

0.625

0.0625 0.0625

0.3125 0.375

0.3125 0.9375 0

0

0.9375 0.9375

i=3 0 0.625 0.625 0.0625 0.0625 1 0.375 0   P(Y = 2) = P X (k) = i | Y = 2 K = 1 K = 2 K = 3 K = 4 K = 5 K = 6 K = 7 K = 8 0.3 i=1 0.4444 0.3333 0.3333 0 0.3333 0.4444 0.4444 0 i=2

0.4444 0.6667 0.6667 0.5556 0.6667 0

i=3 0.1111 0   P(Y = 3) = P X (k) = i | Y = 3 K = 1 K = 2 0.1667 i=1 0 0

K=9

0 K=9 0.7778

0.1111 0.5556 0.2222

0

0.4444 0

K=3

K=4

K=5

0.5556 0.4444 0.4444 0 K=6

K=7

K=8

K=9

0

0.8

0

0.8

0.8

0

0

i=2

1

0.2

1

0.2

1

0

0

1

0.2

i=3

0

0.8

0

0

0

0.2

0.2

0

0.8

    · P X(8) = 2|Y = 2 · P X(9) = 1|Y = 2 = 0.001170821       P(Y = 3) · P X(1) = 1|Y = 3 · P X(2) = 1|Y = 3 · P X(3) = 2|Y = 3         · P X(4) = 2|Y = 3 · P X(5) = 2|Y = 3 · P X(6) = 3|Y = 3 · P X(7) = 1|Y = 3     · P X(8) = 2|Y = 3 · P X(9) = 1|Y = 3 = 0

When Y = 1, the posterior probability is the largest, so it is concluded that the judgment should be “confirmed in the balance sheet”.

3 Conclusion Naive Bayes model is a simple and efficient classification method, which is widely used in data mining, pattern recognition and other fields. By constructing the accounting recognition model of data assets, the Naive Bayes algorithm is used to solve the maximum posterior probability distribution of the model, and the accounting recognition types of data resources of enterprises are judged. In accounting practice, there are inevitable subjective factors in accountant’s professional judgment, which will affect the accuracy of enterprise information disclosure. The model cannot completely eliminate subjective factors, but this method can open up a new idea for accounting confirmation of enterprise data resources. Accountants can use this method to gradually improve the model through learning the experience of accounting experts, which can not only reduce the probability of errors in accounting judgment, but also improve the speed and accuracy of judgment. In addition, the research ideas of this paper also have some reference value for the recognition methods of other types of assets.

References 1. Li, R.: Research on big data asset recognition and measurement. Master dissertation of Xi’an University of Technology (2017)

Data Asset Model Construction

485

2. Li, Y., Ni, S.: Research on accounting recognition and measurement of data assets. J. Hunan Univ. Financ. Econ. (8), 82–90 (2017) 3. Tang, L., Li, S.: Research on data assets accounting. China Certif. Public Account. (2), 87–89 (2017) 4. Li, Z., Tan, X.: Big data asset accounting recognition, measurement and report. Financ. Account. Commun. (10), 58–60 (2018) 5. Shangguan, M., Bai, S.: Analysis of big data asset accounting treatment. Financ. Account. (22), 46–49 (2018) 6. Yu, Y.: Confirm that big data assets boost the development of new economy. J. Financ. Account. Mon. (23), 52–56 (2020) 7. Zou, Z.: Analysis of asset attributes of enterprise big data. Friends Account. (12), 7–12 (2017) 8. Zhang, J., Wei, Y., Song, X.: Research on accounting treatment and information presentation of enterprise data assets. Account. Econ. Res. (34), 3–15 (2020) 9. He, Y., Zhang, P.: Enterprise information resource accounting recognition and measurement framework optimization. J. Financ. Account. Mon. (17), 91–97 (2019) 10. Huang, S.: Seven laws of information resources and their confirmation and measurement. J. Financ. Account. (4), 3–9 (2020) 11. Tan, M.: On the conceptual development and theoretical framework of data assets. J. Financ. Account. (10), 87–95 (2021)

Cultural Creative Products Based on Information Processing Technology Xiaofei Zhou(B) and Can Chen Graduate School, Gachon University, Seongnam-Si, Gyeonggi-Do 130123, Korea [email protected]

Abstract. Today, with the rapid exchange of information, the world is developing rapidly. People can get more convenient and fast information resources, not only reflected in daily life, but also used in entertainment learning. With the development of information technology, people pay more attention to spiritual growth as well as material life. Cultural and creative (CAC) industries have the cultural core of “focus on the present”, “fresh”, and even “interesting”, so they are more popular. CAC industry is an emerging industry with creativity and creativity as the core under the development of economic globalization, which emphasizes the development and marketing through technology, creativity and industrialization. Since the term “cultural creativity” was put forward in Britain in 1997, the world has begun to attach importance to the development of cultural creative industries and cultural creative products. With the rising consumer demand, China’s CAC industries have also achieved considerable development. To promote local culture, and then improve the influence of the city. The purpose of this paper is to study the CAC industries or products under the background of information technology. From the UK to different countries in the world, the market of CAC industries is expanding, and the employees of CAC industries are also increasing. With the vigorous development of information technology, CAC industries and products will also usher in greater development. Keywords: Information technology · Cultural creativity · Creative products · Creative industries

1 Introduction The CAC industry under the background of IT is a sunrise industry at the forefront of the times, and the academic circles are also in the process of expanding branch areas and updating ideas accordingly, and rarely reach consensus [1–4]. In the area of research on the unique attributes of CAC industries that are different from traditional industries, Christiaan pays attention to the strong externalities and marginal cost inefficiency of knowledge industries, which makes the analysis of knowledge industries and related CAC industries require new ideas and methods [5]. Loots E also believes that the common experience and thinking of creative industries and general industries are different [6]. Si also expressed similar views. He believes that the set of research methods that © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 S. Sun et al. (Eds.): ICSINC 2021, LNEE 895, pp. 486–494, 2022. https://doi.org/10.1007/978-981-19-4775-9_60

Cultural Creative Products

487

we have accepted is not suitable for the research on CAC industries [7]. It also summarizes the main differences between traditional industries and creative industries: first, traditional industries are dominated by large-scale enterprises, while creative industries are dominated by small and medium-sized enterprises; second, traditional industries revolve around factories, while creative industries revolve around projects; third, traditional industries are led by entrepreneurs, while creative industries are led by consumers; fourth, production is the key to traditional industries’ value-added, and creative industries reap value at the consumer end of the value chain; fifth, traditional industries are clearly defined. Creative industries are scattered among other industries [8, 9]. This article uses the comparative method many times. Compare the definitions of CAC industries in various countries, compare the classification of primary, secondary, and tertiary industries with resource industries, compare existing theories supporting CAC industries, compare classical, neoclassical, and new economic theories and models that promote economic growth, compare the classification of CAC industries in Beijing and the relationship with the information service industry lay the foundation for the interpretation and abstraction of new theories [10].

2 CAC Industries or Products in the Context of IT 2.1 Digital Mass Communication of CAC Industries or Products in the Context of IT (1) E-book As an information carrier that can integrate words, pictures and sounds, e-books have had a great impact on the traditional publishing industry since its emergence. “The 2019–2020 China Digital Publishing Annual Report pointed out that mobile publishing has always occupied the position of the main force in digital publishing” [11]. The future direction of digital publishing should be mobile publishing with mobile terminals such as mobile phones as reading terminals, which is a general trend in the future [12]. 1) Advantages of e-books 2) Development trend of e-book industry E-books still have a lot of room for development in the future. In the final analysis, there are three factors that affect the development of e-books: one is the audience’s (readers) demand for e-books (ie content), the second is the medium of e-books itself (platform provider), and the third is the progress and relatedness of science and technology. As a new type of media, these three points are also the focus of all links in the industry chain. (2) Newsletter Since the birth of radio and television and their increasing development, the “big brother” newspaper is no longer “leading” in the field of news transmission, and even newspaper readers have been “robbed” by television. However, the emergence of novel electronic newspapers will enable the newspaper industry to regain its glory in the field of news dissemination.

488

X. Zhou and C. Chen

(3) Multimedia broadcasting (network video) 1) The function of network video Record and store. Online video can record a certain fact, activity, and life experience in a timely, accurate, objective, true and complete manner. The facts reflected by online video are clearer than the facts reflected by text, pictures and other means, and have the nature of news clues.  S online video clips are often adopted by traditional media. Network video can be shot by mobile phones, digital cameras, DV and other different equipment, uploaded and stored via broadband network or mobile Internet, and downloaded and watched via PC, mobile phone, MP4 and other clients, which is very convenient. 2) Interaction and sharing. Online video emphasizes the concept of openness and sharing. Netizens can use a large number of tools to share online videos with each other, such as e-mails, blogs, instant messaging, microblogs, etc. Influential online videos can also become the focus of public opinion and become hotspots or public events. Online video users can also interact with netizens through dig, leave messages, comments, reposts, and recommendations, etc., to exchange personal opinions and opinions with netizens. 3) New media on the Internet. Online video advocates the concept of user contribution content, giving users the right to speak, and becoming a typical “selfmedia”. Network video can be spread to any corner of the world with the help of the Internet. The spread of network video has the characteristics of intuitiveness, immediacy, fast speed and wide range. 2.2 Internet Innovative Applications of CAC Industries or Products in the Context of IT In the context of the rapid development of information technology, the network CAC industry presents vigorous vitality, especially the innovative application of the Internet. The user of Internet is not only the audience, but also a kind of media. They are two-way relations and interact with each other. Because of the increase in feedback opportunities, the audience has the ability to control the communication. The sharing of information resources highlights the value of information. The mutual integration of media also promotes the generation of special and customized needs. Interactivity has given birth to news portals and online games. Diversified content including, Weibo, IM, and social networks, and promoted the upgrading of traditional cultural industries. (1) News portal for innovative applications in the context of IT Portal website is the distribution center of information. As the Internet enters China, the portal is the most basic Internet application and has the widest user group. On the one hand, portals control the attention of the largest audience, and the influence of portals is huge; on the other hand, with the development of IT in recent years, portals are also actively integrating emerging applications, and personal and self-organizing trend. The portal’s powerful ability to integrate other Internet applications and huge conversion capabilities make the portal a well-deserved network hub.

Cultural Creative Products

489

(2) Online games of innovative applications under the background of IT China has become the world’s largest online game market, accounting for one-third of the global market share. Online games are also one of the most profitable areas in China’s new media industry. The rapid development of the online game industry has brought a series of economic and cultural benefits to China. Online games not only promote the great development of the entire Chinese economy, but also promote the continuous innovation of the IT technology industry, and contribute to stimulating China’s domestic demand and solving employment. (3) Weibo of innovative applications under the background of IT Weibo is the latest popular application. Among all kinds of popular applications, Weibo can always enter the “Top Ten” list. Weibo brings a new Internet experience. It has both the instant communication function of instant messaging tools and the interaction function of social networking sites. Blog adopts the same content mechanism as SMS and multiple publishing platforms or channels, and Weibo has been enthusiastically sought after by netizens. (4) Instant messaging tools for innovative applications under the background of IT Instant messaging is currently one of the most widely used Internet applications. It is widely welcomed by users for its convenience, speed, low cost, high efficiency, high privacy, simple use, and rich functions. Instant messaging tools have become another important communication tool besides mobile phones and fixed phones in work and life. Many users have two or two of instant messaging tools such as WeChat, QQ, Big Ant, Youdu Instant Messenger, Ruliu (formerly Baidu HI), Skype, Sina UC, MSN, Dingding, Enterprise WeChat, 360 Zhiyu, etc. Condensed on instant messaging tools-most personal relationships such as family, classmates, friends, colleagues, customers, etc., a large part of daily communication is also maintained through instant messaging tools, so instant messaging tools have become a lot of people. An indispensable part of life, there are many innovative applications in this field. As IoT and new coding technology develop, instant messaging tools become more and more convenient [13–15]. Instant messaging tools make people’s communication easier than ever. 2.3 Analysis of Location Selection to Aggregation of CAC Industries or Product Practitioners in the Context of IT It is not feasible to study location selection and agglomeration with completely different analytical methods. This is because if the agglomeration does occur after the location selection of the economic activity subject, then the factor that affects the location selection of the economic activity subject must be the concentration of the economic activity subject. It is not feasible to study location selection and agglomeration with completely different analytical methods. This is because if the agglomeration does occur after the location selection of the economic activity subject, then the factor that affects the location selection of the economic activity subject must be the concentration of the economic activity subject. Regarding “fashion gatekeepers” and ordinary CAC industry practitioners as two different types of talents, it is assumed that when they live alone in different cities, their respective numbers evolve in accordance with the law of logic and are available

490

X. Zhou and C. Chen

to each group. The creativity used is fixed (the term “creative depletion” exists, so this assumption is reasonable). Let PF (t) and PP (t) be the numbers of the two types of talents; rF and rP are their respective inherent growth rates; NF and NP are their respective maximum numbers. For the “fashion gatekeeper” group, there are: dPF PP = rF PF (1 − ) dt NP

(1)

Among them, (1 − NPPP ) means that the “fashion gatekeeper” constantly consumes their own creativity, which has caused the reduction of creative capacity in the minds of this group and hindered their own growth. NPPP represents the number of ideas supplied to this group consumed by a unit of “fashion gatekeeper” (set the total amount of ideas to 1). “Fan gatekeepers” can exist alone, and ordinary CAC industry practitioners must rely on them to survive. At the same time, the existence of the latter can also provide certain benefits for the development of the former (this is because interpersonal communication contributes to the generation of creativity). In this way, the evolutionary law of the number of “fashion gatekeepers” becomes: PP PP dPF = rF PF (1 − + aF ) dt NP NP

(2)

Among them, aF means that the creativity provided to the unit number of “fashion gatekeepers” by the ordinary CAC industry practitioners per unit is aF times the amount of creativity consumed by the group of “fashion gatekeepers” per unit.

3 CAC Industries or Product Research Experiments Under the Background of IT 3.1 Research Objects Investigate the status quo of CAC industries or product agglomeration in domestic and foreign cities, mainly in London, New York, Beijing, Shanghai and other places. The British Blair’s decision has proved to be wise and effective. In 2000, 70% of these companies are concentrated in the software and computer service industry, music, and visual performance industries. The United States is a large country in the CAC industries, and has formed a complete industrial chain, market and consumer groups. Among the 400 richest companies in the United States, more than 70 are directly engaged in the production of CAC products. Beijing’s “Several Opinions on Accelerating Beijing’s Cultural Development” promulgated in 1996 proposed to “vigorously develop cultural industries and make them one of Beijing’s pillar industries. Beijing has become an important cultural industry base in the country.”

Cultural Creative Products

491

3.2 Data Sources The data comes from the website of the British Ministry of Culture, Media and Sports and is compiled by the author.

4 Experimental Analysis of CAC Industries or Product Research Under the Background of IT 4.1 Employment Distribution and Output Value of CAC Industries in London As a world-class city, London has a diverse, mature and international CAC industry structure, and is the center of the British CAC industry. See Table 1 for details. Table 1. Composition of added value of London’s creative industries Industry

Creative industry added value

Creative industry per capita labor output (pounds)

London GVA (million pounds)

London

Absolute value

Proportion (%)

Advertising and marketing

3631

10.5

54773

Building

1349

3.9

49924

Handicraft

159

0.5

61799

Design

947

2.7

42375

8633

25.0

84392

10777

31.1

93914

5341

15.4

89669

Museums, galleries and libraries

601

1.7

36754

Music, performance and visual arts

3163

9.1

42171

34601

100

555771

Film, television, video, radio and photography IT, software and computer services publishing Publishing

Total

492

X. Zhou and C. Chen

35 30

80000

25

60000

20

40000

15 10

20000

Proportion (%)

London GVA (million pounds)

Absolute value 100000

5 0

0

Industry Fig. 1. The composition of the added value of London’s creative industries

It can be seen from Fig. 1 that there are currently about 700,000 people in the City of London engaged in CAC industries-related jobs. The output value created by these people accounts for 15% of the total London economy, and their number accounts for 20% of the employment in London. The annual output value of the CAC industry is around 3 billion pounds. Simply looking at the above figures cannot show the degree of concentration of CAC industries in London. Only by comparing these figures with the UK as a whole can London’s advantages in CAC industries be revealed. The population of London accounts for approximately 12% of the total population of the United Kingdom. But London has 40% of the UK’s art facilities, 70% of music and recording studios, 90% of music events, 70% of film and television productions, 46% of advertising, 85% of fashion designers, and 27% of buildings. In the past 10 years, fashion designers have been the industry with the highest growth in London, with an increase of 88%; artists, commercial artists and graphic designers have increased by 71%; actors, entertainers, stage managers, program producers and directors have increased by 43%; the overall growth of CAC freelancers increased by 81%. 4.2 Distribution of Employment Population and Output Value by Industry in Beijing’s CAC Industries The operation of Beijing’s CAC industries in recent years is shown in Fig. 2:

increments

Cultural Creative Products

1000 800 600 400 200 0 -200

2016

2017

2018

2019

493

2020

Industry Fig. 2. Beijing’s CAC industries in recent years

It can be seen from the above chart that Beijing’s CAC industries have shown a sustained growth trend, both in terms of the added value of the CAC industries and the number of practitioners. In 2018, the city’s cultural, sports and entertainment industries grew by 29.4%, which was much higher than the average level of the tertiary industry of 11.7%, becoming the fastest growing industry in the tertiary industry in Beijing, and its added value has exceeded that of wholesale and retail. Real estate, business services, transportation, etc. The CAC industries have become Beijing’s pillar industries worthy of the name. Another manifestation of the development of Beijing’s CAC industry leading the country is the clustering characteristics of the industry in Beijing.

5 Conclusions The essence of the cultural creative industry is an industry centered on creativity and knowledge. Cultural products with product attributes involve two types of “ordinary production” and “creative production” in the production process. The characteristics of the emerging CAC industries determine the identity of their production and consumption subjects, and the identity of the dissemination and reception subjects. It is worth mentioning that the production and consumption subjects are identical but not completely overlapped, and so are the dissemination subjects. The particularity of the new CAC industry makes the various relationships between the subject and the object, from the source to the terminal, from the increase in value to the change of the institutional structure, and then to the form of the emerging CAC industry, which are unique and rich, and are based on information. New systems and mechanisms for technologically emerging CAC industries.

References 1. Kim, Y.: Making “Creative” movement: transformation of urban culture and politics in Bandung, Indonesia. Geogr. Rev. Japan Ser. B 90(1), 17–25 (2017)

494

X. Zhou and C. Chen

2. Moalosi, R., Setlhatlhanyo, K.N., Sealetsa, O.J.: Cultural memory, an asset for design-driven innovation within the creative industries sector: lessons for design education. Des. Technol. Educ. 21(2), 9–22 (2016) 3. Nurhasan, N.: Integration model of creative telematics industries with manufacturing industry in West Jawa province. J. Sampurasun Interdiscip. Stud. Cult. Herit. 3(2), 91 (2017) 4. Liu, C.H.S.: Examining social capital, organizational learning and knowledge transfer in CAC industries of practice. Tour. Manage. 64, 258–270 (2018) 5. De Beukelaer, C.: Toward an ‘African’ take on the CAC industries? Media Cult. Soc. 39(4), 582–591 (2017) 6. Loots, E., Witteloostuijn, A.V.: The growth puzzle in the creative industries: or why creatives and their industries are a special case. Revue de l’Entrepreneuriat 17(1), 39 (2018) 7. Si, S.: A report on Beijing’s CAC industries media clusters. Glob. Media China 1(4), 412–421 (2016) 8. Kos Koklic, M., Kukar-Kinney, M., Vida, I.: Three-level mechanism of consumer digital piracy: development and cross-cultural validation. J. Bus. Ethics 134(1), 15–27 (2014). https:// doi.org/10.1007/s10551-014-2075-1 9. Alencar, D.G., Santos, M., Guissoni, R.: Creative economy, cinema and tourism: a study about the movie “Os Xeretas” in the city of Castro/ Paraná – Brazil. Revista de Turismo Contemporâneo 9(1), 104–125 (2020) 10. O’Connor, J., Gu, X., Vickery, J.: Teaching the CAC industries: an international perspective. Arts Humanit. High. Educ. 18(2–3), 93–98 (2019) 11. Wu, Y.: Design of tourism CAC products based on regional historical and cultural elements. In: E3S Web of Conferences, vol. 251, no. 6, p. 03004 (2021) 12. Emran, S.J., Ifteakha, R.K., Uddin, M.A.: “Exploring the Unexplored: Determinants of Creative Industry” introduction. Dhaka Univ. Stud. Part B 70(1), 133–147 (2019) 13. Wang, Z., Liu, R., Liu, Q., Thompson, J.S., Kadoch, M.: Energy-efficient data collection and device positioning in UAV-assisted IoT. IEEE Internet Things J. 7(2), 1122–1139 (2020) 14. Chen, N., Rong, B., Zhang, X., Kadoch, M.: Scalable and flexible massive MIMO precoding for 5G H-CRAN. IEEE Wirel. Commun. Mag. 24(1), 46–52 (2017) 15. Nessa, A., Kadoch, M., Rong, B.: Fountain coded cooperative communications for LTE-A connected heterogeneous M2M network. IEEE Access 4, 5280–5292 (2016)

Research on Online Teaching Model Mining Based on Network Database Baoquan Men(B) and Yuan Liu Henan Vocational College of Agriculture, Zhengzhou 451450, Henan, China [email protected]

Abstract. The development of information technology and the promotion of information technology in the field of education provide technical support for education reform and of training. The way the network teaches is something new for people, and is accepted by people who are familiar with it. In order to study the impact of teaching on the network and to improve the thinking and capacity of information technology in higher education, this document collects data from the database based on the management of e-teaching, using case analysis, analysis of documents and other methods, establishes a networked learning method and reads and analyses a large number of relevant libraries through the library research method. The results show that networked teaching can effectively improve students’ technical thinking and skills. Students’ technical thinking and skills are about 30% greater than traditional teaching methods, and their sense of cooperation reaches 0.8, which resolves the contradiction between population and progress, solves the polarisation and transformation of retarded students, it fully demonstrates and develops their personality, and places higher vocational education and teaching in the course of quality education. This shows that the management of network teaching plays a good role in improving the IT culture and the ability of higher education training. Keywords: Online teaching · Vocational education · Thinking and ability · Teaching methods

1 Introduction In the basic education stage of our country, the success in mathematics education is obvious to all. With the development of the times, our country proposes education reform centered on the promotion of quality education. The new round of curriculum reform is exactly some measures to implement quality education. The Ministry of Education requires the relevant services to dynamically boost the innovation of the primary education curriculum, adjust and reform the education model, to build a new education model that meets the requirements of quality education [1]. In today’s society, education reform has had a major impact on education. Education experiences changes in the cognitive structure of pupils, learning methods, interpersonal relationships and self-evaluation. However, due to the restriction of test-oriented education, the current junior high school mathematics teaching model has not undergone © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 S. Sun et al. (Eds.): ICSINC 2021, LNEE 895, pp. 495–501, 2022. https://doi.org/10.1007/978-981-19-4775-9_61

496

B. Men and Y. Liu

a fundamental change, and the teaching model of many schools and teachers has not changed. It is still an old and rigid traditional teaching model, and the phenomenon of emphasizing conclusions and neglecting process is very serious [1]. The way the network teaches mobilises a variety of information, makes students interested in learning, makes students willing to participate in class activities, brain storms, teamwork and problem solving, and ultimately achieves the goal of improving their capacities in all aspects. In the use of this teaching model for mathematics, the various intelligences of students have been improved, and each student has made progress in mathematics learning. Online teaching mode has been widely used in foreign countries, but it is still a new educational concept in my country. Although it is gradually being accepted in the field of education in my country, some educational scholars have conducted research on it, but diversified teaching modes are used in subject teaching [2]. Professor Huang believes that higher vocational education can increase advanced training, develop a new curriculum content and increase pupils’ interest in the curriculum [3]; Xu Jinguang believes that the problem of higher education is due to a lack of interaction with students in combination with their personal characteristics [4]; Chen Zhiwei believes that under the current background of comprehensive emphasis on education, the courses, evaluation methods and categories of higher vocational education should be reviewed. Carry out a comprehensive reform and optimization of teaching in colleges and universities, and make great progress in teaching in colleges and universities [5]. Using the methods of analysing literature and researching the questionnaire, this document constructs a pluralistic system, in which teachers of higher education doctoral programmes, i.e. university teachers and physical education students, participate in the evaluation and investigation. Through the questionnaire survey, we get the current problems that students and teachers think of the current higher vocational education, and put forward related solutions to these problems, which can enrich the research on higher vocational education and teaching by related scholars, which has certain theoretical research significance.

2 Higher Vocational Education Methods for Online Teaching Management We can also call the online learning environment, which is based on computer technology. It generally includes learning resources and an online learning platform. Compared with offline traditional classrooms, the online learning environment is freer in time and space. The types are more abundant, and resources can be quickly shared. Various platforms and resources provide a platform and help for teachers and students to conduct online learning [6]. With the faster and faster development of online learning, more and more people are beginning to learn through online mode. Almost every online learning platform has collected data and information from many users. Since the types of these data are not the same, these data can be integrated and applied to the teaching field. Educational data mining refers to the screening of valuable data on the platform, so that complex data has value, which can have a clearer understanding of students, propose corresponding

Research on Online Teaching Model Mining

497

measures for student education, and establish better teaching models [7]. The platform has a variety of data information, including all the data in the school’s teaching process to students [8]. Most of the university courses are independent of each other, and each subject needs to be reasonably arranged. In this process, it takes a lot of time to conduct detailed analysis. Only in this way can we make course arrangements [9]. University teaching requires scientific teaching methods. For long-term development, university curriculum systems need to be scientific and practical. Moreover, physical education is also an indispensable course. Students’ sports awareness can achieve self-improvement through ideological education, which is a very important part of university education. We can use the following formula to illustrate: 1  βij (z)u(kT + tj−1 ) + v(kT + ti ) α(z) r

y(kT + ti ) =

(1)

j=1

Can be transformed into: α(z) = 1 + α1 z −1 + α2 z −2 + ... + αn z −n

(2)

βij (z) = βij0 + βij1 z −1 + βij2 z −2 + ... + βijn z −n

(3)

Its function s(kT +ti−1 )i = 1, 2, ..., r −1 is to move the sampling signal s(kT +ti−1 ) in time backward by 1 non-uniform sampling interval, and a new transfer function model is proposed: y(kT + tt ) =

Bi (δ) u(kT + ti ) + v(kT + ti ) Ai (δ)

(4)

3 Online Teaching Management Higher Vocational Education Experiment 3.1 Subject In the online learning process of students, students mainly use QQ groups for online learning, and the teacher divides the whole class into groups. Each group will elect a leader and set up a discussion group to facilitate the study and discussion of each member of the group. Prior to class, the teacher uploads the work list of this class to the QQ group file, and provides students with the corresponding microfiche, learning websites and other resources. 3.2 Establish a Model Evaluation Index System The real result usually requires observation of facts. Generally speaking, in the evaluation mode, there are three levels of evaluation results, among which the decomposition and distinction between levels will be carried out. Among them, the results of the first two levels of evaluation are more subjective and cannot be the final result of the evaluation. The three-level evaluation results must be meticulous, calculable, and scientifically based, and can become a powerful basis for evaluation.

498

B. Men and Y. Liu

3.3 Determine the Evaluation Weight The importance of related factors and useful data need to be expressed by indicator weights. In the entire evaluation model, each indicator has a corresponding weight corresponding to it. In other words, the weights are different when the indicators are the same. The index weight is referred to as the weight for short, which can be represented by a lowercase letter a. The index weight value is greater than 0 and less than 1, and the weight of the first level index adds up to 1, that is, 0 < a < 1 and a-1. 3.4 Statistics The data processing process in this article is carried out on the SPSS19.0 tool, and the method of inspection is bilateral inspection. When the p value is less than 0.05, it is defined as dominant. If the data check result shows a normal distribution, the data contained in each group is checked by a double-T method, and each group is checked by a single sample. If the value is positive, it indicates that the distribution is insufficient, and two samples need to be checked.

4 Online Teaching Management Higher Vocational Education Experimental Analysis We first divide the students into groups, collect their current information technology thinking and abilities from different groups of students, quantify them through templates, and compare them separately. The specific statistics are shown in Table 1: Table 1. Student abilities Test score

Interest intensity

Thinking ability

Calculate ability

Information Technology

P

team 1

2.14

2.14

2.43

2.17

2.22