Intelligent Communication, Control and Devices: Proceedings of ICICCD 2020 (Advances in Intelligent Systems and Computing) [1st ed. 2021] 9811615098, 9789811615092

Citation preview

Advances in Intelligent Systems and Computing 1341

Sushabhan Choudhury R. Gowri Babu Sena Paul Dinh-Thuan Do   Editors

Intelligent Communication, Control and Devices Proceedings of ICICCD 2020

Advances in Intelligent Systems and Computing Volume 1341

Series Editor Janusz Kacprzyk, Systems Research Institute, Polish Academy of Sciences, Warsaw, Poland Advisory Editors Nikhil R. Pal, Indian Statistical Institute, Kolkata, India Rafael Bello Perez, Faculty of Mathematics, Physics and Computing, Universidad Central de Las Villas, Santa Clara, Cuba Emilio S. Corchado, University of Salamanca, Salamanca, Spain Hani Hagras, School of Computer Science and Electronic Engineering, University of Essex, Colchester, UK László T. Kóczy, Department of Automation, Széchenyi István University, Gyor, Hungary Vladik Kreinovich, Department of Computer Science, University of Texas at El Paso, El Paso, TX, USA Chin-Teng Lin, Department of Electrical Engineering, National Chiao Tung University, Hsinchu, Taiwan Jie Lu, Faculty of Engineering and Information Technology, University of Technology Sydney, Sydney, NSW, Australia Patricia Melin, Graduate Program of Computer Science, Tijuana Institute of Technology, Tijuana, Mexico Nadia Nedjah, Department of Electronics Engineering, University of Rio de Janeiro, Rio de Janeiro, Brazil Ngoc Thanh Nguyen , Faculty of Computer Science and Management, Wrocław University of Technology, Wrocław, Poland Jun Wang, Department of Mechanical and Automation Engineering, The Chinese University of Hong Kong, Shatin, Hong Kong

The series “Advances in Intelligent Systems and Computing” contains publications on theory, applications, and design methods of Intelligent Systems and Intelligent Computing. Virtually all disciplines such as engineering, natural sciences, computer and information science, ICT, economics, business, e-commerce, environment, healthcare, life science are covered. The list of topics spans all the areas of modern intelligent systems and computing such as: computational intelligence, soft computing including neural networks, fuzzy systems, evolutionary computing and the fusion of these paradigms, social intelligence, ambient intelligence, computational neuroscience, artificial life, virtual worlds and society, cognitive science and systems, Perception and Vision, DNA and immune based systems, self-organizing and adaptive systems, e-Learning and teaching, human-centered and human-centric computing, recommender systems, intelligent control, robotics and mechatronics including human-machine teaming, knowledge-based paradigms, learning paradigms, machine ethics, intelligent data analysis, knowledge management, intelligent agents, intelligent decision making and support, intelligent network security, trust management, interactive entertainment, Web intelligence and multimedia. The publications within “Advances in Intelligent Systems and Computing” are primarily proceedings of important conferences, symposia and congresses. They cover significant recent developments in the field, both of a foundational and applicable character. An important characteristic feature of the series is the short publication time and world-wide distribution. This permits a rapid and broad dissemination of research results. Indexed by DBLP, EI Compendex, INSPEC, WTI Frankfurt eG, zbMATH, Japanese Science and Technology Agency (JST). All books published in the series are submitted for consideration in Web of Science.

More information about this series at http://www.springer.com/series/11156

Sushabhan Choudhury · R. Gowri · Babu Sena Paul · Dinh-Thuan Do Editors

Intelligent Communication, Control and Devices Proceedings of ICICCD 2020

Editors Sushabhan Choudhury Department of Electrical and Electronics Engineering University of Petroleum and Energy Studies Dehradun, Uttarakhand, India Babu Sena Paul Institute for Intelligent Systems University of Johannesburg Johannesburg, Gauteng, South Africa

R. Gowri Department of Electrical and Electronics Engineering University of Petroleum and Energy Studies Dehradun, Uttarakhand, India Dinh-Thuan Do Department of Computer Science and Information Engineering Asia University Taichung, Taiwan

ISSN 2194-5357 ISSN 2194-5365 (electronic) Advances in Intelligent Systems and Computing ISBN 978-981-16-1509-2 ISBN 978-981-16-1510-8 (eBook) https://doi.org/10.1007/978-981-16-1510-8 © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2021 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

The Department of Electrical and Electronics Engineering, University of Petroleum and Energy Studies, Dehradun, has organized the Fourth International Conference on Intelligent Communication, Control, and Devices (ICICCD 2020) from November 27th to 28th 2020. The conference focuses on the integration of intelligent communication systems, control systems, and devices related to all aspects of engineering and sciences. ICICCD 2020 aims to provide an opportune forum and vibrant platform for researchers, academicians, scientists, and industrial practitioners to share their original research work, findings, and practical development experiences. The proceedings will be published in the Advances in Intelligent Systems and Computing (AISC) book series of Springer. The general aim of the conference is to promote international collaboration in education and research in all fields and disciplines of engineering. ICICCD 2020 is an international forum for those who wish to present their projects and innovations, having also the opportunity to discuss the main aspects and the latest results in the field of education and research. The organizing committee is extremely grateful to the authors from India and abroad who had shown tremendous response to the call for papers. Nearly 80 papers were submitted from the researchers, academicians, and students on wide areas of four parallel tracks such as intelligent communication, intelligent control, intelligent devices, and sustainable technology in power system. We are obliged to our Honorable Chancellor Dr. S. J. Chopra; Vice Chancellor Dr. Sunil Rai; Dean Research & Development Dr. D. K. Avasthi; Dean (School of Engineering) Dr. Kamal Bansal for their confidence they have invested on us for organizing ICICCD 2020. We extend our thanks to all faculty members, external reviewers, and staff members with different committees for organizing the conference and making it a grand success. Dehradun, India Dehradun, India Johannesburg, South Africa Taichung, Taiwan

Sushabhan Choudhury R. Gowri Babu Sena Paul Dinh-Thuan Do v

Contents

Improvement of QOS by Optimizing 2.4 GHZ RF Network Parameters for Oil Pipeline Management . . . . . . . . . . . . . . . . . . . . . . . . . . . . Chavala Lakshmi Narayana, Rajesh Singh, and Anita Gehlot

1

Transient Stability Enhancement via Sliding Mode Control for Lower Order Triangular System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Amrendra Kumar Karn, Salman Hameed, and Mohammad Sarfraz

11

A Convolution Neural Networks and IoT-Based Approach to Surveillance System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Shreyash Gondane, Adesh Thakare, Chaitanya Deshpande, and Omprakash Gupta Prevent Road Accident Using Intelligent Device . . . . . . . . . . . . . . . . . . . . . . Amrit Suman, Chiranjeev Kumar, and Preetam Suman Choice of Suitable Height of Substrate in the Designing of Microstrip Antenna for X-band Communication . . . . . . . . . . . . . . . . . . . R. K. Chaurasia and S. K. Nippani

25

35

45

Comparative Analysis of Multi-slot Circular Patch Antenna . . . . . . . . . . . Praful Ranjan, Prasanthi Kumari Nunna, and Bhupesh Chandra Joshi

53

Active Power Curtailment in PV Array Under LVRT Condition . . . . . . . Jyoti Joshi, Anurag Kumar Swami, and Vibhu Jately

65

Autonomous Tomato Harvester Using Robotic Arm and Computer Vision . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Siddhant Salvi, Vivekanand Sahu, Rohit Kalkundre, and Omkar Malwade

75

Impact of Inclination of Path Profiles on the Performance of Electric Vehicles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Indu Krishnakumar and Aswatha Kumar Mattur

91

vii

viii

Contents

Design of Environment-Friendly FIR Filter and Its Implementation on FPGA for Green Communication . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103 Anoop Raghuvanshi, Kuber Singh Mehra, and Indra Singh Bisht Atomistic Simulation Studies of the Mechanical and Thermal Properties of Silver Nanowires as Interconnects for Nanoelectronics Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111 S. K. Joshi, D. P. Singh, and Santosh Dubey Mobility Management Scheme During Offloading in Mobile Cloud Computing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123 Robin Prakash Mathur and Manmohan Sharma Multiuser Detection for MIMO-OFDM System Using Binary Spotted Hyena Optimizer in UWA Communication . . . . . . . . . . . . . . . . . . . 133 Md Rizwan Khan, Bikramaditya Das, and Bibhuti Bhusan Pati Estimation of Solar Radiation at Farasan Island with Two-Step ANN Concepts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143 Arun Kumar Singh, Vikas Pandey, and Rupendra Kumar Pachauri Development of Matlab/Simulink Library for Unsupported Microcontrollers, Case Study: STM32F407 . . . . . . . . . . . . . . . . . . . . . . . . . . 153 Nguyen Van Khanh, Ho The Anh, Nguyen Huynh Anh Duy, Pham Tran Lam Hai, Nguyen Van Muot, Tran Thanh Hung, and Nguyen Chi Ngon A Novel Bistatic LIDAR Device with 1570 nm Centre Wavelength Diode for Detection of Plant Disease . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167 Hai Pham, Khanh Nguyen, Ngon Nguyen, Hung Tran, and W. Genthe FPGA-Based Power Optimized CAM Design Using LVCMOS18 and High-Speed Low Voltage Digitally Controlled Impedance . . . . . . . . . 177 Reeya Agrawal, Manish Kumar, Manish Gupta, and Vinay Kumar Deolia Design of Chili Fruit Flipping Mechanism for Identification of the Damages Caused by Diseases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185 Quoc-Khanh Huynh, Chi-Ngon Nguyen, Wei-Chih Lin, Hong-Phuc Vo-Nguyen, Phuong Lan Tran-Nguyen, Dang-Khanh-Linh Le, and Van-Cuong Nguyen Performance Analysis of Pentuple Micro-optical Asymmetric Ring Resonator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195 S. Vamsi Krishna, Suman Ranjan, and Sanjoy Mandal MPPT Control of Hydrokinetic Energy Conversion System . . . . . . . . . . . . 207 Anuj Kumar Pandey, Irfan Ahmed, and Pooja Sharma

Contents

ix

Using Adaptive Reduced Power Subframes to Mitigate Inter-Cell Nosiness in Heterogeneous Networks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217 Arun Kumar Singh, Vikas Pandey, and Ranjan Mishra Solution of Economic Load Dispatch Using Flower Pollination Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 227 Praveen Kumar Mahto, Kumari Sarwagya, Suman Ranjan, and Upendra Prasad Fault Detection in Overhead Lines Using IoT-Enabled Fault Passage Indicator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 239 Monika Yadav, Prayanshu Verma, and Devender Kumar Saini Linear State Estimation for Power Systems Based on PMU Measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 249 Biswaranjan Mishra, S. S. Thakur, Pankaj Rai, and D. K. Tanti Improvement of PID Controllers by Recurrent Fuzzy Neural Networks for Delta Robot . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 263 Minh Le Thanh, Luong Hoai Thuong, Pham Thanh Tung, and Chi-Ngon Nguyen Vision-Based Measurement of Leaf Dimensions and Area Using a Smartphone . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 277 Chanh-Nghiem Nguyen, Dang-Khoa Thach, Quoc-Thang Phan, and Chi-Ngon Nguyen Long-Range Radio and Vision Node Based Waste Management System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 291 Shaik Vaseem Akram, Rajesh Singh, and Anita Gehlot Predicting the COVID-19 Cases in India . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 299 Arpit Jain, Abhinav Sharma, and S. Nitisha Bharathi Experimental Evaluation of the Performance of Diaphragm Spring Clutch of a Four-Stroke Multi-cylinder Petrol Engine Under Dry Friction Conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 313 Shriram Dravid, Jitendra Yadav, Santosh Kumar Kurre, Adesh Kumar, and Roushan Kumar Cognitive Radio-Based Spectrum Allocation Method for Next Generation Cellular Networks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 321 Rajarao Manda and B. V. Sudhakar Reddy Investigation of Analog Parameters and Miller Capacitance Affecting the Circuit Performance of Double Gate Tunnel Field Effect Transistors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 335 Deepak Kumar, Shiromani Balmukund Rahi, and Piyush Kuchhal

x

Contents

Systematic Framework for Early Fire Detection and Smart Evacuation Using loRa—A Long-Range and Low-Power Communication Protocol . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 351 Gajanand Birajdar, Rajesh Singh, and Anita Gehlot The Development and Implementation of Competency-Based ITS for Learners with Learning Disability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 359 Akanksha Bisht and Neelu Jyothi Ahuja Analysis of Fuzzy Reliability of the System Using Intuitionistic Fuzzy Set . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 371 Akshay Kumar, Mangey Ram, Nupur Goyal, Soni Bisht, Sunil Kumar, and R. P. Pant Reliability Analysis Using Fuzzy Analytical Hierarchy Process (FAHP) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 379 Nupur Goyal, Shristi Kharola, Akshay Kumar, Mangey Ram, Seema Saini, and Pradeep Bedi High Resolution Detection and Localization of Targets in Noisy Environment of Compressed Sensing Radar . . . . . . . . . . . . . . . . . . . . . . . . . 389 Sudha Hanumanthu and P. Rajesh Kumar Exploration of Magnitude Squared Coherence for Assessment of More Functionally Associated Region of the Brain to the Heart of Healthy Subjects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 399 Rajesh Polepogu and Naveen Kumar Vaegae Sign Language Recognition Using Convolutional Neural Networks . . . . . 415 Roshin Mary Varghese, Simran Siddharth, Joel Biju, Sukanya Dutta, Arnav Aggarwal, and Naveen Kumar Vaegae Author Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 427

About the Editors

Dr. Sushabhan Choudhury received his B.E. and M.Tech. from NIT Silchar and Tezpur University, respectively, and Ph.D. degree from UPES, Dehradun. He has rich teaching and industry experience. His research interest includes wireless sensor networks, embedded system, robotics, automation and control, IoT (Node-MCU, ESP8266), nanotechnology, artificial intelligence and machine learning. He has published more than 50 research articles in international/national journals and conferences. He has filed 13 patents and authored several books. He is selected as an outstanding scientist of the twenty-first century by Cambridge Biographical Centre. He is presently working as Professor and Head of the Department of Electrical and Electronics Engineering, School of Engineering UPES. Dr. R. Gowri received her M.Sc. and M.Tech. from Madurai Kamraj University and Delhi University in 1995 and 1998, respectively, and doctoral in 2005. She has been associated with academics and industries as Design Engineer for the past 22 years since 1998. She has the passion to work in the field of designing microwave passive and active components, communication and antenna designs. She has published around 40 publications in international/national journals and conferences and 4 books. She is a member of IEEE (MTT, APS) and a member of IETE. She has successfully completed five research projects funded by ISRO and DRDO. Prior to joining UPES, she got experience in both academic and administration in various capacities as Dean/Director and HOD in Graphic Era University, Dehradun. She has been continuously guiding scholars at various levels. Dr. Babu Sena Paul received his B.Tech. and M.Tech. degree in Radio Physics and Electronics from the University of Calcutta, West Bengal, India, in 1999 and 2003, respectively. He was with Philips India Ltd from 1999 to 2000. From 2000 to 2002, he served as Lecturer of Electronics and Communication Engineering Department at SMIT, Sikkim, India. He received his Ph.D. degree from the Department of Electronics and Communication Engineering, Indian Institute of Technology Guwahati India. He has attended and published over sixty research papers in international and national conferences, symposiums and peer-reviewed journals. He has successfully supervised several postgraduate students and postdoctoral research fellows. He joined xi

xii

About the Editors

the University of Johannesburg in 2010. He has served as Head of the Department at the Department of Electrical and Electronic Engineering Technology, University of Johannesburg, from April 2015 to March 2018. He is currently serving as Associate Professor and Director to the Institute for Intelligent Systems, University of Johannesburg. He was awarded the IETE Research Fellowship. He is a life member of IETE and a member of IEEE. Dinh-Thuan Do (a Senior Member, IEEE) received the B.S., M.Eng. and Ph.D. degrees from Vietnam National University (VNU-HCMC), in 2003, 2007, and 2013, respectively, all in communications engineering. He was Visiting Ph.D. Student with the Communications Engineering Institute, National Tsing Hua University, Taiwan, from 2009 to 2010. Prior to joining Ton Duc Thang University, he was Senior Engineer with Vina Phone Mobile Network, from 2003 to 2009. His publications include over 100 SCIE/SCI-indexed journal articles, over 45 SCOPUS-indexed journal articles and over 50 international conference papers. He is the sole author of one textbook and one book chapter. His research interests include signal processing in wireless communications networks, NOMA, full-duplex transmission and energy harvesting. Dr. Thuan was a recipient of Golden Globe Award from Vietnam Ministry of Science and Technology in 2015 (Top 10 most excellent scientist nationwide). He is currently serving as Editor of Computer Communications (Elsevier), Associate Editor of EURASIP Journal on Wireless Communications and Networking (Springer) and Editor of KSII Transactions on Internet and Information Systems.

Improvement of QOS by Optimizing 2.4 GHZ RF Network Parameters for Oil Pipeline Management Chavala Lakshmi Narayana, Rajesh Singh, and Anita Gehlot

Abstract System empowered computerized advancements are giving new and invigorating opportunities to build the accessibility of devices with the ultimate objective of current mechanization. ZigBee is a mechanized far-off the development that is used for oil pipeline management. In this paper, the oil pipeline management utilizing ZigBee has been sent, investigated, and proposed novel architecture. The boundaries for QoS of the framework as the estimations, i.e., throughput, Traffic Rcvd/Sent, MAC delay, retransmission rate, and so forth are examined. Keywords ZigBee · QOS · Oil pipeline · WSN · Simulations

1 Introduction These days wireless sensor network advances are becoming well known for a few sorts of utilizations. Wireless sensor network (WSN) is oneself sorting out specially appointed multi-bounce organize that comprises of spatially disseminated sensor and hubs [1]. As of late WSN has been utilized in numerous zones in horticulture, industry, household and clinical condition observing, etc. [2]. Zigbee gives low data transfer rate and secure data transfer. ZigBee was intended to give high information throughput with less power utilization. ZigBee low force utilization limits transmission separations to 100 m relying upon power yield and environmental qualities [3]. In this paper, we proposed a methodology of oil pipeline management using Zigbee communication protocol and optimization of Zigbee parameters using riverbed simulator. C. L. Narayana (B) · R. Singh · A. Gehlot Research Scholar, Department of SEEE, Lovely Professional University, Punjab, India e-mail: [email protected] R. Singh e-mail: [email protected] A. Gehlot e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2021 S. Choudhury et al. (eds.), Intelligent Communication, Control and Devices, Advances in Intelligent Systems and Computing 1341, https://doi.org/10.1007/978-981-16-1510-8_1

1

2

C. L. Narayana et al.

1.1 Literature Review Pipelines have been utilized to move water, oil, and gas for several years. Normally they serve private networks, mechanical destinations, and business focuses dependably and quietly. As the pipeline foundation ages, more mishaps cause fatalities, natural, and human damages [12]. It has been demonstrated that oil and gas pipeline systems are the most conservative and most secure mean of shipping rough oils and they satisfy a popularity for proficiency and unwavering quality [4, 5]. Be that as it may, as shipping unsafe substances utilizing miles-long pipelines has gotten mainstream over the globe in late decades, the possibility of basic accidents because of pipeline failures increments [6]. The reasons for the failures are either deliberate or material failure, etc. damages [7, 7], prompting pipeline disappointment and along these lines bringing about irreversible harms which incorporate budgetary losses and extraordinary ecological contamination, especially when the spillage isn’t detected in a convenient manner [9, 9]. The reasons for pipeline damages shift. Figure 1 shows a pie outline that represents measurements of the significant reasons for pipelines failures which incorporate pipeline corrosion, human carelessness, and surrenders during the procedure of establishment and erection work, and imperfections happening during the assembling procedure and outside elements [11].

2 Proposed Methodology In [13] author proposed detection of third-party damages (oil theft, building collapses near the pipeline, natural calamities, etc.) of an oil pipeline using IoT technology, the proposed architecture is divided into three sections, pipeline data transmitter, safety operation, and end-user as shown in Fig. 2. In the pipeline monitoring section,

Fig. 1 Pipeline failure statistics [12]

Improvement of QOS by Optimizing 2.4 GHZ RF …

3

Fig. 2 Architecture of oil pipeline management

all the parameters of the pipeline are monitored and by using IoT sensors and send data to microcontroller for analysis whether the information collected from sensors is normal or abnormal and displayed in LCD. If any abnormal data found, Zigbee communication protocol is activated and the safe operation section. In the safety operation section, we can find two subsections, i.e., drone and hooter sections. Figure 3 shows the hooter system, which is used to alert the local pipeline monitoring people. The Zigbee receives the signal and sent to the controller that gives instructions to the hooter driver which is activated and gives alarm.

4

C. L. Narayana et al.

Fig. 3 Hooter system

Fig. 4 Drone section

Simultaneously, the Zigbee sends the data to the drone section receiver and the controller sends the signals to activate the drone for pipeline inspection as shown in Fig. 4. Finally, the end-user section is used to gather data from the pipeline section using gateway via 2.4 GHz RF module and sent controller. Using Wi-Fi module the data is transformed to the application server as represented in Fig. 5.

2.1 Simulation for Optimizing Zigbee Module ZigBee has three sorts of gadgets; router, coordinator, and end device. ZigBee coordinator plays important role in network. It organizes various hubs in the network. ZigBee coordinator might be separated to work the system keeping up helpful (QoS) [14]. Routing is a significant boundary which routes packets in network. With ACK

Improvement of QOS by Optimizing 2.4 GHZ RF …

5

Fig. 5 Gateway and End-user section

Fig. 6 Simulation environment

mechanism tree topology of ZigBee performs exceptionally in street light observation mechanism [15]. Here, Riverbed Modeler is utilized for structuring and usage of the network. The scenario analysis gives the performance of various parameters of QoS. Table1 shows the simulation parameters of Zigbee.

6 Table 1 Simulation parameters

C. L. Narayana et al. Zigbee framework

Specifications

Number of Coordinators

1

Number of routers

5

Number of End devices

4

Power

−0.85

Topology

Tree

Packet size

1024 kbs

Range

3 Results and Discussions We create a Zigbee scenario with routers, end devices, and coordinator with a TREE topology as shown in Fig. 5. The analysis is done for some QOS parameters simulation, i.e., MAC delay, retransmission rate, throughput, traffic sent, traffic received, and data dropped rate. From simulation results TREE topology shows better performance. In this scenario, from Fig. 7 data dropped rate shows good performance. In beginning of transmission the dropped rate is high where in later stage it gradually decreased to zero and the average data rate is 20 bits/sec. In Fig. 8, MAC delay gives a good performance where maximum 0.1 s is the delay in the beginning later stage; there is no delay in transmitting data, and retransmission attempts gave 0.5 packets almost negligible as shown in Fig. 9. Constant throughput Fig. 7 Data Dropped (bits/sec)

Fig. 8 Media Access Delay (sec)

Improvement of QOS by Optimizing 2.4 GHZ RF …

7

Fig. 9 Retransmission Attempts

received entire timeframe with an average of 18,500 bits/sec as shown in Fig. 10. Figures 11 and 12 represent the consistent performance in MAC Management Traffic sent data which is 1200 bits/sec. However improvement is needed in traffic received which shows 4500 bits/sec. Finally Table. 2 shows the performance chart Fig. 10 Throughput

Fig. 11 Management Traffic Rcvd

Fig. 12 Management Traffic Sent

8 Table 2 Performance chart

C. L. Narayana et al. Parameter

Average data

Data Dropped Rate (bits/Sec)

20.5

Throughput (bits/Sec)

18,500

MAC delay ( Seconds)

0.0451

Retransmission Attempts (Packets)

0.5

Traffic Rcvd (bits/Sec)

4500

Traffic Sent (bits/Sec)

1200

of Zigbee simulations.

4 Conclusion Thus the paper gives the proposed methodology of pipeline management using Zigbee communication protocol. Finally, all the results from simulation show the scenario which is optimized for Zigbee communication protocol that gives higher throughput, lesser data dropped, less MAC delay, consistent in management of traffic sent data, and fewer improvements needed in traffic received data. Future work includes LPWAN networks (LoRa, Sigfox, NB-IoT, etc.) for monitoring of long-distance oil pipeline and unmanned aerial vehicles (UAV’s) for oil pipeline inspections.

References 1. Karl, H., Willig, A.: Protocols and Architectures for Wireless Sensor Networks. John Wiley & Sons (2007) 2. Estrin, D., Girod, L., Pottie, G., Srivastava, M.: Acoustics, Speech, and Signal Processing. ICASSP’01. IEEE International Conference on Instrumenting the World with Wireless Sensor Networks, 4, 2033–2036 (2001) 3. Yang, C.T., Chen, S.T., Chang, C.H., Den, W., Wu, C.C.: Implementation of an environmental quality and harmful gases monitoring system in cloud. J. Med. Biol. Eng., 1–14 (2018) 4. Boaz, L., Kaijage, S., Sinde, R.: An overview of pipeline leak detection and location systems. In Proceedings of the 2nd Pan African International Conference on Science, Computing and Telecommunications, Arusha, Tanzania,; IEEE: Piscataway, NJ, USA (2014) 5. Xiao, Q., Li, J., Sun, J., Feng, H., Jin, S.: Natural-gas pipeline leak location using variational mode decomposition analysis and cross-time–frequency spectrum. Measurement 124, 163–172 (2018) 6. Jia, Z., Wang, Z., Sun, W., Li, Z.: Pipeline leakage localization based on distributed FBG hoop strain measurements and support vector machine. Optik 176, 1–13 (2019) 7. Ajao, L.A., Adedokun, E.A., Nwishieyi, C.P., Adegboye, M.A., Agajo, J., Kolo, J.G.: An antitheft oil pipeline vandalism detection: embedded system development. Int. J. Eng. Sci. Appl. 2, 55–64 (2018)

Improvement of QOS by Optimizing 2.4 GHZ RF …

9

8. White, B., Kreuz, T., Simons, S.: Midstream. In: Klaus, B., Rainer, K. (eds.) Compression machinery for oil and gas, pp. 387–400. Gulf Professional Publishing, Houston, TX, USA (2019) 9. Arifin, B., Li, Z., Shah, S.L., Meyer, G.A., Colin, A.: A novel data-driven leak detection and localization algorithm using the Kantorovich distance. Comput. Chem. Eng. 108, 300–313 (2018) 10. Mokhatab, S., Poe, W.A., Mak, J.Y.: Raw gas transmission. In Handbook of Natural Gas Transmission and Processing, 2nd ed, pp. 103–176. Gulf Professional Publishing: Waltham, MA, USA (2012) 11. Bolotina, I., Borikov, V., Ivanova, V., Mertins, K., Uchaikin, S.: Application of phased antenna arrays for pipeline leak detection. J. Pet. Sci. Eng. 161, 497–505 (2018) 12. U.S. Department of Transportation, Pipeline and Hazardous Materials Safety Administration, Final Report No. 12–173, “Leak Detection Study—DTPH56–11-D- 000001”, Dr. David Shaw, Dr.Martin Phillips, Ron Baker, EduardoMunoz, Hamood Rehman, Carol Gibson, Christine Mayernik, December 10 (2012) 13. Chavala Lakshmi Narayana, Dr. Rajesh Singh, Dr. Anita Gehlot.: Design and development of a system to detect third-party damages of an oil pipeline using IoT. Int. J. Adv. Sci. Technol. 29(05), 9676–9682 (2020) 14. Biddut, M.J.H., Islam, N., Karim, M.M., Miah, M.B.A.: An analysis of QoS in ZigBee network based on deviated node priority. J. Elect. Comput. Eng. (2016) 15. Lavric, A., Popa, V., Males, C., Finis, I.: A Performance Study of ZigBee Wireless Sensors Network Topologies for Street Lighting Control Systems. International Conference on Selected Topics in Mobile and Wireless Networking (iCOST), pp. 130–133 (2012)

Transient Stability Enhancement via Sliding Mode Control for Lower Order Triangular System Amrendra Kumar Karn , Salman Hameed, and Mohammad Sarfraz

Abstract Transient steadiness and voltage stability are vital conditions for the solid and secure activity of the power system framework. In the power system framework, the security control plans ought to be hearty concerning deficiency area, various types of flaw and diverse stacking conditions, and the initial condition of all the parameters. Sliding mode control has a wide application range regarding this issue. In this paper, a novel sliding mode control architecture is formulated for a triangular cascaded type system. Usually, the triangular type system with a strict feedback system with cascaded variation topology is solved using an adaptive back steeping control approach. Here in this paper, the proposed architecture of sliding mode control is applied for transient stability enhancement for the SMIB model on excitation side control. Further, the robustness of the proposed control scheme is tested and verified in the condition of different symmetrical asymmetrical faults, etc. at a different location and compared with the traditional AVR–PSS control scheme. The proposed control scheme gives a far better result in all aspects and possesses the property of the ubiquitous controller. Keywords Sliding mode control · Transient stability enhancement · Excitation control · Robustness

1 Introduction There are a lot of papers on transient stability enhancement for SMIB and multimachine system. In various literature works, it is shown that rotor transient stability can be improved in many ways like governor side control, excitation side control, and by incorporating the FACTS devise like SVR, STATCOM, SSSC [1, 3, 4], etc. In a few decades, many authors focused on transient stability enhanced by a multi-side coordinated approach like PSS and excitation side control [11] or by static VAR A. K. Karn (B) · S. Hameed · M. Sarfraz Department of Electrical Engineering, Zakir Hussain College of Engineering, Aligarh Muslim University, 202002 Aligarh, India e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2021 S. Choudhury et al. (eds.), Intelligent Communication, Control and Devices, Advances in Intelligent Systems and Computing 1341, https://doi.org/10.1007/978-981-16-1510-8_2

11

12

A. K. Karn et al.

control or by STATCOM [6, 12]. For a multi-machine system, generally, decentralized control is preferred for better equipment level control and good performance [8, 12, 15]. Sliding mode control has been implemented for a single-machine and multimachine stability enhancement [9]. Many authors used the nonlinear observer before applying back steeping control to make the control robust with the different conditions of faults and power system component state [13]. Nonlinear systems can be linearized by feedback linearization technique by Lie derivate. A suitable Lyapunov energy function basis observes the stability of the nonlinear system. Back stepping is a method created in 1990 to configuration settling controls for an uncommon class of nonlinear dynamical frameworks called severe input frameworks worked from numerous subsystems, as shown in Eqs. (1–3). In this control system, there is a recursive blend philosophy. It intends to build the state criticism control law and the related Lyapunov work at the same time. This control conspire gives a fruitful pathway for a multivariable nonlinear framework with a heartiness’ property by improving the coefficients [2, 4, 7, 8, 10–12]. The difficulty and complexity arising are that the selection or optimization of Lyapunov energy coefficients that come recursively, and excessive parameterization is there [11, 12]. Second, this methodology requires a predefined model that describes the system dynamics well, so indifferent cases condition a single predefined model will give non-satisfactory results. The back steeping problem has been tried to fill in [16] by including the sliding surface error in the Lyapunov energy function. As mentioned above, the problem of transient stability and voltage regulation is solved by many authors by including the FACTS device from various nonlinear control techniques in [23]. Authors apply a new technique in [17] by the principle of immersion and invariance. This control technique is widely used in many kinds of systems for robust adaptive control [18–21]. Robust control of nonlinear systems has always been a curious task for control engineers. The robustness of any control system has a manifold aspect like sensitivity to parameter variation modeling error, an un-modeled parameter of the system, or encountering noise in the system. In this paper, the main two tasks are carried; the first is the development of the new sliding mode control architecture, and the second, its implementation in the transient stability enhancement by excitation side control and its robustness verification. As in Sect. 2 of the paper, this paper’s organization is developing a new sliding mode control architecture for strict feedback triangular system. In Sect. 3, there is the application of the proposed sliding mode control on the SMIB model. Section 4 illustrates the simulation of the SMIB model in different conditions and produces results. Section 5 discusses the results for testing the robustness of the above, and last, Sect. 6 gives the conclusion of this paper.

2 Lower Order Triangular System and Proposed Sliding Mode Control Algorithm The lower order triangular system is represented as follows in equation form

Transient Stability Enhancement via Sliding Mode …

13

x˙ = (x) + ϒ(x)z 1

(1)

z 1 = 1 (x, z 1 ) + ϒ1 (x, z 1 )z 2

(2)

z n = 2 (x, z 1 , . . . ..z n ) + ϒ2 (x, z 1 , . . . . . . z n )u 1

(3)

·

·

Here n = 3, where z 3 = u 1 , (x, z) ∈ R 2n is system state, u ∈ R is control input, and f i , gi smooth double differentiable functions. · Differentiating Eq. (1) and putting the value of z 1 from (2) into (4), we get (6) and (7) ·

·

·

x¨ = (x)x˙ + ϒ(x) z 1 x˙ + ϒ(x) z 1 ·

·

x¨ = (x) + ϒ(x) z 1 + ϒ(x){1 (x, z 1 ) + ϒ1 (x, z 1 )z 2 } ·

(4) (5)

x¨ = x˙ A + ϒ(x) z 1

(6)

  · · where A =  (x) + ϒ(x) z 1

(7)

Again differentiating Eq. (4) and rearranging all terms, we get (9) and (10)     ·· ·· · · · · ... x = x˙  (x)x˙ +ϒ (x) z 1 x˙ + ϒ (x) z 1 + x¨  (x) +ϒ (x) z 1 ·

·

··

+ ϒ(x) z 1 +ϒ(x) z 1

(8)

  ·· ·· · · dϒ1 (x, z 1 ) d f 1 (x, z 1 ) ... x = x˙  (x)x˙ +ϒ (x) z 1 x˙ + ϒ (x) z 1 +ϒ(x) + g(x) z2 + dx dx     · · · d1 (x, z 1 ) dϒ1 (x, z 1 ) ·· · x  (x) +ϒ (x) z 1 + ϒ (x) +ϒ(x) + g(x) z2 z1 + dz 1 dz 1 ·

ϒ(x)ϒ1 (x, z 1 ) z 2

 ϒ · · ·· ... x = x˙ B + x C + z 1 D + z 2 E

(9) (10)

  ·· ·· · · dg1 (x, z 1 ) d f 1 (x, z 1 ) + ϒ(x) z2 where B = f (x)x˙ +ϒ (x) z 1 x˙ + ϒ (x) z 1 +ϒ(x) dx dx   · · C=  (x) +ϒ (x) z 1

14

A. K. Karn et al.

  · d f 1 (x, z 1 ) dϒ1 (x, z 1 ) D = ϒ (x) +ϒ(x) + ϒ(x) z 2 and E = +ϒ(x)ϒ1 (x, z 1 ) dz 1 dz 1 (11) The sliding mode regulator (SMC) plan starts from picking a sliding surface, which characterizes wanted framework elements σ(x) = Sx = 0. SMC intends to drag the state to drive and afterward to stay on the sliding surface. For this current, SMC’s framework direction comprises of two stages: an arriving at the stage by u_eq and a sliding stage by. It is critical to recognize the sliding surface σ(x) = 0, which is a period-free complex in state space, and the exchanging capacity μ(t), which is a period subordinate capacity of the framework state’s position to the sliding surface. Here state factors are taken as follows x˙ = x1

(12)

x1 = x2 = x¨

(13)

·

Since for a system of the kind described by Eqs. (1), (2), and (3), we have to control the only one variable x and the remaining variable z i are intermediate variables. Here we will construct a sliding surface such that it possesses variables given by Eqs. (12) and (13). Now let us consider a sliding surface σ as to x1 + ax2 where a is constantly chosen depending upon space vector trajectory to be converge. Theorem: According to Lyapunov theory of stability for nonlinear system  represented as follows. x˙ = (x) + ϒ(x)u the Lyapunov energy function ϑ > 0, ∀ϑ > 0 for all x ∈ R such that ϑ˙ < 0 then system  is stable at x = 0 infinite time. For stability studies of sliding mode control law, reader may refer to dedicated works of literature. Now constructing a Lyapunov energy function ϑ in terms of sliding surface σ as follows Lyapunov energy 1 2 σ =ϑ 2

(14)

From Lyapunov criteria of stability ⇒ σσ˙ < 0 dϑ dt

So

σ˙ = +k −k

0

(15)

Transient Stability Enhancement via Sliding Mode …

15

Now the sliding surface σ and σ˙ in detail terms from Eqs. (1), (6), and (10)   · σ = x˙ + a x¨ = (x) + ϒ(x)z 1 + a x˙ A + ϒ(x) z 1

(16)

  · · · ... σ˙ = x¨ + a x = x˙ A + ϒ(x) z 1 +a x˙ B + xC ¨ + z1 D + z2 E

(17)

·

·

Here z 2 term contains input manipulated variable now putting the value of z 2 from Eq. (3) σ˙ = x˙ A +

· ϒ(x) z 1

 +a x˙ B + xC ¨ +

· z 1 D +(2 (x, z 1 , z 2 )

 + ϒ2 (x, z 1 , z 2 )u 1 )E (18)

For the sliding mode reaching condition σ˙ = 0, while solving it from Eq. (17), we get ⎡

⎢ ⎣



−

· x˙ A+ϒ(x) z 1 a

·

E

u eq =

⎤

−x˙ B+xC+ ¨ z1 D

⎥ − 2 (x, z 1 , z 2 )⎦

ϒ2 (x, z 1 , z 2 )

(19)

Now for remain staying condition on the sliding surface ⎡⎧

⎫ ⎪ ⎪ ⎪ ⎪ ⎪ E ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ϒ2 (x,z 1 ,z 2 ) ⎪   ⎪ · ⎪ σ < 0 or (x) + ϒ(x)z 1 + a x˙ A + ϒ(x) z 1 < 0 ⎬

⎫ ⎨ k−x˙ A+ϒ(x) z·1 · ⎬ −x˙ B+xC+ ¨ z1 D a ⎭

⎢⎩ ⎢ ⎢ ⎣

u sl = ⎡⎧

⎫ ⎨ −k−x˙ A+ϒ(x) z·1 · ⎬ −x˙ B+xC+ ¨ z1 D a ⎭

⎢⎩ ⎢ ⎢ ⎣

=

⎤

⎥ ⎥ −2 (x,z 1 ,z 2 )⎥ ⎦

⎤

⎪ ⎪ ⎪ ⎪ ⎪ E ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ϒ2 (x,z 1 ,z 2 ) ⎪   ⎪ · ⎪ σ > 0 Or (x) + ϒ(x)z 1 + a x˙ A + ϒ(x) z 1 > 0 ⎭ ⎥ ⎥ −2 (x,z 1 ,z 2 )⎥ ⎦

Now final control u 1 of Eq. (3) for sliding on the surface σ as u t = u eq + u sl Now from Eqs. (18) and (19), finally

(20)

16

A. K. Karn et al.

⎡

⎢ ⎣



−

· x˙ A+ϒ(x) z 1 a

E

ut =

· sign(k)−x˙ A+ϒ(x) z 1 a

⎥ − 2 (x, z 1 , z 2 )⎦

·

⎤

−x˙ B+xC+ ¨ z1 D

E

+ sign(k) = +k where −k

⎤

ϒ2 (x, z 1 , z 2 )

⎡  ⎢ ⎣

·

−x˙ B+xC+ ¨ z1 D

⎥ − 2 (x, z 1 , z 2 )⎦

ϒ2 (x, z 1 , z 2 ) 

(21)

when σ < 0 . when σ > 0

3 SMIB Model and Application of Sliding Mode Algorithm Framework The single-machine infinite bus SMIB system is represented in Fig. 1. All the governing equations of rotor dynamics are represented by the Eqs. (22) to (30), where all the symbols that have the usual meaning are shown in Table 1. δ˙ = ω(t) − ω0 ω˙ =

ω −D ˙ + 0 (Pm − Pe ) (δ) 2H 2H

(23)

Vs E q (t) sin δ(t) xd

(24)

Pe = E q (t) =

xds (xds − xds  ) E q (t) − Vs cos δ(t)  xds xds 

Fig. 1 SMIB model representation

(22)

(25)

Transient Stability Enhancement via Sliding Mode …

17

E f = K c μ f (t) · E q (t) =

(26)

·

  E f (t) E q (t) Vs  +  xd − xd sin δ(t)ω(t) +  Td xd Td

(27) ·

After differentiating (23) and substituting the terms of E q (t) and E q (t) from (25) and (26), we obtained the expression as follows ·

Pe =

1 1  Pe +  v f (t) Td0 Td0

 where Td0 =

 xds Td0 xds

Vs  Vs sin δ(t)[K c μ f (t) + Td0 ((xd − xd ) sin δ(t)ω(t))] xds xd Vs Vs E q (t) cos δ(t)v f (t) = sin δ(t)[K c μ f (t) + xds xds Vs  Vs  E q (t) cos δ(t) + Td0  ((x d − x d ) sin δ(t)ω(t))] + xd xds

(28) (29)

v f (t) =

(30)

Equations (21), (22), and (29) represent the lower order triangular model, as discussed in Sect. 2 of the SMIB system of this paper. Relating the variables of the SMIB model with Eqs. (1), (2), and (3), we get the following functional parameter of the SMIB model. Here the value of a = 0.1 and k = + -1 is taken. −D ω0 ω(t), ϒ1 (x, z 1 ) = − , 2H 2H    1 Vs2 z 2 = Pe 2 (x, z 1 , z 2 ) =  Pe + sin δ(t) xd − xd sin δ(t)ω(t)  Td0 xds xd

(x) = −ω0 , g(x) = 1, z 1 = ω(t), 1 (x, z 1 ) =

+

1 Vs Vs E q (t) cos δ(t), ϒ2 (x, z 1 , z 2 ) =  sin δ(t)K c , u 1 = μ f (t) xds Td0 xds

(31)

Now for obtaining the desired rotor angular stability control as proposed in Sect. 2, the controller coefficients are obtained by substituting the value of the different functions from Eq. (31) in Eq. (11) A = 0, B =

α(t) −D ω0 −D , C = 0, D = ,E = − 2H (−ω0 + ω(t)) 2H 2H

(32)

Hence the sliding surface as in Eq. 16 for transient stability for SMIB is as following   · σ = x˙ + a x¨ = (x) + ϒ(x)z 1 + a x˙ A + ϒ(x) z 1

18

A. K. Karn et al.

= −ω0 + ω(t) + aω(t)

(33)

The final excitation control from Eq. (21) and of (31) μ f (t) is as follows

μ f (t) =

   Vs2 Vs H 1  2 α(t) a D − T  Pe − xds x  sin δ(t) x d − x d sin δ(t)ω(t) − xds E q (t) cos δ(t) d0 d

+

1 Vs sin δ(t)K c  x ds Td0

   Vs2 Vs 1  2 (sigKa−α(t)) H D − T  Pe − xds x  sin δ(t) x d − x d sin δ(t)ω(t) − xds E q (t) cos δ(t) d0 d 1 Vs sin δ(t)K c  x ds Td0

(34)

4 Simulation and Case Study SMIB model is taken here with two parallel transmission lines with their conventional parameters, and their various constant t. The model is simulated undertaken from ref 25 two conditions; first with constant mechanical power input to the turbine and excitation control with an automatic voltage regulator (AVR) and power system stabilizer (PSS) with parameter setting value as in ref [1] for transient stability and second with constant mechanical power proposed sliding mode excitation control under different conditions. The control architecture is schematically represented in Fig. 2. As per the control scheme, there are some online directly measured variables since this control architecture is developed for robust performance. So for testing the robust performance concerning transient stability enhancement, various conditions are applied in simulation condition to the SMIB model in the following manner.

Fig. 2 Control architecture schematic diagram

Transient Stability Enhancement via Sliding Mode …

19

Remark 1: All the manipulated variables that are to be applied for the control framework like α(t) rotor angular acceleration, Pe electromechanical torque exerted by the rotor, and δ(t) the rotor angle are readily accurately available for application. Case studies in different conditions: During the simulation for 5 s three-phase short circuit occurs from time 2 to 2.2 s at the beginning end of transmission line 2, and observations are plotted in Figs. 3, 4, and 5. Again in the next case during the simulation for 5 s single-phase short circuit occurs from time 2 to 2.2 s at the beginning end of transmission line 2, and observations are plotted in Figs. 6 and 7, now again simulation for 5 s. Again transmission line 2 is isolated from time 2 to 2.2 s, and observation is plotted in Figs. 8 and 9. At last, during the simulation for 5 s, three-phase short circuit occurs from time 2 to 2.2 s at the midpoint end of transmission line 2. Observation is plotted in Fig. 10.

Fig. 3 Rotor (torque) angle in three-phase short circuit condition at the start point of line 2

Fig. 4 Generator terminal voltage in three-phase fault condition at the start point of line 2

20

Fig. 5 Excitation during three-phase fault

Fig. 6 Rotor (torque) angle in single-phase short circuit

Fig. 7 Generator terminal voltage in single-phase fault

A. K. Karn et al.

Transient Stability Enhancement via Sliding Mode …

Fig. 8 Rotor (torque) angle in second-line isolation condition

Fig. 9 Generator terminal voltage in second-line isolation

Fig. 10 Rotor (torque) angle in three-phase short circuit condition at midpoint of line 2

21

22

A. K. Karn et al.

5 Discussion The rotor oscillation margin becomes very less in all conditions by observing Figs. 3 to 10. The rotor angle attains a new equilibrium point. The rotor angle smoothly achieves a constant angle under applied sliding mode excitation control than the AVR + PSS control. The terminal voltage is at the generator end during the transition condition is fluctuation under the lower limit in the sliding mode control excitation control than the AVR + PSS condition.

6 Conclusion The transient stability enhancement for SMIB is executed from the excitation side in this paper. The model taken for SMIB in this paper follows the lower triangular matrix topology. The novel control algorithm is developed on the basic philosophy of sliding mode control that is robust concerning the system parameter. The developed novel control framework is applied in simulation mode for the determination of its feasibility and viability. Simulation is performed under the condition of three phases, single-phase short circuit, and line outage also. As transient stability is not considered in the condition of symmetrical and asymmetrical fault in earlier reference literature and cannot be performed by conventional backstopping and h∞ techniques, this paper also considers these cases. The observation found is compared with the conventional excitation control by AVR + PSS arrangement with stander parameter settings as in ref [1] in different conditions. The control algorithm proposed in this paper gives better performance in all situations.

References 1. Rahman Khezri: Hassan Bevrani AVR and PSS coordinated based fuzzy approach for transient stability enhancemen. 23rd Iranian Conference on Electrical Engineering, pp. 1659–1664 (2015) 2. Karimi, A., Feliachi, A.: Decentralized adaptive backstepping control of electric power systems. Electr. Power Syst. Res. 78(3), 484–493 (2008) 3. Sirine Essallah, Adel Bouallegue, Adel Khedher: Integration of automatic voltage regulator and power system stabilizer: small-signal stability in DFIG-based wind farms. J. Modern Power Syst. Clean Energy (2019) 4. Salma Keskes, Nouha Bouchiba, Souhir Sallem, Larbi Chrifi-Alaoui, MBA Kammoun: Transient stability enhancement and voltage regulation in SMIB power system using SVC with PI controller. 6th International Conference on Systems and Control (2017) 5. Zhang, C. et al.: An advanced adaptive backstepping control method for STATCOM. 26th Chinese Control Decis. Conf. CCDC 2014, pp. 1822–1827 (2014) 6. Lei, B., Fei, S., Zhai, J., Wu, X.: Nonlinear adaptive dynamic surface control of static var compensator for improving power system transient stability. Chinese Control Conf. CCC, pp. 343–347 (2013)

Transient Stability Enhancement via Sliding Mode …

23

7. Karimi, A., Feliachi, A.: Decentralized extended-backstepping control of power systems. 2006 IEEE Power Eng. Soc. Gen. Meet. PES (2006) 8. Huerta, H., Loukianov, A.G., Cañedo, J.M.: Multimachine power-system control: integral-SM approach. IEEE Trans. Ind. Electron. 56(6), 2229–2236 (2009) 9. Dou, C.X., Yang, J., Li, X., Gui, T., Bi, Y.: Decentralized coordinated control for a large power system based on transient stability assessment. Int. J. Electr. Power Energy Syst. 46(1), 153–162 (2013) 10. Sun, L.Y., Liu, Y.: Nonlinear adaptive back stepping controller design for static VAR compensator. Chinese Control and Decision Conference, Xuzhou, China, pp. 3013–3018 (2010) 11. Lee, S.S., Li, S.Y., Park, J.K.: Nonlinear adaptive back stepping controller design for power system stabilizer in multi-machine power systems. 2008 American Control Conference, Washington, USA, pp. 2504–2509 (2008, June) 12. Dib, W., Kenné, G., Lamnabhi-Lagarrigue, F.: An application of immersion and invariance to transient stability and voltage regulation of power systems with unknown mechanical power. Proc. IEEE Conf. Decis. Control, pp. 7837–7842 (2009) 13. Bouzid, Y., Siguerdidjane, H., Bestaoui, Y.: Hierarchical autopilot design based on immersion & invariance and nonlinear internal model tracking controllers for autonomous system. IFACPapersOnLine 49(5), 103–108 (2016) 14. Fidan, B., Bayezit, I.: Immersion and invariance based adaptive control of three dimensional autonomous vehicle maneuvers. IFAC Proc. 1(PART 1), 89–93 (2010) 15. Mattioni, M., Monaco, S., Normand-Cyrot, D.: Digital stabilization of strict feedback dynamics through immersion and invariance. IFAC-PapersOnLine 48(11), 1074–1079 (2015) 16. Hristea, E., Siguerdidjane, H.: Stabilization of a magnetic suspension by Immersion and Invariance and experimental robustness study, 44, no. 1 PART 1. IFAC (2011) 17. Wang, L., Kellett, C.M.: Adaptive nonlinear regulator design via immersion and invariance. IFAC-PapersOnLine 52(16), 592–597 (2019) 18. Barosy, S., Ili´c, M.D.: On transient stability and voltage regulation through MIMO feedback linearizing control of generator and energy storage. Proc. Am. Control Conf. 2015-July, pp. 4320–4326 (2015) 19. Akhrif, O.: Application of a multivariable feedback linearization scheme for rotor angle stability and voltage regulation of power systems. IEEE Trans. Power Syst. 14(2), 620–628 (1999) 20. Kanchanaharuthai, A., Chankong, V., Member, S., Loparo, K.A.: Transient stability and voltage regulation in multimachine power systems Vis- à -Vis STATCOM and battery energy storage. IEEE Trans. Power Syst., pp. 1–13 (2014) 21. Zhao, P., Yao, W., Wen, J., Jiang, L., Wang, S., Cheng, S.: Improved synergetic excitation control for transient stability enhancement and voltage regulation of power systems. Int. J. Electr. Power Energy Syst. 68, 44–51 (2015) 22. Keskes, S., Bouchiba, N., Sallem, S., Chrifi-Alaoui, L., Kammoun, M.: Transient stability enhancement and voltage regulation in SMIB power system using SVC with PI controller. 2017 6th Int. Conf. Syst. Control. ICSC 2017, pp. 115–120 (2017) 23. Liu, H., Hu, Z., Song, Y.: Lyapunov-based decentralized excitation control for global asymptotic stability and voltage regulation of multi-machine power systems. IEEE Trans. Power Syst. 27(4), 2262–2270 (2012) 24. Cong, L., Wang, Y.: Coordinated control of generator excitation and STATCOM for rotor angle stability and voltage regulation enhancement of power systems. IEE Proc. Gener. Transm. Distrib. 149(6), 659–666 (2002) 25. Cong, L., Wang, Y., Hill, D.J.: Coordinated control design of generator excitation and SVC for transient stability and voltage regulation enhancement of multi-machine power systems. Int. J. Robust Nonlinear Control 14(910), 789–805 (2004) 26. Cong, L., Wang, Y., Hill, D.J.: Coordinated control design of generator excitation and SVC for transient stability and voltage regulation enhancement of multi-machine power systems. Int. J. Robust Nonlinear Control 14(910), 789–805 (2004) 27. Wang, K., Ouyang, Z., Krishnan, R., Shu, L., He, L.: A game theory-based energy management system using price elasticity for smart grids. IEEE Trans. Ind. Informatics 11(6), 1607–1616 (2015)

A Convolution Neural Networks and IoT-Based Approach to Surveillance System Shreyash Gondane, Adesh Thakare, Chaitanya Deshpande, and Omprakash Gupta

Abstract Recognition of faces is of great importance for applications in the real world such as securities, human–computer interfaces, and traffic. As opposed to conventional machine learning methods, neural network models have also demonstrated improved outcomes in the case of processing precision in face detection. Surveillance system provides a close examination of a region, or the suspect in specific. The paper implements a novel system with the Internet of Things (IoT) capabilities for real-time surveillance using Haar-cascade detection integrated with a convolutional neural network for recognizing facial attributes to identify and allow access. The system presents 96.3% accuracy with real-time response through SMS or mail. Keywords Surveillance systems · Image processing · Machine learning · Face recognition · Haar Cascade · Convolution neural networks · Object detection

S. Gondane (B) · O. Gupta Department of Electronics & Telecommunication, Vishwakarma Institute of Technology, Pune, India e-mail: [email protected] O. Gupta e-mail: [email protected] A. Thakare · C. Deshpande Department of Computer Engineering, Vishwakarma Institute of Technology, Pune, India e-mail: [email protected] C. Deshpande e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2021 S. Choudhury et al. (eds.), Intelligent Communication, Control and Devices, Advances in Intelligent Systems and Computing 1341, https://doi.org/10.1007/978-981-16-1510-8_3

25

26

S. Gondane et al.

1 Introduction Intelligent surveillance systems have become more relevant and commonly used in these times as the growing demand for safety and protection in many outdoor areas including automotive applications, crowded public areas, and human activity surveillance. The growing need for security has underscored the use of smart technology in monitoring systems. There are different techniques commonly, ranging from satellite imagery to CCTV, used for tracking and location estimation. Although satellite imaging is often used by the military industry and is costly and used on a broader scale, CCTV covers restricted areas and is comparatively cheaper, but requires continuous surveillance by additional workers to detect suspicious activity [1]. The amount of data collected daily for video surveillance is increasing exponentially. Scalability and usability are becoming very important when dealing with this huge amount of knowledge. A major driving force in this area has been the availability and cost of high-resolution. surveillance cameras, along with the increasing need for remote-controlled protection. A rising volume of information raises the cost for processing and tagging this information for quick retrieval afterward [2]. To meet this demand, many researchers are working on finding enhancements and better approaches in video analytics, which is video data semantic analysis, to minimize running time and overall monitoring system costs. Only essential data are collected from the available video feeds in smart surveillance systems [3] and passed on for further analysis. This can save both time and storage space and allow video data retrieval easier, as the use of smart tags requires considerably less scanning through. Basically, a surveillance system is used against criminals in order to prevent a crime. The image capturing from cameras with video analysis is the most important part of surveillance systems because we can train those cameras through various algorithms based on embedded systems [4]. In addition, using the concept of the Internet of Things in surveillance systems will make the world safer and better, as we can access specific data or information even though we are somewhere else. One of the aims of developed surveillance applications is to automatically identify people for safety purposes. Facial recognition systems based on face image analysis may provide an efficient solution for improving monitoring and control capabilities of remote human operations without requiring the participants to indulge or have expertise [5]. Some of the difficult problems in image processing is the identification of faces from a video sequence in an uncontrolled environment. Therefore, the development of a Convolutional Neural Networks-based intelligent video surveillance system is to analyze dynamic system, face tracking, image modeling, and similar techniques [6]. Within this study, the latest surveillance algorithms for tracking a person which include the use of face recognition image processing techniques often involve the use of IoT where the camera is used to capture people and the image is sent to the owner via email. The effect is facial recognition to track a person’s entry according to his/her attribute and the system lets the owner access or deny entry. These will also record a person’s data while entering the frame and leaving the frame that

A Convolution Neural Networks and IoT-Based …

27

can be used for security purposes. In this paper, we are going to build a surveillance system integrating concepts of the Internet of Things with multiple image processing algorithms and convolutional neural networks for facial recognition. The paper is divided into three sections. Section 2 explains in details the proposed model, its workflow, the algorithm used for each sub-system, and implementation of the Internet of Things. Section 3 emphasizes the system implementation in real world and the results inferred from the model.

2 Proposed Model The proposed surveillance system incorporated the fundamental concepts of image processing and the Internet of Things to build a robust and reliable model with optimum effectiveness. The model performs under the flow of various algorithms as given in Fig. 1 which are explained in more detail below. Pre-processing: The current frame picture was applied after computer recording with color channel switching and the Gaussian blur algorithm as Eq. 1.

G(x, y) =

1

√ e− σ 2π

x2+y2 2σ2

(1)

Here x gives the measure of distance in x-axis from the origin, y gives the measure of distance in y-axis from the origin, and σ gives the standard deviation of the gaussian distribution. The frame difference between the source frame and the current frame is then determined using the cv2.absdiff function. The cv2.absdiff aids in calculating the absolute difference in the two arrays of images between the pixels. By doing that, we’ll be able to capture only the pixels of the moving objects. We’ll need to transform our pictures and grayscale to use cv2.absdiff. The dilation of threshold images was

Fig. 1 System flowchart

28

S. Gondane et al.

detected and processed in a simple sequence, depending on the frame difference. The binary threshold is a method used here to reduce noise as the simplest method. After that, the functions of cv2.findContours run to output those separate shapes. Face Detection: Viola–Jones algorithm is a helpful tool for facial detection [7]. This algorithm is generally not only limited for facial detection but can also be used for many rigid structured object detection tasks. Until the discovery and implementation of Haar Cascade, other models and artifacts occurred that suited algorithms with exceptionally high precision, such as transforming the scale-invariant functions. Such algorithms show a strong performance but due to their long processing periods, they cannot be extended to real-time detection. Whereas, the Haar cascade is a machine learning-based technique, where multiple positive and negative images train a cascade function. For other pictures, it is also used to track artifacts. The algorithm consists of four stages: collection of hair characteristics, development of essential pictures, training with AdaBoost, and classification of cascades. Haar-like features: Haar-like features are a set of two-dimensional (2D) hair functions which encode objects’ appearance. We include multiple rectangular areas embedded in the template. As in the following equation, the function value f of a Haar-like system that has k rectangles is obtained:

f=

Σk w(i).μ(i) i=1

(2)

where μ(i) gives the mean pixel intensity of the ih rectangle. The rectangular mean is given by μ. Integral Image: The integral image is given by.

Σ i(xt , yt ) ii(x, y) = xt ,yt ≤x,y

(3)

where i(x, y) gives gray scale intensity at pixel (x, y). Using this for any rectangular area ABCD, the summation of intensity pixels can be computed using (x, y)



s ABC D i(x, y) = ii(D) + ii(A) − ii(B) − ii(C)

(4)

This computational benefit has allowed multi-scale detection functionality to be scaled at no extra expense, because it needs the same set of activities given the complexity (Fig. 2).

A Convolution Neural Networks and IoT-Based …

29

Fig. 2 Haar-like features calculation

Adaboost: The Adaboost is used for both feature selection and classifier training. Here the slow learner is programmed to pick the function that better distinguishes the weighted negatives and positives. A weak classifier h(x, f, p, θ) is defined as.

h(x, f, p, θ) =

1 if pf (x) < pθ . 0 otherwise

(5)

where f is a function of the vast collection with various measurements of the Haar-like features, p gives the polarity; where f is the threshold and x is a size 24 × 24 pixel training subframe. Throughout the machine learning literature, such poor classifiers that classify specific features may be regarded as decision stups. Face recognition: Traditional neural networks use a completely connected design in which each neuron in one layer binds to all the neurons in the next layer. A fully connected architecture is ineffective when it comes to the image data processing as: – A conventional neural network can produce multiple factors for any ordinary picture with a large number of pixels which can contribute to overfitting. – The model will be rather heavy on computing. – It can be difficult to improve its performance by analyzing data and evaluating and optimizing the design. The neurons in one level do not connect to all the neurons in the next layer of a Convolutional Neural Network (CNN) as opposed to a fully connected neural network. A convolutional neural network, however, employs a three-dimensional structure where a certain region or image attribute is explored by each set of neurons. lSimilar relationships are filtered by CNNs, thus yielding the test phase low computationally achievable. One of the standard CNN architectures has been shown in Fig. 3. CNN’s framework comprises convolutional, shared, rectified linear units and completely linked layers. The convolution neural network provides the number of encoded face which provides a description of the face to differentiate the details from

30

S. Gondane et al.

Fig. 3 Traditional Neural Network (left) versus Convolutional Neural Network (right) architecture

individuals. When we move two separate photos of the same person, it should display identical outputs for both images, while the network will display quite different outputs for the two images when you pass images of two different people. It demonstrates that it is important to train the neural network to automatically recognize various features of faces and to determine numbers based on that. Padding: Sometimes we apply padding to the image to take into consideration the pixels on the sides. As a standard, we pad with zeros and denote the padding parameter with p, which describes the number of elements placed on each of the image’s four sides. Stride: The stride is the move that the convolutionary product takes. A big phase causes the performance to reduce, and vice-versa. The stride parameter is denoted by s. Convolution: If we’ve described the stride and the padding, we can describe the result of the convolution between a tensor and a filter. We may now formally describe the convolution product on a volume after having previously defined the convolution product on a 2D matrix which is the sum of the element-wise component. In general, an image can be interpreted mathematically as a tensor of the following dimensions: dim(image) = (n H, nW, nC)

(6)

where nH denotes the size of height, nW is the size of width, and nC is the number of channels. Pooling: It is the step of downsampling characteristics of the picture by summing up the detail. The procedure is done via each channel and thus it influences just the measurements (nH, nW) and leaves nC unchanged. We implement a pooling after performing a sequence of convolutions accompanied by activation functions and repeat this phase for a certain period of time. These operations enable the extraction of image features that will be fed to a neural network identified by the completely linked layers that are also routinely followed by the activation functions. The key concept is to reduce nH and nW and raise nC while running across the network more thoroughly.

A Convolution Neural Networks and IoT-Based …

31

IoT SMS and Mail: The IoT part of the project includes sending alert SMS and sending a captured image to the monitoring person which will find out whether to allow access to that unknown person or not. This thing can be done in Open CV using the Twilio server platform. It allows us to receive incoming messages which relate interactivity between user. We can even send the SMS by detecting if any person in the database has a criminal record. For sending an outgoing message, the messages list resource URI receives an HTTP POST.

3 System Implementation and Results The database considered makes use of multiple angles of the individual’s face for better training in the facial recognition system as shown in Fig. 4. The cameras are mounted from the top view for security and monitoring in such a way it covers all the area. The image is being captured using the camera with real-time data given to the images from the database, and then analyzed the data. The importance of this research is to detect a person and then identify with the concurrent database for a match. The images captured detects a person using a bounding box as shown in Fig. 5 for further processing to detect the facial attributes with the database which is used by the surveillance system. The outputs of the bounding box processing have been shown in Fig. 6a & b along with the image convolution matrix in Fig. 6c. In this project, we got the desired output of face recognition with an accuracy of 96.3% from the training set of the dataset. Accuracy can be further improved by k-crossvalidation using around 2000–2500 images as a testing set. Current settings for this project include the data of 1500 images as a training set and 750 images as a testing set resulting in an accuracy of recognizing images with 96.3%. In detail, if a person comes to the front of the door, the camera captures the image, object detection takes place where we come to know the arrival and leaving time of a particular person as shown in Fig. 7b. Parallely, face recognition also takes place with the captured image. If the face is matched with our database, then captured

Fig. 4 CNN architecture for feature learning and classification

32

Fig. 5 Dataset

Fig. 6 Bounding box for object detection

Fig. 7 a Delta output b Threshold output c Convolution output image matrix

S. Gondane et al.

A Convolution Neural Networks and IoT-Based …

33

Fig. 8 a Face recognition output b Surveillance details

image is sent to the owner’s mail and SMS is also sent to the owner as “person is known” and particular information is given to the owner via SMS. If the face is not matched with our database that means the face is unknown as shown in Fig. 7a. In this case, the image is sent to the owner and particularly message is also transferred via SMS as “person is unknown” and the owner can take particular decision to open the door or not (Fig. 8).

4 Conclusion The surveillance framework is an appropriate tool to track criminal activity. Traditional systems are effective and have little maintenance but the energy usage is more as the machine is constantly switched on. The proposed smart monitoring system based on IoT offers energy conservation by turning on the power of the machine, based on the event of a particular entity. The system will allow the specific person’s entries and for unknown entities, the system will take decision according to the database. Depending on the decision, the intruder will be captured, a notification on the owner’s smartphone via SMS or mail if the intruder is unidentified through the model proposed.

34

S. Gondane et al.

References 1. Zhang, T., Liu, S., Xu, C., Lu, H.: Mining semantic context information for intelligent video surveillance of traffic scene. IEEE Trans. Ind. Inf. 9(1), 149–160 (2013) 2. Mishra, P.K., Saroha, G.P.: A study on video surveillance system for object detection and tracking. 2016 3rd International Conference on Computing for Sustainable Global Development (INDIACom), New Delhi, pp. 221–226 (2016) 3. Papageorgiou, C.P., Oren, M., Poggio, T.: A general framework for object detection, in: ICCV ’98: Proceedings of the International Conference on Computer Vision, Washington, DC, USA, pp. 555–562 (1998) 4. Cuimei, L., Zhiliang, Q., Nan, J., Jianhua, W.: Human face detection algorithm via haar cascade classifier combined with three additional classifiers. In Proceedings of the IEEE 13th International Conference on Electronic Measurement and Instruments, Yangzhou, China, pp. 483–487 (2017, 20–22 October) 5. Yang, M., Yu, K.: Real-time clothing recognition in surveillance videos. In Proceedings of the 18th IEEE International Conference on Image Processing, Brussels, Belgium, pp. 2937–2940 (2011, 11–14 September) 6. Hu, G., et al.: When face recognition meets with deep learning: an evaluation of convolutional neural networks for face recognition. 2015 IEEE Int. Conf. Com- put. Vis. Work., pp. 384–392 (2015) 7. Lienhart, R., Maydt, J.: An extended set of Haar-like features for rapid object detection. In ICIP ’02: Proceedings of the International Conference on Image Processing, pp. 900–903 (2002)

Prevent Road Accident Using Intelligent Device Amrit Suman, Chiranjeev Kumar, and Preetam Suman

Abstract Transport is one of the pillars of any country’s economy: every year, government and car/vehicle manufacturers spend a lot of money on road safety. But unfortunately, crashes on roads cause various problems for the government. The main factors that cause fatal crashes, grievous injuries, and deaths are of varied nature. The type of roads, the structure of ways, weather conditions, and the driver’s behavior are a few pertinent factors studied, and the relationship is identified. This paper describes a device that is capable of locating crashes early. If a collision occurs, the device can recognize and send information to the nearest hospital and police station. There are two acoustic signals analyzed, vehicle crash sound and human distress call (generated during crashes/accidents). For recognition of both sounds, this paper describes an algorithm based on acoustic signal processing. The algorithm was tested in the lab. It was found robust, with an efficiency of 94.7% in detecting a collision. Keywords Fatal crashes · Grievous Injury · Road structure · Road type · Location of impacts

1 Introduction An incident that occurs in a way that is open for public movements is known as a Road Traffic Accident. Generally, one or more persons and a moving vehicle are involved, and a minimum of one person being injured or killed in this incident. An incident can happen between cars and vehicles, vehicles and pedestrians, vehicles A. Suman (B) · C. Kumar Computer Science & Engineering Department, IIT (ISM), Dhanbad, India e-mail: [email protected] C. Kumar e-mail: [email protected] P. Suman Information Technology, Jaipuria Institute of Management, Lucknow, India e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2021 S. Choudhury et al. (eds.), Intelligent Communication, Control and Devices, Advances in Intelligent Systems and Computing 1341, https://doi.org/10.1007/978-981-16-1510-8_4

35

36

A. Suman et al.

and animals, and vehicles and geographical or architectural obstacles. Road traffic accidents are a human tragedy. Premature deaths, injuries, loss of productivity, etc. are the results of vehicles accident. World Health Organization (WHO) reports that the road accident is the primary cause of death, and about 1.3 million people meet to fatalities, and about 20–50 million are injured every year due to crashes on the road. Road accidents are a significant reason for death among people. And between the 5–29 years of children and young people are dying every year due to the road accident. Road crashes have become a public issue, and mortalities are from every segment of society. These fatalities were caused by several factors involving the designing of roads, road users’ behavior, and drivers of vehicles, so road safety demands an all-inclusive approach to identify solutions to prevent the cause and fatalities. Various analysts and researchers examined a lot of data for road safety issues. According to Geethu et al. [1], the number of motor vehicles on the road increases daily and a significant cause of accidents, congestions, delays, etc. on the road traffic. Urban areas of developing countries are more affected due to traffic congestions. The authors provide data on vehicle characteristics in different scenarios. The authors also provide data analysis of pedestrian traffic at intersections and junctions. Harnam et al. [2] observed that below the age of 16 get involved in fatal road accidents with the percentage of 59, and the male candidate percentage is 83.1. According to the authors, more than half of road accidents occurred in the winter season between 12–4 PM. And most of the victims are pedestrians with 61% and then cyclist with 13.6%. Sanjay et al. [3] provide an analysis of road accidents in Patna city. The city roads are congested and encroached by other activities. According to the authors, every year, the number of road accidents increases in the city area. Forty-five people were killed per 100 casualties in the year 2000. The percentage of deaths of pedestrians is exceptionally high. A new bypass road on the national highway is the most prone location in the city. In Thokchom et al. [4], according to the author, every day, nearly 1.3 million lost their lives every year, and 20–50 million people injured at the same time. Authors say that by the year 2030, road accidents will be the top 5 causes of death. The methodology in the study is based on secondary data collected from the Road safety cell, Transport department, Uttar Pradesh, India, for the year 2012–2018. The findings indicate that there is a significant relationship observed between the type of crashes and a particular factor.

1.1 Road Accidental Deaths and Injuries in India There has been an alarming situation in accidental deaths on Indian roads over the years. It can be observed that the highest rates of fatal crashes are 19,364 in the year 2018. And the lowest rates of fatal crashes are 13,293 in the year 2012, which shows that the deadly crashes increase year by year. The high degree of grievous injury in

Prevent Road Accident Using Intelligent Device

37

2018 is 14685 and in 2012 is 11497, which reflects that the unfortunate injury also increases year wise. The total numbers of deaths due to collisions are the highest in the year 2018 which is 22256 and in 2012 which is 16149. The minimum number of fatal crashes, grievous, and deaths due to collisions are registered in 2012 year. But the number of vehicles is lower than in the year 2018. Due to the increasing number of cars, traffic will increase year by year. It can be observed the high rate of fatal crashes is in National Highways (38,613) and state Highways (36,474), whereas the minimum number of fatal crashes occurred in expressways (3713). A similar pattern is observed in the case of grievous injury and deaths due to collisions. The severe injury cases are in state highways (32,567) and national highways (29,980) and the least in expressways (3573). The deaths due to crashes are the highest in National Highways (42,483) and State Highways (43,653), and it is the least in expressways (5227). The new vehicles (age less than five years) have high chances of collisions compared to older vehicles. This may be due to modern vehicles being able to get up and go fast. Almost in every parameter, the speed is having a direct or indirect relationship with the type of crashes. In sunny and clear weather, the maximum number of fatal crashes occurred. The number is 39493 and followed by the rainy season (16,534) and foggy and Misty (13,658). In the case of grievous crashes vs. weather conditions, a similar pattern is observed where the maximum incidence happened during sunny and bright weather (34,353) followed by rainy season (15,008) and foggy and misty weather conditions (11,260). It can be observed that the maximum numbers of deaths due to crashes are in clear weather (44,908) as compared to rainy (21,020) and foggy weather (16,298).

2 Proposed Solution 2.1 Intelligent Device An intelligent device that can detect the accident is proposed, and this device can be mounted inside the vehicle. As shown in Fig. 1, the device can be mounted near steering through the OBD [5] port of the vehicle using CAN bus [6]. The OBD can also provide the status of the engine and other peripheral readings. The figure also shows a wireless sensor (connected to the device through Bluetooth) on wheels to measure the tire’s air pressure and temperature. The device will be equipped with various elements, as shown in Fig. 2. These elements are the following: Microphone sensor [7]: This sensor will sense acoustic signals in the environment; it will record the acoustic signal of the engine and human distress call (during mishappening). Accelerometer [8]: This sensor will record the movement and direction of the vehicle. It will send recorded data to the microcontroller.

38

A. Suman et al.

Fig. 1 Hardware placement

Fig. 2 Hardware device architecture

Temperature sensor [9]: It will record the temperature of the environment and the engine’s temperature. It will be helpful to identify if the engine is getting overheated or the vehicle got fire. Gas sensor [10]: It can sense the presence of methane gas inside the vehicle. Signal conditioning: It will amplify the signal and also filter the noise in the movement. GPS [11]: It will record the location of the vehicle. Bluetooth: It will provide connectivity of devices to sensors. OBD port [5]: It will provide connectivity of the device to the vehicle. GSM module [11]: It will help store all the readings on cloud storage and send messages during crashes. Microcontroller: It is the heart of the device and takes all the decisions based on all the sensors’ input. In this device, the Cortex-M ARM microcontroller is used. Buffer: Buffers were used to store the digital signal. In this architecture, two buffers were used. The concept of using two buffers is to capture and process the

Prevent Road Accident Using Intelligent Device

39

signal without any loss of signal. One buffer will record the signal, and another buffer will process the signal on the processor simultaneously. The processor unit processes the stored signal through algorithms to recognize the event. The 8-bit low-power ATmega328P processor [12] has been used for decision-making. Figure 4 shows the connected components, and these are Arduino, GSM 800 module, GPS module, microphone breakout board, temperature module, and accelerometer. Sensors and microcontrollers were connected, and each module works properly. Many other sensors can be added to the device for various readings.

2.2 Mobile Application A mobile application has been developed to manage the device. It keeps all the records stored by devices through sensors. There are various features provided by the app like display the exact location of the crash or the exact location of the vehicle in the case of the missing vehicle and in the case of any unwanted happenings that occur like a hijacking. There are various activities proposed by the application given in the following: i. ii. iii. iv. v.

Identify the event message to be transmitted. Send the exact location, along with the type of event (collision or fire). Automatic call generation into the nearest police station, hospital, and ambulance. Inform you about the status of vehicles peripheral. Keep track of information for the last seven days.

Figure 3 shows screenshots of the developed application for android. There are six options available in APP: nearest police station, emergency contact, services, operations, vehicle locations, and about.

3 Implementation of the Algorithm on Microcontroller This section describes the algorithms that were implemented on microcontroller for acoustic signal recognition. The sound of the vehicle collision and human distress call may be the same so that the intelligent device can not recognize the actual accident. For analysis and training of intelligent devices, collected 800 signals and 500 signals of a vehicle collision and human distress call and analyzed in MATLAB. These signals were analyzed in the time and frequency domain. Figure 4 shows the spectrogram of a vehicle collision with other vehicles. Here xaxis shows the time domain, and the y-axis shows the signal strength. Results reflect that there is a sudden rise in signal when vehicles collide with each other and after that, the signal goes poor.

40

Fig. 3 Mobile application module

Fig. 4 spectrogram of the signal for a collision of the car

Fig. 5 spectrogram of wheels of car skid and collision.

A. Suman et al.

Prevent Road Accident Using Intelligent Device

41

Figure 5 reflects the spectrogram of the collision of a car after a skid. The x-axis reflects the time domain and the y-axis represents the strength of the signal. A sudden rise in signal is observed when car skids and after the collision of vehicles. The signal goes poor after a rise in both cases. According to Fig. 6 spectrogram, a car is about to collide after the skid. The xaxis reflects the time domain, and the y-axis reflects the signal strength. After the collision, a sudden rise in the signal can be observed after this signal goes poor. Figure 7 shows the spectrogram of a vehicle collision on a small level. Here the x-axis is time, and the y-axis is signal strength. A sudden rise in the signal can be observed at the time of the collision. Figure 8 shows the spectrogram of the car-braking skid. Here the x-axis is time, and the y-axis is signal strength. When brake skids, a constant rise in signals can be observed. After analyzing various signals and the literature, 10 statistical parameters were identified that are very helpful to recognize the vehicle accident [13, 14]. The parameters are the Minimum peak of signal, Maximum height of signal, Zero Crossing Rate (ZCR), Root mean square (RMS), Standard Deviation, Spectral Centroid, Spectral Rolloff, Spectral Flux, Spectral Crest Factor, and Spectral Entropy. The implementation of the recognition algorithm above-mentioned features was used and stored in the master database for each signal. The early implementation and testing were done in MATLAB. The flow of the process is shown in Fig. 9. The algorithm was

Fig. 6 Spectrogram of the car skidding about to crash

Fig. 7 Spectrogram of signal for a small collision of the car

42

A. Suman et al.

Fig. 8 Spectrogram of signal for car-braking skid

implemented on the microcontroller after testing on MATLAB. The pseudo-code is the following.

4 Results The algorithm is designed for use in a real environment to save a life. Therefore, that algorithm was tested in a regress manner. To create a realistic setting, the speaker produced recorded acoustic sounds of accidents and distress calls in the lab. There were three combinations of datasets created for training and testing. Both the training and testing datasets were exclusive in all three cases. The results noted for all the three datasets for vehicle crash sound and distress calls are in Table 1. The intelligent device was also tested for each event to transmit the message. During experimentation, it was found that the device works well and can be implemented in vehicles.

5 Conclusion As the population is increasing, the number of vehicles is also growing, which causes a lot of accidents. There is a lot of infrastructure and wealth loss in every severe accident. As per the observation and interpretations of graphs, road crashes, deaths, and injuries in Uttar Pradesh, India vary according to the time segments, age of the vehicle, types of roads, and weather conditions. An intelligent device has been presented to detect accidents. This smart device uses acoustic signal processing. Vehicle collision and distress calls are the two types of acoustic signals collected and analyzed for crash recognition. Results show that the implemented algorithm is very efficient, more robust, and reliable. The intelligent device was tested, and it can recognize the crash. It also transmits the message to the nearest police station and hospital after identifying the vehicle’s collision.

Prevent Road Accident Using Intelligent Device

43

Fig. 9 Flow diagram of the process Table 1 Efficiency of the algorithm for dataset 1

Class

Efficiencydataset 1

Efficiencydataset 2

Efficiencydataset 3

Crash sound

94.7

93.6

92.5

Distress call

88.7

91.2

89.8

44

A. Suman et al.

References 1. Geethu Lal et al.: Sustainable traffic improvement for urban road intersections of developing countries: a case study of Ettumanoor, India. Sci. Direct, Procedia Technol. 25, 115–121 (2016) 2. Harnam Singh et al.: Fatal road traffic accidents among young children. J. Indian Acad Forensic Med. 32(4) 3. Sanjay et. al.: Road accident analysis: a case study of Patna city. Publ. Urban Trans. J. 2(2), 60–75. 4. Thokchom et. al.: Road traffic accidents in India: an overview. Int. J. Clin. Biomed. Res. Sumathi Publications © (2018) 5. Eric Walter, Richard Walter: On-Board Diagnostics (OBD) Background and Standards. In Data Acquisition from Light-Duty Vehicles Using OBD and CAN, SAE, pp. 25–37 (2018) 6. Eric Walter, Richard Walter: Data acquisition from light-duty vehicles using OBD and CAN. In Data Acquisition from Light-Duty Vehicles Using OBD and CAN, SAE, pp. i-xv (2018) 7. https://www.rhydolabz.com/sensors-other-sensors-c-137_148/sound-detectionsensor-mod ule-p-2306.html. 8. https://www.rhydolabz.com/sensors-accelerometers-c-137_138/triple-axisaccelerometer-mod ulemma7341-p-1078.html. 9. https://www.rhydolabz.com/sensors-weather-sensors-c-137_147/dht11temperaturehumidity-sensor-module-p-2044.html. 10. https://www.rhydolabz.com/sensors-gas-sensors-c137_140/lpg-gas-sensormodule-v2-p1104.html. 11. https://www.rhydolabz.com/wireless-gsm-gprs-c- 130_185/gsmgprs-ttl-uartmodemsim900-p1076.html 12. https://www.microchip.com/wwwproducts/en/ATmega328P 13. Preetam Suman, Ashwaray Raj, Pramila Choudhary, M. Radhakrishna: Detection of soil digging using pattern recognition of acoustic signals. Advances in the field of health and Environment Safety, ISBN: 9811071225, pp. 35–45, Springer (2017) 14. Sheikh Fahad Ahmad, Deepak Kumar Singh: Automatic detection of tree cutting in forests using acoustic properties. J. King Saud Uni. Comp. Inf. Sci., ISSN 1319–1578 (2019)

Choice of Suitable Height of Substrate in the Designing of Microstrip Antenna for X-band Communication R. K. Chaurasia and S. K. Nippani

Abstract The current research aims in developing a square-shaped microstrip antenna, and the focus is on the role of the height of the substrate on the performance. In this work, both substrate and patch dimension have been taken in square shape and the height of the substrate has varied as per the commercially availability in the market. Here, we observed improvement in bandwidth with good return loss. The impedance is also high at that 1.57 mm of the substrate height, which has high bandwidth and return loss. The bandwidth of 560 MHz and impedance of 44 ohms is achieved. This validates the choice of the height of the substrate to find the designing of compact patch antenna meant to operate in the X-band of frequency ranging from 8 to 12 GHz. The proposed microstrip antenna can be used in many wireless applications within the X-band. Keywords Microstrip · Return loss · Impedance · Bandwidth · X- band

1 Introduction Microstrip antennas find their use exclusively in many of the current devices in the wireless communication system. An antenna is a key component in many of the wireless applications. Microstrip patch antenna fulfills most of the commercial requirements for wireless communication. The functionality of wireless devices has been increased recently, where the size of electronic circuits is shrinking drastically. The size of the antennas being used, for most of the applications, is also shrinking. The very basic form of the microstrip antenna can be constructed using a substrate of dielectric material [1, 2]. The shape of the substrate has been taken either as rectangular or square, to simplify the analysis, with a conducting patch, mainly R. K. Chaurasia Department of Electronics and Communication Engineering, ICFAI University, Jaipur, India S. K. Nippani (B) Department of Physics, School of Engineering, University of Petroleum & Energy Studies, Dehradun, India e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2021 S. Choudhury et al. (eds.), Intelligent Communication, Control and Devices, Advances in Intelligent Systems and Computing 1341, https://doi.org/10.1007/978-981-16-1510-8_5

45

46

R. K. Chaurasia and S. K. Nippani

radiating, on one side and a non-radiating ground plane on another side. Based on the simplification in fabrication, analysis, and performance prediction, the shape of the radiating patch can be simple or complex geometrical shapes. The simple microstrip antenna undergoes with narrow bandwidth. The low bandwidth makes it unsuitable for many applications in wireless communication. Numerous methods have been considered and found that the impedance bandwidth of the rectangular microstrip patch antenna may be increased, and the choice of the height of the substrate could be one of them [3, 7].

2 Antenna Design The designed antenna is meant for operating in the X-band of frequency and the original central frequency chosen around 11 GHz, which falls in the X-band region (8.0 GHz–12 GHz) [8]. The designed antenna must show its best performance around the resonant frequency as specified [9]. The two widely used substrate materials for the antenna design are RT Duriod and FR4 epoxy material. RT Duriod, having a dielectric constant of 2.2, provides very sharp resonance. The other material is FR4 epoxy, whose dielectric constant is 4.4. The thickness of the substrate is normally selected on the greater side to add the volume and enhance the bandwidth [9]. Also, the antenna mustn’t be bulky, otherwise spurious radiation occurs that accounts for more losses, so the thickness of the dielectric substrate is taken from 1.0 to 2.0 mm. Microstrip feedline is the feeding method chosen to feed the source and this is the most convenient form of feeding methods [10]. The microstrip patch dimensions are calculated as per the below equations [11, 12].  c × Width, W = 2f c Length, L = 2f



εr + 1 εr − 1 + 2 2

2

(1)

(εr +1 )



h 1 + 12 W

−1  2

− 2L

(2)

Here, εr is the substrate’s dielectric constant, and f is the resonant frequency. For this study, RT Duroid substrate of εr = 2.2 is taken. The height or thickness (h) of the rectangular microstrip antenna is the height of the substrate since the two metallic layers are very thin in height. The equation guarding the relationship between the resonant operating frequency, the height, and dielectric constant is given as h ≤ 0.06

c √ 2π f εr

(3)

Choice of Suitable Height of Substrate …

47

3 Results Concerning the commercially available height of the substrate and using the above equations, the dimension of the microstrip antenna comes out to be 10.77 mm and 8.08 mm, with a substrate height of 1.54 mm. Since variation in substrate height is ranging between 1 and 2 mm and working on the fixed dimension of the antenna for the fulfillment of the objective, the shape is kept square-shaped with each side measuring 9.5. The dimension of the ground plane is 18 by 18 mm. The simulation has been carried out by using the finite element method-based HFSSv13 tool. Now, substrate height is changed from 117 to 1.77 mm. as per availability of the substrate commercially. The simulated results of reflection coefficient with frequency (a measure of bandwidth) for different heights of the substrate are plotted and shown in Figs. 1, 2, 3 and 4. Fig. 1 Bandwidth and reflection coefficient with 1.17 mm of height

Fig. 2 Bandwidth and reflection coefficient with 1.37 mm

48

R. K. Chaurasia and S. K. Nippani

Fig. 3 Bandwidth and reflection coefficient with 1.57 mm of height

Fig. 4 Bandwidth and reflection coefficient with 1.77 mm of height

Figures 1, 2, 3 and 4 reveal that the maximum bandwidth of 560 MHz is noticed at the substrate’s height of 1.57 mm with a peak at 10.8 GHz. While the maximum reflection coefficient of −38 dB is achieved at the substrate height of 1.77 mm with a bandwidth of 480 MHz. Therefore, it is concluded that there is a tradeoff between bandwidth and reflection coefficient on varying the substrate height. The simulated results for bandwidth and reflection coefficient with height substrate are presented in Table 1.

Choice of Suitable Height of Substrate … Table 1 Bandwidth and reflection coefficient at a different height for the proposed antenna

49

Height (mm) Bandwidth (MHz) Reflection Coefficient (dB) 1.17

410

−18

1.37

510

−22

1.57

560

−25

1.77

480

−38

4 Impedance Calculation Using Transmission Line Model The transmission line model is considered simpler among all other models, as it offers a reasonable behavior of the microstrip antenna [13, 14]. The equivalent representation of the rectangular patch antenna is considered as a parallel-plate 3-port transmission line model as in Fig. 5. The input admittance of an antenna is given as [15] Yin =

Yc2 + Ys2 − Ym2 + 2Ys Yc coth(γ L) − 2Ym Yc cshh(γ L) Ys + Yc coth(γ L)

(4)

On solving, generalized solution of input impedance is found to be [14]    Yc + jY L tan(β L)   Z in = Z c  Y L + jYc tan(β L) 

(5)

The value of Zin is equal to ZL, when the normalized value of βL is equal to π radian. This means that in the extreme case of optimization, the line impedance value and the input impedance values are identical. The normalized value of the characteristic impedance is found by solving equation (5), and is given as Zc =

√ 120π/ εe f f W h

+ 1.393 + 0.667ln( Wh + 1.444)

Fig. 5 Equivalent circuit of the microstrip antenna

(6)

50

R. K. Chaurasia and S. K. Nippani

Characteristic impedance plot of different heights, as commercially available, using Eq. (6) is plotted in Figs. 6, 7, 8 and 9, and the results are discussed herewith. With 1.17 mm of substrate height, the impedance maximum value is at 31 ohms, and the impedance maximum value increases to 37 ohms with 1.37 mm of substrate height. When the substrate height is 1.57, maximum impedance is observed and it is 44 ohms, a value very near to line impedance of 50 ohms. The impedance maximum value decreases to 39 ohms with 1.77 mm of substrate height. The outcomes are briefed in the below-mentioned table (Table 2). The proximity of the two (bandwidth and impedance) values is observed at 1.57 mm of substrate height. This validates that this height is of maximum use for all design purposes.

Fig. 6 Impedance plot of 1.17 mm height

Fig. 7 Impedance plot of 1.37 mm height

Choice of Suitable Height of Substrate …

51

Fig. 8 Impedance plot of 1.57 mm height

Fig. 9 Impedance plot of 1.77 mm height

Table 2 The impedance at different heights for the proposed antenna

Height (mm)

Impedance (Ohm)

1.17

31

1.37

37

1.57

44

1.77

39

5 Conclusion The proposed microstrip square-shaped antenna showed a good bandwidth of the order of 560 MHz. It exhibits a maximum return loss of 38.5 dB at the substrate height of 1.57 mm. Also at this height, the impedance is maximum, 44 ohms, and very close to the line impedance. This validates the most purposeful value of the height of the substrate, as per the commercial availability, in the design procedure of the antenna.

52

R. K. Chaurasia and S. K. Nippani

References 1. Mishra, R.: An Overview of Microstrip Antenna. HCTL Open Int. J. Technol. Innov. Res. (IJTIR) 21(2), 1–35 (2016) 2. Chaurasia, R.K., Vishal Mathur, R.L. Pareekh, Tamsir, Vineet Srivastava: A computational modelling of micro strip patch antenna and its solution by RDTM. Alexerdria Eng. J. 57(3), 1877–1881 (2018) 3. Garg, R., Bhartia, P., Bahl, I., Ittipiboon, A.: Microstrip Antenna Design Handbook, Artech House (2001) 4. Chaurasia, R.K., Vishal Mathur: Enhancement of bandwidth for square of micro strip antenna by partial ground and feedline technique. Asia Pacific J. Eng. Sci. Technol. 3(1), 49–53 (2017) 5. Dong-Zo Kim, Wang-Ik Son, Won-Gyu Lim, Han-Lim Lee, Jong- Won Yu: Integtated planar monopole antenna with microstrip resonators having band-notched characteristics. IEEE Trans. Antennas Propagation. 58, 2837–2842 (2010) 6. Mishra, R., Jayasinghe, J., Mishra, R.G., Kuchhal, P.: Design and performance analysis of a rectangular microstrip line feed ultra-wide band antenna. Int. J. Signal Proc. Image Proces. Pattern Recogn. 9(6), 419–426 (2016) 7. Kushwaha, R.S., Srivastava, D.K., Saini, J.P., Dhupkariya, S.: Bandwidth enhancement for microstrip patch antenna with microstrip line feed. 2012 Third International Conference on Computer and Communication Technology (ICCCT), pp. 183, 185 (2012) 8. Coulibaly, T., Denidni, A., Boutayeb, H.: Broadband microstrip-fed dielectric resonator antenna for X-band applications. IEEE Antennas Wirel. Propag. Lett. 7, 341–345 (2008) 9. Su, S.W., Wong, K.L., Tang, C.L.: Band-notched ultra-wideband planar monopole antenna. Microwave Opt. Technol. Lett. 44, 217–219 (2005) 10. Mishra, R., Mishra, R.G., Kuchhal, P.: Analytical study on the effect of dimension and position of slot for the designing of Ultra Wide Band (UWB) Microstrip Antenna. 5th IEEE International Conference on Advances in Computing, Communications and Informatics (ICACCI), pp. 502– 507 (2016) 11. Lee, K.F., Luk, K.M., Tong, K.F., Shum, S.M., Huyn, T., Lee, R.Q.: Experimental and simulation studies of the coaxially Fed U–slot rectangular patch. IEEE Proce. Microwave Antenna propag. 144(5), 354–358 (1997) 12. Mishra, R., Kuchhal, P., Kumar, A.: Effect of height of the substrate & width of the patch on the performance characteristics of microstrip antenna. Int. J. Elect. Comp. Eng. 5(6), 1441–1445 (2015) 13. Balanis, C.A.: Antenna theory—analysis and design. Wiley-Inter Science (2012) 14. Gupta, K.C., Garg, R., Bahl, I., Bhartia, P.: Microstrip lines and slot lines. Artech House, 2nd ed (2001) 15. Pues, H., van de Cappelle, A.: Accurate transmission-line model for the rectangular microstrip antenna. Proc. IEE. 131(6) (1984)

Comparative Analysis of Multi-slot Circular Patch Antenna Praful Ranjan, Prasanthi Kumari Nunna, and Bhupesh Chandra Joshi

Abstract This paper is based on a circular patch antenna with multi-slot, designed and compared with different slots, on FR4 substrate with εr = 4.4, and h = 1.6 mm. The simple circular Patch antenna operating at 6.1 GHz. Again we have cut different slots, then it operates in dual-frequency 6.1, 9.6 GHz (C band and X band) and triple frequency 6.1, 14.8, 16.3 GHz (C band and Ku band). The simulation is done using HFSS™ v12. The dual frequency antenna achieves −39.7596 dB at 6.1 GHz and − 14.9487 dB at 9.6 GHz return loss and triple frequency antenna achieves return loss −25.5677 dB at 6.1 GHz, −21.4460 dB at 14.8 GHz and −19.9484 dB at 16.3 GHz. The impedance matching is achieved near 50 with better VSWR. Keywords Multiband · Voltage standing wave ratio · Circular patch · FR4 · S11 Radiation pattern

·

1 Introduction The field of microstrip antenna is getting advanced due to the different wireless technological advancements and research. There are several applications of microstrip antenna in the current scenario; some are wireless networking devices, laptops, mobiles, defense, etc. For designing of patch antenna, different feeding techniques are used, the most popular are microstrip feed, coaxial feed, proximity and aperture coupling [1–4]. The slot antenna is a new method that eliminates the limitation of narrow bandwidth and provides the capability of resonating at more than one frequency. Circular patch antenna is mostly used after rectangular patch, because its biggest advantage

P. Ranjan (B) · B. C. Joshi Department of ECE, THDC-IHET, Tehri Uttarakhand, India e-mail: [email protected] P. K. Nunna UPES Dehradoon, Dehradoon, India © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2021 S. Choudhury et al. (eds.), Intelligent Communication, Control and Devices, Advances in Intelligent Systems and Computing 1341, https://doi.org/10.1007/978-981-16-1510-8_6

53

54

P. Ranjan et al.

is that unlike rectangular patch, we only required the radius of patch. It can be used in an array form with the same source and different source, single as well [5–9]. This paper compares the result of a simple circular patch antenna with other circular patch antennas that contains two-rectangular slots, and one more antenna that contains three rectangular slots [10–13].

2 Circular Patch Antenna Structure Design The designing part of the antenna can be concluded using Eqs. (1 and 2). The feeding mechanism is the microstrip line feed, with 50 impedance [14–16]. The circular shape patch antenna structure is designed on dielectric constant = εr 4.4 (FR4 substrate) where thickness h = 1.6. Figure 1 clearly shows the simple design of circular patch antenna with different parameters (length, width, etc.) Actual dimensions are taken with reference to Fig. 1 for simulation, are shown in Table 1. The simulated design of this antenna is shown in Fig. 2. F=

Fig. 1 Simple circular patch with microstrip feed

8.791 × 109 fr √εr

(1)

Comparative Analysis of Multi-slot Circular Patch Antenna Table 1 Design parameters of antenna

Parameters

55 Unit(mm)

Radius of patch (a)

6.828

Length of substrate (l)

32.44

Width (w)

28.772

Length of strip line (l’)

16

Width of strip line (c)

1

Internal cut (b)

2

Fig. 2 Simulated structure of circular patch antenna

a = 

F     1 + ε2h ln π2hF + 1.7726 rF

(2)

where fr = Resonant frequency (taken in Hz), εr = Dielectric Constant, h = Thickness of the substrate (taken in cm), and a = Radius of Patch. The actual dimension for simulation at a frequency of 6.1 GHz is given in Table 1. Now the same circular patch antenna is designed with different slots. Due to the different slots, it gives the characteristics of resonating at more than one frequency (dual band). Simulated designs are shown in Fig. 4 (with 2 rectangular slots) and Fig. 5 (with 3 rectangular slots). The positions and dimensions of slots are given in Fig. 3 and mentioned in Table 2.

56

P. Ranjan et al.

Fig. 3 Geometry for dimensions and positions of different slots

Fig. 4 Simulated structure of circular shape patch antenna with 2-rectangular slots

3 Results of Simulated Antennas The simulated part of all patch configuration is shown in the following figures, the results contain Return Loss (db), VSWR, Z- parameter, and Radiation pattern (of gain). Simulated result of simple circular patch antenna in shown in Figs. 5 and 6. Simulated result for 2-cut and 3-cut circular patch is shown in Figs. 7 and 8. Results of simulated antenna with three rectangular slots shown in Figs. 9 and 10. The comparisons of all three circular patch antennas are shown in Table 3.

Comparative Analysis of Multi-slot Circular Patch Antenna

Fig. 5 Results of simple circular patch antenna a Return loss b VSWR curve

57

58 Table 2 Dimensions for different slot

P. Ranjan et al. Parameters

Unit (mm)

L1

2.5

L2

1

L3

7

L4

2.5

L5

1

4 Conclusions The design of a circular structure patch antenna with multi slots has been simulated. The result of different circular patch antenna shows that due to the 2-slots it operates in dual frequency (C and X band), and due to 3 slots it operates in triple frequency(C and Ku band). Designed antenna achieved VSWR < 2, Gain = 4.237 dB and dual frequency antenna achieves −39.7596 dB at 6.1 GHz and −14.9487 dB at 9.6 GHz return loss and triple frequency antenna achieves return loss −25.5677 dB at 6.1 GHz, −21.4460 dB at 14.8 GHz and −19.9484 dB at 16.3 GHz. Applications of the designed antenna are FTA satellite, radar application, and TV.

Comparative Analysis of Multi-slot Circular Patch Antenna

Fig. 6 Results of simple circular patch antenna a Z-Parameter b Radiation pattern (gain)

59

60

Fig. 7 Results of circular patch antenna with 2-rectangular slots a S11 b VSWR

P. Ranjan et al.

Comparative Analysis of Multi-slot Circular Patch Antenna

61

Fig. 8 Results of the simulated antenna with two rectangular a Z-Parameter b Radiation pattern (gain)

62

P. Ranjan et al.

Fig. 9 Results of simulated circular antenna with 3-rectangular slots a Return loss b VSWR curve

Comparative Analysis of Multi-slot Circular Patch Antenna

63

Fig. 10 Results of circular patch antenna with three rectangular slots a Z-Parameter b Radiation pattern (gain)

64

P. Ranjan et al.

Table 3 Result of simulated antenna Structure

Return loss (dB)

VSWR

Gain (dB)

Resonating frequency s (GHz)

Resonating frequency bands

Simple

−25.17

1.11

4.34

6.1

C band

With 2 rectangular slots

−39.75, − 14.948

1.026, 1.435

4.30

6.1, 9.6

C, X band

With 3 rectangular slots

−25.56, − 21.446, − 19.94

1.11, 1.185, 1.223

4.237

6.1, 14.8, 16.3 C and Ku band

Reference 1. Balanis, C.A.: Antenna Theory and Design. Wiley (2004) 2. Pozar, D.M.: Microwave Engineering, 3rd edn. Wiley (2004) 3. Pandey, S., Markum, K.: Design of circular shape microstrip patch antenna for C-band application. IJARCST 4 (2016) 4. Ranjan, P., Tomar, G.S., Gowri, R.: Metamaterial Loaded Shorted Post Circular Patch Antenna. Int. J. Signal Process. Pattern Recognit. 9(10), 217–226 (2016) 5. Pandey, K., Kumar, A.: Comparative analysis of different angles of CPW fed UWB rectangular slot triangular patch antenna. IJARCCE 3 (2014) 6. Stutzman, W.L., Thiele, G.A.: Antenna Theory and Design, 2nd edn. Wiley, New York (1998) 7. Garg, R., Bartia, P., Bahl, I., Ittipiboon, A.: Microstrip Antenna Design Handbook, pp. 1–68, 253–316. Artech House Inc. Norwood, MA (2001) 8. Kirar, A.S., Jadaun, V.S., Sharma, P.K.: Design of circular Microstrip patch antenna for dual band. IJECCT 3 (2013) 9. Thaker, N.M., Ramamoorthy, V.: A review on circular microstrip patch antenna with slots for C band applications. IJSER 5 (2014) 10. Ponkia, A.V., Dwivedi, V.V., Chaudhari, J.P.: Dual band circular shaped slotted microstrip patch antenna. IJAP (2012) 11. Yaghjian, A.D., Best, S.R.: Impedance, bandwidth, and Q of antennas. IEEE Trans. Antennas Wave Propag. 53(4), 1298–1324 (2005) 12. Rahmat-Samii, Y., Colburn, J.S.: Patch antennas on externally perforated high dielectric constant substrates. IEEE Trans. Antennas Propag. 47(12), 1785–1794 (1999) 13. Zakaria, N.A., Sulaiman, A.A., Latip, M.A.A.: Design of a circular microstrip antenna. In: International RF and Microwave Conference, IEEE (2008) 14. Kumar, G., Ray, K.P.: Broad Band Microstrip Antennas. Artech House (2003) 15. Kraus, J.D., Marhefka, R.J.: Antennas for all Application, 3rd edn 16. Gowri, R., Ranjan, P., Prashanti, N.: Design of double sided metamaterial antenna for mobile handset applications. In: SPIN 2014, pp. 675–678. IEEE (2014). ISBN 978–1–4799–2866–8/14

Active Power Curtailment in PV Array Under LVRT Condition Jyoti Joshi, Anurag Kumar Swami, and Vibhu Jately

Abstract Under LVRT, GCPV inverters aid the grid by injecting reactive power according to the grid standards whenever there is a short-term dip in the grid voltage. To ensure that the inverter output currents are limited during this period, inverter power capacity is continuously updated. Hence, active power curtailment is essential to avoid overvoltage in DC link capacitor, limit the current, and disconnection of PV inverters. To reduce this real power, PV array should operate at a non-MPPT point on the P–V curve. This non-MPPT point can operate on the LHS or RHS of the P–V curve. This paper analyses the behavior response of the GCPV inverter while switching from MPPT to non-MPPT operating mode under unbalanced grid voltage conditions. The MATLAB/Simulink results show that the steady-state response time to reach the non-MPPT point is highly dependent on the location of the non-MPPT point of operation. Keywords GCPV · MPP · LVRT · THD · APC

1 Introduction Among all other renewable resources, solar PV is gaining more popularity because of its numerous advantages like cleanliness, ease of availability, low cost, etc. In MPPT, maximum power from PV is most commonly extracted by perturb and observe algorithm [1]. This is due to its ease of implementation [2]. The PV inverter is the most vital component of GCPV systems. The inverter controller converts the DC power extracted using the MPPT algorithm to AC power and is injected into the Grid [3]. Moreover, inverter controllers should be designed to ensure that accurate active and reactive power is injected from the inverter into the grid during LVRT. LVRT J. Joshi (B) · A. K. Swami Department of Electrical Engineering, G. B. P. U. A. & T., Pantngar 263145, India e-mail: [email protected] V. Jately MCAST Energy Research Group, Institute of Engineering and Transport, Malta College of Arts, Science and Technology, Paola, Malta © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2021 S. Choudhury et al. (eds.), Intelligent Communication, Control and Devices, Advances in Intelligent Systems and Computing 1341, https://doi.org/10.1007/978-981-16-1510-8_7

65

66

J. Joshi et al.

Fig. 1 Challenges under LVRT

is a condition in which, the GCPV inverter is required to remain synchronized with the utility for a brief period dip in the grid voltage. During this period, the inverter injects P and Q into the utility according to the specified standards. As shown in Fig. 1, under LVRT, when unbalanced voltage occurs, the control of the PV inverter should be such that as to ameliorate the following issues; (1)

(2) (3)

to accurately determine the amplitude of voltage sag. This is achieved by using a PLL. Several researchers have proposed different types of PLL to accurately determine the sag under asymmetrical grid voltage conditions [4–7]. Generation of a current reference to inject P and Q into the utility [8–11]. DC-link voltage control for reducing overvoltage in dc capacitor and to attain power balance in the circuit [12–15].

Under severe grid faults, some LVRT grid codes only require the injection of Q (100%) and during this period the injected P to the grid is zero. In [16] and [17], only the Q is delivered to the grid. However, the schemes suffer from oscillations in both P and Q. In [18], only P is delivered to the grid. This strategy minimizes the peaks in the injected grid currents at the cost of an increase in total harmonics distortion (THD). According to the updated LVRT grid codes, the inverter is required to inject P and Q to the grid [19]. In [20], a flexible controller is developed to maintain the currents to a safe limit and inserting both P and Q into the utility. However, this control strategy requires phase angles to evaluate current commands, which enhances the complexity of the scheme. In [21], a strategy that also injects P and Q into the grid is proposed. This controller injects the maximum rated current into the grid irrespective of the voltage sag and also helps in reducing the oscillations in active power. Under normal operating conditions, the GCPV inverter delivers the Pinj to the grid according to the active power reference (Pref ) and Pref > Pinj . When a dip in grid voltage is seen, the inverter capacity is minimized to limit the current amplitude. In this case, the power imbalance will occur in the system if maximum power is extracted and P∗ < Pinj . Hence the MPPT operation is deactivated. The voltage across the DC link capacitor exceeds if the MPPT operation continues. This situation results in a power imbalance in the system and may give rise to the disconnection of the GCPV inverter. In [12], the authors suggested an approach where P is curtailed. Three cases

Active Power Curtailment in PV Array Under LVRT Condition

67

Fig. 2 Determination of points for non-MPPT mode

have been discussed in the presented scheme; (1) short-circuiting the PV (2) opencircuiting the PV and (3) reduced power is inserted into the grid. The first two cases P injected into the grid is zero. Whereas, in the third case P is reduced. For reduction of P from PV array, the operation should be in non-MPPT mode. The operation should move either to the left side (point C) or to the right side (point B) as shown in Fig. 2. Mirhosseini et al. [12] has considered the movement of the operating point from A to B. Hence the non-MPPT operation is on the RHS of PV curve. In [11], the authors have also suggested using point B for active power curtailment (APC). The authors in [12] have used point C to achieve better current control capability and reduced duty cycle value. Since there are two possible points of operation in the non-MPPT mode, a detailed investigation is vital to evaluate the behavior response of the system operating at these non-MPPT operating points. The authors have compared steady-state response time and power oscillations at these two possible non-MPPT operating points to determine the most efficient region of operation in non-MPPT mode. Section 2 gives the overall description of the investigated system. Section 3 entails an overview of the controller design. Section 4 shows the results obtained in MATLAB/Simulink environment and Sect. 5 gives a brief conclusion.

2 System Description In this section, a three-phase two-stage GCPV system, during symmetrical and asymmetrical condition is investigated. As shown in Fig. 3, a boost converter is used in a two-stage system and coupled with the inverter via a dc-link capacitor. The use Fig. 3 Block diagram of the investigated system

68

J. Joshi et al.

of two-stage system is beneficial as there is a separate control for MPPT which is implemented using a boost converter. Mainly there are three frames of reference that can be used to design the inverter control; natural reference frame (a-b-c), synchronously rotating reference frame (dq), and stationary reference frame (α-β). This paper has used α-β reference frame as it eliminates the application of PLL. The grid voltages are converted to α-β frame (SRF) as in (1).  vαβ =

vα vβ



⎡ ⎤    va 2 1 −1/2 −1/2 ⎣ ⎦ √ √ = vb 3 0 3/2 − 3/2 vc

(1)

where vα , vβ are the voltages in the SRF and va , vb , vc are the grid voltages in the a-b-c frame of reference. The voltage and current under unbalanced grid condition is expressed as V = V+ + V− + V0

(2)

I = I+ + I− + I0

(3)

where V+ , V− , V0 and I+ , I− , I0 are the positive, negative and zero sequence component of voltage and current. Since a three-phase three-wire system is taken, hence zero sequence component will not be considered for further analysis. The apparent power S is written as S = v.i∗ = P + jQ

(4)

where v and i are the voltage and current vectors in the SRF, and P, Q are the active and reactive power, respectively. Under normal operating conditions, the three-phase voltages are balanced and hence, no negative sequence component and oscillatory components are present in P and Q. On the contrary, the voltage and currents contain negative sequence components under unbalanced grid conditions. Hence (4) can be re-written as



∗ − + − S = vαβ · i∗αβ = v+ αβ + vαβ · iαβ + iαβ +∗ + −∗ − +∗ − −∗ = v+ αβ · iαβ + vαβ · iαβ + vαβ · iαβ + vαβ · iαβ

(5)

where v+ αβ

    1 1 −q − 1 −q = vαχ vαχ and vαβ q 1 2 q 1

(6)

Active Power Curtailment in PV Array Under LVRT Condition

69

In (6), q = e−jπ/2 is a 90° lagging phase-shifting operator applied to the time domain. Further



∗

+∗ + + + + + + + + + + + (7) = v+ v+ α iα + vβ i β + j vβ iα − vα iβ αφ · iαβ = vα + jvβ · iα + jiβ From (7) the constant and oscillating terms for P and Q can be obtained.

3 Controller Design By equating the reactive power oscillations to zero, the current reference generating equations for proper injection of P and Q, according to the grid code is obtained as − v+ α − vα



Pref iαP = −2 −2 + v−2 v+2 − v + v α α β β

(8)

− v+ β − vβ



Pref iβP = −2 −2 v+2 − v−2 α + vβ α + vβ

(9)

− v+ α⊥ + vα⊥



Qref iαQ = − −2 −2 −2 v+2 α⊥ + vβ⊥ − vα⊥ + vβ⊥

(10)

− v+ β⊥ + vβ⊥



Qref iβQ = − −2 −2 −2 v+2 − v + v + v α⊥ β⊥ α⊥ β⊥

(11)

where iαP , iβP , iαQ and iβQ are active and reactive currents in SRF, respectively, and Pref and Qref are reference active and reactive powers, respectively. Further, vα⊥ = − vβ and vβ⊥ = vα . As mentioned earlier, when there is a dip in the voltage the inverter capacity reduces, to limit the current. Hence the updated inverter capacity (UIC) is obtained as √ √ VP − VN S (12) UIC = VB where vp = vα +2 + vβ +2 , vN = vα −2 + vβ −2 and S is the apparent power of GCPV inverter. Moreover, the injected reference reactive power, according to the severity of voltage sag as per grid code is

70

J. Joshi et al.

⎧ ⎨

Qref = 0 if Vp.u.  > 0.9 Qref = S ∗ 1.5 ∗ 0.9 − Vp.u. if 0.2 < Vp.u. < 0.9 ⎩ Qref = 1.05 ∗ S if Vp.u. < 0.2

(13)

The per unit (p.u.) voltage in (13) is given as Vp.u. =

 v2α +v2β vB

(14)

Further the maximum active power (Pmax ), which inverter can inject during sag, is expressed as Pmax =



UIC2 − Q2

(15)

Under voltage sag condition Pmax is continuously compared with Pref and if Pmax > Pref . The inverter continues to deliver the same amount of active power. As soon as the Pmax < Pref, under voltage sag the MPPT is terminated and to prevent overvoltage across DC capacitor, the point of operation is moved to the non-MPPT region as shown in Fig. 2. For non-MPPT mode, the new duty cycle is obtained as Dnew =

Pmax Dmpp Pmpp

(16)

where Dnew is the new duty cycle for operation in non-MPPT mode. Pmpp and Dmpp are the MPP power and duty cycle. The controller is in Fig. 4. Pref is obtained as the output of PI controller and Qref is calculated from (13) as per grid codes given in [22]. The current controlling is achieved by employing a proportional resonant (PR) controller. As shown in the figure, the MPPT operation is achieved by using perturb and observe algorithm. Under voltage sag conditions, the APC is performed by moving the point of operation either to LHS or RHS of PV curve and a new duty cycle is determined using (16).

4 Simulation Results A 2 kW PV system is considered for this case study. Initially, at t = 0 s, the grid voltages are balanced and the PV system is working at the MPP. At t = 5 s, phase A amplitude is reduced to 0.5 pu, while B and C phases remain unchanged. When a dip is observed in grid voltage, a non-MPPT mode is activated by the controller. A pu voltage sag of 0.85 is obtained using (14). The controller updates the inverter

Active Power Curtailment in PV Array Under LVRT Condition

71

Fig. 4 Block diagram of the controller

capacity using (12). For this case study, the UIP obtained is 66.6%. Qref is obtained from the grid standards using (13) and is 150 kVAr. The active power (Pmax ) is then obtained using (15) as 1.3 kW. The MPPT operation is terminated and the reduction of active power from 2 to 1.3 kW is obtained with the help of APC block and the duty ratio corresponding to the lower power is evaluated by using (16). The designed model was tested for two case scenarios: (a) (b)

When the point of operation is shifted LHS of MPP When the point of operation is shifted to RHS of MPP

For the first case study, the duty of the boost converter is varied in such a way that the operating point is shifted to LHS of MPP to extract 1.3 kW power. For the second case study, the duty of the boost converter is decreased to ensure the operating point is shifted to LHS of MPP to obtain same power from the PV array. The results obtained in Fig. 5 clearly demonstrate that when the point of operation is on LHS of MPP, low steady-state power oscillations are observed as compared to large oscillations when the point of operation is on RHS of MPP. This is because of the large slope in the PV curve on RHS of MPP, meaning, a small variation in voltage results in a large variation in power when operated on RHS operating point. On the other hand, poor dynamic response was observed when the point of operation is moved to LHS of MPP as compared to right. This is because the shifting from MPP to the left non-MPPT operating point takes a longer time. This trade-off can be improved by selecting a flexible power point tracking technique that adaptively reduces the step size to reach the non-MPPT point of operation.

72

J. Joshi et al.

Fig. 5 Simulation results in MATLAB/Simulink environment showing PV output when a operating under left side, b operating under right side; injected c active power, d reactive power e inverter currents, f grid voltage and g DC link capacitor voltage for both cases

The inserted powers for both case studies are the same and is depicted in Fig. 5c, d. It can be seen that real power is free from oscillations while the injected reactive power contains oscillations having twice of the grid frequency. The inverter injected currents and grid voltages for both cases are in Fig. 5e, f, respectively. The injected current are unbalanced in both scenarios but are nearly sinusoidal with a very low total harmonic distortion (THD). Finally, the DC link capacitor voltage is in Fig. 5g, which delineates an efficient DC link control loop as it contains very small oscillations.

Active Power Curtailment in PV Array Under LVRT Condition

73

5 Conclusion This paper compares and analyzes the performance of PV system working under non-MPPT mode. The two possible case scenarios for active power curtailment were implemented. A poor dynamic and good steady-state response was observed when the non-MPPT point of operation is on the left side, whereas the right non-MPPT point of operation can be quickly reached on account of large power oscillations. It was determined that there exists a quid pro quo between response time and power oscillations when selecting the non-MPPT region of operation. A flexible power point tracking technique is generally suggested to efficiently operate under balanced and unbalanced grid voltages to efficiently operate under low-voltage ride through conditions.

References 1. Jately, V., Arora, S.: An efficient hill-climbing technique for peak power tracking of photovoltaic systems. In: 7th IEEE Power India International Conference (PIICON), pp. 1–5. IEEE, India (2016) 2. Jately, V., Arora, S.: Development of a dual-tracking technique for extracting maximum power from PV systems under rapidly changing environmental conditions. Energy 133, 557–571. I (2017) 3. Xue, Y., Chang, L., Bordonau, J.: Topologies of single-phase inverters for small distributed power generators: an overview. IEEE Trans. Power Electron. 19 (2004) 4. Rodríguez, P., Teodorescu, R., Candela, I., Timbus, A.V., Liserre, M., Blaabjerg, F.: New positive-sequence voltage detector for grid synchronization of power converters under faulty grid conditions. In: 37th IEEE Power Electronics Specialists Conference, pp. 1–7. IEEE, Spain (2006) 5. Golestan, S., Guerrero, J.M., Vidal, A., Yepes, A.G., Doval-Gandoy, J.: PLL with MAF-based prefiltering stage: small-signal modeling and performance enhancement. IEEE Trans. Power Electron. 31, 4013–4019 (2016) 6. Golestan, S., Guerrero, J.M., Vasquez, J.C.: Three-phase PLLs: a review of recent advances. IEEE Trans. Power Electron. 32, 1894–1907 (2017) 7. Golestan, S., Monfared, M., Freijedo, F.D.: Design-oriented study of advanced synchronous reference frame phase-locked loops. IEEE Trans. Power Electron. 28, 765–778 (2013) 8. Rodriguez, P., Timbus, A.V., Teodorescu, R., Liserre, M., Blaabjerg, F.: Flexible active power control of distributed power generation systems during grid faults. IEEE Trans. Indus. Electron. 54, 2583–2592 (2007) 9. Camacho, A., Castilla, M., Miret, J., Borrell, A., de Vicuña, L.G.: Active and reactive power strategies with peak current limitation for distributed generation inverters during unbalanced grid faults. IEEE Trans. Indus. Electron. 62, 1515–1525 (2015) 10. Guo, X., Liu, W., Lu, Z.: Flexible power regulation and current-limited control of the gridconnected inverter under unbalanced grid voltage faults. IEEE Trans. Indus. Electron. 64, 7425–7432 (2017) 11. Afshari, E., et al.: Control strategy for three-phase grid-connected PV inverters enabling current limitation under unbalanced faults. IEEE Trans. Industr. Electron. 64, 8908–8918 (2017) 12. Mirhosseini, M., Pou, J., Agelidis, V.G.: Single- and two-stage inverter-based grid-connected photovoltaic power plants with ride-through capability under grid faults. IEEE Trans. Sustain. Energy 6, 1150–1159 (2015)

74

J. Joshi et al.

13. Tang, C., Chen, Y., Chen, Y.: PV power system with multi-mode operation and low-voltage ride-through capability. IEEE Trans. Indus. Electron. 62, 7524–7533 (2015) 14. Jain, C., Singh, B.: An adjustable DC link voltage-based control of multifunctional grid interfaced solar PV system. IEEE J. Emerg. Selected Topics Power Electron. 5, 651–660 (2017) 15. Ding, G., et al.: Adaptive DC-link voltage control of two-stage photovoltaic inverter during low voltage ride-through operation. IEEE Trans. Power Electron. 31, 4182–4194 (2016) 16. Rodriguez, P., Luna, A., Hermoso, J.R., Etxeberria-Otadui, I., Teodorescu, R., Blaabjerg, F.: Current control method for distributed generation power generation plants under grid fault conditions. In: 37th Annual Conference of the IEEE Industrial Electronics Society, pp. 1262– 1269. IECON, Melbourne (2011) 17. Miret, J., Castilla, M., Camacho, A., Vicuña, L.G.D., Matas, J.: Control scheme for photovoltaic three-phase inverters to minimize peak currents during unbalanced grid-voltage sags. IEEE Trans. Power Electron. 27, 4262–4271 (2012) 18. Camacho, A., Castilla, M., Miret, J., Guzman, R., Borrell, A.: Reactive power control for distributed generation power plants to comply with voltage limits during grid faults. IEEE Trans. Power Electron, 29, 6224–6234 (2014) 19. BDEW, Technical Guideline-Generating Plants Connected to the Medium Voltage NetworkGuideline for Generating Plants Connection to and Parallel Operation with the MediumVoltage, https://www.bdew.de/internet.nsf/id/A2A 20. Lee, C., Hsu, C., Cheng, P.: A low-voltage ride-through technique for grid-connected converters of distributed energy resources. IEEE Trans. Industry Appl. 47, 1821–1832 (2011) 21. Sosa, J.L., Castilla, M., Miret, J., Matas, J., Al-Turki, Y.A.: Control strategy to maximize the power capability of PV three-phase inverters during voltage sags. IEEE Trans. Power Electron. 31, 3314–3323 (2016) 22. Wang, Y., Yang, P., Yin, X., Ma, Y.: Evaluation of low-voltage ride-through capability of a two-stage grid-connected three-level photovoltaic inverter. In: 17th International Conference on Electrical Machines and Systems (ICEMS), pp. 822–828. IEEE, Hangzhou (2014)

Autonomous Tomato Harvester Using Robotic Arm and Computer Vision Siddhant Salvi, Vivekanand Sahu, Rohit Kalkundre, and Omkar Malwade

Abstract The number of people willing to work in the agricultural sector is decreasing. Also, the cost of labour is increasing making it harder for small farmers to gain profits. With the decreasing number of farmers in India, there is a shift to find more autonomous solutions in the agricultural sector. Harvesting of fresh tomatoes is very essential from the farmer’s point of view and requires precision. Harvesting of tomatoes has always been a manual job requiring a large amount of labour. Making use of robotic arm in such application increases precision, reduces labour and time. We have used a camera module for the detection of tomatoes that captures the image of the tomatoes. Image processing algorithms were developed to determine the sizes and location of mature tomatoes including the ones occluded by leaves and branches. On detection of tomatoes, the robotic arm is activated and the robotic arm then picks the tomato and places it inside the container which is attached to the body of the robot. An end effector sub-system is used to pick and place the tomatoes. Once the tomatoes are picked up by the robot an acknowledgment is sent to the farmer via the GSM module. Subsequently, the robot sprays the pesticides on the tomato crops for a better life of the tomatoes. Keywords Image processing · Robotic arm · End effector system · Raspberry Pi 3 · Ultrasonic sensor · Internet of things · Thermal sensor · Camera module · Microcontroller

1 Introduction The main idea of the project is to create a wireless autonomous robot that detects red mature tomatoes and places the tomatoes inside the container attached to its body. The system includes a driving module, power module, microcontroller unit, camera module and a conveyor wheel system. Image processing is used to train the system, S. Salvi (B) · V. Sahu · R. Kalkundre · O. Malwade Bachelors of Engineering in Electronics and Telecommunication, D.J. Sanghvi College of Engineering, Mumbai, India e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2021 S. Choudhury et al. (eds.), Intelligent Communication, Control and Devices, Advances in Intelligent Systems and Computing 1341, https://doi.org/10.1007/978-981-16-1510-8_8

75

76

S. Salvi et al.

capture the mature tomatoes. For image processing, a camera module is used which is integrated with Raspberry Pi (microcomputer). Various sensors were integrated with the microcontroller for determining the characteristics of the farm such as the moisture content in the soil, the surrounding temperature, etc. This data related to the farm is sent to the farmer remotely via the GSM module. The robot was controlled using a DC motor which was driven by the basic electronic circuit whose inputs were programmed using a microcontroller unit like Arduino. We have restricted the motion of robotic arm in two dimensions. For this purpose, a rack and pinion gear system was used. The end effector of the robotic arm is connected to this system. The rack and pinion are motored using the servo motor. The servo motor is programmed using the Arduino according to the requirement of the project. The chassis of the robot and the placement of the electronics system was done in such a way that the entire weight of the robot was concentrated at the centre. The CG of the robot was located at the exact geometrical center which ensured symmetrically balanced turning and also prevented the toppling of the robot. The entire robot was powered using a 12 V lithium polymer (li-po) battery.

2 Literature Review The team researched the projects similar to the desired application, for better implementation and solid design of the robot. Various research papers tackling the problems related to the design of the robot and its arm actuation were read by the team. The factors required for determining the motion of the robot, the design of the end effector system, the weight calculation of the system were done on the basis of various researches done in this field. Our team analysed previously developed robotic systems used for the purpose of harvesting tomatoes [1, 2]. Various papers included the design of a three-dimensional robotic arm that would be used to pick the tomato present in space [3]. The end effector system uses a scissor to cut the branches on which the tomatoes grow and the tomato is effectively dropped into the basket attached to the body. A camera system is used for the detection of tomatoes in the field. The camera module is used for colour as well as shape detection. Image processing techniques are used for the application of detecting the tomatoes and subsequently moving the robot to the place where the tomatoes grow. The robot used a 12-V battery for providing power to the system. Lithium-Iron (LiFePO4) battery was used for powering the system. For extracting the position of the tomatoes in the field, a vision algorithm was developed. This information was used to guide a tomato harvesting robot. The robot collected one tomato at a time. The success rate of the robot was found to be 60%. Object detection and tracking is a widely researched domain with many useful and real-life applications. The main aim of object detection systems is to detect the desired objects in the image along with its localization in the real world and classification. Vision-based software systems also use artificial intelligence and machine learning to further improve the accuracy in the detection and tracking with the help of machine learning algorithms like Scale-Invariant Feature Transform (SIFT), Random

Autonomous Tomato Harvester Using Robotic Arm and Computer Vision

77

Sample Consensus (RANSAC) and Convolutional Neural Network (CNN). Using neural networks requires powerful hardware and access to a huge database, which results in a complicated system. Such systems are harder to run remotely and to implement practically in the real world. Another approach is to use purely image processing methods like colour detection with the help of different colour spaces like RGB (Red, Green and Blue), CMY (Cyan, Magenta and Yellow), HSL (Hue, Saturation and Luminance), HSV (Hue, Saturation and Value), etc. and shape detection using morphological transformation methods like dilation and erosion. The advantage of using such an approach is the ease of remotely processing in real time on smaller systems at the cost of reduced precision to some extent.

3 Motivation Farmers are an important part of the survival of our various societies because they provide food and fibre that nourishes and clothes us. They make responsible use of natural resources and utilize both primitive and very advanced technologies to accomplish this. In order to make their life simpler, this project was selected. The amount of labour and time required for harvesting is huge and also the efficiency of harvesting gets reduced because of human body fatigue and human errors. To reduce these errors the robot is developed, which could replace the farmers in the field. The farmer can sit back and enjoy the fruits of their own hard work with the help of this robot.

4 Problem Definition 4.1 Farmers Put in a Lot of Time and Labour with no Use of Technology Agriculture has been the most crucial occupation in the world. 60% of people are linked with this occupation. there haven’t been any technological advancements in this field. Even today farmers primarily rely upon labour for harvesting crops. A large amount of time is wasted in the process of harvesting. A lot of manpower is consumed by this tedious task. Farmers and labourers are constantly exposed to pesticides which are harmful to their health. The farmers are unaware of the temperature, the moisture content and the humidity of the farm which are the essential aspects of growing mature tomatoes. Health of the farmer is at stake as they have to work during extreme climatic conditions. Also, the efficiency required to remove the mature tomatoes is less as it may include human error such as squeezing the tomatoes while plucking them.

78

S. Salvi et al.

4.2 Manpower Shortage There is a shortage of labour in the agricultural industry due to the shift of newer generations to secondary and tertiary occupations. Such a shortage leads to an increase in the cost of the labour. This results in a higher cost of production leading to inflation. Also, it becomes less profitable for the farmers who are the backbone of the Indian economy.

4.3 Lack of Products for the Indian Market India is one of the largest producers and exporters of agricultural products and 70% population works in this sector. Intensive agriculture is practiced in the country and the land is distributed among a lot of owners thus individual holding of land is very less. As a result, heavy agricultural machinery is not suitable for the Indian market. Also, the Indian farmer can rarely afford such farming equipment.

5 Proposed Solution The tomato harvester detects the tomatoes present in the farm with the help of the camera module. The image is captured by the camera and the captured image is then sent to the Raspberry Pi model, which processes the image and determines the shape and colour of the object captured in the frame and compares it with the data set with which the Raspberry Pi is trained. Once the tomatoes are detected a low-level signal is sent to the microcontroller which stops the motion of the robot. On detection of tomatoes by the Raspberry Pi, the data is sent to the microcontroller (Arduino). This microcontroller is then used to actuate the motors which help the robotic arm reach the desired location. For this, there is a dedicated hardwired circuit that manages the speed of the motors by using the concept of PWM and using MOSFET as a switch. Also turning the Tomato Harvester at the desired angle can be done by switching the rotation direction of the wheels. This is achieved by using a relay that switches the power terminals of the motor. The actuation involves the movement of the rack and pinion gear system which is driven with the help of servo motors. The servo motors are geared motors used to cause rotary actuation. The back and forth movement of the rack helps the robotic arm move in a two-dimensional plane. The end effector subsystem of the arm is also controlled using a servo motor. These motors are programmed according to our needs. The servo motor on the effector causes the opening and closing of the claw. This mechanism helps in plucking the tomatoes. We have restricted the motion of the robotic arm to two dimensions to reduce various difficulties. The closing of the claw is activated once the distance between the end effector and the tomato is less than a threshold value. This distance

Autonomous Tomato Harvester Using Robotic Arm and Computer Vision

79

Fig. 1 Overview of the system

is measured with the help of an Ultrasonic sensor. When the distance between the object is less than a threshold value, a high pulse is given to the servo mounted on the end effector. This causes the servo to rotate and the claw is accordingly closed. Another application of the robot is to use a GSM module for transmitting the data wirelessly to the farmer. The wireless data transmitted includes the temperature and humidity of the surroundings as well as the moisture content present in the soil. This helps the farmer know the ambient condition of the farm and the moisture content helps the farmer know whether watering of crops is required or not. Thus, proper care of the crops can be taken with the help of these sensors. After the collection of all the tomatoes in the farm, an acknowledgement message is sent to the farmer using the GSM module so that the farmer could collect all the gathered tomatoes from the basket (Figs. 1 and 2).

6 Project Description 6.1 Motors 6.1.1

DC Motor

The DC motor is a transducer that converts electrical energy into mechanical energy. These are most commonly used for driving wheels in robotic systems. Two separate DC motors are used on both the rear wheels of the robot. The purpose of using two

80

S. Salvi et al.

Fig. 2 Detailed working of the system

separate motors is to provide an independent driving mechanism to the wheels and for giving it a turning feature [1, 2]. The module used had a gear reduction system which reduced the RPM (Revolutions Per Minute) of the wheels and since the power it receives is constant the torque provided by the motor gets increased (Power = Torque * RPM). The diameter of the wheel was chosen with respect to our speed requirement. The speed chosen was 0.01– 0.03 m/s. By this, we could calculate the diameter of the wheel as follows (r*RPM = v*60). We used a conveyor belt that was connected to the left and right wheels separately. This was done in order to keep the driving force at the center [4, 5].

6.1.2

Servo Motor

In order to achieve precise control or angular or linear position, velocity, and acceleration, a servo motor is used [3, 6]. A servo motor is a rotary or linear actuator. It consists of a suitable motor coupled to a sensor for position feedback. Servo motors

Autonomous Tomato Harvester Using Robotic Arm and Computer Vision

81

Fig. 3 A 2 axis robot arm manipulators

are used in closed-loop control systems. The servo motor is controlled using an Arduino. Servo motors are being used in the system to actuate the rack and pinion gear movements thus causing the linear motion of the robotic arm. Also, servo motor is used to control the movement of the end effector subsystem of the robotic arm. The servo motor used in the system is a 360-degree rotational motor. The servo motors are controlled by PWM (pulse width modulation) signals. The servo responsible for the movement of rack and pinion is a 360-degree rotational and the one connected at the end effector part of the robotic arm can rotate for only 270° (Fig. 3) [4, 5].

6.2 Motor Driving Circuit Since the motion of the robot is in a two-dimensional plane it has to take turns and thus the direction of the wheel has to be reverted. For this, we used a DPDT (Double Pole Double Throw) relay, a MOSFET and a resistor. The Arduino controls the MOSFET which acts like a controlled relay driver. Whenever there is a need to take a turn the Arduino gives a high signal to the gate of the MOSFET and thus the relay gets actuated causing reverting of the power terminals of the motor. This provides the motor a rotation in the reverse direction. When the right wheel is given forward rotation and the left wheel is given backward rotation the robot turns left and when the left wheel is given forward rotation and the right wheel is given backward rotation the robot turns right. The angle of rotation is controlled by the time duration for which the robot is kept in this configuration. The explanatory circuit diagram is as shown. The speed modulation was done by using the concept of PWM (Pulse Width Modulation). The T(ON) and T(OFF) were altered in the code according to our need which in turn changed the duty cycle. The circuitry needed for this involved a resistor, a MOSFET and controlling signal provided by the Arduino (Fig. 4).

82

S. Salvi et al.

Fig. 4 Motor driving circuit

6.3 Raspberry Pi 3 We used the Raspberry Pi 3 model of the credit card sized computer from the Raspberry Pi foundation. The Raspberry Pi 3 is about 50% faster than its predecessor the Pi 2. It includes built-in WIFI and Bluetooth Low Energy connectivity, making it truly an IoT-ready device. The Raspberry Pi 3 comes with a quad-core ARM CortexA53 processor. It is described as having ten times the performance of a Raspberry Pi 1. The purpose of using Raspberry Pi 3 was to initiate faster image processing.

6.4 Wireless Data Transmission GSM module is a fundamental IoT gadget. A GSM module is used for transmitting the data acquired from the sensor and also the acknowledgement message to the farmer or to any associated person handling the farm. The GSM module is used for simple messaging or calling. Once all the data is acquired the GSM module sends the message or calls the associated person. Subsequently, the farmer can analyse the data sent and accordingly manage the condition of the farm. The data sent includes the moisture content of the soil and the surrounding temperature. Accordingly, the farmer can make changes to the farm. For example, if the moisture content of the soil is not sufficient enough for the tomatoes to grow, the water irrigation system can be turned on. This is how the GSM module is helpful for wirelessly transmitting the data.

Autonomous Tomato Harvester Using Robotic Arm and Computer Vision

83

6.5 Sensors 6.5.1

Raspberry Pi Camera Module

The Raspberry Pi camera module is a 5 MP camera designed especially for the Pi. It offers a resolution of 1080p at a framerate of 30 frames/sec. The camera module is cheap and robust and is the best option for image processing on the Pi.

6.5.2

Ultrasonic Sensor

Using the ultrasonic waves, the distance is calculated by this sensor. The sensor consists of two heads, one for transmitting the ultrasonic waves and one for receiving the waves reflected from the object. The sensor emits the ultrasonic waves which is then reflected from the object surface, which are captured by the receiving head of the sensor. Accordingly, the distance is calculated. The purpose of using this sensor is that it helps in calculating the distance based on which the servo actuation is initiated. Also, this sensor is used to find the distance between the obstacle and the robot and accordingly the robot can switch its way.

6.5.3

Contactless Temperature Sensor

Infrared temperature sensors sense electromagnetic waves in the 700–14,000 nm go. While the infrared range stretches out up to 1,000,000 nm, IR temperature sensors don’t quantify over 14,000 nm. The sensor senses the electromagnetic rays emitted by the object whose temperature is to be determined. The sensor emits infrared rays centring on the object and the transmitted rays are detected by the photodetectors for determining the temperature. These photodetectors convert that waves into an electrical signal, which is relative to the infrared vitality radiated by the article. Since the produced infrared vitality of any item is corresponding to its temperature, the electrical sign gives a precise perusing of the temperature of the article that it is pointed at. We used a temperature sensor to measure the temperature of the surroundings and also the temperature of the robotic system. If the system gets heated to a great extent, the system would shut down. The data is wirelessly sent to the farmer using the GSM module.

6.5.4

Humidity and Moisture Detection Sensor

For nominal operation of the electronic system, electrostatic sensitive devices and high voltage devices, humidity also proves to be one for factors disturbing the normal operation of the electronics. Hence, sensing, measuring, monitoring and controlling humidity is a very important task. Irrigation techniques like drip irrigation need

84

S. Salvi et al.

accurate moisture content for plants. In order to ensure that the tomatoes produced are fresh and mature watering of crops is important. Watering needs to be done when the moisture content in the soil is less. This sensor helps in determining the moisture content of the soil and accordingly the crops can be watered.

6.6 Image Processing and Tomato Detection The image processing was done on the Python 3 integrated with Open CV 3.4 on the Raspberrian OS [7, 8]. NumPy library was used to convert the image into arrays and thus making it processable. We create the trackbars in order to change the ranges to detect a specific colour in real time. The image filtration is done by applying an appropriate mask over it that ensures that only the desired colour gets highlighted and the rest of the colours are suppressed. The shape detection algorithm involved the finding of the contours in the beginning followed by the polygon analysis. Later we improve even more the detection, removing all the smalldots detected, which are just noise. If a contour has 3 points is a triangle, 4 points a rectangle and between 10 and 20 points is a circle. The analysis is in such a way that if the number of sides is greater than 10 then it is considered as a circle (Fig. 5) [8–10]. Python: Python is a high-level, general-purpose programming language. It supports multiple programming paradigms like object-oriented, procedural, functional and many more. It offers a simple syntax for the developers. The advantage of using Python is that it has multiple libraries for specific applications. This makes a popular choice. OpenCV: It stands for Open-Source Computer Vision and is a library of programming functions for real-time image and video processing. It has multiple functions for image processing and is compatible with Python. Colour Detection: The Pi camera captures the images at 30 frames per second. To detect the colour, a mask is applied on the captured frame such that only the red part is highlighted. The Fig. 5 Object detection flowchart

Autonomous Tomato Harvester Using Robotic Arm and Computer Vision

85

captured image is in the RGB colour space and is converted into the HSV colour space [7, 11]. This change helps in highlighting the single colour and blocking all other colours from the image. The mask is further divided into two types, one for the lower intensity levels of red and one for the higher intensity levels of red. Trackbars are created and the intensity values of the red pixels in the lower HSV and upper HSV colour spaces are varied to define an optimal value of the mask [9]. As the image is converted to the HSV colour space only the red part in the image is highlighted. Upon optimization, satisfactory results are obtained. The mask-based approach helps to highlight the red tomatoes (Figs. 6 and 7) [7, 11–14]. Shape Detection: In image processing, detecting the shapes is done with the help of contours. Contours are basically all the points along the boundary of an object in the image with the same colour or intensity. Masking the image helps to improve the accuracy of contour detection as the image is now converted to a binary one and there are only two colour levels. This helps in realizing a sharper edge whose detection becomes easy. Fig. 6 Image processing and object detection

86

S. Salvi et al.

Fig. 7 Raspberry Pi 3 and the camera setup

Upon detecting the contours, the number of vertices plays an important role in identifying whether the object of interest is a square or a rectangle or a circle [13]. There is an available function in OpenCV to find contours in the image. We can define the contours by only the vertices points without storing all the pixels along the boundary, this process is used in almost all shape detection applications and is known as contour approximation. CHAIN_APPROX_SIMPLE is the argument that is passed for contour approximation. RETR_TREE is the argument that is passed in this function and helps to create hierarchy in the contours. For improving the accuracy of contours morphological transformation is used on the image. The concept of erosion is used to make the contours accurate and improve object detection and tracking (Fig. 8) [12, 13]. Software Flowchart: Optimization of results: The optimal values obtained for the hue, saturation and value for the lower and upper values of the intensity level of red depends on the lighting conditions. To make sure that the results are accurate there should be ambient lighting conditions. The following results were obtained and the values for the trackbar were set. The optimal results obtained are as follows (Figs. 9 and 10):

Autonomous Tomato Harvester Using Robotic Arm and Computer Vision

87

Fig. 8 Flowchart of software part

7 Limitations The robot designed can be used to perform all the basic tasks required in the profession of farming but always there is a scope for improvement. There are places where we have limited our application and can be further improved. One of the major problems is the movement of the end effector. The movement of the robotic arm is restricted to only two dimensions. The robotic arm can make movements in the forward and backward direction. Also, it is difficult for the system to capture the tomatoes occluded in the leaves. These tomatoes are not captured by the camera and thus are not plucked by the robotic arm. Another problem is the waterproofing of the system. During the rainy season, the electronics of the system may get damaged in the presence of water. Also, the system would stop functioning once the battery is discharged.

88

S. Salvi et al.

Fig. 9 Detection of tomato in frame

Fig. 10 Mask for colour detection

The project uses image processing techniques for detecting the shape and the colour. The RGB is converted to the HSV colour space and the mask is applied. Masking has the disadvantage that for smoother images that is an image with a smaller number of edges smaller changes in noise results in visual distortions. Also, it results in coronas or overdrive artifacts which appear due to sharp transitions in a spurious image. The limitations of active contours are that they neglect minute details

Autonomous Tomato Harvester Using Robotic Arm and Computer Vision

89

in an image in the process of reducing the energy which results in the reduction of accuracy. Also, the overall method implemented in the project is simple due to the fact that the area of application for such a system does not require higher accuracy.

8 Future Scope The field of robotics and machine learning is going to be the future. Robots are going to replace humans in almost each and every domain. 70% of the world’s population has occupied themselves in the field of agriculture. Also, farming requires a lot of labour and time. A lot of manpower is required for farming. The system proposed by us can help solve all these problems but there is always a scope for improvement. The future scope includes, how the system can be effectively improved by overcoming the limitations. The robotic arm can be made to access three dimensions of the space. This would cause movement in all three directions and all the points in the space can be accessed. The system can be trained more efficiently using machine learning and artificial intelligence and can be used to perform various tasks. The system can be made waterproof and thus the system can be used during extreme weather conditions. We can use solar panels and hence the issue of battery discharging can be resolved. Also, the robot can be used for removing the weed (extra waste crops) and sowing of the seed operation can also be performed. The use of Artificial Intelligence (AI) and Machine Learning (ML) in object recognition and its tracking has significantly increased. Neural networks are trained on large datasets with the help of various algorithms to improve the accuracy of object detection. Also, with machine learning algorithms multiple objects can be detected at a time. One can also train the software in such a way that the tomatoes can be plucked on the basis of analysing the different stages of ripening. Conventional neural networks are widely being used in object tracking in autonomous vehicles and many more autonomous applications. Planetary rovers used for exploration are made autonomous with the help of image processing and artificial intelligence and are used to improve the accuracy of the kind of analysis which is performed. Signal processing and its analysis is the basis of image processing and with an increase in the computing power of the hardware more complex and sophisticated systems are being developed with the help of artificial intelligence. Furthermore, such systems can also be used to monitor the health of the crop along its growth stage to ensure maximum productivity. The domain of image processing and machine vision has an immense number of applications in real life.

90

S. Salvi et al.

9 Conclusion We aim to come up with a system which would suffice the need for farming in every corner of the world and reduce the workload of the farmers. Also, we aim to increase the efficiency in terms of harvesting and introduce as few errors as possible. The time taken and the amount of labour required for farming can thus be reduced thereby reducing the manpower and the workload of farmers. Also, the analysis of the farm can be done which can be proven beneficial for the farmers. Farmers can remotely access the field and perform desired operations on the farm. Our aim is to make an object detection and tracking system that provides a significant amount of accuracy to the user. We have tried to make the system simple yet robust such that it finds application in many other real-life applications. The above system works satisfactorily for the intended application of tracking and detecting tomatoes. As discussed above the system can be improved with the help of AI. The project is made so that it also stays affordable for the farmers and is relevant to their application. Furthermore, we aim to improve the project by improving the accuracy of the solutions discussed above.

References 1. de la Peña, C.: The tomato harvester. Boom Spring 3(1) (2013) 2. Taqi, F., Al-Langawi, F., Abdulraheem, H., El-Abd, M.: A cherry-tomato harvesting robot 3. Ling, P., Ehsani, R., Ting, K.C., Chi, Y.-T., Ramalingam, N., Klingman, M.H., Draper, C.: Sensing and end-effector for a robotic tomato harvester 4. Liang, J., Zhang, G., Chen, X., Wang, D.: Development of two degrees of freedom deterministic parallel robotic arm unit 5. Thomas, R.O., Rajasekaran, K.: Remote control of robotic arm using Raspberry Pi 6. Yenorkar, R., Chaskar, U.M.: GUI based pick and place robotic arm for multipurpose industrial applications 7. Benkhaled, I., Marc, I., Lafon, D.: Colour contrast detection in chromaticity diagram: a new computerized colour vision test 8. Gupta, B., Chaube, A., Negi, A., Goel, U.: Study on object detection using Open CV—Python 9. Liu, X., Dai, B., He, H.: Real-time object segmentation for visual object detection in dynamic scenes 10. Chandan, G., Jain, A., Jain, H.: Real time object detection and tracking using deep learning and OpenCV 11. Veerapur, K.A., Bhat, G.V.: Colour object tracking on embedded platform using Open CV 12. Low, C.H., Lee, M.K., Khor, S.W.: Frame based object detection - an application for traffic monitoring 13. Bhardwaj, A.: Shape size colour estimation of object using OpenCV 14. Celebi, E., Smolka, B.: Advances in low-level color image processing 15. Raghunandan, A., Raghav, P., Aradhya, H.V.R.: Object detection algorithms for video surveillance applications

Impact of Inclination of Path Profiles on the Performance of Electric Vehicles Indu Krishnakumar and Aswatha Kumar Mattur

Abstract Electric vehicles are recognized to be beneficial in different aspects, are considered environmentally friendly as well. They are preferred for futuristic greener driving scenarios over the common internal combustion engines. This study being a part of research on electric vehicles presents elaborate modeling steps required for electric vehicle design. The study is from the context of a system engineering outlook considering control system aspects through simulation approach using MATLAB/Simulink. The observations and results of this study are based on the impact of inclination on the performance of electric vehicle. It is intended to provide a better understanding of the need for continuously monitoring path profiles for improved vehicle handling. Concepts in this work are presented in a manner to help students and researchers working in the field of Electric Vehicle Design and Development. Keywords Electric vehicle · Modeling · Simulation · Inclination · Path profiles · Control system

1 Introduction Electric Vehicles (EV) are slowly finding their significance in the automobile forefront due to their emission-free driving capabilities. EV Research is known to be spread across different areas due to its multidisciplinary nature. Research work may vary from motor characteristics-related studies, battery functional studies, vehicle stability analysis, dynamics, energy and battery management systems, chemical characterizations for battery cells. There still exists a need to understand the fundamentals of EV design involved from an embedded and automotive electronics perspective. I. Krishnakumar (B) · A. K. Mattur Electronics and Communication Engineering Department, Christ (Deemed To Be University), Bengaluru, India e-mail: [email protected] A. K. Mattur e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2021 S. Choudhury et al. (eds.), Intelligent Communication, Control and Devices, Advances in Intelligent Systems and Computing 1341, https://doi.org/10.1007/978-981-16-1510-8_9

91

92

I. Krishnakumar and A. K. Mattur

Several controllers or control system modules function as a separate entity or in unison with several other modules of EV for its enhanced performance. Even though works in the past focus on EV-based control techniques for different applications, this research article focuses on considering ambiguities arising from a researcher’s perspective. Hence, this research work tries to incorporate the essentials required to perceive basic concepts required for designing an EV. Step-by-step approach is used starting from mathematical modeling to validation of simulation results as per the EV design. Here in this article, the importance of inclination of path profile and its effect on the EV performance has been highlighted in detail. Different case studies involving inclination changes and their effect on the EV sub-system parameters are illustrated and explained.

1.1 Fundamentals of Electric Vehicle Before modeling and simulation, basic modules or sub-systems of an EV need to be perceived. Basic EV sub-systems can be as shown in Fig. 1. The main modules of an EV include the powertrain, motor, battery, controller, wheels, and brakes. Controller monitors analyze the responses of different modules, check for unpleasant conditions, react for better performance. EV mathematical modeling is carried out in this work using equations of motions. The following section elaborates on the previous research works under vehicle control and their latest applications. The mathematical modeling and simulation approach used in this study are elaborated in the following sections. Besides, based on the

Fig. 1 EV subsystems

Impact of Inclination of Path Profiles …

93

design, responses for varying inclinations are also discussed in detail. In the final section of the article, conclusions drawn from the study conducted are summarized.

2 Literature Review on EV Research Several researchers have focused on designing and developing control techniques, controllers for the improvement of EV performance. In the EV segment, one of the latest works by [1] considers designing a controller in order to maintain a stable velocity by monitoring the torques and steering angle alone using sliding mode control algorithm. It seems to be different from the idea of using the model predictive control technique [2] by combining the road gradient parameter with vehicle velocity for the purpose of energy management system. An alternative approach is adopted in [3], where vehicle controlling is achieved through the vehicle mass estimator. Furthermore, multimodel-based controller [4] design was studied for a distributed drive electric vehicle with a focus on steering control and performance. In [5], a GPS-based tracking is enabled to the electric vehicle for continuously monitoring the vehicle for energy consumption. Few other notable contributions by considering road-related parameters for vehicle control are tabulated as shown in Table 1.

3 Modeling and Simulation The modeling block diagram involved in this design simulation is illustrated in Fig. 2. EV design consists of three different sub-systems namely Input Drive Subsystem, EV Dynamic subsystem and the Motor and Battery Sub-system. A feedback controller is designed to continuously monitor the vehicle behavior and act according to the situation. Based on drive cycle input, the acceleration and deceleration commands are continuously fed based on pedal movements to the EV Dynamic sub-system. EV Velocity and Displacement are derived from the same, fed to driveline sub-system (includes the motor and battery). Motor torque, Vehicle speed and Battery State of Charge (SOC) are measured and fed back to the input drive system. Based on the situation, the controller is designed to respond and actuate for better performance. EV Performance variation when the vehicle of mass ‘m’ Kg, travels through an inclined path is being analyzed in this work. MATLAB/Simulink software is used for this purpose. EV Acceleration and EV Velocity are the parameters used in the model to calculate the force required to move the EV. Whenever a vehicle moves, inertial and body frames are considered in the design process. Furthermore, EV performance along an inclined plane is evaluated based on acceleration/deceleration, motor torque, motor speed, current drawn by battery. Effect of tire forces is also significant in the performance of the vehicle. Whenever EV travels on an inclined plane, normal direction will always be perpendicular to EV.

94

I. Krishnakumar and A. K. Mattur

Table 1 Literature on EV research from previous years Year of Purpose of study publication/authors

Key insights from work Inferences

Hojin et al. [6]

Proposes model predictive control design for an all-wheel-drive EV

Simulink-CarSim-based modeling approach is used Planar full vehicle model with tire and wheel estimates

Wang et al. [7]

Design of controller by cascaded control approach for improving the lateral stability of EV

Sliding mode control, Feedback correction with Yaw moment control sliding mode control, along enhancement with with horizon optimization torque allocation, 2 DOF vehicle model, tire model, independently actuated wheels

Bian et al. [8]

Proposing the concept of speed planning with friction estimation for EV’s

Friction coefficient estimation with a focus on tire model, Speed planning and control, Processes involved in the injection of torque

Vehicle control while travelling on curved roads is critical due to the reduced coefficient of friction. Speed control is a feasible solution in curved paths

Nguyen et al. [9]

Elaborates on various road surface based response estimation strategies used in Vehicles as a review article

Temporal spatial functions, classification based on surface irregularities, response-based approach, data-driven methods

Areas of Research for Future: 1. Active-suspensionbased studies 2. Machine learning with Artificial Neural Networks 3. Fleet vehicle approach 4. Connected Vehicle Studies

Huang et al. [10]

Proposes the multi-model-based control system design for electronic vehicle

Lateral stability analysis through direct yaw moment control of distributed drive EV, Nonlinear Controller design

Reduced computations due to better controller switching policy Improved control accuracy due to the online tuning approach

Andrei et al. [11]

Improving the road interpretations by enhanced ABS in regenerative mode

In-wheel motor drive EV, fuzzy logic controller concepts, Simulator-based study concepts

Comparative analysis on different braking models with simulations

Lateral and longitudinal control with all-wheel drive EV. Model predictive controller design showed better effectiveness in performance

(continued)

Impact of Inclination of Path Profiles …

95

Table 1 (continued) Year of Purpose of study publication/authors

Key insights from work Inferences

Menhour et al. [12] Model-free approach for longitudinal and lateral control of EV is proposed

Nonlinear algebraic estimation, 3 DOF vehicle model, Software-in-loop concept

Accuracy of model-free approach is higher Highlights the need to think a little deviated from the complex driving dynamics analysis approach

Aalizadeh and Asnafi [13]

Proposes controller design considering uncertain road condition with particle swarm optimization approach

Adaptive neural network control approach, Optimization algorithm, different maneuvers settings were tried for yaw rate estimation

Vehicle handling and control improvement by combining the traditional controller design with artificial neural network

Aksjonov et al. [14]

Develop a control algorithm for vehicle stability control

Anti-Lock Brake Systems (ABS), Electronic Stability Program, Concepts of Fuzzy Logic, 10 Degree of Freedom (DOF) Vehicle Model

Performance visualization by Simulator software, Parameterized for a sports utility vehicle with different braking maneuvers

Luo et al. [15]

Proposes the concept of integrating adaptive cruise control for energy management analysis on hybrid EV

Controller design using Non-linear prediction, Adaptive cruise control for Energy efficiency estimation, Inter-Vehicle Dynamics

Fuel economy, traffic safety, and comfort of ride with the integrated control technique. Experimental validation was carried out for its results

Fig. 2 Block diagram for simulation

96

I. Krishnakumar and A. K. Mattur

The mathematical approach follows the below-mentioned set of equations: Longitudinal force on the vehicle: F1 = F f 1 + Fr 1 − D f 1 − F f s1 − Fr s1 + G f x .

(1)

Normal force on the vehicle: F2 = D f 2 − F f s2 − Fr s2 + G f z .

(2)

When the path inclination angle ‘Ø’ changes, tractive forces along the x- and z-axis changes. The total force on the vehicle due to the road load is represented as Ft. Ft = m ∗ d 2 (x)/dt 2 + Fr .

(3)

     Fr = r1 + r2 ∗ (d(x)/dt) + r3 ∗ d 2 (x)/dt 2 + m ∗ g ∗ sin Ø .

(4)

Total Power: Pt = Ft ∗ (d(x)/dt).

(5)

   Pr = Fr ∗ d(x) dt .

(6)

ax = F1 /m−((d(β)/dt) ∗ z).

(7)

az = F2 /m−((d(β)/dt) ∗ x).

(8)

d(β)/dt = Ty /I y .

(9)

Acceleration is given by

where

Table 2 summarizes the variables used throughout this work in detail. Based on the equations, the EV design for interpreting the performance of the EV on the inclined path was simulated using MATLAB/Simulink.

4 Results and Discussions Following the simulation approach in MATLAB/Simulink, the EV performance variations when a vehicle moves along an inclined path is illustrated in this section of

Impact of Inclination of Path Profiles …

97

Table 2 Nomenclature of variables ax

Acceleration along x-axis

az

Acceleration along z-axis

Ø

Path inclination

Ft

Total tractive force

Fr

Force on the road

F1

Longitudinal force

F2

Normal force

β

Pitch

x, z

Displacements along x-axis and z-axis

g

Gravitational force

m

Vehicle Mass

tr

Response time

Ty

Torque along y-axis I y

Df 1 , Df 2

Drag forces along x- and z-axis

F fs2, F rs2

Normal suspension r 1 forces for front and rear wheels

Rolling resistance r 2 coefficient

Driveline and rolling resistance coefficient

r3

Aerodynamic drag coefficient

Total power

Power due to road surface

Moment of Inertia F f 1, F r1 along y-axis

Longitudinal forces on front and rear wheels

F fs1, F rs1 Longitudinal Gfx , Gfz suspension forces for front and rear wheels

Gravitational forces on x-, z-axis

Pt

Pr

the research article. Three different path inclinations are tested in this study. Case 1 represents 0°, Case 2 for 5° and Case 3 for 10°. The responses are divided into three parts: Vehicle dynamics response, Electrical system response and finally the Battery behavior. Amongst all the EV dynamics parameters, response plots shown in Fig. 3, depicts acceleration, deceleration, drive cycle input and displacement of the vehicle in motion.

Fig. 3 Response for path inclination = 0°

98

I. Krishnakumar and A. K. Mattur

Electrical system plots in the same figure include responses of motor torque, motor speed, front-wheel, and rear-wheel forces. Furthermore, the battery State of Charge (SOC), battery current, and battery power are plotted together as a set amongst the battery parameters. Visible variations are observed with the change in the path inclination. Three different cases are tried out to understand the performance variation. For case 1, where path inclination is 0°, the acceleration demand is relatively low as expected when compared to the cases where the inclinations are higher. The vehicle follows the expected drive cycle profile and reaches the target destination as shown in the displacement plot below the drive cycle variations. The motor torque variations can be seen satisfying the acceleration and deceleration requirements as the case may be during the motion. It can be seen that the motor speed follows a similar pattern with respect to time while meeting the drive cycle pattern exactly. With minimal variations in the battery current drawn for vehicle motion, it is obvious from the plots that the Battery SOC drops from the initial 80% to only around 72% at the end of this vehicle motion. Similarly, the EV behavior at higher inclination is illustrated in Fig. 4. For case 2, with an inclination angle of 5°, clearly the plot for acceleration goes up with a constant positive offset to satisfactorily follow the required drive cycle variations profile. The vehicle reaches the target in the scheduled time interval. However, the expected offset in motor torque is also observed as shown in the plots. The battery current variations can be seen to reach higher peaks thereby affecting a considerable drop in the SOC. Compared to 80–72% drop in Case 1, this Case 2 plots show a drop in SOC from 80 to 42% even though the EV reaches the target by satisfactorily following the drive cycle. Obviously, the expected increase in the battery current drawn and power requirement are also observed as shown in the plots. In case 3, as shown in Fig. 5, case 3 assumes a still higher inclination of 10° for

Fig. 4 Response for path inclination = 5°

Impact of Inclination of Path Profiles …

99

Fig. 5 Response for path inclination = 10°

the path. In this case, the acceleration demands are much higher, and therefore the demand is seen for motor torque as well. Even in this case, the driving cycle profile is met satisfactorily and the vehicle reaches destination distance as per schedule. However, an expected much higher current is drawn from the battery thereby impacting the SOC to drop from 80% to a very low value of 11%. The respective changes in the power plot are also as expected. Table 3 provides a detailed comparison of parameters (in terms of range) like motor torque, acceleration, deceleration, forces on wheels, battery current, and battery SOC when subjected to different inclinations. The observations bring out the actual variations which can be expected in an EV performance in terms of Torque, Speed, Battery Currents, and Battery SOC. It can be noted that, along with the satisfying profiles of drive cycle, variations in EV displacement are considered within the design. In addition to all these, the plots also illustrate the variations in the forces acting on the front and rear wheel axes in this typical study considering front-wheel drive system. Table 3 Simulation parameters and their range Parameters

Accel./Decel. (m/s2 )

Variation in Variation in motor front- wheel torque (Nm) force (N)

Variation in rear- wheel force (N)

Battery current (A)

Battery SOC (%)

Case 1: 0°

0.4/0.4

+130 to −90 ±2500

0 to −500

+100 to −70

80–71

Case 2: 5°

0.7/0.2

+220 to −45 +4000 to − 5000

+2500 to − 4000

+175 to −25

80–42

Case 3: 10°

1.0/0.0

0 to 230

0

0–230

80–11

0 to +5000

100

I. Krishnakumar and A. K. Mattur

5 Conclusion This research article tries to touch upon the basics of an electric vehicle design from a control system perspective. Recent advancements and notable works in the area of electric vehicle research were briefly touched upon. Mathematical modeling steps involved in EV design have been explained, followed by a simulation approach using MATLAB/Simulink. The impact of path profile inclination on the electric vehicle performance parameters was keenly observed and discussed. From the case studies, it was perceived that, as the inclination of a path changes, significant changes are observed in vital electric vehicle dynamics, battery state of charge, motor torque, speed parameters. Moreover, similar studies and analyses with respect to electric vehicle design and development with varying complexities will be continued in the future too. Acknowledgements The authors would like to sincerely thank Christ (Deemed to be University) for providing the necessary facilities for carrying out research.

References 1. Liang, Z., Zhao, J., Dong, Z.: Torque vectoring and rear-wheel-steering control for vehicle’s uncertain slips on soft and slope terrain using sliding mode algorithm. IEEE Trans. Veh. Tech. 69, 3805–3815 (2020) 2. Guo, J., He, H., Sun, C.: ARIMA-based road gradient and vehicle velocity prediction for hybrid electric vehicle energy management. IEEE Trans. Veh. Tech. 68, 5309–5320 (2019) 3. Sun, S., Zhang, N., Walker, P., Lin, C.: Intelligent estimation for electric vehicle mass with unknown uncertainties based on particle filter. IET Int. Trans. Sys. 14, 463–467 (2020) 4. Shi, K., Cheng, D., Yuan, X.: Interacting multiple model-based adaptive control system for stable steering of distributed driver electric vehicle under various road excitations. ISA Trans. 103, 37–51 (2020) 5. Liu, K., Yamamoto, T., Morikawa, T.: Impact of road gradient on energy consumption of electric vehicles. J. Trans. Res. Part D: Trans. and Env. 54, 74–81 (2017) 6. Hojin, J., Beomjoon, P., Seibum, C.: Model predictive control of an all-wheel drive vehicle considering input and state constraints. Int. J. Auto. Tech. 21, 1–10 (2020) 7. Wang, Y., Zhenpo, W., Lei, Z., Mingchun, L., Jingna, Z.: Lateral stability enhancement based on a novel sliding mode prediction control for a four-wheel-independently actuated electric vehicle. IET Int. Tran. Sys. 13(1), 124–133 (2019) 8. Bian, C., Tong Z., Guodong Y., Liwei X.: Integrated speed planning and friction coefficient estimation algorithm for intelligent electric vehicles. Algorithms 12(2) (2019) 9. Nguyen, T., Bernhard, L., Yiik D.W.: Response-based methods to measure road surface irregularity: a state-of-the-art review. Euro. Trans. Res. Rev. 11(1) (2019) 10. Huang, G., Xiaofang Y., Ke, S., Xiru, W.: A BP-PID controller-based multi-model control system for lateral stability of distributed drive electric vehicle. J. Fran. Ins. Elsevier Ltd.356(13), 7290–7311 (2019) 11. Andrei, A., Valery, V., Klaus, A., Eduard, P.: Design of regenerative anti-lock braking system controller for 4 in-wheel-motor drive electric vehicle with road surface estimation. Int. J. Auto. Tech. 19(4), 727–742 (2018)

Impact of Inclination of Path Profiles …

101

12. Menhour, L., Brigitte, D.A.-N., Michel, F., Dominique, G., Hugues, M.: An efficient modelfree setting for longitudinal and lateral vehicle control: validation through the interconnected pro-SiVIC/RTMaps prototyping platform. IEEE Trans. Int. Trans. Sys. 19(2), 461–475 (2018) 13. Aalizadeh, B., Asnafi, A.: Combination of particle swarm optimization algorithm and artificial neural network to propose an efficient controller for vehicle handling in uncertain road conditions. Neu. Com. Appl. 30(2), 585–593 (2018) 14. Aksjonov, A., Klaus A., Valery V.: Design and simulation of the robust ABS and ESP fuzzy logic controller on the complex braking maneuvers. App. Sci. 6(12) (2016) 15. Luo, Y., Tao, C., Shuwei, Z., Keqiang, L.: Intelligent hybrid electric vehicle ACC with coordinated control of tracking ability, fuel economy, and ride comfort. IEEE Trans. Int. Trans. Sys. 16(4), 2303–2308 (2015)

Design of Environment-Friendly FIR Filter and Its Implementation on FPGA for Green Communication Anoop Raghuvanshi, Kuber Singh Mehra, and Indra Singh Bisht

Abstract In this method, we design a FIR filter with the help of DSP slice Spartan 7 of FPGA family. When the filter is operating for higher clock frequencies like here as 50 GHz, it dissipates more power. At low clock frequencies comparatively, less energy is wasted for DSP, Leakage, input–output and total power. So, this range of frequency and the specific load can be used as an environmentally friendly device, which will emit less energy into the atmosphere. There is 51% less energy dissipated when the filter operates from 20 to 2 pF for a clock frequency of 1 GHz. There is 53% less leakage power dissipation is found at a clock frequency of 10 GHz by adjusting the load from 20 to 2pF. Keywords FIR filter · Capacitance · DSP · FPGA · Vivado · Power dissipation · Leakage power

1 Introduction FPGA a computer chip full of open gates is efficient for FIR filter design as they offer custom, fully parallel algorithms. It adapts to changing standards or specifications and offers design changes in the last minutes too. We are using here Spartan 7, which is built on 28 nm technology. The package used is fgga484, which has 484 I/O pin counts, 250 available IOBs, the LUT elements are 32,600 with 65,200 flipflops and 75 block RAMs.

A. Raghuvanshi (B) · K. S. Mehra · I. S. Bisht Birla Institute of Applied Sciences, Bhimtal 263136, India e-mail: [email protected] K. S. Mehra e-mail: [email protected] I. S. Bisht e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2021 S. Choudhury et al. (eds.), Intelligent Communication, Control and Devices, Advances in Intelligent Systems and Computing 1341, https://doi.org/10.1007/978-981-16-1510-8_11

103

104

A. Raghuvanshi et al.

Due to the non-feedback structure of FIR, they are highly stable and also offer exact linear phase, these are the two great advantages due to which they are in huge demand as compared to IIR filter. Here we are using eighth order FIR Gaussian LPF. Here we proposed that by reducing the capacitance of the output load, the power dissipation also decreases. For this kind of a FIR filter, we find that: Power dissipation α Load Capacitance.

In order to make less heat dissipation circuit, one can reduce the load capacitance. Here the emphasis is given to adjust the capacitance and to attain the FIR filter with minimum power dissipation at 0 pF. In Sect. 2, the past related work on this domain is discussed. In Sect. 3, we see the effect of variation in the load capacitance of FIR filter. In Sect. 4, the results are concluded for various iterations. Section 5 suggests the future scope for this research.

2 Literature Review FIR filter is implemented on Kintex-7 FPGA, which uses voltage scaling to design an energy-efficient Gaussian filter [1]. In [2], it is proposed that dynamic power consumed by a digital CMOS circuit is directly proportional to switching activity and payload capacitance. Redundant Binary Signed Digit (RBSD) number system enables fast computing by reducing carry propagation time [3]. There is no change in clock power, logic power, signal power and DSP power on capacitance scaling. But there is a significant reduction in IOs, leakage and total power of FIR filter on 28 nm Kintex-7 FPGA [4]. The computational structures are proposed [5] based on the theory of fast algorithms for short linear convolutions, which are suitable for the implementation L-block digital filters. In [6], without sensing the capacitor voltages, the sensorless capacitor voltage balancing control (SCVBC) is proposed by which the total number of the feedback signals is saved. FPGA-based system is used to implement the SCVBC digitally. An iterated short convolution (ISC) algorithm that is based on the mixed-radix algorithm and fast convolution algorithms is proposed in [7]. A new hardware efficient fast parallel finite-impulse response (FIR) filter structure is achieved by transposing the ISC-based linear convolution structure. Large amount of hardware cost is reduced by this, especially for the case when the length of the FIR filter is large. Different methods are adopted in [8] to make the filter energy efficient. Methods used in this are capacitance scaling and frequency scaling. Scaling of the output load from 5000 to 0 pF shows the effect of output load on on-chip and off-chip power consumption of FIR filter in [9], using 20 nm technology-based FVA1156 package and Kintex-7 family. A parallel FIR filter structure was designed in [10], in which an adjacent coefficient sharing-based sub-structure sharing technique is proposed, which is used to reduce the hardware cost of parallel FIR filters. FIR filter is implemented in [11] on 28 nm FPGA and the effect of in-built mechanism of Air Flow controller is analyzed.

Design of Environment-Friendly FIR Filter …

105

3 Load Capacitance and Its Role in Energy Efficiency of FIR Filter Operating the FIR for large values of capacitive load increases the total on chip power dissipation. The dynamic power is almost half when operating the filter from 20 to 2 pF and total on chip power is also reduced (Table 1). The variation of the on-chip power dissipation is shown for the load variation. For 1 GHz of clock frequency almost constant difference is obtained for dynamic and static power (Fig. 1). The analysis of the data in Table 2 is done at a clock frequency of 1 GHz of the FIR filter. The leakage power is reduced to 53% when operating the filter at 2 pF. So, huge amount of energy is saved. The I/O power dissipation is reduced to 43%. A clear view of the power profile is seen in Fig. 2, which suggests that on chip power is very high for 20 pF capacitance load. For 10 GHz clock when the device was operated for 20 pF load, the junction temperature reaches 85.3 °C, which is very high temperature and may lead to damage to the device. The junction temperature is within acceptable limits up to 15 pF of load at the same clock frequency. At 2 pF of load, the junction temperature was found to be at 54.8 °C for 10 GHz. The data in Table 3 are obtained when the FIR filter is operated at a clock frequency Table 1 Power dissipation for 1 GHz clock frequency at various load capacitances Capacitance (pF)

Dynamic power (Watt)

I/O power (Watt)

Leakage power (Watt)

Total power (Watt)

2

1.022

0.812

0.071

1.094

10

1.489

1.278

0.072

1.561

15

1.78

1.57

0.073

1.853

20

2.072

1.862

0.073

2.145

Fig. 1 Variation in the power dissipation for various load capacitances at 1 GHz clock frequency

106

A. Raghuvanshi et al.

Table 2 Power dissipation for 10 GHz clock frequency at various load capacitances Capacitance (pF)

Dynamic power (Watt)

I/O power (Watt)

Leakage power (Watt)

Total power (Watt)

2

10.222

8.118

0.095

10.317

10

14.887

12.784

0.121

15.008

15

17.803

15.7

0.144

17.948

20

20.179

18.616

0.176

20.896

Fig. 2 Variation in the power dissipation for various load capacitances at 10 GHz clock frequency

Table 3 Power dissipation for 50 GHz clock frequency at various load capacitances Capacitance (pF)

Dynamic power (Watt)

I/O power (Watt)

Leakage power (Watt)

Total power (Watt)

2

51.109

40.591

0.485

51.594

10

74.437

63.919

0.485

74.922

15

89.017

78.499

0.485

89.502

20

103.597

93.079

0.485

104.082

of 50 GHz. Here it is seen, there is no change in the leakage power of the device for the said range of the load adjustment. Also, the junction temperature is gone beyond the optimum value when operating the device at 50 GHz clock frequency, which might lead to damage to the device. So, it is not suggested to operate the filter for such a high clock frequency. The device for this clock frequency gets overheated. Here it is seen that operating the filter for 50 GHz continuously at the capacitive load of 20 pF leads the junction temperature to reach 125 °C, which is a very high value for any device and is beyond its bearing limit. For almost at any load for a clock frequency of 50 GHz, the junction temperature was found to be 125 °C. The plots for the different values of Table 3 are mapped in Fig. 3. From this, it is

Design of Environment-Friendly FIR Filter …

107

Fig. 3 Variation in the power dissipation for various load capacitances at 50 GHz clock frequency

analyzed that the static power is very low for almost all of the combinations. Table 4 listed the data reveal that for the same load by varying the clock frequency of the filter, there is a significant increase in the dynamic power of the filter. As in Fig. 4, the dynamic power dissipation of the device is shown in the graphical manner. The value keeps on increasing the load from 2 to 20 pF. Table 4 Dynamic power dissipation for various clock frequencies with variation in load capacitance Clock frequency (GHz)

2 pF

10 pF

15 pF

20 pF

1

1.022

1.489

1.78

10

10.222

14.887

17.803

20.179

50

51.109

74.437

89.017

103.597

Fig. 4 Dynamic power at different clock frequencies with load variation

2.072

108

A. Raghuvanshi et al.

As the FPGA consists of configurable logic blocks also known as CLBs, input– output block and there are reconfigurable interconnect also has fixed function logic blocks (Tables 5 and 6). As shown in Fig. 5, there is a significant increase in the graph, is observed, which is also higher for 50 GHz as compared to 1 GHz of operating clock frequency. This can be controlled by using 2 pF load and found minimum at 1 GHz. The Verilog code for the FIR filter is written according to the design flow as mentioned in Fig. 6. The junction temperature was found to be 28.2 °C, 29.5 °C, 30.4 °C and 31.2 °C for the load values of 2 pF, 10 pF, 15 pF and 20 pF, respectively, which are the accepted values for the designed filter. These are the optimized values for the filter design. The total power is actually the combination of on chip dynamic, static and input– output powers. The optimum results are obtained at a clock frequency of 1 GHz and at a load capacitance of 2 pF. Table 5 I/O power dissipation for various clock frequencies with variation in load capacitance Clock frequency (GHz)

2 pF

10 pF

15 pF

20 pF

1

0.812

1.278

10

8.118

12.784

15.7

18.616

50

40.591

63.919

78.499

93.079

Table 6 Device utilization summary of FIR filter on FPGA

1.57

1.862

Resource

Utilization

Available

LUT

173

32,600

0.53

FF

64

65,200

0.1

I/O

28

250

Fig. 5 I/O power at different clock frequencies with load variation

Utilization (%)

11.2

Design of Environment-Friendly FIR Filter …

109

Fig. 6 Design flow for the Verilog code

4 Conclusion We have seen here the design and analysis of FIR filter, which is done with the help of Vivado design suite. The Verilog code is generated for the said filter, which is synthesized and implemented in the design suite. The various power reports of the design are observed, which are mentioned here. From these analyses, it is found that operating the filter at the lowest possible clock frequency and at low value of capacitive loads, the power dissipation profile of the filter is found very low. On increasing the clock frequency, the power dissipation increases. Also, at higher values of load capacitance, more power is radiated into the atmosphere. At load capacitance of 2 pF and at 50 GHz of clock frequency, the optimum energy is radiated into the atmosphere by the device. Also, the junction temperature is within the acceptable limit of 29.2 °C at 2 pF and at 1 GHz. By this, the device life is also increased.

5 Future Scope Here we have implemented the design of the FIR filter on Spartan 7 FPGA DSP slice of FPGA family [12], which will be environmentally friendly. The technique can also be used to design and implement other circuits with energy efficiency like UART, packet counter, Router, Ethernet media access counter, key generator, for error correction and even more. We have used here Spartan 7 FPGA, this design can be re-implemented with other FPGA like Kintex and Vertex.

110

A. Raghuvanshi et al.

References 1. Kumar, T., Pandey, B., Das, T.: Voltage scaling based energy efficient FIR filter design on FPGA. Int. J. Curr. Eng. Technol. 200–203 (2014) 2. Wang, X., Yu, M.: Power research of JPEG circuits in FPGA. in: Seventh International Conference on Intelligent Information Hiding and Multimedia Signal Processing, pp 206–208 (2011) 3. Behl, A., Gokhale, A., Sharma, N.: Design and implementation of fast booth-2 multiplier on Artix FPGA. In: International Conference on smart sustainable Intelligent Computing and Applications under ICITETM2020, pp 140–148 (2020) 4. Pandey, B., Kumar, T., Das, T., Yadav, R., Pandey, O.J.: Capacitance scaling based energy efficient FIR filter for digital signal processing. In: IEEE International Conference on Reliability Optimization & Information Technology (ICROIT) (2014) 5. Jose, I.: ACHA, computational structures for fast implementation of l-path and l-block digital filters. IEEE Trans. Circ. Syst. 36, 805–812 (1989) 6. Chen, H.-C., Liao, J.-Y.: Design and Implementation of Sensor less Capacitor Voltage Balancing Control for Three-Level Boosting PFC. IEEE Trans Power Electronics 29, 3808–3817 (2014) 7. Cheng, C.: Hardware efficient fast parallel FIR filter structures based on iterated short convolution. IEEE Trans Circuits Syst 51, 1492–1500 (2004) 8. Pandey, B., Jain, A., Kumar, A., Kumar, P., Hussain, A., Levy, J., Chowdhry, B.S.: Energy efficient and high–performance FIR filter design on Spartan–6 FPGA. In: 3C Tecnología Glosas de innovación aplicadas a la pyme, pp 381–449 (2019) 9. Pandey, B., Pandey, N., Kaur, A., Akbar Hussain, D.M., Das, B., Tomar, G.S.: Scaling of output load in energy efficient FIR filter for green communication on ultra-scale FPGA. Wireless Pers Commun (2018) 10. Parker, D.A., Parhi, K.K.: Low-area/power parallel FIR digital filter implementations. J. VLSI Sig. Process. 76–92 (1997) 11. Pandey, B., Kumar, T., Das, T., Islam, S.M.M., Kumar, J.: Thermal mechanics based energy efficient FIR filter for digital signal processing. Appl. Mech. Mater. 612, 65–70 (2014) 12. Xilinx Spartan-7 FPGAs Data Sheet: DC and AC Switching Characteristics, https://www.xil inx.com/support/documentation/data_sheets/ds189-spartan-7-data-sheet.pdf

Atomistic Simulation Studies of the Mechanical and Thermal Properties of Silver Nanowires as Interconnects for Nanoelectronics Applications S. K. Joshi, D. P. Singh, and Santosh Dubey Abstract The increasing demand for more compact electronic devices is stimulating the researchers to develop nanoscale circuit components. Nanowires are molecularscale connection wires, which carry current inside these small circuits. The present work examines the deformation and melting behavior of silver nanowires with classical molecular dynamics simulations. Large Atomic/Molecular Massively Parallel Simulator (LAMMPS) has been used to simulate silver nanowires of different dimensions at selected temperatures. The deformation behavior is investigated by applying uniaxial stress to the nanowire, and recording the resulting strain in it. The nanowires are also subjected to incremental heating up to 1600 K to understand their melting characteristics. The change in the potential energy with temperature has been used to understand the melting properties of the silver nanowires. It has been found that the melting temperatures of all the nanowires, investigated in this exercise, are lower than their bulk counterparts. Keywords Molecular dynamics simulation · LAMMPS · Deformation · Melting

1 Introduction The ever-increasing demand for efficient and development of miniaturized nanoelectronic devices has attracted noticeable research activity in recent years. Metallic nanowires, due to their attributes like large surface area, high electrical and thermal conductivity, high yield stress, and elastic modulus compared to their bulk counterparts, find myriad applications in optoelectronics devices, nano-electromechanical device fabrication, nanocomposite materials, sensors for detecting toxins of biological and chemical nature, etc. [1–5]. Specifically, metallic nanowires are found to be the most promising constituents for future nanoscale and integrated circuit-based devices. S. K. Joshi · D. P. Singh · S. Dubey (B) Department of Physics, University of Petroleum and Energy Studies, Dehradun 248007, Uttarakhand, India e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2021 S. Choudhury et al. (eds.), Intelligent Communication, Control and Devices, Advances in Intelligent Systems and Computing 1341, https://doi.org/10.1007/978-981-16-1510-8_12

111

112

S. K. Joshi et al.

Metallic nanowires are being used for nano-MOSFETs (metal oxide semiconductor field effect transistor), which helps in excellent electrostatic gate control for reducing the short channel effects [6]. Nanowires are also of great interest in photovoltaics, due to the attributes like larger surface area, high aspect ratio, enhanced antireflection effect (which increases their capacity of light absorption), and their capability to direct light absorption with specifically designed nanowire networks. Such properties become quite useful in solar cells where one is interested in enhancing the absorption of incident energy by the solar panels [7, 8]. Logic-gate structures like OR, AND, and NOR gates, etc. have been developed by nanowire junction arrays. These gates have substantial gain and may be used to implement the routine computation tasks [9]. In addition, nanotubes have also been researched quite extensively for their use as photon waveguide in quantum dot/quantum effect well photon logic arrays. In these arrays, photons are guided inside the nanotube, whereas electron moves on outside shell [10]. Several nanocomposite materials also utilize metal nanowire networks as additives to enhance the strength of flexible nanoelectronic devices [11]. Silver nanowires (AgNWs), in particular, offer a higher yield strength as compared to the bulk silver [12], which is attributed to the absence of dislocations in these nanostructures. This property makes their use appropriate in applications where mechanical stability is crucial (like in soft wearable electronic devices) [13, 14]. AgNW networks have potential to replace various transparent electrodes (TEs) used nowadays like Indium Tin Oxide (ITO) and Fluorine doped in Tin Oxide (FTO) due to various advantages related to their synthesis and stability properties. First, there are a variety of methods available for the synthesis of AgNWs. This diversity in synthesis further widens their compatibility with different substrate materials. Second, silver is more stable mechanically and chemically, as compared to other metals. In addition, compared with carbon nanotubes, AgNWs can create network with better electrical conductivity. Molecular Dynamics (MD) is an efficient computational technique to simulate the nanostructures and understand their properties. Recently, MD has been used extensively for the study of tensile [15–22] and melting properties [17, 23–28] of the nanowires of Cu, Ni, Sn, Ti, Au, etc. However, only a few articles investigating the tensile/melting properties of silver nanostructures using molecular dynamics are available in the literature. Alam et al. [29] studied the fracture mechanism of a single and polycrystal AgNW and observed that the polycrystalline nanowire shows a significant amount of ductility. Shi et al. [30] monitored the effect of superheating on Ni-coated AgNW with MD simulation. This article aims at providing the simulation results on material property considerations of AgNW, which may be useful for a user planning to design a nanoelectronic system. Published research on the size dependence of the tensile and melting properties for AgNWs is not widely available. Flexibility is an important factor for transparent electrodes in soft electronics, for example in wearable sensors. AgNWs have comparatively better mechanical properties, as compared to other nanowires; however, in order to use them in various applications, the size

Atomistic Simulation Studies of the Mechanical …

113

and temperature-dependent material property data are required. Melting characteristics of the nanowires are important because due to high current density inside a nanoscale interconnect, joule heating may result in a significant increase in temperature. Increase in temperature may also result in increased material diffusion, which may initiate local discontinuities in the nanowire geometries, and eventually localized hot spots develop. These hot spots may further increase material diffusion. An accurate monitoring of these physical mechanisms and the resulting melting process in nanowires is, therefore, important for their subsequent design and development of nanoelectronic systems. Current work offers a computational approach to monitor mechanical and thermal properties of AgNWs, keeping in mind their prospective use in nanoelectronics. In the present work, the deformation and melting behavior of [101] oriented AgNWs have been investigated using LAMMPS MD simulator [31]. All the visualization of strained/heated nanowires has been done using Open Visualization Tool (Ovito) [32]. Section 2 deals with the simulations methodologies and results are detailed in Sect. 3.

2 Simulation Methodology In the present work, Classical Molecular Dynamics simulator LAMMPS is used to model the properties of AgNWs. In molecular dynamics, Newton’s equations of motion are solved to observe the time evolution of interacting particles. For ith atom of the system, having mass m i , the force Fi may be given as: → ri d 2− − → Fi = m i dt 2 − → −−→ Fi = −∇i V , − → here, the force Fi is derived from the average potential energy V (r); r = r1 , r2 , . . . , r N represents the positions the various atoms. The pair interactions have been modeled using Embedded Atom Method (EAM) potential function [33]. Periodic boundary conditions are applied along the length of the nanowire and fixed boundary conditions are applied along the remaining two directions. All the atomistic simulations were performed considering NVT ensemble in a temperature range of 300–1600 K.

114

S. K. Joshi et al.

3 Results and Discussion The results section has been divided into two subsections: in the first subsection, the deformation properties of the nanowires under the uniaxial tension have been presented. The effect of parameters like temperature and diameter on the deformation properties of the nanowires has been subsequently discussed. In the second subsection, the melting properties of AgNWs are presented. All the visualization of strained/heated nanowires has been done using Open Visualization Tool (Ovito).

3.1 Deformation Properties The simulation parameters used to study the deformation properties have been displayed in Table 1. The nanowire is equilibrated approximately for 250 ps, before applying the strain. Subsequently, a strain of 10−9 /s is applied along the length of the nanowire (y-axis), and the longitudinal stress is calculated. Figure 1 shows stress–strain curve for the AgNWs having diameters in the range Table 1 Simulation details for AgNW deformation study

Parameter

Value

Length

Infinite

Diameters

1–3 nm

Temperature range

300–800 K

Strain rate

109 s−1

Time step

0.005 ps

Fig. 1 Stress versus strain curves for AgNWs of diameters 1–3 nm, at 300 K

Atomistic Simulation Studies of the Mechanical …

115

Fig. 2 Lateral cross section snapshots of atomic configuration at various strains a 1.0% b 9.0% c 11% d 15% e 17% f 20.0%

1–3 nm. The simulation of the nanowires deformation has been done at a temperature of 300 K; other simulation parameters are given in Table 1. As shown in this graph, for 2 nm nanowire, the yield stress is obtained at around 1.4 GPa, corresponding to a strain of 0.09; beyond which the plastic zone is developed. The effect of uniaxial strain may be better understood from the snapshots captured at various stages of the simulations (Fig. 2). For small strain values, the deformation is not visible since the system is still in elastic regime. As strain increases, the deformation in the nanowire may be observed; in this case, the slipping of atomic layers in [111] direction may be visible from the figure. During this process, many atoms can be seen accumulating in the neck region. Figure 3 gives the stress–strain graph for a nanowire of diameter 2 nm, in the temperature range of 300–800 K. It is observed that at a constant strain rate, the yield stress decreases as we increase the temperature of the nanowire. There is no significant change in the yield strain (Table 2). As the temperature is raised from 300 to 800 K, the decrease in the yield stress (w.r.t. the value at 300 K) has been found to be more than 53%, whereas decrease in yield strain (w.r.t the strain at 300 K) is only 38%. Decrease in the strength of the nanowires at higher temperatures may be due to the faster material recovery over the dislocation nucleation rate [34].

3.2 Melting Properties The effect of cross-section and length on the melting properties of AgNWs has also been studied. The simulation parameters used in this study are given in Table 3. It can clearly be seen that the potential energy (PE) curve (see Fig. 4) passes through three different stages.

116

S. K. Joshi et al.

Fig. 3 Stress versus strain curves for Silver nanowire at different temperatures

Table 2 Yield stress and yield strain at different temperatures for AgNW of 2 nm diameter

Table 3 Simulation parameters for AgNW deformation

Temperature (K)

Yield stress (GPa)

Yield strain

300

1.38

0.09

400

1.16

0.084

500

0.98

0.082

600

0.84

0.079

700

0.76

0.068

800

0.64

0.055

Parameter

Value

Length

20 nm

Diameter

1–3 nm

Temperature range

300–1800 K

Heating rate

3 × 10−3 Kelvin/ps

Time step

0.005 ps

Stage-1: In this stage, as the temperature of the AgNW is increased gradually from 300 to 1075 K, their potential energy also increases. Stage-2: The second stage started at temperature 1075 K and ended at the temperature of 1175 K. In this stage, the rate of potential energy increase

Atomistic Simulation Studies of the Mechanical …

117

Fig. 4 Temperature versus Potential Energy graph for the AgNW of diameter 2 nm

  ∼ d(PE) , where T is the temperature is more than in the first stage. Therefore, dT 1075 K may be taken as the starting point for the surface pre-melting process. The steepest part of the curve, starting from 1125 K, may be considered as the start of the quasi-liquid state, where the physical state of the nanowire seems to be in between liquid and solid state. Stage-3: This stage starts at a temperature higher than 1125 K, when the slope of the graph decreases. However, the slope is still higher slightly, as compared to that for the first stage. All these stages can also be recognized from the cross-sectional and lateral snapshots of the nanowires captured at these temperatures (Fig. 5). The melting temperature of the AgNW corresponds to the point in the potential energy vs temperature diagram (Fig. 4), where the potential energy changes abruptly due to the heating process. The effect of nanowire size on its melting behavior has been shown in Fig. 6. From this figure, we see that the potential energy increases as we decrease the diameter of the nanowire, which might be due to increase in surface energy with decreasing dimensions [22]. It is further observed that the melting point of AgNWs decreases with decrease in the diameter; however, these values are smaller as compared to the bulk Silver melting point [24, 28, 35, 36]. The variation in the melting temperature of AgNWs with the diameter may be understood from Fig. 7. The graph gives a straight-line fit of the form.

118

S. K. Joshi et al.

Fig. 5 Effect of heating on a Silver nanowire (cross-sectional and side views) of 2 nm diameter, at a 300 K b 500 K c 800 K d 1000 K e 1500 K

Fig. 6 Temperature versus Potential energy graph for the AgNWs of different diameters

Tm = T0 +

k , D

Here, T0 = 1254.5 is the extrapolated value of the melting temperature for infinite diameter (bulk), which is near to the actual melting temperature for the bulk silver; k = −357.24, which shows the linear dependence of Tm with (diameter)−1 , a behavior well-known for metal nanoclusters [28]. The experimental value of melting temperature for AgNWs of 5 nm [37] is close to the extrapolated value for the same in the present work.

Atomistic Simulation Studies of the Mechanical …

119

Fig. 7 Melting Temperatures of AgNWs as a function of 1/diameter

4 Conclusions In summary, LAMMPS with embedded atom method potentials is used to study AgNWs of different dimensions. Since temperature affects the mechanical and thermal properties largely, the present study is performed to understand the deformation and melting characteristics of AgNWs, which may find applications in nanoelectronics. The stress–strain graphs indicate that with increase in the temperature, yield stress decreases, indicating a decrease in the strength of the nanowire. Visualizations of strained wire indicate slip along [111] plane. Finally, melting temperature of AgNWs is found to depend on the diameter: with increase in the diameter, the melting temperature is found to increase, which may eventually approach toward the bulk value. Acknowledgements Authors thank University of Petroleum and Energy Studies, Dehradun (India) for providing the necessary facilities to conduct this work. D. P. Singh thanks to the Department of Science and Technology, Government of India for the financial support through project number ECR/2017/000641 under ECRA.

120

S. K. Joshi et al.

References 1. Feng, X.L., He, R., Yang, P., Roukes, M.L.: Very high frequency silicon nanowire electromechanical resonators. Nano Lett. 7(7), 1953–1959 (2007). https://doi.org/10.1021/nl0 706695 2. Kondo, Y., Takayanagi, K.: Gold nanobridge stabilized by surface structure phys. Rev. Lett. 79, 3455–3458 (1997) 3. Burt, D.P., Wilson, N.R., Weaver, J.M., Dobson, P.S., Macpherson, J.V.: Nanowire probes for high resolution combined scanning electrochemical microscopy atomic force microscopy. Nano Lett. 5(4), 639–643 (2005). https://doi.org/10.1021/nl050018d 4. Cui, Y., Wei, Q.Q., Park, H.K., Lieber, C.M.: Nanowire nanosensors for highly sensitive and selective detection of biological and chemical species. Science 293, 1289–1292 (2001) 5. Patolsky, F., Timko, B.P., Zheng, G., Lieber, M.: Nanowire-based nanoelectronic devices in the life sciences. MRS Bull. 32(2), 142–149 (2007). https://doi.org/10.1557/mrs2007.47 6. Cui, Y., Zhong, Z., Wang, D., Wang, W.U., Lieber, C.M.: High performance silicon nanowire field effect transistors. Nano Lett. 3, 149–153 (2003) 7. Lei, T., Peng, R.X., Song, W., Hong, L., Huang, J.M., Fei, N.N., Ge, Z.Y.: Bendable and foldable flexible organic solar cells based on Ag nanowire films with 10.30% efficiency. J. Mater. Chem. A. 7, 3737–3744 (2019). https://doi.org/10.1039/c8ta11293b 8. Singh, M., Rana, T.R., Kim, S., Kim, K., Yun, J.H., Kim, J.: Silver nanowires binding with sputtered ZnO to fabricate highly conductive and thermally stable transparent electrode for solar cell applications. ACS Appl. Mater. Interfaces 8(20), 12764–12771 (2016). https://doi. org/10.1021/acsami.6b01506 9. Huang, Y., Duan, X., Cui, Y., Lauhon, L.J., Kim, K., Lieber, C.M.: Logic gates and computation from assembled nanowire building blocks. Science 1313–1317 (2001) 10. Reimer, M.E., Bulgarini, G., Akopian, N., Hocevar, M., Bavinck, M.B., Verheijen, M.A.: Bright single-photon sources in bottom-up tailored nanowires. Nat. Commun. 3, 737 (2012) 11. Wang, S., Shan, Z., Huang, H.: The mechanical properties of nanowires. Adv. Sci. 4(4) (2017). https://doi.org/10.1002/advs.201600332 12. Lucas, M., Leach, A., McDowell, M., Hunyadi, S., Gall, K., Murphy, C., Riedo, E.: Plastic deformation of pentagonal silver nanowires: comparison between AFM nanoindentation and atomistic simulations. Phys. Rev. B 77(24) (2008). https://doi.org/10.1103/physrevb.77. 245420] 13. Sannicolo, T., Cabos, M., Simonato, C.J., Bellet, D.: Metallic nanowire-based transparent electrodes for next generation flexible devices: a review. Small 12(44), 6052–6075 (2016) 14. Kim, S., Kim, B., Im, I., Kim, D., Lee, H., Nam, J., Chung, H.K., Lee, H.J., Cho, S.M.: Employment of gold-coated silver nanowires as transparent conductive electrode for organic light emitting diodes. Nanotechnology 28, (2017). https://doi.org/10.1088/1361-6528/aa7ba0 15. Wu, H.A., Soh, A.K., Wang, X.X.: Strength and fracture of single crystal metal nanowire. Key Eng. Mater. 261–263, 33–38 (2004) 16. Liang, W., Zhou, M.: Response of copper nanowires in dynamic tensile deformation. Proc. Inst. Mech. Eng. Part C J. Mech. Eng. Sci. 218(6), 599–606 (2004) 17. Jing, Y., Meng, Q., Zhao, W.: Molecular dynamics simulations of the tensile and melting behaviours of silicon nanowires. Physica E 41(4), 685–689 (2009). https://doi.org/10.1016/j. physe.2008.11.006 18. Branício, P.S., Rino, J.P.: Large deformation and amorphization of Ni nanowires under uniaxial strain: a molecular dynamics study. Phys. Rev. B 62(24), 16950–16955 (2000) 19. Wen, Y.H., Zhu, Z., Shao, G.F., Zhu, R.Z.: The uniaxial tensile deformation of Ni nanowire: atomic-scale computer simulations. Physica E 27(1–2), 113–120 (2005) 20. Koh, S.J.A., Lee, H.P., Lu, C., Cheng, Q.H.: Molecular dynamics simulation of a solid platinum nanowire under uniaxial tensile strain: temperature and strain-rate effects. Phys. Rev. B, 72(8), Article ID 085414 (2005) 21. Gall, K., Diao, J., Dunn, M.L.: The strength of gold nanowires. Nano Lett. 4(12), 2431–2436 (2004)

Atomistic Simulation Studies of the Mechanical …

121

22. Joshi, S.K., Pandey, K., Singh, S. K., Dubey, S.: Molecular dynamics simulations of deformation behaviour of gold nanowires. J. Nanotechnol. Article ID 5710749, 5 (2019) https://doi.org/10. 1155/2019/5710749 23. Ridings, K. M., Aldershof, T. S., Hendy, S. C.: Surface melting and breakup of metal nanowires: theory and molecular dynamics simulation. J. Chem. Phys. 150(9) (2019). https://doi.org/10. 1063/1.5086435 24. Dalgic, M., Senturk, S., Domekeli, U.: Molecular dynamics simulation studies on melting of Sn nanowires. J Optoelectron. Adv. Mater. 13(11–12), 1570–1576 (2011) 25. Wang, B., Wang, G., Chen, X., Zhao, J.: Melting behavior of ultrathin titanium nanowires. Phys. Rev. B Condens. Matter Mater. Phys. 67(19), 4–7 (2003). https://doi.org/10.1103/Phy sRevB.67.193403 26. Bilalbegovi´c, G.: Structures and melting in infinite gold nanowires. Solid State Commun. 115(2), 73–76 (2000) 27. Kang, J., Hwang, H.: Molecular dynamics simulation study of the melting of ultrathin copper nanowires 40(5), 946–948 (2002) 28. Wen, Y.H., Zhu, Z.Z., Zhu, R., Shao, G.F.: Size effects on the melting of nickel nanowires: a molecular dynamics study. Physica E 25(1), 47–54 (2004). https://doi.org/10.1016/j.physe. 2004.06.048 29. Alam, M.F., Rahman, M.S.: Fracture mechanism of single and polycrystal silver nanowire: computational study. Int. J. Math. Trends Technol. 54(6), 471–476 (2018). https://doi.org/10. 14445/22315373/ijmtt-v54p557 30. Shi, D.W., He, L.M., Kong, L.G., Lin, H., Hong, L.: Superheating of Ag nanowires studied by molecular dynamics simulations. Modell. Simul. Mater. Sci. Eng. 16(2), (2008). https://doi. org/10.1088/0965-0393/16/2/025009 31. Plimpton, S.: Fast parallel algorithms for short-range molecular dynamics. J. Comp. Phys. 117, 1–19 (1995) 32. Daw, M.S., Baskes, M.I.: Embedded-atom method: derivation and application to impurities, surfaces, and other defects in metals. Phys. Rev. B 29(12), 6443–6453 (1984) 33. Stukowski, A.: Visualization and analysis of atomistic simulation data with OVITO–the open visualization tool, IOP publishing. Model. Simul. Mater. Sci. Eng. 18(1), 015012(2010) 34. Alarifi, H.A., Atis, M., Hu, A., Yavuz, M., Zhou, Y.: Determination of complete melting and surface premelting points of silver nanoparticles by molecular dynamics simulation. J. Phys. Chem. C. 117, 12289–12298 (2013) 35. Chushak, Y.G., Bartell, L.S.: Melting and freezing of gold nanoclusters. J. Phys. Chem. B 105(47), 11605–11614 (2001). https://doi.org/10.1021/jp0109426 36. Yang, G., He, B., Li, R., Qu, X.: Molecular dynamics study on the physical properties of gold and silver alloys. In: 6th International Conference on Energy, Environment and Sustainable Development, ICEESD. Advances in Engineering Research, vol. 129. Zhuhai, 11–12 Mar 2017 37. Gordon, E., Karabulin, A., Matyushenko, V., Sizov, V., Khodos, I.: Stability and structure of nanowires grown from silver, copper and their alloys by laser ablation into superfluid helium. Phys. Chem. Chem. Phys. 16(46), 25229–25233 (2014). https://doi.org/10.1039/c4cp03471f

Mobility Management Scheme During Offloading in Mobile Cloud Computing Robin Prakash Mathur and Manmohan Sharma

Abstract Mobility is a fundamental aspect of mobile cloud computing. The mobile device gets connected to the cloud or edge server through the intermediate cellular network during roaming for the offloading process. The seamless connectivity with the network is required for the mobile device so that it can remain connected to the cloud or edge server and completes the process of offloading effectively. In this work, the seamless mobility scheme has been proposed, which is based on the heterogeneous network of Wi-Fi and 4G network. The proposed scheme is based on the fourth-order Markov model for mobility prediction and received signal strength (RSS) of the network nearest to the next predicted move of the device. The proposed scheme performs better as compared to the SINR-based handoff mechanism on the basis of the number of handoffs and dropped count during device mobility in urban, semi-urban, and rural areas. Keywords Mobility management · Mobile cloud computing · Offloading · Vertical handoff

1 Introduction Mobile cloud computing (MCC) is an emergent field with the convergence of two major areas—cloud computing paradigm and mobility [1]. It provides a solution of computational offloading to the device having low processing power and limited memory capacity. In the process of offloading, the data that need to be processed are sent to the cloud or edge servers from the mobile device, get computed on the servers and once the computation is done, the mobile device receives the computation results R. P. Mathur (B) Department of Computer Science & Engineering, Lovely Professional University, Phagwara, Punjab, India e-mail: [email protected] M. Sharma Department of Computer Application, Lovely Professional University, Phagwara, Punjab, India e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2021 S. Choudhury et al. (eds.), Intelligent Communication, Control and Devices, Advances in Intelligent Systems and Computing 1341, https://doi.org/10.1007/978-981-16-1510-8_13

123

124

R. P. Mathur and M. Sharma

from the cloud or edge server. The offloading process in MCC may use heterogeneous types of wireless networks, which may include wireless LAN (WLAN) and cellular services like 3G, 4G services, and even 5G services in the nearby future. Various issues get raised when the offloading application runs like availability of connectivity, the energy level of mobile device, and availability of the cloud or edge servers. The different types of mobile services [2] are available to the mobile device like Bluetooth, Wi-Fi, 2G/3G/4G services. The transitions among these services are getting possible by a concept of vertical handoff. The problems in mobile cloud computing are similar to mobile computing such as the issues [3, 4] related to handoffs, network delays, bandwidth, and limited battery energy. In the case of computational offloading, the mobile device or mobile nodes (MN) roams around different access networks like a mobile device may initially start some cloud services in the 4G network and can commit offloading process in the Wi-Fi network due to its mobility. Handoff is a process when a mobile device changes its network from one to another. While in mobility, the mobile nodes initiate the process of handoffs for seamless connectivity. Handoff can introduce packet loss and long delays and hence can affect the cloud services further. Many applications in the modern computing environment are cloudspecific like gaming applications, healthcare services, natural language processing (NLP)-based applications, and computer vision. The mobile device can perform computations while roaming and may require cloud services for offloading purposes. The seamless transitions among networks can be either horizontal or vertical. When a device moves from one network to another without changing the network type, then the process is called horizontal handoff, and if it changes network, then it is called vertical handoff. Heterogeneous networks (Hetnets) have various types of features like data rates, received signal strength (RSS), network capacity, bandwidth, and coverage span. The mobile device perceives these features and decides to select the best available network in its current location. The mobility of the device has a greater impact on the process of offloading. While the user is in a moving state, the probability of changing the network is high. For the flawless process of offloading, the transition among the cellular network must be smooth and handoff must be minimized so that the mobile device remains connected with the cloud server. To address the issues of seamless connectivity, a technique has been proposed for reducing the number of handoffs and the dropped rate, which is based on the fourthorder Markov model for location prediction and RSS-based scheme for handoff decision. The paper is arranged in different sections where in segment 2, the related work in mobility in mobile cloud computing is discussed and the proposed mobility scheme for the offloading process is discussed in segment 3. Results and discussion are presented in segment 4, which is trailed by a conclusion in segment 5.

2 Related Work Different solutions have been proposed by researchers that address mobility during computational offloading. M2C2 [5] has proposed a multihoming mechanism where

Mobility Management Scheme During Offloading …

125

device probes for network and cloud network during mobility. In this work, the best cellular network and cloud network are selected based on the RSSI and cloud ranking process. Clonecloud [6], MAUI [7], ThinkAir [8], and Cuckoo [9] have proposed an effective and standard solution for computational offloading in mobile cloud clouding but have not considered mobility as a factor in their work. The user mobility model [10] has been proposed where user mobility is predicted using the 2nd order Markov chain but has worked on a homogeneous type of network only. A handoff scheme [11] in mobile cloud computing has been proposed, which focuses on saving the device power but during offloading the bandwidth must be above 4 MB for avoiding the handoff delay. Alnezari and Rikli [12] presented the handoff mechanism and offloading strategy in mobile cloud computing using the fuzzy logic model approach and worked on 3G and WLAN environments. Hani and Dichter [13] presented a robust five-layer service-oriented architecture that can perform seamless handoff in WiMAX and helps in reducing the bandwidth and power consumption. In [14, 15], researchers have presented the fuzzy approach which is based on multi-criteria and multi-attribute handover decisions respectively. Sanaei et al. [16] has discussed the challenges and issues faced in mobile cloud computing when heterogeneity in the network is introduced. The complexity gets increases when the mobile device roams in 3G, Wi-Fi, and WiMAX networks. The authors in [17–19] discussed the user device mobility and the prediction of the location in the next movement. The mobility in mobile cloud computing is still an open area where a lot of research is still required and hence, mobility is incorporated in the computational offloading in this work.

3 Proposed Mobility Scheme During Computational Offloading Computational offloading is a complex problem in mobile cloud computing. The offloaded task initiated by the mobile device during the offloading process reaches the cloud server through the heterogeneous cellular network or Wi-Fi network. In Fig. 1, the offloading mechanism is shown where the mobile device may roam among different networks and the mobility of the device compels it to choose the network, which is having higher signal strength through the process of network probing. Further, cloud probing is also required for selecting the best cloud for executing the offloaded application component. In this work, an assumption is made that mobile device has fixed the cloud server based on the technique used in [5] which is Simple Additive Weighting (SAW). The contribution to this work is to devise a mechanism for network probing where an optimal network can be searched and the device remains connected to the cloud server. It is aimed to have less number of handoff and dropped rate happen so that seamless connectivity can be realized without interrupting the cloud service.

126

R. P. Mathur and M. Sharma

Fig. 1 Computational offloading of task during mobility in mobile cloud computing

The proposed technique aims to provide a mobility scheme where the number of handoffs can be minimized and further the handoff dropped can be reduced. In this work, the COST-231-Hata path loss model [23] is considered along with the pathway mobility model in various mobile environments like urban, semi-urban, and rural locations to simulate the cellular model of the offloading device (Fig. 2). In this scenario, the mobile device roams in the different states or locations, which are represented with the Markov model. In Fig. 2, the generic case of n state Markov model is presented where a device moves among different states, and the probability of moving to the next state is updated after every movement in the cellular or Wi-Fi area. P{Sn+1 = w|Sn = qn , Sn−1 = qn−1 . . . S1 = q1 , S0 = q0 }

(1)

The mobility of the device is a stochastic process that is random and the device can move to any state in state space. State space is an area that is having network

Fig. 2 N state Markov chain model for mobility prediction in mobile cloud computing

Mobility Management Scheme During Offloading …

127

coverage or Wi-Fi access point. The mobile device performs the transition from one state to another based on the transition matrix. The probability of transition matrix gets updated after every movement initially and the process of learning is undergone. The transition matrix for N states will be as:

In the first-order Markov chain, the model forecasts the next action by only seeing the last action made by the user. In the current scenario, the prediction of the next location movement accessed by a mobile device is the problem consideration. The Markov model consists of user mobile location as states and the fourth-order Markov model is considered, which means the next state can be predicted based on the previous four states. It has been seen that a higher-order Markov chain [20] increases the accuracy of the prediction; there is an increase in the number of states also. So, to manage the trade-off, the fourth-level Markov model is applied for location prediction and further higher order may make a mobile application more complex and could affect the battery drainage also during the decision-making for handoff.

3.1 Proposed Technique In this paper, the fourth-order Markov chain-based mobility scheme is proposed, which aims to reduce the number of handoffs and dropped rate also. The mobile device is connected initially with either Wi-Fi or cellular network. Once the device starts the offloading process, it may start moving from one location to another location. The problem has considered the pathway mobility model [21] in the scenario, which has been implemented in a different cellular environment like urban, semiurban, and rural locations and compared with SINR-based handoff mechanism [25]. The cloud server is assumed to be stationary and connected to the cellular service in the current scenario. When the device starts moving, it probes for the network or access point having strong signal strength. The network probing scheme is proposed as: Procedure: (Network probing for candidate list of BS/AP and network selection) Finding the list of the predicted base station and access point based on the current location of the user equipment using a fourth-order Markov model Input: Base station threshold Access point threshold

128

R. P. Mathur and M. Sharma

D, database of all access point and base station Output: L, list of candidate base station and access point in D Method: 1. User equipment is connected to base station or access point that is connected with cloud. 2. Candidate list = {} 3. For each access point and base station in D, do 4. Repeat 5. Compare the current RSS with the existing Base station and Access point RSS 6. Update candidate list with those BS and AP whose RSS is above the threshold value in available area 7. until no change 8. If (User equipment Current RSS < threshold value), Select BS/AP from the candidate list based on the predicted location of user equipment derived using 4th order Markov model and perform offload 9. Else continue offload at current location

4 Performance Evaluation 4.1 Experiment Settings In this work, a scenario is assumed where the mobile device is connected to the cloud server through various base stations and access points. In Table 1, the frequency of the base station is considered as 2100 MHz for all towers as the 4G network is considered in the work. The height of the base stations (Hbs), transmitted power (Pt), transmitted antenna (Gt), and connector loss (A) is also considered for urban, semi-urban, and rural while defining the base stations [22] during this work. The work presented also shows the comparison of techniques based on the cellular area like urban, semi-urban, and rural. The cellular network’s received signal strength (RSS) can be computed as: Table 1 Base station features based on the cellular environment

Urban

Sub-Urban

Rural

Frequency Base station (F)

2100 MHz

2100 MHz

2100 MHz

Height of base station (Hbs)

30

34

38

Transmitted power (Pt)

43

46

48

Transmitted antenna (Gt)

18

18

18

Connector loss (A)

2

2

2

Mobility Management Scheme During Offloading … Table 2 WLAN access point features

129 Urban

Constant power loss (L)

147 dB

Path loss exponent (n)

3 dB

Shadow fading (S)

2

Transmit Power (Pt)

1dBm

RSS = Transmit_ Power (Pt) + Transmit_ Antenna (Gt)-Path_ loss (PL) -connect_ loss (A).

(2)

The pathloss model expressed for the cellular network COST-231-Hata model [23] Path_loss (dB) = A + B log10 (d) + C

(3)

A = 46.3 + 33.9 log10 (carrier_freq) −13.28 log10 (BS_height) − a (Mobile_height)

(4)

B = 44.9−6.55 log 10 (BS_height)

(5)

In Eq. 3,

and the value of C is 0 for rural and suburban areas and 3 for urban areas. In Table 2, various parameters are presented which are used in WLAN simulation setting. The path loss [24] in WLAN is represented in dB as: P L(W L AN ) = constant power loss(L) + 10(path loss exponent n) log(d) + Fading effect(S).

(6)

The RSS for WLAN is articulated in dBm as: RSS (WLAN) = Trasmit_ Power (Pt) − Path_ loss (PL).

(7)

4.2 Results and Discussion During the implementation of the strategy, the numbers of user equipment were considered in the range of 100–500 and the three environments were considered i.e. urban, sub-urban, and rural.

130

R. P. Mathur and M. Sharma

Fig. 3 Comparison of the number of handoff in proposed technique with SINR-based approach in Urban, Sub-Urban, and Rural

Fig. 4 Comparison of the number of handoffs dropped in proposed technique with SINR-based approach in Urban, Sub-Urban and rural environment

Mobility Management Scheme During Offloading …

131

Figure 3 shows the handoff comparison between the two different strategies. Based on the simulation, it is founded that the proposed handoff strategy based on location prediction performs better as compared with SINR-based handoff. The numbers of handoffs are less in three different environments when compared with SINR based handoff. The lower handoff count depicts seamless connectivity with the cloud server and flawless offloading process. Handoff dropped count depicts the scenario when user equipment does not get the required signal strength during handover and loses the connectivity with the base station of the cellular network. The results shown in Fig. 4 interpret that the proposed technique reduces the dropped rate in the cellular network and altogether reduction in dropped rate provides seamless connectivity of the mobile device to the cloud server. In all three cellular environments, the proposed technique has performed well in handoff dropped count.

5 Conclusion A mobility-based offloading scheme in MCC has been proposed in this work where the emphasis is given to reduce the number of the handoff of the mobile device and also the handoff dropped count. The fourth-order Markov model has been devised for predicting the next location of the user equipment. The technique will enable the user equipment to remain connected to the cellular network or Wi-Fi network which at the end gets connected to the cloud or edge server for completing the computational offloading task. The work includes the various mobility environments like urban, semi-urban, and rural for implementing the proposed work and it has been found that it has performed better than the compared SINR-based approach. In future work, 5G and 6G networks will be explored where the handoff mechanism will be studied and implemented for providing relevant ground for the computational offloading process in mobile cloud computing.

References 1. Song, W., Su, X.: Review of mobile cloud computing. In: 3rd International Conference on Communication Software and Networks, Xi’an, pp 27–29 (2011) 2. Dinh, H.T., et al.: A survey of mobile cloud computing: architecture, applications, and approaches. Wireless Commun. Mobile Comput. 13, 1587–1611 (2013) 3. Bahl, P., Han, R.Y., Li, L.E., Satyanarayanan, M.: Advancing the state of mobile cloud computing. In: Proceedings of the Third ACM Workshop on Mobile Cloud Computing and Services, pp 21–28. ACM (2012) 4. Fernando, N., Loke, S.W., Rahayu, W.: Mobile cloud computing: a survey. Future Gener. Comput. Syst. 29(1), 84–106 (2013) 5. Mitra, K., Saguna, S., Åhlund, C., Granlund, D.: M2C2: a mobility management system for mobile cloud computing. In: IEEE Wireless Communications and Networking Conference (WCNC), pp 1608–1613 (2015)

132

R. P. Mathur and M. Sharma

6. Chun, B., Ihm, S., Maniatis, P., Naik, M., Patti, A.: Clonecloud: elastic execution between mobile device and cloud. In: Proceedings of the Sixth Conference on Computer Systems, EuroSys’11, pp 301–314. New York, NY, USA, ACM (2011) 7. Cuervo, E., Balasubramanian, A., Cho, D., Wolman, A., Saroiu, S., Chandra, R., Bahl, P.: Maui: making smartphones last longer with code offload. In: Proceedings of the 8th International Conference on Mobile Systems, Applications, and Services, MobiSys’10, pp 49–62. New York, NY, USA, ACM (2010) 8. Kosta, S. et al.: ThinkAir: dynamic resource allocation and parallel execution in the cloud for mobile code offloading. In: Proceedings IEEE INFOCOM, pp 945–953 (2012) 9. Kemp, R., Palmer, N., Kielmann, T., Bal, H.: Cuckoo: a computation offloading framework for smartphones. MobiCASE (2010) 10. Lee, K., Shin, I.: User mobility model based computation offloading decision for mobile cloud. J. Comput. Sci. Eng. 9 (2015) 11. Qi, Q., Liao, J., Cao, Y.: Integrated multiple services handoff in mobile cloud computing. In: Global Information Infrastructure Symposium (GIIS 2013), pp 1–3 (2013) 12. Alnezari, A.S., Rikli, N.: Achieving mobile cloud computing through heterogeneous wireless networks. Int. J. Commun. Netw. Syst. Sci. 10,107–128(2017) 13. Hani, Q.B., Dichter, J.P.: Energy-efficient service-oriented architecture for mobile cloud handover. J. Cloud Comput. 6, 1–13 (2017) 14. Ali, T., Saquib, M: Analysis of an instantaneous packet loss based vertical handover algorithm for heterogeneous wireless networks. IEEE Trans. Mobile Comput. 13(5), 992–1006 (2014) 15. Sgora, C.A., Gizelis, A., Vergados, D.D.: Network selection in a WiMAXCWiFi environment. Pervasive Mobile Comput. 7(5), 584–594(2011) 16. Sanaei, Z., Abolfazi, S., Gani, A., Buyya, R.: Heterogeneity in mobile cloud computing: taxonomy and open challenges IEEE Commun. Surveys Tutor. 16, 369–392(2013) 17. Bhattacharjee, D., Rao, A., S., C., Shah, M., Helmy, A.: Empirical modeling of campus-wide pedestrian mobility observations on the USC campus. In: IEEE 60th Vehicular Technology Conference, vol. 4, pp 2887–2891 (2004) 18. Su, W., Lee, S., & Gerla, M.: Mobility prediction in wireless networks. In: MILCOM 2000 Proceedings of the 21st Century Military Communications. Architectures and Technologies for Information Superiority, pp 491–495(2000) 19. Liu, T., Bahl, P., Chlamtac, I.: Mobility modeling, location tracking, and trajectory prediction in wireless ATM networks. IEEE J. Sel. Areas Commun. 16, 922–936 (1998) 20. Bai, F., & Helmy, A.: a survey of mobility models in Wireless Adhoc Networks (2004) 21. Deshpande, M., Karypis, G.: Selective Markov models for predicting Web page accesses. ACM Trans. Internet Techn. 4, 163–184 (2004) 22. Akinwole, B.: Comparative analysis of empirical path loss model for cellular transmission in rivers state (2013) 23. Singh, Y.: Comparison of Okumura, Hata and COST 231Models on the basis of path loss and signal strength. Int. J. Comput. Appl. 59 (2012) 24. Ayyappan K, Dananjayan P.: RSS measurement for vertical handoff in heterogeneous network. J. Theor. Appl. Inform. Technol. 989–994 (2008) 25. Ayyappan, K., Narasimman, K., Dananjayan, P.: SINR based vertical handoff scheme for QoS in heterogeneous wireless networks. In: International Conference on Future Computer and Communication, Kuala Lumpar, pp 117–121 (2009)

Multiuser Detection for MIMO-OFDM System Using Binary Spotted Hyena Optimizer in UWA Communication Md Rizwan Khan, Bikramaditya Das, and Bibhuti Bhusan Pati

Abstract Multiple-Input Multiple-Output Orthogonal Frequency Division Multiplexing (MIMO-OFDM) system is a promising solution to fulfill the demand of highspeed transmission rate and high quality of service for future Underwater Acoustic (UWA) communication. However, Multi-Access Interference (MAI) due to the interference from co-channel users at the receiver of MIMO-OFDM is considered as main source limiting the system capacity. Therefore, Multiuser Detection (MUD) technique is needed to suppress the effect of MAI at the receiver of the MIMOOFDM system. In this research, MUD is achieved using the global and local optimal solution of binary spotted hyena optimizer (BSHO) multiuser detector algorithm in an underwater environment. The transceiver model of the MIMO-OFDM system in underwater is implemented using World Ocean Simulation System (WOSS). The bit error rate (BER) performance of the proposed BSHO multiuser detector algorithm is analyzed using MATLAB simulation and also compared with conventional detectors and with metaheuristic approach such as binary particle swarm optimization (BPSO) in the UWA network. The proposed BSHO metaheuristic algorithm outperforms in BER analysis, which results in the best optimal solution toward MUD over existing detectors and the existing metaheuristic approach in the UWA network. Keywords Underwater acoustic (UWA) · Multiple-Input Multiple-Output (MIMO) · Orthogonal frequency division multiplexing (OFDM) · Binary spotted hyena optimizer (BSHO) · Multiuser detection (MUD) · Multi-Access interference (MAI)

Md R. Khan · B. Das (B) Department of Electronics and Telecommunication Engineering, VSS University of Technology, Burla, Odisha, India e-mail: [email protected] B. B. Pati Department of Electrical Engineering, VSS University of Technology, Burla, Odisha, India © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2021 S. Choudhury et al. (eds.), Intelligent Communication, Control and Devices, Advances in Intelligent Systems and Computing 1341, https://doi.org/10.1007/978-981-16-1510-8_14

133

134

Md R. Khan et al.

1 Introduction The need for high-quality service in underwater wireless communication arises in scientific, military, and civilian applications including the exchange of data for environmental monitoring in underwater wireless sensor networks [1]. Specifically, highspeed transmission rate and high quality of service in underwater are the demands for data transmission among devices such as buoy to autonomous underwater vehicle (AUV), AUV to AUV [2], and particularly those employing wireless links. But limited bandwidth due to low frequency of propagation, Intersymbol Interference (ISI) due to large multipath spread [3], and MAI due to the interference from cochannel users in the multiuser environment are the challenging problems in the UWA communication. OFDM has been used in multiband [4, 5] when combined with MIMO system diminishes the effect of ISI. Therefore, MIMO-OFDM is considered as a promising solution to fulfill the ever-increasing demand for bandwidth, efficiency, and performance of the UWA communication. MUD is needed at the receiver of the MIMOOFDM system to overcome the degradation caused by MAI in multiuser environment in the UWA communication. In a multiuser environment, a maximum-likelihood (ML) detector was first proposed in [6] for optimal error probability performance, but the complexity increases exponentially with the number of users. Therefore, metaheuristic algorithms are considered as an efficient optimization tool nowadays to solve various engineering optimization problems. Metaheuristic algorithms such as hybrid Firefly algorithm based on an evolutionary algorithm [7] and Hybrid GA (a combination of Simulated Annealing (SA) and PSO) [8] are implemented in MIMOOFDM system to obtain optimal solution for MUD. Metaheuristic algorithms [9, 10] are implemented for various engineering problems in various application areas in underwater communication. The novelty in this article is that a binary version of Spotted Hyena Optimization (SHO) algorithm called BSHO is used to minimize the objective function of maximum likelihood (ML) metric to obtain optimal solution towards MUD at the receiver of the MIMO-OFDM system in the UWA communication. The hunting and social behavior of spotted hyenas is used to motivate the metaheuristic SHO algorithm [11]. The minimization of the objective function toward optimal solution is carried out using both local and global optimal solutions of BSHO algorithm. The optimal solution is obtained using the position coordinate vectors (PCVs) of spotted hyenas within the search space in BSHO algorithm. The rest of article is organized as follows. Section 2 represents the problem formulation along with the contribution. Section 3 demonstrates the proposed BSHO algorithm-based MUD. Section 4 demonstrates result and discussion that illustrating the proposed scheme effectiveness over conventional detection schemes and the existing metaheuristic approach. Concluding remarks are drawn finally in Sect. 5.

Multiuser Detection for MIMO-OFDM System …

135

2 Problem Formulation and Contribution of This Study Consider a underwater MIMO-OFDM system shared by s number of users where s m 1, 2, 3… S equipped with S number of transmitting hydrophone antenna and L number of receiving hydrophone antenna. m number of subcarriers where m m 1, 2, 3….M is used to divide the frequency band. Channel matrix that is a MIMO channel impulse response for s number of users is represented as ⎡

Hs,m

s,m h s,m 1,1 h 1,2 s,m ⎢ h h s,m ⎢ 2,1 2,2 =⎢ . .. ⎣ .. . s,m h h s,m L ,1 L ,2

⎤ · · · h s,m 1,S s,m ⎥ · · · h 2,S ⎥ .. .. ⎥ . . ⎦ · · · h s,m L ,S

(1)

where h s,m L ,s represent the channel gain from sth transmitting antenna to Lth receiving antenna on subcarrier m. Further, the channel matrix H is decomposed using SVD as in [12] and is represented as H = U DV H . The received signal on subcarrier m for an underwater MIMO OFDM system is represented as rm =

S 

u s,m ds,m

√

ps,m xs,m + ηm = Um Dm



Pm xm + ηm

(2)

s=1

T

where rm = r1,m , r2.m , . . . . . . r L ,m defines signals received from L number of antennas on subcarrier m. ηm = Complex Gaussian noise with variS T

ance is equal to σ2. xm = xs,m = x1,m , x2,m . . . . . . . . . x S,m represents s=1

symbol sequences after modulation. Um =

S

u s,m = [u 1,m, u 2,m, . . . . . . , u S,m ]

s=1

represents left singular vector in SVD. Dm = diag(d1,m, d2,m, . . . . . . , d S,m ) √ pm = represents the eigen value having non-negative singular value. √ √ √ diag ( p1,m , p2,m , . . . . . . . . . p S,m ) represents the transmit power level. MUD at the receiver has been performed using MF detector and linear MUD technique such as ZF detector and MMSE detector. The output from MF detector, ZF detector, and MMSE detector is mathematically expressed as: −1

√ ymM F = UmH rm , ymZ F = Rm−1 ymM F , ymM M S E = Rm + σ 2 (Dm pm )−2 ymM F where Rm = UmH Um and ymM F is the output of MF detector. Further, to enhance the performance toward MUD, the optimal ML multiuser detector minimizes the bit error probability or suppress the MAI at the expense of

136

Md R. Khan et al.

higher complexity to estimate the transmitted symbols for all users. Therefore, a suboptimal detector called BSHO multiuser detector is proposed in this research to minimize the objective function with a comparable BER performance to that of existing detectors. The optimal solution for the ML metric for a combination of the transmitted symbol is expressed as 2   S    √   Hs,m vs,m ps,m xs,m  x˜m = arg min rm −   xm ∈ {−1,+1} s=1   

(3)

Objective f unction

where • denotes the Frobenius norm or Euclidean norm. The contributions of the research are summarized as follows: • WOSS is used to implement the MIMO-OFDM system in the UWA communication in which the underwater channel response is decomposed using singular value decomposition (SVD). • A metaheuristic BSHO multiuser detector algorithm is used at the receiver of the MIMO-OFDM system for optimal solution towards MUD through successful MAI cancellation. • The BER performance of the proposed scheme is analyzed and also compared with that of MF detector, ZF detector, MMSE detector, and at last with metaheuristic approaches such as BPSO in the UWA network.

3 Proposed BSHO Algorithm-Based MUD The BSHO multiuser detector algorithm is adjusted as follows 1.

2.

Each user from s active and independent users in the MIMO-OFDM system allows to transmit one bit per subcarrier at a time to set xs,m in (3). The element of each hyena’s position coordinate vector (PCV) is set as x˜s,m where m represents the mth hyena and s represents the sth element of the hyenas in PCV. Each T

hyena has s elements, composes as xm = x˜1,m , x˜2,m , . . . . . . x˜s,m that denotes the transmission of s users data at a time for a particular PCV. For the PCV of each hyena which is associated with s number of elements, we let S dimensional hypercube space for the encircling behavior of spotted hyena. As the transmitted information by each user is either be +1 or 0 and hence what value sth user transmitted in an S dimensional hypercube space will be decided by the PCV of spotted hyena in that binary position and is mathematically represented as follows Cluster formation in BSHO is mathematically represented as [13]

Multiuser Detection for MIMO-OFDM System …

C = Bk + Bk+1 + Bk+2 + . . . . . . . . . + Bk+N

137

(4)

where Bk represents the position of remaining spotted hyena in the search space during hunting behavior of spotted hyena. For MUD, each cluster formation is represented as C = [x1 , x2 , x3 , . . . . . . xm ]

(5)

T

where xm = x˜1,m , x˜2,m , . . . . . . x˜s,m . The value for each hyenas PCV in a S dimensional hypercube space is evaluated as  1 Sig(x˜s,m ) ≥ rand I +1 x˜(s,m) = (6) dim 0 other wise where dim represents the dimension, I represents the iteration number, Sig represents the sigmoid function, and rand is the random number drawn from uniform distribution [−1, 1]  Sig(x˜s,m ) =

 1 . 1 + exp(−(x˜s,m ))

(7)

The mathematical modeling of encircling, hunting, and attacking prey behavior of spotted hyena in this research is done in the same way as in [11] and with same parameter value as in [11]. Steps of BSHO algorithm for MUD Step 1

Step 2 Step 3 Step 4 Step 5 Step 6 Step 7 Step 8 Step 9 Step 10

Initialize the population for spotted hyenas Bi (i = 1, 2, 3…, m), Select the values for initial parameters of BSHO: Y, Z, è, and N, iteration number I = 0, Set the dimension of hypercube Bi spotted hyenas are explored in the S dimensional hypercube to generate Bi number of PCVs each with x˜s,m elements. Compare the Euclidean norm in (3) using PCVs from Bi spotted hyenas. Choose the PCV as the best search agent that gives minimum Euclidean norm. Till satisfactory result, define optimal solutions for the cluster using (4). Update the PCVs for other search spotted hyena. Adjust the search space if any search spotted hyena goes out of the boundary in the given hypercube search Space. Evaluate the updated Euclidean norm in (3) through updated PCVs and for a better solution update best search agent. Update the cluster C è using (4) to updated spotted hyena PCVs as in step 5. If stopping criteria satisfied then stop else return to step 4. Output the best optimal solution for MUD.

138

Md R. Khan et al.

Y, Z, represents the coefficient vector, è is decreased linearly from 5 to 0, N represents the number of optimal solution.

4 Result and Discussion The simulation results are carried out to demonstrate the proposed BSHO algorithm performance at the receiver of the MIMO-OFDM system in the UWA communication. An underwater network is set up using WOSS and the data analysis is performed using MATLAB. The list of parameters to set an underwater network using WOSS is summarized in Table 1. In Fig. 1, the BER performance as a function of SNR is evaluated for spotted hyena population size of 16, number of users is equal to 4 and by setting the signal to noise (Eb/No) ratio for the desired user is at 10 dB. From Fig. 1, it is observed that the BER performance of the proposed BSHO algorithm gives better performance towards MUD over the MF detector, ZF detector, MMSE detector, and existing BPSO metaheuristic approach. The poor performance of MF detector is due to the presence of MAI and noise with the desired signal as mathematically explained in Sect. 2. As the output of ZF detector contains only noise term as mathematically explained in Sect. 2, therefore it gives better performance over MF detector. The performance is further improved using MMSE detector but MMSE detector depends on received signal amplitude. At last, the proposed BSHO algorithm with number of iterations is equal to 500 and PCV evaluating function as sigmoid is compared with the existing BPSO algorithm with inertial weight of 0.72, cognitive and social coefficients of 1.8 and 2, respectively, range of bit velocity [−3, + 3], and PCV evaluating function as sigmoid. The convergence rate of the proposed BSHO algorithm is almost equal to that of BPSO algorithm for equal number of iteration but lower BER in case of Table 1 Parameters for underwater network set up

Parameter

Value

Number of subcarriers (M)

16

Number of bits per OFDM symbol interval

16

Modulation type

BPSK

Center frequency

36 kHz

Noise type

Complex Gausssian

Transmission distance

200 m

Depth from water surface

500 m

Subcarrier spacing

42.22 Hz

Number of transmitter

4

Number of receiver

4

Signal interval

10.3 ms

Multiuser Detection for MIMO-OFDM System …

139

Fig. 1 BER performance as a function of SNR

the proposed BPSO algorithm is due to its better local and global search ability. The percentage error between theoretical and simulated results for the proposed scheme is 0.08%. Figure 2 gives the performance for BER for the proposed scheme as a function of the number of users by setting the signal to noise (Eb/No) ratio for the desired user is at 10 dB. From Fig. 2, it is observed that BER increases with the increase in the number of users. This is due to an increase in the value of MAI with an increase in number of users at the receiver of the MIMO-OFDM system in UWA network. Even though BER increases with the number of users still the proposed BSHO algorithm support eight number of users to achieve a target BER of 10−3 , whereas BPSO support seven number of users and ZF, MMSE detector support almost three number of users.

5 Conclusion This paper investigated a metaheuristic BSHO multiuser detector algorithm towards MUD at the receiver of the MIMO-OFDM system implemented in an underwater

140

Md R. Khan et al.

Fig. 2 BER performance as a function of the number of users

environment using WOSS simulation system. The proposed BSHO scheme provided the best optimal solution through both global and local search ability in the search space to minimize the objective function towards MUD in the UWA communication The BER performance of the proposed BSHO multiuser detector as a function of SNR and the number of users is obtained and compared with the existing MF detector, ZF detector, MMSE detector, and existing BPSO metaheuristic algorithm. Among all, the proposed BSHO scheme outperforms as BER decreases to least than that of MF detector, ZF detector, MMSE receiver, BSHO, and BPSO approach in the UWA network.

References 1. Qu, F., Wang, Z., Yang, L., Wu, Z.: A journey toward modeling and resolving doppler in underwater acoustic communications. IEEE Commun. Mag. 54(2), 49–55 (2016) 2. Das, B., Subudhi, B., Pati, B.B.: Cooperative formation control of autonomous underwater vehicles: an overview. Int. J. Autom. Comput. 13, 199–225 (2016) 3. Khan, M.R., Das, B., Pati, B.B.: Channel estimation strategies for underwater acoustic (UWA) communication: an overview. J. Franklin Inst. 357(11), 7229–7265 (2020) 4. Das, B., Tiwari, S., Das, S.: Performance study of discrete wavelet packet based MB-OFDM system for short range indoor wireless environment 2011. In: International Conference on Devices and Communications (ICDeCom), pp 1–5. Mesra (2011)

Multiuser Detection for MIMO-OFDM System …

141

5. Das, B., Das, S., Das, ChS: Efficacy of multiband OFDM approach in high data rate ultra wideband WPAN physical layer standard using realistic channel models. Int. J. Comput. Appl. 2(2), 81–87 (2010) 6. Verdu, S.: Minimum probability of error for asynchronous Gaussian multiple-access channels. IEEE Trans. Inf. Theory 32(1), 85–96 (1986) 7. Kumar, B.S., Kumar, K.R.S.: Multiuser detection in MIMO-OFDM wireless communication system using hybrid firefly algorithm. Int. J. Eng. Res. Appl. 4(5), 176–183 (2014) 8. Lu, H., Yang, G., Hu, F., Sun, D.: Performance analysis of a multi-user MIMO-OFDM system based on a hybrid genetic algorithm. J. Eng. Sci. Technol. 9(6), 95–102 (2016) 9. Jiang, R., Wang, X., Cao, S., Zhao, J., Li, X.: Joint compressed sensing and enhanced whale optimization algorithm for pilot allocation in underwater acoustic OFDM systems. IEEE Access. 95779–95796 (2019) 10. Qu, C., Gai, W., Zhong, M., Zhang, J.: A novel reinforcement learning based grey wolf optimizer algorithm for unmanned aerial vehicles (UAVs) path planning. Appl. Soft Comput. 89, (2020) 11. Dhiman, G., Kumar, V.: Spotted hyena optimizer: a novel bio-inspired based metaheuristic technique for engineering applications. Adv. Eng. Softw. 114, 48–70 (2019) 12. Zhang, Y.J., Lataief, K.B.: An efficient resource allocation scheme for spatial multiuser access in MIMO/OFDM systems. IEEE Trans. Commun. 53, 107–116 (2005) 13. Kumar, V., Kaur, A.: Binary spotted hyena optimizer and its application to feature selection. J. Ambient Intell. Human Comput. 10, 1–21 (2019)

Estimation of Solar Radiation at Farasan Island with Two-Step ANN Concepts Arun Kumar Singh, Vikas Pandey, and Rupendra Kumar Pachauri

Abstract The availability of information about solar radiation characteristics, particularly solar radiation predictions, is important for efficiently designing solar energy systems. Solar radiation information available in Saudi Arabia with this study, a new two-step artificial neural network (ANN), is proposed to estimate both the daily average and hourly solar radiation at Farasan Island, KSA (Kingdom of Saudi Arabia). The input parameters for the daily average solar radiation estimation are the location and time required, along with five selected monthly meteorological parameters that predict for the subsequent month. The selected meteorological parameters are temperatures, relative humidity, and precipitation. These three selected meteorological parameters (temperatures, relative humidity, and precipitation) are used them for measurement. The accuracy of the proposed method is comparable to previous studies with an average R2 of 96.80% for the daily average solar radiation estimate and 94.33% for the hourly solar radiation estimate. Keywords Daily average solar radiation energy forecasting · Global solar radiation hourly solar radiation multilayer feed-forward

1 Introduction An artificial neural network (ANN) is a computational model to perform tasks like prediction, classification, decision making, etc. It consists of artificial neurons. These artificial neurons are a copy of human brain neurons. Neurons in the brain pass the A. K. Singh Saudi Electronic University, Riyadh, Saudi Arabia e-mail: [email protected] V. Pandey Babu Banarasi Das University, Lucknow, Uttar Pradesh, India e-mail: [email protected] R. K. Pachauri (B) University of Petroleum and Energy Studies, Uttarakhand Dehradun, India e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2021 S. Choudhury et al. (eds.), Intelligent Communication, Control and Devices, Advances in Intelligent Systems and Computing 1341, https://doi.org/10.1007/978-981-16-1510-8_15

143

144

A. K. Singh et al.

signals to perform the actions. Similarly, artificial neurons connect in a neural network to perform tasks. The connection between the artificial neurons is called weight. As of November 2019, Saudi Arabia has utilized only 0.028% of its abundant 536 GW solar energy potential through solar energy systems [1]. A lack of official solar radiation measurements could be one of the factors contributing to the slow development of solar energy systems, as optimal designing becomes more difficult without this information [2]. This problem could be overcome by generating an accurate solar radiation estimator for a specific location. Farasan Island is the largest island of the Farasan Islands, in the Red Sea. It is located some 50 km offshore from Jizan, the far southwestern part of Saudi Arabia. It is located at around 16°42 21 N 41°59 0 E. The main town on the island is Farasan. Estimates of solar radiation have been used in various optimal solar energy designs. The annual solar radiation calculation has been used to determine the ideal location to build a photovoltaic system [3, 4]. The monthly total, or daily average, solar radiation is required for the optimal sizing of solar panels or storage systems [5–7], while determining panel installation angles calls for the hourly solar radiation [8]. The artificial neural network (ANN) method for estimation is one of the most popular methods for daily average prediction due to its high accuracy [9, 10]. While a wider range of methods have been used to estimate hourly solar radiation, ANNbased methods have also been applied with accurate results. Those methods include multilayer perceptron [11], deep learning [12], adaptive neuro-fuzzy [13, 14], Jordan recurrent [15] and comparison of various methods [16]. However, those aforementioned methods cannot be used for estimating solar radiation in Saudi Arabia because they require meteorological input parameters. The objective of this study is to build a solar radiation estimator using an ANNbased method and the meteorological data types available in Saudi Arabia. Because no hourly weather information is available, the hourly solar radiation estimator relies on the daily average solar radiation in the same month. Therefore, a two-step ANN is required. The first ANN (ANN-1) predicts the daily average solar radiation using the selected monthly weather parameters, while the second ANN (ANN-2) predicts the hourly solar radiation using the output of ANN-1. Both the training and the validation of the ANNs were performed using data from the cities of Farasan and Jizan, which are located on Farasan Island. The cities were selected because they have the largest electricity demand in Saudi Arabia. The results of this research can be used to help solar energy system designers to build high-efficiency photovoltaic systems in Saudi Arabia, especially on Farasan Island.

2 Research Method In this study, a two-step ANN is proposed to estimate the amount of solar radiation at Farasan Island. ANN-1 estimates the amount of the daily average solar radiation, which is then used by ANN-2 to estimate the hourly solar radiation in the same month.

Estimation of Solar Radiation at Farasan Island …

145

Fig. 1 Configuration of the proposed method (https://techvidvan.com/tutorials/artificial-neural-net work/)

The configuration of the two-step ANN is depicted in Fig. 1. The meteorological data for training both networks were from 2018 and 2019 from the city of Jizan, the far southwestern part of Saudi Arabia.

2.1 ANN-1 Design ANN-1 estimates the daily average solar radiation by using the inputs of month, geographical position, and meteorological data of daily average minimum, maximum, and average temperature, relative humidity, and precipitation. These meteorological data were selected because forecast these data for the subsequent month. Thus, using these meteorological data allows for the estimation of the daily average solar radiation for the following month. The geographical positions of latitude and longitude were included because the position of the sun, and thus the solar radiation, varies for different locations in the same month. The daily solar radiation data for training ANN-1 were taken from the NASA Power database. The data were converted into the daily average solar radiation, with each day in a defined month assumed to have equal solar radiation. Detailed information about the input–output parameters is presented in Table 1.

146

A. K. Singh et al.

Table 1 Parameters of ANN-1 Parameter

Notation

Latitude

ϕ

Unit

Min.

Max.

Ave.

−7.25

−6.21

−6.73

Longitude

ρ

106.85

112.75

109.8

Month number

n

1

12

6.5

Minimum temperature

Tn

°C

23.05

27.4

26.23

Maximum temperature

Tx

°C

30.36

36.37

33.1

Average temperature

Ta

°C

27.53

31.38

28.83

Relative humidity

RH

%

61.54

82.43

73.8

Precipitation

RR

mm

0

21.52

5.29

Daily average radiation

Gm

kWh/m2

4.09

6.79

6.16

The number of hidden neurons in ANN-1 was chosen by comparing the regression value (R) during training for various numbers of neurons in the hidden layer. The candidates for the number of the hidden neurons are determined by the following rules [17] is shown in Eq. (1, 2) as: h≥

2 i +o 3

(1)

h < 2i

(2)

where h is the number of the hidden neurons, i is the number of the inputs, and o is the number of outputs. With eight inputs and one output, the number of hidden neurons is between 6 and 15. ANN-1 was trained 10 times for each number of hidden neurons. The comparison of the highest R for each number of hidden neurons is shown in Fig. 2. According to this analysis, the preferred number of hidden neurons for ANN-1 is 15. The training regression line of the preferred ANN structure is shown in Fig. 3. The feed-forward

97 Regression value (%)

Fig. 2 Comparison of regression values for ANN-1

96 95 94 93 92 91 90

6

7

8 9 10 11 12 13 Number of hidden neurons

14

15

Estimation of Solar Radiation at Farasan Island … Regression value (%)

Fig. 3 Comparison of regression values for ANN-2

147

99.7 99.6 99.5 99.4 99.3 99.2 99.1 99.0

4

5 6 7 8 Number of hidden neurons

9

Levenberg–Marquardt method was chosen to train ANN-1 due to its speed and stable convergence [18]. The iteration-stopping criteria were based on the minimum mean squared error (MSE), which is calculated in Eq. (3) as follows: MSE =

1 n 1 n ek2 = (tk − yk )2 k=1 k=1 n n

(3)

where n is the total number of samples of the output parameter, k is the index of the output samples, e is the error of the output, t is the target value of the output from the data, and y is the output value obtained from the calculation of the ANN.

2.2 ANN-2 Design ANN-2 estimates hourly solar radiation based on the daily average solar radiation, geographical position, month, and local hour. The advantage of this method is that it does not require any hourly weather parameters as input, which are not measured and recorded by NREP. The raw data were converted into the average of hourly solar radiation in each month. Detailed information about the input–output parameters is listed in Table 2. Using the same method as described in Sect. 2.1, the number of hidden neurons for ANN-2 was also determined by comparing R values for candidates determined using Eqs. (1) and (2). The range of the number of hidden neurons was between four Table 2 Parameters of ANN-2 Parameter

Notation

Local time hour

th

Hourly solar radiation

Gh

Unit Wh/m2

Min.

Max.

5

18

0

998.03

Ave. 11.5 391.9

148

A. K. Singh et al.

and nine, and nine was selected as the number of hidden neurons due to having the highest R, as shown in Fig. 3.

3 Results and Discussions Validation of the proposed method was completed by comparing the estimated results from each ANN with the measured solar radiation in Farasan for January through July 2020. Although the comparison was not performed for a full year, the selected months include portions of the rainy season (October–April) and the dry season (May–September) on Farasan Island. The accuracy of each ANN is measured by the coefficient of determination R2 , which is calculated in Eq. (4) as follows:    G Meas − G Pr ed X 100% R2 = 1 − G Pr ed

(4)

where Gmeas is the measured value of the solar radiation and Gpred is the estimated value of the solar radiation.

3.1 Accuracy of Daily Average Solar Radiation Estimation The accuracy of the proposed method was validated by a minimum value of R2 of 97.86%, as shown in Table 3. A higher value of R2 indicates that the estimated value is more accurate; a 100% value can be achieved if all the estimated values are equal to the measured values. The estimation accuracy for Jizan is higher than for Farasan, which indicates that the solar radiation in Jizan is more predictable and is more strongly correlated with the input parameters than in Farasan. The accuracy of the daily average solar radiation estimation on Farasan Island might be increased if other meteorological parameters are involved. Previous studies Table 3 Daily average solar radiation estimation accuracy City

n

Gm-meas

Gm-pred

R2

City

Gm-meas

Gm-pred

R2

Jizan

1

5.08

5.6

99.55%

Farasan

4.46

5.51

97.86%

2

4.96

5.6

4.41

5.51

3

4.64

4.67

4.68

4.57

4

4.81

4.66

4.64

5.14

5

4.61

4.63

4.13

4.93

6

4.88

5.04

4.44

4.83

7

4.92

4.90

4.69

4.51

Estimation of Solar Radiation at Farasan Island … Table 4 Comparison of the proposed daily average solar radiation estimation with previous studies

149

Author

Method

Location

Average of R2 (%)

Proposed

ANN

Saudi Arabia

98.70

Marzouq et al. ANN [21]

Morocco

97.16

Gurlek and Sahin [22]

ANN

Turkey

96.88

Teke and Yildirim [23]

Cubic Regression

Turkey

95.60

have indicated that sunshine duration and cloud cover percentage significantly correlate with the amount of solar radiation [19, 20]. Compared with the R2 of the daily average solar radiation estimates from previous studies, the accuracy of the proposed method is comparable, as shown in Table 4.

3.2 Accuracy of Hourly Solar Radiation Estimation The measured and estimated hourly solar radiation was compared on an hourly basis for an entire day from 12 AM to 11:59 PM. However, in Jizan, the earliest sunrise is between 5 AM and 6 AM and the latest sunset is between 5 PM and 6 PM for the entire year. Farasan has a wider range of sunrise and sunset times, with the earliest sunrise occurring between 5 AM and 7 AM and the latest sunset occurring between 5 PM and 7 PM. The R2 for the estimated hourly solar radiation for each month is listed in Table 5. For the hourly solar radiation, the estimation accuracy for Farasan is higher than Jizan, meaning that the correlation between the daily average solar radiation and hourly solar radiation in Farasan is higher than in Jizan. Table 5 Hourly solar radiation estimation accuracy City

N

R2 (%)

City

R2 (%)

Jizan

1

99

Farasan

93

2

93

98

3

93

98

4

96

98

5

98

98

6

95

98

7

93

98

150

A. K. Singh et al.

4 Conclusion The goal of this study was to propose a method to accurately estimate both the daily average and hourly solar radiation in Saudi Arabia, where the official measurements are not available from NREP. The proposed method not only estimates the radiation profile for each month based on the previous history but also can be used to predict daily average and hourly solar radiation for the subsequent month, using meteorological parameter inputs available from NREP. Due to the unavailability of hourly weather parameter measurements, the estimation of hourly solar radiation is only based on the daily average solar radiation and the time and location in which the estimation is required. The proposed method is also capable of estimating the solar radiation for multiple cities, at least on Farasan Island, by employing latitude and longitude as input parameters. The accuracy of the proposed method has been validated by calculating the R2 and comparing it with the results from previous studies in the same field of study. Because the proposed method produced a greater error in the peak of the rainy season than in the dry season, it may be possible to improve the proposed method by using a clustering method that treats the dry and rainy seasons separately.

References 1. Singh, A.K., Sharma S.D.: Digital era in the kingdom of Saudi Arabia: novel strategies of the telecom service providers companies. Webology 17(1), 227–245 (2020). https://doi.org/ 10.14704/web/v17i1/a219, ISSN: 1735-188X 2. Singh, A.K.: An intelligent reallocation of load for cluster cloud environment. Int. J. Innovative Technol. Exploring Eng. (IJITEE) 8(8) (2019). ISSN: 2278-3075 3. Singh, A.K.: Texture-based real-time character extraction and recognition in natural images. Int. J. Innovative Technol. Exploring Eng. (IJITEE) 8(8) (2019). ISSN: 2278-3075 4. Singh, A.K.: A wireless networks flexible adoptive modulation and coding technique in advanced 4G LTE. Int. J. Inform. Technol. 11(1), 55–66 (2019). ISSN: 2511-2104 (Print) 2511-2112 (Online), indexed in Springer. https://doi.org/10.1007/s41870-018-0173-5, https:// link.springer.com/article/10.1007/s41870-018-0173-5 5. Mahdi, B.D.E., Abdelatif, H., Zafrane, M.A.: An interactive approach for solar energy system: design and manufacturing. Int. J. Electr. Comput. Eng. 10(5), 4478–4489 (2020) 6. Calcabrini, A., Ziar, H., Isabella, O., Zeman, M.: A simplified skyline-based method for estimating the annual solar energy potential in urban environments. Nat. Energy 4(3), 206–215 (2019) 7. Abass, A.Z., Pavlyuchenko, D.A.: The exploitation of western and southern deserts in Iraq for the production of solar energy. Int. J. Electr. Comput. Eng. 9(6), 4617–4624 (2019) 8. Al Riza, D.F., Gilani, S.I.U.H., Aris, M.S.: Standalone photovoltaic systems sizing optimization using design space approach: Case study for residential lighting load. J. Eng. Sci. Technol. 10(7), 943–957 (2015) 9. Yao, C., Chen, M., Hong, Y.Y.: Novel adaptive multi-clustering algorithm-based optimal ess sizing in ship power system considering uncertainty. IEEE Trans. Power Syst. 33(1), 307–316 (2017) 10. Azmi, A.N., Bin Salim, N., Khamis, A.B.: Analysis of an energy storage sizing for gridconnected photovoltaic system, Indones. J. Electr. Eng. Comput. Sci. 16(1), 17–24 (2019)

Estimation of Solar Radiation at Farasan Island …

151

11. Kurniawan, A., Shintaku, E.: Determining the optimal inclination and orientation angles of solar panels installed on ship. Int. J. Renew. Energy Res. 10(1), 45–53 (2020) 12. Marzouq, M., El Fadili, H., Lakhliai, Z., Zenkouar, K.: A review of solar radiation prediction using artificial neural networks. In: International Conference on Wireless Technologies, Embedded and Intelligent Systems (2017) 13. Kurniawan, A., Shintaku, E.: Estimation of the monthly global, direct, and diffuse solar radiation in Japan using artificial neural network. Int. J. Mach. Learn. Comput. 10(1), 253–258 (2020) 14. Notton, G., Voyant, C., Fouilloy, A., Duchaud, J.L., Nivet, M.L.: Some applications of ANN to solar radiation estimation and forecasting for energy applications. Appl. Sci. 9(1) (2019) 15. Jallal, M.A., Chabaa, S., Zeroual, A.: A new artificial multi-neural approach to estimate the hourly global solar radiation in a semi-arid climate site. Theor. Appl. Climatol. 139(3–4), 1261–1276 (2020) 16. Hussein, H.S.: Global solar radiation prediction using a combination of subtractive clustering algorithm and adaptive neuro-fuzzy inference system: a case study. J. Eng. Sci. Technol. 15(3), 1652–1669 (2020) 17. Salisu, S., Mustafa, M.W., Mustapha, M., Mohammed, O.O.: Solar radiation forecasting in Nigeria based on hybrid PSO-ANFIS and WT-ANFIS approach. Int. J. Electr. Comput. Eng. 9(5), 3916–3926 (2019) 18. Firdaus, A.A., Yunardi, R.T., Agustin, E.I., Putri, T.E., Anggriawan, D.O.: Short-term photovoltaics power forecasting using Jordan recurrent neural network in Jizan, Telkomnika (Telecommunication Comput. Electron. Control. 18(2), 1089–1094 (2020) 19. Khosravi, A., Koury, R.N.N., Machado, L., Pabon, J.J.G.: Prediction of hourly solar radiation in Abu Musa Island using machine learning algorithms. J. Clean. Prod. 176, 63–75 (2018) 20. Heaton, J.: Artificial Intelligence for Humans, vol.3. In: Deep Learning and Neural Networks. Heaton Research Inc., Chesterfield (2015) 21. Kurniawan, A., Shintaku, E.: A neural network-based rapid maximum power point tracking method for photovoltaic systems in partial shading conditions. Appl. Sol. Energy 56(3), 157– 167 (2020) 22. Bouchouicha, K., Bailek, N., Mahmoud, M.E.S., Alonso, J.A., Slimani, A., Djaafari, A.: Estimation of monthly average daily global solar radiation using meteorological-based models in Adrar, Algeria. Appl. Sol. Energy 54(6), 448–455 (2018) 23. Koca, A., Oztop, H.F., Varol, Y., Koca, G.O.: Estimation of solar radiation using artificial neural networks with different input parameters for Mediterranean region of Anatolia in Turkey. Expert Syst. Appl. 38(7), 8756–8762 (2011)

Development of Matlab/Simulink Library for Unsupported Microcontrollers, Case Study: STM32F407 Nguyen Van Khanh, Ho The Anh, Nguyen Huynh Anh Duy, Pham Tran Lam Hai, Nguyen Van Muot, Tran Thanh Hung, and Nguyen Chi Ngon Abstract A Matlab/Simulink library for an unsupported microcontroller is developed in this study. The library is implemented by two main steps. First, analyzing and creating Simulink blocks for common peripherals of the microcontroller by using S-Function that combines embedded C language and Target Language Compiler (TLC) language to implement the mechanism for automatic transferring from a peripheral block to a C-embedded program. Second, proposing an algorithm in TLC language to generate automatically an embedded project that links all generated C-files, compile it to binary program, and write to the memory of the target microcontroller. The embedded project is run on the Keil C version 4.0 and the lowcost STM32F4Discovery board developed by STMicroelectronics. The experimental results have shown that the proposed library was designed successfully, generated codes are optimistic, and executed exactly on the target development board. The method in this study can be potentially applied to other microcontrollers. N. Van Khanh (B) · N. H. A. Duy · P. T. L. Hai · N. Van Muot · T. T. Hung · N. C. Ngon College of Engineering Technology, Can Tho University, Can Tho, Vietnam e-mail: [email protected] N. H. A. Duy e-mail: [email protected] P. T. L. Hai e-mail: [email protected] N. Van Muot e-mail: [email protected] T. T. Hung e-mail: [email protected] N. C. Ngon e-mail: [email protected] H. T. Anh College of Mechanical Engineering, Can Tho University of Engineering and Technology, Can Tho, Vietnam e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2021 S. Choudhury et al. (eds.), Intelligent Communication, Control and Devices, Advances in Intelligent Systems and Computing 1341, https://doi.org/10.1007/978-981-16-1510-8_16

153

154

N. Van Khanh et al.

Keywords Code generation · Simulink · TLC language · S-function · RTW

1 Introduction Microcontrollers have been manipulating in many embedded applications because of many useful peripherals, low energy consumption, and very low-cost. Besides the conventionally embedded system, microcontrollers have been applying to implement the controller algorithms designed and simulated by using special design tools. This process was used to apply on a specific microcontroller circuit board integrated into a specific software environment. Nowadays, two main methods have been applied to implement embedded controllers. The first is the whole embedded controller is writing manually by using C or assembly language. This method usually produces optimal controllers, but it spends a lot of time coding and debugging the controller source code. The second is the embedded controller is simulated and designed by using software integrated with a microcontroller card supporting full access to its peripherals. The big advantage of this method is that the controller is implemented commonly by using functional blocks and generating directly to embedded code and does not require any knowledge about the electronic circuits of this card. However, this method completely depends on the existing library and the cost is very high for a powerful microcontroller. An example is Matlab/Simulink developed by Mathworks that supports many simulated and designed tools in many fields including control engineering. Many embedded cards are integrated on Matlab/Simulink such as Texas Instrument C2000, C5000, xPC, Real-time target, and so on depending on the Matlab version. Some organizations have developed and integrated new Simulink libraries for their hardware platform as third-party libraries. The first well-known library package is the Simulink Support Package for Arduino Hardware that integrated Arduino hardware, an open-source hardware platform, to Simulink [1]. This library enables users can manipulate Simulink blocks to communicate with the Arduino circuit through a USB connection, supports many fundamental peripherals such as digital inputs/outputs, analog inputs, PMW outputs, serial connection, and other useful peripherals. The other library is the Simulink Support Package for Raspberry Pi hardware that integrated a Raspberry Pi mini-computer board into Simulink [2]. It is suitable for applications that require large memory, enhance peripherals (camera interface, Video Graphics Array (VGA) display), and high-speed performance. This package can be applied to create applications for Raspberry Pi minicomputer directly on the Simulink environment which supported many digital peripherals such as digital inputs/outputs, audio input, camera input, User Datagram Protocol (UDP). One common feature of the above libraries is that they allow users to create applications by choosing and connecting functional blocks in a specific manner. This is very convenient, but users completely depend on presupported blocks and unable to create a custom one when having demand.

Development of Matlab/Simulink Library …

155

To be more flexible, the solution combining manually between Simulink and embedded programming was recommended by [3–6]. With this solution, users who develop an application will follow two main steps. First, designing and simulating the core function of the system by using Simulink. Second, using the Real-Time Workshop (RTW) and Embedded Coder feature to transfer the Simulink system to embedded C language and write an additional embedded C program to link the generated codes with the inputs and outputs of the target microcontroller to control the real objects. This solution can be applied for any microcontroller platforms, which support standard C language; however, users must have the ability to analyze the microcontroller architecture, peripheral structures, and write C program to configure and control them for an expected application. The mechanism of the automatic code generation of Mathwork has been analyzing to customize and generate embedded code for an unsupported microcontroller. Lamberský et al. have been successfully created a library containing some functional blocks and the ability of automatic code generation for an unsupported hardware platform, Cerebot MX7 board [7]. This study proposed a solution that combined TLC language [8] and S-Function to make Simulink blocks for the Cerebot MX7 board’s peripherals (i.e. LEDs, UART, ADC, and OLED2) and create automatic code generation protocol embedded projects. Their library can help users writing the program to control the Cerebot MX7 board completely by using the Simulink environment. This study proved that using TLC language and S-Function can control the code generation protocol of RTW and Embedded Coder to generate an embedded program for any target microcontrollers. However, this library has applied the basic code generation mechanism of Simulink that using default templates the monolithic scheduler for embedded tasks, it is difficult to manipulate the advanced features of the microcontrollers. This study proposes a general procedure for joining TLC language and S-Function to create a Simulink library for some common peripherals of an unsupported microcontroller. The Simulink application using the designed library can be fully transformed into an embedded project by clicking one button. The code generation mechanism and important file templates are fully modified. Users can also customize these templates to be suitable for their interesting platforms. All processes are demonstrated on the low-cost ARM Cortex M4 microcontroller that integrated a very strong hardware floating-point and digital signal processing unit. It promises to increase significantly the performance of the target embedded system.

156

N. Van Khanh et al.

2 Architecture of Designed Library This paper aims to develop a library that includes some Simulink blocks of an unsupported microcontroller. The target platform is the ARM Cortex M4 microcontroller built-in STM32F4 Discovery board [9], a very cheap and high-performance development board. Matlab RTW and Embedded Coder features are combined to automatically generate the embedded codes for the selected board. Users only drag and drop Simulink blocks, connect them in a specific manner to design the features for an embedded system, and convert to an embedded C language project to be compiled by an embedded C compiler. The general architecture of the designed library is proposed in Fig. 1. The principal components of it are Simulink blocks of microcontroller peripherals, and some modules wrote by TLC language to control the code generation mechanism of RTW and Embedded Coder. Output C-embedded project is compatible with Keil C version 4.0, an IDE environment for the target microcontroller.

Fig. 1 The architecture of the designed library

Development of Matlab/Simulink Library …

157

Each Simulink block comprised three main parts, including C full inline MEXfile, TLC program file, and S-Function of definition block. The C full inline MEX-file defines all information of Simulink block such as inputs/outputs name, type and with of data inputs/outputs, initial sample time, and other configured parameters of the block. The TLC file is a program written in the TLC language and developed by Mathworks. This program can access inputs/outputs, parameters of a block to set up the mechanism of embedded C language generating of this block. The generated mechanism depends on the hardware and C language of the target platform. The S-function blocks are the user interface of the library. They define parameters, link MEX-file, TLC-file, and parameters of a block in the code generation process of Embedded Coder. Besides blocks, a Keil C project can also be created automatically via the designed library by executing a TLC language program. At first, the main file of the C program is generated to initialize the interrupt feature of a Timer module for generating sampling time and create an interrupt service routine for executing the generated code of the Simulink model. After that, a Keil C project is created by changing some XML tags to link all generated files.

3 Design Procedures 3.1 Implementation of the Custom Blocks Simulink block is a user interface part of a Simulink library providing visual information of the current block such as inputs/outputs, parameters, and sample time. The symbol of custom blocks will be presented in the Simulink tree after installing which can be dragged and dropped by users, configured parameters as built-in blocks. A custom block is created by the following procedures. First, studying the hardware architecture of the interested peripheral to configure and manipulate this peripheral using C language. Next, figuring out necessary parameters, inputs/outputs, and suitable data types for creating the Simulink block, writing a C S-Function program to declare them and converting the program to a MEX-file using Matlab MEX compiler. Two formats of MEX files are available based on the Matlab version, including MEX32 and MEX64. Then, creating an S-Function block that links to S-Function and the name of the block’s parameters must be matched with defined parameters in the previous step and a TLC file to control the code generation of this block. Users must have the ability to configure and manipulate the peripherals of the microcontroller in this step. Finally, making a simple Simulink model with a recent block, compiling it and verifying the output C code, and modifying again the block files if the result does not satisfy.

158

N. Van Khanh et al.

3.2 Embedded C Code Generation C files generated from the Simulink block of the microcontroller peripherals in a specific model must be linked with a compiler to convert it to binary code which will be written into the program memory of the microcontroller. With recent libraries, a project of an Integrated Development Environment (IDE) can be created directly by running some TLC scripts. The format of the project should be investigated to figure out the method of creating and modifying the project’s data. In our study, Keil C version 4.0 was used as an IDE of the compiler that manages the project information by XML files, so it is easy to create the project by modifying some tags in these files. The project is generated automatically from Simulink by clicking one button and users must open it to compile and write binary code into the program memory of the microcontroller separately. Three TLC programs were developed to automatically generate the embedded project from Simulink. The first is stm32f407vg_main.tlc which is a custom template to create the content of the main C-program. This file is called by my_file_process.tlc, a custom template based on the built-in template for controlling the code generation process of the Simulink model, during the generation process. The second is keil_project_gen.tlc that finds all generated files and other required files to create the Keil C project by changing necessary XML tags. The last is keil_project_opt_gen.tlc that contains XML tags for configuring other parameters of the Keil C project, such as programmer, the algorithm of the selected programmer.

3.2.1

Generation of Main Program

The content of the main program is fully customized by using the stm32f407vg_main.tlc file. The following listings describe a part of the TLC program in this file that generating the main function of the C program named ert_main.c (i.e. default name of Matlab/Simulink). This C program includes header files of all used peripherals, initializing a Timer interrupt for calling an Interrupt Service Routine code periodically based on the sample time of the Simulink model. A part of TLC program for generating main C program content

Development of Matlab/Simulink Library …

159

static void SampleTime_Config(void); int main(void) { %\ SampleTime_Config(); while(1){ /*Others user task here*/ } return 0; } //sample time configure static void SampleTime_Config(void){ if (SysTick_Config(SystemCoreClock * %)) { while(1); /*handle wrong clocking parameter*/ } } //System tick handler void SysTick_Handler(void){ rt_OneStep(); }

This TLC code is written in C language structure with some customized information wrote in TLC language to retrieve the information from the compiled database of the Simulink model. Two important information were customized in this program. The first is the structure of the main function, which may be changed depending on the compiler. The second is the sample time of the system. This is a crucial parameter and it must be matched between the Simulink model and embedded system. In the Simulink model, the sample time value was stored in the system variable named CompileModel.FixedStepOpts.FixedStep. This variable was used to get the sampling value and configure the system timer (SysTick) of the ARM microcontroller to precisely produce the sampling time of the embedded system. It is noted that the rt_OneStep function contains the whole generated C code of the Simulink model, and it will be called every time the system interrupt occurred. The users must guarantee that the executive time of this function does not exceed the sampling duration.

3.2.2

Generation of Keil C Project

Keil C manages the project by using the XML language. The parameters of the project are stored in separate XML tags. Four parameters must be modified when creating the Keil C project, including project file name, project name, the paths to header files, the name list of C source files in the project. Other information has been set as the default values. The project file name is the Simulink model’s name retrieved from the Simulink system variable CompiledModel.Name. This study used the Cortex-M Software

160

N. Van Khanh et al.

Interface Standard (CMSIS) DSP library to import the definitions of ARM microcontrollers [10]. Two basic files were used that are system_stm32f4xx.c and startup_stm32f4xx.s. The most difficult approach in this study is finding all generated C files and adding them to the project. The number of generated C files depends on the Simulink models. By analyzing all system variables, we found that the number of generated files including header and source files is retrieved by the LibGetNumSourceFiles function. The name and type of the files are got by the LibGetSourceFileSection function. The source file is marked as SystemBody property to distinguish from the header file. The foreach loop in the following listings is applied to scan and add all generated source files to the generated project. The algorithm of finding and adding all generated files into Keil C project %foreach fileIdx = LibGetNumSourceFiles() %assign FileName = LibGetSourceFileSection(fileIdx, ”Name”) %assign FileName = LibGetSourceFileSection(fileIdx, ”Type”)== "SystemBody"?”source”:”header” %if FileType == "source"

% 1 ..\%_ert_rtw\ %

%endif %endforeach

4 Designed Library 4.1 Library Installation The designed library was installed in some Matlab versions including 2009b, 2012b, 2014a, 2015a, and 2015b. After installation, the library has appeared in the Simulink Library Browser window as a pre-installed library. Figure 2 presents this library in Matlab version 2015a named CTUProj Library with seven function blocks, including General-purpose inputs/outputs (GPIOs), Universal Synchronous/Asynchronous Receiver/Transmitter (USART) inputs/outputs, Pulse-width modulation (PWM) outputs, Quadrature Encoder Interface (QEI) inputs. Each block indicates some default parameters, they will be updated in the design window after changed by users. These peripherals are very common in the embedded system and control engineering.

Development of Matlab/Simulink Library …

161

Fig. 2 CTUProj library in Matlab version 2015a after installing

4.2 Embedded C Code and Keil C Project Auto-generation features were demonstrated in this section. One simple Simulink model was created to read the stage of a push-button connected to Pin_0 of GPIOA and display its stages directly to a LED at Pin_12, and concurrently send the inversed logic to LED at Pin_13 of GPIOD. Simultaneously, a LED connected to Pin_14 of GPIOD is blinking under the signal from a Pulse generator (Fig. 3). The model was tested with 0.01 ms of sample time; the sample times of all blocks are similar to model sample time. If one block has a larger sample time, it must be multiplied with the model sample time. After compiling the Simulink model, all developed TLC programs will be executed automatically to generate the C main file (ert_main.c), model C source file (example2.c), Keil C project (example2.uvproj), and other C source files and header files. Figure 4 presents the content of the main file generated by the previous algorithm. The interrupted duration of the C program (i.e. value in the red box) is matched with the sample time of the Simulink model. The C code of the Simulink model was generated in example2_step() function and executed periodically by an Interrupt Service Routine SysTick_Handler() function. Figure 5 shows the compiled output of the Simulink model. The configuration code of the input and output blocks of GPIOs was configured separately (Fig. 5a). They have applied a similar algorithm that is manipulating the logical operations

162

N. Van Khanh et al.

Fig. 3 STM32F4 Discovery board and the Simulink model

Fig. 4 ert_main.c file

AND (&) and OR (|) with a constant to set or clear the corresponding bit of special function registers. The constant was calculated by the TLC language based on the configuration of the users. Figure 5b depicts the sequence of generated code of block in the Simulink model. Each block was generated separately and activated by the system timer interrupt.

Development of Matlab/Simulink Library …

163

Fig. 5 The generated C code of Simulink model: initialization of GPIO (a) and access the logical values of GPIO (b)

Figure 6 indicates the structure of the Keil C project of the Simulink model. The algorithm of the TLC program has found all generated source files and combines them into a Keil C project. With this project, users can compile it and write the binary program into the program memory of the microcontroller in the STM32F4 Discovery board to execute it in real-time.

164

N. Van Khanh et al.

Fig. 6 Structure of the project tree in Keil C version 4.0

5 Conclusion In this study, procedures to create a Simulink library for an unsupported microcontroller have been proposed. The designed library has been integrated some common peripherals of the target microcontroller, including GPIO, ADC, PWM, QEI, and USART. Users can develop their embedded applications directly in the Simulink environment without any further coding. It reduces the developing time of embedded applications and easy to take the advantage of Simulink in designing the application for embedded systems. The designed library has only integrated some basic peripheral blocks. However, with the same mechanism and all templates of the generation process demonstrated, more advanced blocks can be developed in the future. Acknowledgements This study was supported by the Can Tho University, the project number is T2016-11.

References 1. Berisch, G., Donath, H., Heinrich, F., Huebener, I., Lipke, T., Mende, T., Schriefer, M., Schulte, H., Zajac, M.: Design and development of a low cost rapid control prototyping system applied to an air suspension system. In: Proceedings of the 9th IFAC Symposium Advances in Control Education, Russia, pp. 194–199 (2012) 2. Horak, K., Zalud, L.: Image processing on raspberry Pi for mobile robotics. Int. J. Signal Process. Syst. 4(6), 494–498 (2016)

Development of Matlab/Simulink Library …

165

3. Vitale, V., Centioli, C., Iannone, F., Panella, M., Pangione, L., Sabatini, M., Zaccarian, L., Zuccalà, R.: A Matlab based framework for the real-time environment at FTU. Fusion Eng. Des. 82, 1089–1093 (2007) 4. Zuo, W., Li, Y., Wang, F., Hou, X.: A new design method of automotive electronic real-time control System. Phys. Procedia 25, 1126–1132 (2012) 5. Lambersky, V.: Model based design and automated code generation from Simulink targeted for TMS570 MCU. In: 5th European DSP Education and Research Conference (EDERC), pp. 225–228 (2012) 6. Khanh, N.V.: A framework for transferring algorithms designed on Matlab/Simulink to arm microcontroller embedded systems. Can Tho Univ. J. Sci. 4, 36–45 (2016) 7. Lamberský, V., Vejlupek, J., Sova, V., Grepl, R.: Development of Simulink blockset for embedded system with complex peripherals. In: Proceedings of the 2014 International Conference on Electronics and Communication Systems II (ECS’14), Prague, Czech Republic, pp. 112–117 (2014) 8. Mathworks: Target Language Compiler Reference Guide. The Mathworks Inc. (1999) 9. STMElectronics: Discovery Kit with STM32F407VG MCU. User manual (2017) 10. CMSIS DSP Software Library. https://www.keil.com/pack/doc/CMSIS/DSP/html/. Accessed 28 Oct 2020

A Novel Bistatic LIDAR Device with 1570 nm Centre Wavelength Diode for Detection of Plant Disease Hai Pham, Khanh Nguyen, Ngon Nguyen, Hung Tran, and W. Genthe

Abstract Remote sensing (RS) is capable of acquiring information of an object with non-physical contact, which is applied to monitor crop health for timely mitigation purpose in nutrition and health conditions. A new bistatic Light Detection and Ranging (LIDAR) system is used to early detect crop diseases presented in this paper. The crop canopy anomalies in carbon dioxide (CO2 ) concentration, which is possibly related to disorder in plant photosynthesis, are measured by the proposed bistatic LIDAR system. Particularly, the initial modeling and calculation activities are carried out and discussed in the paper in order to predict the sensor performance. Future experiments are also developed according to the results of this analysis. Keywords Carbon dioxide · Crop diseases · Infrared · Laser · Photosynthesis · LIDAR · Remote sensing

1 Introduction Agricultural crops are affected by a variety of plant diseases, weeds and pests that cause a remarkable yield loss and quality degradations. Pesticides are used to protect crops from weeds, pests and diseases that increase the products’ costs and harm the H. Pham (B) · K. Nguyen · N. Nguyen · H. Tran Can Tho University, Can Tho, Vietnam e-mail: [email protected]; [email protected] K. Nguyen e-mail: [email protected] N. Nguyen e-mail: [email protected] H. Tran e-mail: [email protected] W. Genthe Gardes-du-Corps Straße 14, 14059 Berlin, Germany e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2021 S. Choudhury et al. (eds.), Intelligent Communication, Control and Devices, Advances in Intelligent Systems and Computing 1341, https://doi.org/10.1007/978-981-16-1510-8_17

167

168

H. Pham et al.

surrounding environment with toxic residues. Consequently, the spatial and temporal expansion of weeds, pests and diseases over broad areas are characterized by diversity of RS techniques. A group of plants can be detected by RS compared to the individual plants assessed by normal manual techniques with physical proximity [1]. RS is employed to identify the presence of diseases via alterations in gas exchange, cell structure, pigment, chemical concentrations, nutrient and water uptake. Differences in the spectral and shape characteristics of canopy caused by biotic and abiotic stress agents can be determined and assessed by conventional RS methods. For instance, canopy reflectance variations can be induced by pests and diseases, inducing variances in color and temperature of the canopy, which can be detected across the visible and infrared spectrum [2]. The utilization of RS for assessing crop diseases has been researched since the last century. Specifically, in the 1950s, the presence of certain cereal crop diseases was detected using infrared photographs [3]. In the 1980s, crop conditions were also estimated by aerial photography [4]. In these studies, broad spectral bands were recorded by airborne cameras on analog films. RS technology has evolved significantly ever since. Plant pathogens have been observed and quantified by many spectroscopic and imaging techniques [5–8]. These sensors have been used to identify changes in plant physiology as early detection of disease, i.e. visible (VIS)/multiband spectroscopy [9], multispectral or hyperspectral imaging [10, 11], fluorescence spectroscopy [12], infrared (IR) spectroscopy [13] and fluorescence imaging [14]. Not all pests make change in leaf color, but they all change CO2 sink in/on the canopy, for example, Rice stem gall midge, strawberry mildew, etc. The aim of this paper is, therefore, to develop stationary RS methods to early detect crop diseases by the change of CO2 -flux above a crop canopy. Field-based conditions are planned with many numerical and experimental activities in order to identify the appropriate compound of sensors/systems for particular diseases and support the sensor design and integration on a stationary bistatic LIDAR platform. Subsequent ground test activities will support a full system validation in real conditions and test/training on actual plantations.

1.1 Rice Leaf Blight and Blast Disease Multispectral and Hyperspectral Imaging (MSI/HIS) methods have been developed maturely and used to detect plant diseases in the early states, according to a significant number of literatures. The distinction in the spectral signatures of infected and healthy plant leaves can be detected by these two techniques. However, the similarities of spectral signatures in the early stages of many considerable diseases cannot be identified. Rice leaf blight and blast disease are caused by bacteria and fungi and are the most dangerous on rice cultivation area. Both diseases make the rice leaf loose its green color and consequently reduce its photosynthesis capacity. These two diseases can be occurring at any growth stage during crop season and starting in small group of rice plants (spot) and spread out very rapidly. Due to suitable microclimate,

A Novel Bistatic LIDAR Device with 1570 nm Centre …

169

the spot occurs in the middle of a large rice field, where it is difficult for farmer to access. Therefore, it is necessarily detected as early as possible and remotely, since the diseases may cause server loss of rice yield.

1.2 CO2 Sinks by Plants Sources and sinks of CO2 in a crop are carefully taken into account in this paper. Soil and vegetation respiration processes lead to increase below-canopy CO2 concentrations in night-time [15]. The complex interactions of the soil, vegetation and boundary layer atmosphere are the reasons for dynamic and progressive changes in CO2 concentrations within the canopy and above. The CO2 concentrations around the canopy are very dynamic due to diffusion and turbulence processes. Even though, in field conditions, CO2 concentrations around the canopy are significantly lower than the atmospheric sea-level averages during photosynthetic activity periods (sunlight) and usually higher than average outside these (night-time) [16–20]. Especially, in the area just above the plant canopy, the Leaf boundary layer (LBL) [21], because of the change in CO2 flux that is formed a profile above the plant canopy, variation in CO2 -flux possibly indicate plants stress caused by diseases and other factors. The Leaf Area Index (LAI) is a useful parameter, since it indicates growth stages and the intensity of CO2 -flux [22].

2 Information from HITRAN 2020 There are typically five gases that should be considered in this project in order to measure the CO2 gas concentrations on the field, which are CO2 , CH4 , H2 O, N2 and NH3 . Besides the main CO2 gas, the other four gases are considered as interferences. Figure 1 shows the absorption cross-sections of five gases (downloaded from HITRAN 2020 database) associated with the wavelength ranges between 1538 and 1665 nm in which the CO2 absorption cross-sections are dominant. The absorption cross-section wavelength line interval is identified as 0.475907. It can be seen that the best wavelength lines for measure CO2 concentrations are at 1600, 1576 and 1570 nm.

2.1 Bistatic LIDAR System Concept Light detection and ranging (LIDAR) systems are a practical method for sensing atmospheric constituents remotely due to the characterization of the received signal passed through the medium from the laser transmitter. Molecular species have the properties to absorb the selected wavelength based on the Differential Absorption

170

H. Pham et al.

Fig. 1 Molecules absorption cross-section from HITRAN2020 database

LIDAR (DIAL) measurement principle that is widely used for RS [15, 23]. Nevertheless, the majority of DIAL systems are commonly monostatic that necessitates compound and costly components to identify the aerosol particles with extremely weak backscatter. For this research, we accordingly pay particular attention to the bistatic LIDAR developed by [24], which is a cost-effective solution that can be integrated on small drones for early detection of crop diseases. However, the extension of this method to the agricultural sector was not evaluated and in particular for measuring CO2 concentrations associated with photosynthesis and respiration. The proposed bistatic LIDAR system, which is dedicated to Thorlabs device, is conceptually depicted in Fig. 2. The bistatic LIDAR has transmitter (Tx) and detector (Rx) separately at a certain distance. The Tx of the bistatic LIDAR, which is attached with the collimations lens, normally emits a continuous wave laser beam of precisely known power characteristics. The bistatic LIDAR Rx together with custom-designed telescope measures the amount of laser energy that reaches the Rx, which is less than the emitted LIDAR radiation because atmospheric molecular/aerosol concentrations absorb or scatter the radiation.

2.2 DIAL Measurement Principles CO2 column density is retrieved by DIAL technique that is based on near infrared (NIR) bistatic LIDAR irradiance measurements. In the DIAL technique, laser beams

A Novel Bistatic LIDAR Device with 1570 nm Centre …

171

Fig. 2 Bistatic LIDAR system concept including environmental factors

at two specific wavelengths are emitted from laser source towards a calibrated photodiode detector for each measurement cycle. The first wavelength is selected associated with peak absorption band (on-absorption). The second wavelength is selected according to minimal absorption band (off-absorption), which is close to the first one for allowing a correlation. These two beams are attenuated by the concentration of molecules gases in the cell. The transmitted beams are received by a calibrated photodiode detector, which carries out radiometric measurements. Based on DIAL technique, the molecular scattering (β m ), aerosol absorption (α a ) and aerosol scattering (β a ) are neglected in the measurement. For the molecular absorption (α m ), there are actually two parameters including absorption σa (αm ) and scattering σs (αm ). However, the DIAL measurements mainly focus on molecule vibrational absorption, σa (αm ) associated with particular laser wavelengths. Signal analysis enables to estimate the CO2 concentration. The transmittance coefficient τ is affected by absorption and scattering of molecules and aerosols along the transmission path and can be calculated based on the Beer–Lambert’s law as given by: λ2

¨ 2] − ∫ σCO2 (λ,z,P,T )·dλ·[CO PR X . = e λ1 τλ = PT X

(1)

where PTX is the transmitted beam power [nW]; PRX is the received beam power [nW]; λ1,2 = CWL ∓ FWHM [nm]; CWL = center wavelength [nm]; FWHM = full width 2 half maximum of the laser bandwidth [nm]; σCO2 (λ, z, P, T ) is absorption crosssection as a function of wavelength (λ), transmitted path length (z), pressure (P) and    z  ¨ 2 = ∫ n C O2 ·dz temperature (T ) based on HITRAN2020 database cm2 mol−1 ; CO 0   is the volume density of the CO2 species mol cm−3 and z is the distance along

172

H. Pham et al.

the transmitted path [cm]. Assuming the superposition principle σC O2 (λ) can be expressed to separately account for the attenuation effects of different aerosol and molecular species, as: σCO2 (λ, z, P, T ) =



ψk (λ, z, P, T ) +

k

2   √ α j (λ, z, P, T ) αi (λ, z, P, T ) αn (λ, z, P, T ) + + ··· + ρi ρj ρn

(2)

where ψk (λ, z, P, T ) is the  sum of molar absorption cross-section according with particular wavelength cm2 mol−1 determined by HITRAN2020 Application Programming (HAPI); αn (λ, z, P, T ) is total absorption coefficients of n   Interface molecules cm−1 obtained by Bytran  softwaredepend on HITRAN2020 library; ρn is the total density of n molecules mol cm−3 . Furthermore, measurements based on Eq. (2) introduce measure error avoidance, which is collision-induced absorption (CIA) adapted from [25] in CIA equation of binary mixture of gases. Hence, it is essential to determine these extra impacts to accurately measure the CO2 concentrations. The equation is nonetheless used for most of the system design activities. In particular, the DIAL technique is based on two predefined wavelengths, selected based on the response of measured atmospheric constituent. The first wavelength, λON , is selected in relation to one major vibrational band of the molecular species being measured. The second wavelength, λOFF , is selected in proximity to the first, but outside the vibrational band of the molecular species, in order to maximize the difference in molecular cross-section σCO2 = σCO2 (λON ) − σCO2 (λOFF ). Therefore, the relation between transmitted energy associated with the on-transmission (τ ON ) and the off-transmission line (τ OFF ) for determining the CO2 concentration is adapted from [26] as follow: Z

PRx = PT x · e

−[σCO2 (λ O N )−σCO2 (λ O F F )]·∫ n CO2 (z)dz 0

.

(3)

Hence, for an ON/OFF beam pair with the same path length z and transmitted ¨ 2 over the beam column can be power PT X , the average CO2 volume density CO determined using the following equation:   ¨2 = CO

− ln[τON /τOFF ] ln(τOFF ) − ln(τON )  = z · σCO2 z · σCO2 (λOFF ) − σCO2 (λON )

(4)

where σCO2 (λON ) and σCO2 (λOFF ) are selected in the near-infrared band using the High Resolution Transmission (HITRAN) database 2020 edition, which was also used to determine the cross-sections of other atmospheric constituents, the most important of which is water vapor (H2 O). The selected wavelengths are shown in Fig. 3, where λON is depicted as a dark green rectangle and λOFF is indicated by a light green one, which are associated with FWHM of dedicated Thorlabs laser diode device. At typical atmospheric concentrations, the cross-section of H2 O (green lines) is considerably smaller than CO2 one (red lines). In addition, the spectral bandwidth of the laser source (black dash-dotted line) must be narrow enough to cover lambda λON and

A Novel Bistatic LIDAR Device with 1570 nm Centre …

173

Fig. 3 Selected HITRAN2020 absorption cross-section together with beam ON/OFF according to FWHM of dedicated Thorlabs laser diode device

λOFF separately and any excess in bandwidth is to be considered a measurement error.

2.3 Discussion The expected power is simulated based on Eq. (1) with the length between transmitter and detector set to 100 m and the absorption coefficient contains five gases of CO2 , H2 O, N2 , NH4 and NH3 . In fact, the expected power has been calculated according to the selected λON and λOFF . The minimum CO2 concentration, which is calculated  by post-processing, that detected by the device is 20.78 mg m−3 as shown in Fig. 4, since the calculated Pdiff of 382 nW is about three times greater than the determined  device’s error of 127 nW. The device can be used to detect 411 mg m−3 CO2 that considered as ambient CO2 concentration. Furthermore, we can identify 0.3   is −3 by using the Thorlabs detector PM16-144, since the Pdiff mg m  CO2 differences  with 0.3 mg m−3 interval is 1.33 nW compared to 1 nW from the PM16-144 device datasheet. There are 10 samples associated with the CO2 concentrations ranged from 10 to  500 mg m−3 , which are set for this simulation activity. Figure 5 shows the inversed concentrations indicated by red line that calculated based on the received powers ON and OFF, whereas the fixed concentrations are illustrated by blue line. It is noticed that there are errors between inversed values and fixed values according to the availability of other molecular gases along the transmitted path length. The concentration step is estimated as 0.06 ppm per 1 nW received power. The received noise is negligible due to the calculated Pdiff without CO2 gas involved is 5.23 nW compared with the

174

H. Pham et al.

Fig. 4 Detected powers ON and OFF according to various fixed CO2 concentrations

Fig. 5 Inversed CO2 concentrations from detected powers ON and OFF versus fixed CO2 concentrations

  one of 10 mg m−3 . CO2 gas is 186.54 nW. The proposed bistatic LIDAR technique has another advantage in avoiding the shifted background thanks to the emitted laser beam of ON and OFF, simultaneously.

A Novel Bistatic LIDAR Device with 1570 nm Centre …

175

3 Conclusions A novel bistatic LIDAR device is proposed in this paper together with design concept experiments with 100 m path length from transmitter to detector. Anomalies in the CO2 gas concentrations over the canopy, which is symptomatic of the presence of a disease, can be detected by the bistatic LIDAR device compared with MSI/HIS methods. The experimental characterization and calibration of the system in the lab in a crafted chamber and a rice field as well will be addressed in further works. Ultimately, a stationary bistatic LIDAR system along with air temperature and humidity sensors will be tested in a rice field in various atmospheric conditions.

References 1. Omran, E.-S.E.: Early sensing of peanut leaf spot using spectroscopy and thermal imaging. Arch. Agron. Soil Sci. 63(7), 883–896 (2017) 2. Raikes, C., Burpee, L.: Use of multispectral radiometry for assessment of Rhizoctonia blight in creeping bentgrass. Phytopathology 88(5), 446–449 (1998) 3. Colwell, R.: Determining the prevalence of certain cereal crop diseases by means of aerial photography. Calif. Agric. 26(5), 223–286 (1956) 4. Curran, P.J.: Aerial photography for the assessment of crop condition: a review. Appl. Geogr. 5(4), 347–360 (1985) 5. Mutka, A.M., Bart, R.S.: Image-based phenotyping of plant disease symptoms. Front. Plant Sci. 5 (2014) 6. Fahlgren, N., Gehan, M.A., Baxter, I.: Lights, camera, action: high-throughput plant phenotyping is ready for a close-up. Curr. Opin. Plant Biol. 24, 93–99 (2015) 7. Goggin, F.L., Lorence, A., Topp, C.N.: Applying high-throughput phenotyping to plant–insect interactions: picturing more resistant crops. Curr. Opin. Insect Sci. 9, 69–76 (2015) 8. Mahlein, A.-K.: Plant disease detection by imaging sensors–parallels and specific demands for precision agriculture and plant phenotyping. Plant Dis. 100(2), 241–251 (2016) 9. Chen, B., et al.: Spectrum characteristics of cotton canopy infected with verticillium wilt and inversion of severity level. In: International Conference on Computer and Computing Technologies in Agriculture. Springer (2007) 10. Qin, J., et al.: Detection of citrus canker using hyperspectral reflectance imaging with spectral information divergence. J. Food Eng. 93(2), 183–191 (2009) 11. Shafri, H.Z., Hamdan, N.: Hyperspectral imagery for mapping disease infection in oil palm plantation using vegetation indices and red edge techniques. Am. J. Appl. Sci. 6(6), 1031 (2009) 12. Lins, E., Belasque, J., Marcassa, L.: Detection of citrus canker in citrus plants using laser induced fluorescence spectroscopy. Precis. Agric. 10(4), 319–330 (2009) 13. Deborah, E.P., et al.: Near-infrared spectroscopy for the prediction of disease ratings for Fiji leaf gall in sugarcane clones. Appl. Spectrosc. 63(4), 450–457 (2009) 14. Sankaran, S., et al.: A review of advanced techniques for detecting plant diseases. Comput. Electron. Agric. 72(1), 1–13 (2010) 15. Prueger, J., et al.: Carbon dioxide dynamics during a growing season in midwestern cropping systems. Environ. Manag. 33(1), S330–S343 (2004) 16. Keeling, C.D.: The concentration and isotopic abundances of atmospheric carbon dioxide in rural areas. Geochim. Cosmochim. Acta 13(4), 322–334 (1958) 17. Wright, J.L., Lemon, E.: Photosynthesis under field conditions. IX. Vertical distribution of photosynthesis within a corn crop 1. Agron. J. 58(3), 265–268 (1966)

176

H. Pham et al.

18. Allen Jr., L.: Variations in carbon dioxide concentration over an agricultural field. Agric. Meteorol. 8, 5–24 (1971) 19. Reicosky, D.: Diurnal and seasonal trends in carbon dioxide concentrations in corn and soybean canopies as affected by tillage and irrigation. Agric. For. Meteorol. 48(3–4), 285–303 (1989) 20. Buchmann, N., Kao, W.Y., Ehleringer, J.R.: Carbon dioxide concentrations within forest canopies—variation with time, stand structure, and vegetation type. Glob. Change Biol. 2(5), 421–432 (1996) 21. Vesala, T.: On the concept of leaf boundary layer resistance for forced convection. J. Theor. Biol. 194(1), 91–100 (1998) 22. Lemon, E., Wright, J.: Photosynthesis under field conditions. XA. Assessing sources and sinks of carbon dioxide in a corn (Zea mays L.) crop using a momentum balance approach 1. Agron. J. 61(3), 405–411 (1969) 23. Schotland, R.M.: Errors in the lidar measurement of atmospheric gases by differential absorption. J. Appl. Meteorol. 1988(13), 71–77 (1974) 24. Gardi, A., Sabatini, R., Wild, G.: Unmanned Aircraft Bistatic LIDAR for CO2 Column Density Determination in Metrology for Aerospace (MetroAeroSpace). IEEE (2014) 25. Baranov, Y.I.: On the significant enhancement of the continuum-collision induced absorption in H2 O + CO2 mixtures. J. Quant. Spectrosc. Radiat. Transf. 175, 100–106 (2016) 26. Sabatini, R., et al.: Airborne laser systems for atmospheric sounding in the near infrared. In: Laser Sources and Applications. International Society for Optics and Photonics (2012)

FPGA-Based Power Optimized CAM Design Using LVCMOS18 and High-Speed Low Voltage Digitally Controlled Impedance Reeya Agrawal, Manish Kumar, Manish Gupta, and Vinay Kumar Deolia

Abstract A cache addressable memory is a special memory that performs a single clock-cycle search operation. For packet forwarding and packet classification, CAMs are most wanted in a network router. A high-speed table is needed to implement the CAM application. Reducing power consumption is the primary reason for CAM design. But CAM also has drawbacks, such as low-cost density and high per-bit cost. This paper reviews two stages, i.e. the level of the circuit and the level of architecture. At the circuit level, the analysis is conducted at a low power match line, but three methods for power consumption are architecture level analysis. In this paper, using Xilinx standard VHDL, the power consumption was carried out at different frequencies for different Input/Output requirements. Keywords Content addressable memory (CAM) · Clock power (CP) · Logic power (LP) · Signal power (SP) · Input/output power (I/OP)

1 Introduction CAM is a special type of memory of a computer. It is used in high-speed searching applications. It is composed of conventional semiconductor memory [1, 2]. CAM is also called associative memory. Two search intensive tasks that are done by using CAM internet routers are (a) Packet forwarding and (b) Packet classification as shown in Fig. 1. R. Agrawal (B) · M. Kumar · M. Gupta · V. K. Deolia GLA University, Mathura, India e-mail: [email protected] M. Kumar e-mail: [email protected] M. Gupta e-mail: [email protected] V. K. Deolia e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2021 S. Choudhury et al. (eds.), Intelligent Communication, Control and Devices, Advances in Intelligent Systems and Computing 1341, https://doi.org/10.1007/978-981-16-1510-8_18

177

178

R. Agrawal et al.

Fig. 1 Comparison of power

Clock Power

Logic Power

I/O Power

Power

Signal Power

In this paper, we will discuss the power consumption at different frequencies for different Input/output (I/O) standards and will compare the CP, LP, I/OP, SP, leakage power, and total power. Depending on the results, researchers will see which is more suitable to use as power efficient [3, 4].

1.1 Content-Addressable Memory Binary content-addressable memory has fewer transistors hence it supports reducing the size of the transistors. In binary content-addressable memory, the date zero and one are stored in a memory cell then the input data and the stored data are compared [5]. After comparing both input data and stored data, the address of a location is searched where the associated data are stored. But in the case of Ternary content, addressable memory does not care zero and one data are stored in the memory cell [6]. Ternary content addressable memory is used for finding any of the sets of values in a single searching operation that shares some of the bit values but not all [7]. Content addressable memory contains a routing table. Figure 2 above shows a content addressable memory that implements the lookup. This table contains cells that are arranged into four words horizontally and there are five bits arranged vertically as shown in fig. 3. Core blocks contain storage as well as circuitry for comparison. There are two lines, search line and match line, Fig. 2 Basic forms of content addressable memory

CAM Binary CAM

Ternary CAM

FPGA-Based Power Optimized CAM Design …

179

Fig. 3 Output port address of a routing table in CAM

0110X

101XX

Output Port Address

011XX

10011

where the search line runs vertically, while the match line runs horizontally and tells whether the data that searched are matching the word in the row [8, 9]. The match is shown but the activated match line and non-match are shown by deactivated match line i.e. known as a mismatch in the CAM literature. The match lines are inputs to an encoder, which generates the address corresponding to the match location [10].

2 Literature Review In a fully parallel content, addressable memory is proposed to minimize power dissipation and circuit area using a lightweight, dynamic CAM cell with a stacked condenser structure and a novel hierarchal priority coding [11]. This involves two condensers and five transistors of NMOS . The implementation of frequency scaling dependent energy is proposed in the energy-saving design and implementation of the arithmetic logic unit (ALU) on the 40ηm field-programmable gate array (FPGA) [12]. If the battery voltage or current reaches the threshold voltage, an overhead warning will be raised. With frequency scaling, it is possible to reduce the frequency from 1 THz to 25 GHz [13]. The ultra-scale FPGA design of low voltage data capture interface (LVDCI) I/O standard green image ALU is suggested in [14] to offer low-voltage energy efficiency. The low voltage energy efficient digitally regulated impedance Vedic multiplier architecture is proposed at 28 ηm of FPGA [15]. The I/O norm is most significant in circuit design power dissipation. Researchers should, therefore, choose the Vedic multiplier for the most energy-efficient I/O norm.

3 Power Analysis of Content Addressable Memory If low voltage complementary metal oxide semiconductor (LVCMOS_18) in place of HSLVDCI_18, then it will save 82.92%, 20%, 98.43%, 94.84% of CP, LP,

180

R. Agrawal et al.

leakage power, and total power, respectively, but can save 13.95% of SP when use HSLVDCI_15 in place of HSLVDCI_18 as shown in Table 1 and Fig. 4. If LVCMOS_18 in place of HSLVDCI_18, then it will save 83.13%, 20%, 98.43%, 94.69% of CP, LP, leakage power, and total power, respectively, but can save 4.54% of SP when use HSLVDCI_15 in place of HSLVDCI_18 as shown in Table 2 and Fig. 5. If LVCMOS_18 in place of HSLVDCI_18, then it will save 83.72%, 20%, 98.43%, 94.69% of CP, LP, leakage power, and total power, respectively, but can save 2.22% of SP when use HSLVDCI_15 in place of HSLVDCI_18 as shown in Table 3 and Fig. 6. If LVCMOS_18 in place of HSLVDCI_18, then it will save 83.33%, 15.38%, 98.40%, 93.53% of CP, LP, leakage power, and total power, respectively, but can save 3.33% of SP when use HSLVDCI_15 in place of HSLVDCI_18 as shown in Table 4 and Fig. 7. Table 1 Power consumption on 1.73 GHz frequency S. no.

Device

Clock

Logic

Signal

I/O

Leakage

Total

1.

HSLVDCI_18

0.082

0.010

0.043

0.047

2.882

3.064

2.

LVCMOS_18

0.014

0.008

0.042

0.049

0.045

0.158

3.

HSLDVCI_15

0.082

0.009

0.037

0.047

2.881

3.056

3.5 3

Clock

2.5

Logic

2

Signal I/O

1.5

Leakage

1

Total

0.5 0 HSLVDCI_18 LVCMOS_18 HSLVDCI_15 Fig. 4 Powers with HSLVDCI and LVCMOS I/O standard

Table 2 Power consumption on 1.74 GHz frequency S. no.

Device

Clock

Logic

Signal

I/O

Leakage

Total

1.

HSLVDCI_18

0.083

0.010

0.044

0.048

2.882

3.065

2.

LVCMOS_18

0.014

0.008

0.042

0.049

0.045

0.159

3.

HSLDVCI_15

0.083

0.010

0.037

0.047

2.881

3.057

FPGA-Based Power Optimized CAM Design …

181

3.5 3 Clock

2.5

Logic

2

Signal

1.5

I/O

1

Leakage

0.5

Total

0 HSLVDCI_18 LVCMOS_18 HSLVDCI_15 Fig. 5 Powers with HSLVDCI and LVCMOS I/O standard Table 3 Power consumption on 1.8 GHz frequency S. no.

Device

Clock

Logic

Signal

I/O

Leakage

Total

1.

HSLVDCI_18

0.086

0.010

0.045

0.049

2.882

3.072

2.

LVCMOS_18

0.008

0.008

0.044

0.051

0.045

0.163

3.

HSLDVCI_15

0.010

0.010

0.039

0.048

2.881

3.064

3.5 3

Clock

2.5

Logic

2

Signal

1.5

I/O

1

Leakage

0.5

Total

0 HSLVDCI_18

LVCMOS_18

HSLVDCI_15

Fig. 6 Powers with HSLVDCI and LVCMOS I/O standard Table 4 Power consumption on 2.4 GHz frequency S. no.

Device

Clock

Logic

Signal

I/O

Leakage

Total

1.

HSLVDCI_18

0.114

0.013

0.060

0.067

2.885

3.139

2.

LVCMOS_18

0.019

0.011

0.058

0.069

0.046

0.203

3.

HSLDVCI_15

0.114

0.013

0.052

0.065

2.884

3.128

182

R. Agrawal et al.

3.5 3 Clocl

2.5

Logic

2

Signal

1.5

I/O

1

Leakage

0.5

Total

0 HSLDVCI_18

LVCMOS_18

HSLDVCI_15

Fig. 7 Powers with HSLVDCI and LVCMOS I/O standard

If LVCMOS_18 in place of HSLVDCI_18, then it will save 82.92%, 20%, 98.43%, 94.84% of CP, LP, leakage power, and total power, respectively, but can save 13.95% of SP when we use HSLVDCI_15 in place of HSLVDCI_18 as show in Table 5 and Fig. 8. Table 5 Power consumption on 3.06 GHz frequency S. no.

Device

Clock

Logic

Signal

I/O

Leakage

Total

1.

HSLVDCI_18

0.146

0.017

0.077

0.086

2.884

3.212

2.

LVCMOS_18

0.025

0.014

0.074

0.088

0.046

0.247

3.

HSLDVCI_15

0.146

0.017

0.066

0.084

2.886

3.199

3.5 3 Clock

2.5

Logic

2

Signal

1.5

I/O

1

Leakage

0.5

Total

0 HSLVDCI_18

LVCMOS_18

HSLVDCI_15

Fig. 8 Power with HSLVDCI and LVCMOS I/O standard

FPGA-Based Power Optimized CAM Design …

183

4 Conclusion In this paper, CAM circuit and high capacity CAM have been concluded. The analysis was done at the circuit level of two basic CMOS cells i.e. NOR cell and NAND cell. Furthermore, the power consumption at different frequencies for different I/O standards concludes that using LVCMOS_18 is more advantageous and reliable. LVCMOS_18 at all possible frequencies is proved most power-efficient. If use LVCMOS_18, then the total power consumption of the circuit to 94.84%, 94.69%, 93.53%, 92.31% at 1.73 GHz, 1.8 GHz, 2.4 GHz, 3.06 GHz, respectively, when it compared with the circuit, which is designed using HSLVDCI_18, using Xilinx support, VHDL power is calculated.

5 Future Scope I/O specification is an effective and highly efficient way of reducing circuit power consumption. Therefore, to make such a design minimizing the consumption and making it a circuit power efficiently, it is very important to choose the correct I/O norm. It can be used to maximize the system’s energy performance by using other I/O specifications such as the arithmetic logic unit (ALU), the control unit (CU), the memory unit (MU). In a broad range of applications, CAM is a key element. Power consumption can reduce and increase speed without any loss of stored data by deactivating CAM blocks.

References 1. Kannan, P., Raghunathan, K.S., Jayaraman, S.: Aspects and solutions to designing standard LVCMOS I/O buffers in 90 nm process. In: AFRICON 2007, Windhoek, pp. 1–7 (2007). https:// doi.org/10.1109/afrcon.2007.4401475 2. Agrawal, T., Kumar, A., Priyanka, Aggarwal, P., Tirmizi, S.S.: LVCMOS based 4-bit register. In: 2018 9th International Conference on Computing, Communication and Networking Technologies (ICCCNT), Bangalore, pp. 1–3 (2018). https://doi.org/10.1109/icccnt.2018.849 4059 3. Kumar, T., Pandey, B., Das, T., et al.: Mobile DDR IO standard based high-performance energy efficient portable ALU design on FPGA. Wirel. Pers. Commun. 76, 569–578 (2014). https:// doi.org/10.1007/s11277-014-1725-z 4. Kalra, L., Bansal, N., Saini, R., Bansal, M., Pandey, B.: LVCMOS I/O standard-based environment-friendly low power ROM design on FPGA. In: 2015 2nd International Conference on Computing for Sustainable Global Development (INDIACom), New Delhi, pp. 1824–1829 (2015) 5. Singh, P.R., Pandey, B., Kumar, T., Das, T.: LVCMOS I/O standard-based million MHz highperformance energy-efficient design on FPGA. In: 2013 International Conference on Communication and Computer Vision (ICCCV), Coimbatore, pp. 1–4 (2013). https://doi.org/10.1109/ icccv.2013.6906738

184

R. Agrawal et al.

6. Kumar, A., Pandey, B., Akbar Hussain, D.M., Atiqur Rahman, M., Jain, V., Bahanasse, A.: Low voltage complementary metal oxide semiconductor based energy efficient UART design on Spartan-6 FPGA. In: 2019 11th International Conference on Computational Intelligence and Communication Networks (CICN), Honolulu, HI, USA, pp. 84–87 (2019). https://doi.org/ 10.1109/cicn.2019.8902356 7. Hanzawa, S., et al.: A large-scale and low-power CAM architecture featuring a one-hot-spot block code for IP-address lookup in a network router. IEEE J. Solid-State Circuits 40(4), 853–861 (2005) 8. Pagiamtzis, K., Sheikholeslami, A.: Content-addressable memory (CAM) circuits and architectures: a tutorial and survey. IEEE J. Solid-State Circuits 41(3), 712–727 (2006). https://doi. org/10.1109/jssc.2005.864128 9. Kim, K., Jeong, G.: Memory technologies in the nano-era: challenges and opportunities. In: 2005 International Conference on Integrated Circuit Design and Technology, 2005. ICICDT 2005. IEEE (2005) 10. Sheikholeslami, A., Glenn Gulak, P.: A survey of circuit innovations in ferroelectric randomaccess memories. Proc. IEEE 88(5), 667–689 (2000) 11. Yamagata, T., Mihara, M., Hamamoto, T., Murai, Y., Kobayashi, T.: A 288-kb Fully Parallel Content Addressable Memory Using a Stacked-Capacitor Cell Structure, Dec 1992 12. Pandey, B., et al.: Energy-efficient design and implementation of ALU on 40 nm FPGA. In: 2013 International Conference on Energy-Efficient Technologies for Sustainability (ICEETS). IEEE (2013) 13. Pandey, B., Kumar, R.: Low voltage DCI based low power VLSI circuit implementation on FPGA. In: 2013 IEEE Conference on Information & Communication Technologies (ICT). IEEE (2013) 14. Das, T., et al.: LVDCI I/O standard-based green image ALU design on ultra-scale FPGA. In: 2013 8th IEEE International Conference on Industrial and Information Systems (ICES). IEEE (2013) 15. Goswami, K., et al.: Low voltage digitally controlled impedance based energy efficient vedic multiplier design on 28 nm FPGA. In: 2014 International Conference on Computational Intelligence and Communication Networks (CICN). IEEE (2014)

Design of Chili Fruit Flipping Mechanism for Identification of the Damages Caused by Diseases Quoc-Khanh Huynh, Chi-Ngon Nguyen, Wei-Chih Lin, Hong-Phuc Vo-Nguyen, Phuong Lan Tran-Nguyen, Dang-Khanh-Linh Le, and Van-Cuong Nguyen Abstract An identification model for detecting damages on fresh destemmed chili (capsicum) was developed, but the limitation is that the cameras can only capture from one side. Therefore, this study focused on developing the flipping mechanism, which rotates and flips chili fruits and helps the camcorder to observe and recognize errors on the entire body. The model was designed, fabricated and tested. Experiments were performed on 2400 fruits with 12000 flipping times, at 12 different velocities. The effective flipping rate was up to 96.3%. The fruits destroying percentage was also found to be remarkably low. These results were acceptable. The mechanism would be used to implement and test the identification algorithms in future chili grading systems. Keywords Fresh destemmed chili fruit · Flipping mechanism · Identification · Damages Q.-K. Huynh (B) · C.-N. Nguyen · P. L. Tran-Nguyen · V.-C. Nguyen College of Engineering Technology, Can Tho University, Can Tho City, Vietnam e-mail: [email protected] C.-N. Nguyen e-mail: [email protected] P. L. Tran-Nguyen e-mail: [email protected] V.-C. Nguyen e-mail: [email protected] W.-C. Lin · D.-K.-L. Le School of Medicine, Can Tho University of Medicine and Pharmacy, Can Tho City, Vietnam e-mail: [email protected] D.-K.-L. Le e-mail: [email protected] H.-P. Vo-Nguyen Department of Mechanical and Electromechanical Engineering, National Sun Yat-Sen University, Kaohsiung City, Taiwan e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2021 S. Choudhury et al. (eds.), Intelligent Communication, Control and Devices, Advances in Intelligent Systems and Computing 1341, https://doi.org/10.1007/978-981-16-1510-8_20

185

186

Q.-K. Huynh et al.

1 Introduction Chili (capsicum) is widely grown in Vietnam, especially in the Mekong Delta region with large production and processing capacity [1]. Its nutritional ingredients are really good for health [2, 3]. In order to enhance the value and competitiveness, the processing process, especially stem separation, is essential [4, 5]. The destemming system studies have been carried out. However, their products still need to be classified to achieve higher quality [6, 7]. Various fruit classification systems have been researched and developed. Among them, computer vision, in combination with artificial intelligence, is used to obtain a high successful identification rate [8, 9]. In the previous study, an image recognition algorithm using a convolution neural network (CNN) was built to detect damages caused by diseases that appeared on the destemmed fresh chili fruits, as demonstrated in Fig. 1 [10]. The accurate identification rate obtained between damaged and non-damaged ones was up to 97.5%. As described in Fig. 2, the camera could only capture from one side (the upside),

10 mm

10 mm

10 mm

10 mm

10 mm

Fig. 1 Five types of fresh destemmed chili fruit. a Non-damaged. b Head-damaged. c Middamaged. d Side-view damaged. e Tail-damaged

Fig. 2 Visible zone for identification

Design of Chili Fruit Flipping Mechanism …

187

and the defects on the other side (the downside) could not be detected for classification. This study developed a flipping mechanism that can rotate the invisible zone of the fruit’s body currently located at the downside, flip it upward to observe and check the entire fruit. The proposed model was designed, fabricated, and tested to evaluate the flipping ability.

2 Mechanism Design 2.1 The Identification Ability of CNN Model As shown in [10], damage incurred was identified and classified into one of four groups: head-damaged (Fig. 1b), mid-damaged (Fig. 1c), side-view damaged (Fig. 1d) and tail damaged (Fig. 1e), in which, depending on the fruit position, the damage could be identified as mid-damaged in direct view or side-view damaged in an inclined view. The accuracy identification rate was up to 97.5%, which occurred in non-damaged ones (Fig. 1a) [10]. Thus, this model should be used to identify and classify non-damaged fruits into the finished group to obtain the highest efficiency.

2.2 Design of the Flipping Mechanism The working principle diagram of the proposed flipping mechanism is shown in Fig. 3. This structure consists of a timing belt powered by a stepper motor, and a flipping roller. The outer surface of the roller was in contact and driven by friction with the belt’s surface. Grooves were designed on the rollers with appropriate size in order to position the fruits. Fruits were fed into the grooves from the top of the roller, and rotated to the first camera’s position, where its upper part would be captured and identified. They were then rotated to the attachment position where the belt’s surface came in contact with the roller surface. Here, fruits were kept in the groove and rotated by the rollers to flip the direction gradually. When reaching the detachment position, the downside of the chili’s fruit was turned and flipped upside. Here, the belt was separated gradually from the grooves, and the fruits were also released from the grooves under the effect of gravity and moved along the belt to the second camera position. Thus, the downside damages were flipped to the upside, and the identification algorithm could identify the entire fruit. To ensure the smooth release of the fruits at the detachment position, the grooves were designed in the form of an involute curve. A continuous belt tension mechanism with springs was used to maintain the stability of the system. The grooves were designed to position one fruit. The height of the groove must be enough to limit the

188

Q.-K. Huynh et al.

Fig. 3 Working principle of the flipping mechanism

deformation and avoid damage. Additionally, the width should be enough to allow the free release at the detachment position. The average diameter of the popular chili species grown in the Mekong Delta was measured as 8.8 mm. In regards to the curvature of the fruit’s body, the groove’s size was chosen with a height of 10 mm, a top width of 22 mm, and a bottom width of 8 mm. Its dimensions are presented in Fig. 4. The diameter of the flipping roller was calculated as in (1): d=

n × P × (1 − ε) π

where d: is the outer diameter of the flipping roller (mm); n = 20: is the number of grooves (positive integer); Fig. 4 The dimensions of the grooves on the flipping roller

(1)

Design of Chili Fruit Flipping Mechanism …

189

P = 25 mm: is the desired pitch; ε = 0.02: is the friction sliding coefficient of belt transmission. The relationship between the translational velocity of belt and the productivity was defined in (2): V =

N ×P 3600

(2)

where V: is the translational velocity of the conveyor belt (mm/s); N: is the productivity (fruit/h). The desired pitch must be larger than the top width of the groove and must be a multiple of the timing belt’s pitch. The ε-coefficient was added to compensate for friction sliding. It ensured that the flipping roller rotates exactly one groove while the timing belt runs 25 mm.

3 Fabricating and Testing the Mechanism 3.1 Manufacturing and Fabricating the Model The flipping roller was constructed with PLA plastic and manufactured through the 3D printing method. Its surface was treated to reduce the reflection and the roughness. A general view of the model was fabricated and presented in Fig. 5, and the top view is shown in Fig. 6. A 5-Nm stepper motor drives the main power shaft. The timing belt has a pitch of 5 mm and a width of 80 mm.

3.2 Testing the Flipping Ability The model was run in a continuous mode to evaluate the flipping ability. Twelve samples, each containing 200 fruits, were utilized in this test. Out of these 200 fruits, 100 were diseased, and the remaining 100 were healthy. The fruit’s body length was chosen in the range of 53 ± 3 mm, and the max diameter was 9 ± 1 mm. Every sample was run 5 times at a specific velocity shown in Table 1. Fruits were put on a horizontal flat surface in a natural balanced state as in Fig. 7a. The upper part was marked by a permanent marker. This side was called the “upside”, and the other side was the “downside”. The model was started and the timing belt ran at translational velocity V (mm/s), respectively, with the angular velocity ω (rad/s) as in (3): ω=

2×V d

(3)

190

Q.-K. Huynh et al.

Fig. 5 General view of the flipping mechanism

Fig. 6 Top view of the flipping mechanism

At the input position, each fruit was placed manually into the groove so that the marker will be upwards as shown in Fig. 7b. The fruit was moved to the first camera position for capturing the image, then to the attachment position. It was kept inside the groove and gradually flipped while moving to the detachment position. This transposition of fruit is shown in Fig. 8. The

Design of Chili Fruit Flipping Mechanism …

191

Table 1 Test results Sample

Productivity (fruit/h)

V (mm/s)

Total times

Flipped

Half-flipped

Non-flipped

Success rate (%)

1

3600

25

1000

957

35

8

99.2

2

4800

33.33

1000

969

24

7

99.3

3

6000

41.67

1000

970

21

9

99.1

4

7200

50

1000

971

23

6

99.4

5

8400

58.33

1000

969

23

9

99.2

6

9600

66.67

1000

965

29

6

99.4

7

10800

75

1000

969

24

7

8

12000

83.33

1000

971

19

10

99

9

13200

91.67

1000

946

25

29

97.1

10

14400

100

1000

933

32

35

96.5

11

15600

108.33

1000

961

26

13

98.7

12

16800

116.67

1000

972

18

10

99

99.3

Fig. 7 Difference positions of fruits. a Free. b Input. c Upside. d Flipped. e Half-flipped. f Nonflipped 10 mm

10 mm

10 mm

Fig. 8 The transposition of fruit. a At the attachment. b On moving. c Approaching the detachment

192

Q.-K. Huynh et al.

20 mm

20 mm

20 mm

Fig. 9 Fruit destroyed by the mechanism. a Right after running. b After 24 h. c After 48 h

translational velocity of the timing belt was set in the range of 25 ÷ 116.67 mm/s, respectively, with the productivity from 3600 to 16800 fruit/h. The testing results are described in Table 1.

3.3 Evaluation of the Fruit’s Destroying Caused by the Mechanism There were 20 samples in this test, each containing 10 fruits, which were chosen as in Sect. 3.2. Every sample was fed into the mechanism, respectively, with the flipping time ranging between 1 and 20. After passing the flipping process, they were stored in a cool, well-ventilated house at a certain period of time to ensure a full assessment of possible damages. The evaluation was performed at four points of time: right before running, right after running, 24 and 48 h after running. In detail, the sample number 7 was run seven times, as shown in Fig. 9.

4 Discussion Figure 10 presents the average flipping rate at various velocities. The successful flipping rate is an average of 96.3%, which decreases a little bit to 93.3% at a velocity of 100 mm/s. At the same time, the half-flipped rate is low, around 2.5%. The nonflipped rate changes a little bit, around 1.2%. However, it is doubled to 2.5% at 91.67 mm/s, and almost a triple at 3.2% at 100 mm/s. In this range of velocity, the timing belt was detected to vibrate strongly. This would rotate the fruits after separating from the detachment position. At faster speeds, the incorrect rates come back to the normal average value. The reasons were supposed by the resonance of the system vibration. Its natural frequency would have coincided with the external oscillation frequency caused by the belt and running motor. A better method of stepper motor control may solve this problem. Based on the results in previous work [10], the flipped and half-flipped fruits were both available to identify. Thus, the equivalent success rate would be the total

Design of Chili Fruit Flipping Mechanism …

1.8

2.6

1.9 97.1

3.2

2.4 96.9

2.5

2.9

2.3 97.1

96.5

2.1 97.0

2.3

2.4 96.9

3.5

Non-flipped

95%

97.2

96.1

93.3

94.6

90% 95.7

Success rate

Half-flipped

96.9

Flipped

100%

193

85%

Running velocity (mm/s) Fig. 10 Success rate of the flipping mechanism

of those two cases. In other words, up to 98.2% flipping times can be used to identify the diseased fruit. The non-flipped cases were found on relatively small fruits. They were not positioned well in grooves, thus moved freely. The smaller groove led to more deformation in larger fruits. A flipping roller with a soft elastic rubber layer on the outside would be a better adaptive solution for this problem. The implementation of the identification algorithm into this flipping model will be carried out in the future to evaluate the recognition ability in a real system. The current work focused on analyzing the flipping characteristics, efficiency, and productivity of the mechanism. This helps to clarify the causes of error in further analysis. The investigated velocity range was chosen based on the estimation of the computer system’s recognition speed and the response of the actuators. The running velocity did not have a significant influence in the range of equal or below 4.7 fruit/s (16800 fruit/h). The storage period allows small spoils to decompose, enlarge enough to be observed by visual inspection. Only four fruits were broken, accounting for 2%. They all had a relatively large curvature, thus not entirely positioning inside the groove, those were partly gripped between the top of the groove and the belt, and broken. Increasing groove size will reduce the breakage, but it will also reduce the successful flipping rate. Therefore, these suitable groove dimensions were designed to obtain the minimum total error. In fact, the fruit is only processed once, so that the destroying rate would be significantly lower. It is possible to coat a soft rubber layer on the surface of the rollers to minimize damages, this will also expand the ability of the mechanism over a wider range of fruit sizes. However, a detailed investigation would be carried out in the future.

194

Q.-K. Huynh et al.

5 Conclusion The flipping mechanism helping the identification process has been designed, fabricated and tested. The success flipping rate was obtained up to 96.3%. The destroy rate was significantly low in all test cases and was acceptable for the real system. This structure is effective to support the detection of damages caused by diseases on fresh chili fruits. The study on implementing the identification algorithm into a real model would be performed in the future to evaluate the performance of the entire grading system. Acknowledgements Quoc-Khanh Huynh was funded by Vingroup Joint Stock Company and supported by the Domestic Master/ PhD Scholarship Programme of Vingroup Innovation Foundation (VINIF), Vingroup Big Data Institute (VINBIGDATA), code VINIF.2019.TS.33 and VINIF.2020.TS.109. Thanks to Minh-Vuong Le, Van-Luan Khuu and Thanh-Hieu Do for their enthusiastic participant in this research.

References 1. Loc, V.-T.-T., et al.: Analysis of chili value chain in Dong Thap Province. Can Tho Univ. J. Sci. 38D(2015), 13 (2015) 2. Srivastava, A., et al.: Inheritance of fruit attributes in chilli pepper. Indian J. Hortic. 76(1), 86–93 (2019) 3. Subha, G., et al.: Medicinal properties of chilli pepper in human diet: an editorial. ARC J. Public Health Commun. Med. 2(1) (2017) 4. Shil, S., Mandal, J., Das, S.: Evaluation of postharvest quality of four local chilli (Capsicum frutescens) genotypes of Tripura under zero energy cool chamber 7, 3698–3702 (2018) 5. Tunde-Akintunde, T.Y.: Mathematical modeling of sun and solar drying of chilli pepper. Renew. Energy 36(8), 2139–2145 (2011) 6. Herbon, R.P., et al.: Designing a High Volume Chile De-stemming Machine, Pittsburgh, Pennsylvania. ASABE, St. Joseph, MI, 20 June–23 June 2010 7. Kodali, N.B.: System and method of de-stemming produce. United States Patent and Trademark Office, US9173431B2, Editor, USA (2015) 8. Hakami, A., Arif, M.: Automatic inspection of the external quality of the date fruit. Procedia Comput. Sci. 163, 70–77 (2019) 9. Bhargava, A., Bansal, A.: Fruits and vegetables quality evaluation using computer vision: a review. J. King Saud Univ. Comput. Inf. Sci. (2018) 10. Quoc-Khanh, H., et al.: Identification of the damages caused by diseases on fresh destemmed chili fruits. In: The 12th IEEE International Conference on Knowledge and Systems Engineering, Can Tho City, Viet Nam (2020)

Performance Analysis of Pentuple Micro-optical Asymmetric Ring Resonator S. Vamsi Krishna, Suman Ranjan, and Sanjoy Mandal

Abstract An optical filter is justified by its frequency response and determination of frequency response of a system is carried out by calculating its transfer function. The transfer function of an optical waveguide-based pentuple micro-asymmetric optical RR (PMAORR) is calculated in z-domain. Delay line signal dispensation approach is incorporated to formulate the transmission of the resonator. The proposed RR consists of four unsymmetrical micro-optical rings placed in another larger micro ring. The objective of this work is to meet the ever-expanding demand of enhanced free spectral range (FSR). The measured FSR is 738 THz and correspondingly the crosstalk is virtually reduced to zero. Some of its important characteristics, that are, group delay, dispersion, finesse and Q-factor are also presented in this article. Keywords Optical filter · Ring resonator · Frequency response

1 Introduction In the most recent 20 years, optical MRR has demonstrated its essentialness as described in [1–6]. Optical channels are fundamentally portrayed by their recurrence reaction where a variety of size and stage are estimated with the variety of recurrence. Consequently, it is critical to discover the suitable recurrence reaction of a RR so it tends to be utilized as an optical channel. Execution qualities of twofold RRs with rings of various sizes are recently revealed in [7]. Moslehi et al. built up the z-space correlation for optical components for its utilization in channels [8]. Mandal et al. built up the Z-area model to examine the recurrence reaction of TRR [9], QORR S. V. Krishna Department of EEE, UPES, 248007 Dehradun, India S. Ranjan (B) Department of EE, BIT Sindri, 828123 Dhanbad, Jharkhand, India e-mail: [email protected] S. Mandal Department of EE, IIT (ISM), 826004 Dhanbad, Jharkhand, India © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2021 S. Choudhury et al. (eds.), Intelligent Communication, Control and Devices, Advances in Intelligent Systems and Computing 1341, https://doi.org/10.1007/978-981-16-1510-8_21

195

196

S. V. Krishna et al.

[10] PORR [11], and it is discovered that there is a relating increment in the FSR and smothered cross talk up to 30 dB speaking to that the auxiliary pinnacles got decreased altogether with each expansion in the number of rings [12]. Two methods, for example, chain framework definition and defer line signal handling methods are accessible to process the conveyance of the RR. Postpone line signal handling strategy alongside signal stream diagram technique is executed for our situation to dissect the recurrence reaction range. The unit postponed length in z-area is spoken to Z−1 . Z-transform has the ability to handle the complex system problem, so this technique is applied for describing the signals. The z-transform, H(z), of a signal, h(g), is defined as follows [8]: ∞ 

H (z) =

h(g)z −g

(1)

g=−∞

The size of the ring is chosen such that the span of each ring is an integer multiple of the span of the delay unit [13, 14]: T =

Lu ng c

(2)

here, L u the unit delay span, c represents the velocity of beam and ng denotes the group refractive index and is expressed as n g = n e f f + f o (dn e f f /d f ) fo

(3)

where, neff represents the effective index and f o represents the frequency. The relationship between unit delay and free spectral range (FSR) is represented as: FSR =

c 1 = T L u ng

(4)

From Eq. (4), it is obvious that the trivial the delay period of the unit, the greater the obtained FSR, i.e. the extended FSR. There is a drawback that the maximum FSR is limited by a single ring’s bending radius, while the improved FSR is desirable as an optical filter for its use. Therefore, another technique that will not be constrained by the bending radius of the ring should be implemented to increase the FSR. Therefore, the theory method of Vernier is used to boost the FSR and can be represented as [12, 15]; F S Roverall = L ∗ F S R1 = M ∗ F S R2 = N ∗ F S R3 = O ∗ F S R4 = P ∗ F S R5 (5)

Performance Analysis of Pentuple Micro-optical Asymmetric Ring …

197

where, L, M, N, O and P are the co-prime numbers of the respective MRRs whereas FSR1 − FSR5 is the frequency response of individual rings in order of their increasing sizes. The overall frequency response can also be calculated using the formula [16]; F S R = |L − M|

(F S R1 ∗ F S R2 ) |F S R1 − F S R2 |

(6)

here, F S Ri =

c , where i = 1, 2, 3, 4, 5. 2π n Ri

(7)

2 Modeling of the Proposed PMAORR As soon as two optical waveguides are placed near enough to one another, their evanescent field overlaps with one another and both waveguides are thus coupled and known as the directional coupler is shown in Fig. 1, and the corresponding block is shown in Fig. 2. Here E1i , E2i are the coupler inputs whereas E1o , E2o represent the coupler output. 

E 1o E 2o





C −jS = −jS C



E 1i E 2i



here, C represents the coefficient of common port transmission Fig. 1 Schematic of directional coupling

Fig. 2 Schematic block representation

(8)

198

S. V. Krishna et al.

C=

√

1 − K;

(9)

here, k represents the power coupling ratio −jS = −j

√

K;

(10)

here, S is the coefficient of cross port transmission. Using Eqs. (9) and (10), it can be inferred as C 2 + S2 = 1

(11)

The matrix of the optical coupler, which refers to the directional coupler input and output in Eq. (8) can be modified with an amplitude transmission coefficient ‘q’ and can be represented as Eq. (12) 

E 1o E 2o





C −jS =q −jS C



E 1i E 2i

 (12)

3 Transfer Function-Mason’s Gain Rule As expressed before that for the use of RR, optical channel needs a steady recurrence reaction. To examine the recurrence reaction, its exchange work must be known. In this way, signal stream chart (SFG) strategy or Mason’s benefit rule is utilized to ascertain the necessary exchange capacity of the resonator. The general conveyance can be determined by [17]  Tf =

Tn n 

(13)

where the symbols have their usual meanings as per the Mason’s gain rule. =1−



La +



Lb −



Lc +



Ld −



Le . . .

(14)

Figure 3 shows the proposed PMAORR schematic model consisting of four MRRs with hilter kilter ring radii and is inserted in bigger MRR. The signal stream map of the proposed z-space model that joins all directional couplers is shown in Fig. 4 in which X(z) and Y(z) are the person info and yield of the system. The best possible transmission of the signals can be determined by equating  Eq. (15) i.e. T f  = 1. Equation (15) can be expressed in frequency domain using relation Z = exp( j2π νT ), where v is the optical recurrence. A decision is made

Performance Analysis of Pentuple Micro-optical Asymmetric Ring … Fig. 3 Proposed PMAORR with directional coupler

Fig. 4 The respective Z-transform

199

200

S. V. Krishna et al.

to accomplish the ideal transfer function by expecting K1 and K2 (K2 = K3 = K4 = K5 = K6 = K7 ). As the ideal transfer function can be determined regarding coupling coefficients, the relationship got somewhere in the range of K1 and K2 can be communicated as Eq. (16)

Tf

(15) √ (1 − 2k2 + k22 (1 − k2 (1 − k2 )) + 2k23 − k24 )2 k1 = 1 − (1 − k2 )(2 − 6k2 + 7k22 − k23 − 2k24 + 2k25 − k26 )2

(16)

Variety of crosstalk because of variety of K1 maintaining K2 steady is sketched in one bend whereas the subsequent bend speaks to the variety of crosstalk because of the variety of K2 keeping K1 consistent. The crossing point of the two bends in the sketch gives to the ideal coupling coefficient. The loss due to movement of signals inside the RR is specified by γ = exp(−αL) [9] where “L” is the all out edge of the ring and “α” is the normal ring misfortune per unit length.

4 Characteristic of MRR A.

Group Delay

The negative subsidiary of digression converse of point of move work concerning the precise recurrence is generally called as the gathering delay.

Performance Analysis of Pentuple Micro-optical Asymmetric Ring …

201

 

d −1 Im T f (z)  tan τg = − dω Re T f (z) B.

(17)

Scattering or Dispersion

The gradient of group delay w.r.t. the “ω” is known as dispersion. It tends to be numerically communicated as [18, 19] D= C.

(18)

Finesse F=

D.

d(τg ) dω

FSR FW H M

(19)

Q-Factor

Q-factor determines the serration of the resonance w.r.t. to f 0 [11]: Q=

f0 FW H M

(20)

5 Simulation Results and Discussion In MATLAB program, the proposed MRR frequency response simulation is done with the coupling coefficient K1 = 0.9298 and K2 = K3 = K4 = K5 = K6 = 0.450, which is derived from Fig. 5. For the silicon-on-insulator (SOI) waveguides (5), the Fig. 5 Coupling coefficient variation with crosstalk variation

0.05

Cross talk

0.04 0.03 0.02 0.01 0 0.9305 0.454

0.93 0.452 0.9295

Coupling Coefficient (K1)

0.45 0.929

0.448

Coupling Coefficient (K2)

202

S. V. Krishna et al.

value of index is 1.7 and the encircling tour RR loss is 0.12 dB/cm. The “q” coefficient of amplitude transmission is assumed to be 1. The ring perimeters are 22.227 μm, 28.680 μm, 40.869 μm, 68.832 μm and 100.380 μm in order to maximize their scale and their resonance correspondingly. L = 31, M = 40, N = 57, O = 96 and P = 140. are their resonant numbers. The plot for the frequency response is represented in Fig. 6 in which the obtained FSR is 738 THz. Without weakening the reaction, the crosstalk is practically brought to zero. The lack of resonance is also sustained at a very low value. The characteristic of the proposed PMAORR such as group delay, dispersion and FWHM is shown in Fig. 7–9 respectively. The plot clearly shows that the characteristics are oscillating around the resonant peaks. For the resonator, the group delay is given in Fig. 7, the resonant peaks and Fig. 8 display the dispersion. Figure 9 shows the zoom-in view of the transmission curve. Fig. 6 The simulated result frequency response having FSR of 738 THz

1 0.9

FSR = 738 THz

0.8

Transmittance

0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 0.4

0.6

0.8

1

1.2

1.4

1.6

1.8

2

Frequency (Hz)

Fig. 7 The respective group delay plot

2.2 x 10

15

0

Normalized Group delay

-500

-1000

-1500

-2000

-2500

-3000 0.4

0.6

0.8

1

1.2

1.4

1.6

Frequency (Hz)

1.8

2

2.2 15

x 10

Performance Analysis of Pentuple Micro-optical Asymmetric Ring … Fig. 8 The respective dispersion plot RR

4

x 10

203

-9

Normalized Dispersion

3 2 1 0 -1 -2 -3 -4 0.4

0.6

0.8

1

1.2

1.4

1.6

1.8

Frequency (Hz)

Fig. 9 The magnified view of Fig. 6 to determine FWHM

2

2.2 x 10

15

1 0.9 0.8

Transmittance

0.7 0.6 0.5 0.4

FWHM = 0.75 THz 0.3 0.2 0.1 0

5

5.02

5.04

5.06

5.08

5.1

5.12

5.14

Frequency (Hz)

5.16

5.18

5.2 14

x 10

Comparative assessment of FSR with previously proposed RR-based filters is represented in Table 1. Table 1 Comparative analysis on performance

Model

FSR

TRR [9]

200 GHz

QORR [10]

342.4 THz

QARR [18]

6.6 THz

Proposed PMARR

738.0 THz

204

S. V. Krishna et al.

6 Conclusion Performance analysis of pentuple RR where four little lopsided miniature ring is set into other miniature ring is dissected utilizing Z-space design utilizing signal stream chart in which defer line signal preparing procedure is actualized to figure the exchange capacity of the previously mentioned RR. The FSR acquired is 738.0 THz with crosstalk is practically disposed of. The artfulness got is 984. The proposed fourfold RR has a great arrangement, which can be modified the room use factor and extensively expands the FSR. More modest rings are obliged into a larger RR, which adds to expand the FSR according to Vernier standard and the whole space needed to create the design is limited inside the measurement of the bigger ring. The gathering deferral and scattering trademark are additionally discovered to be protruding around the resounding pinnacles.

References 1. Van, V., Ibrahim, T.A., Ritter, K., Absil, P.P., Jonshon, F.G., Grover, R., Goldhar, J., Ho, P.T.: All-optical nonlinear switching in GaAS-AlGaAs microRRs. Photonics Tech. Lett. IEEE 14(1), 74–76 (2002) 2. Oda, K., Takato, N., Toba, H.: A wide-FSR waveguide double-Rmg resonator for optical FDM transmission systems. J. Lightwave Tech. 9(6), 728–736 (1991) 3. Kominato, T., Ohmori, Y., Takato, N., Okazaki, H., Yasu, M.: RRs composed of GeO2 - doped sillica waveguides. J. Lightwave Tech. 12(10), 1781–1788 (1992) 4. Adar, R., Shani, Y., Henry, C.H., Kistler, R.C., Blonder, G.E., Olsson, N.A.: Measurement of very low-loss silica on silicon waveguides with a RR. Appl. Phys. Lett. 58, 444 (1991) 5. Little, B.E., Chu, S.T., Haus, H.A., Foresi, J.: MicroRR channel dropping filters. J. Lightwave Tech. 15(6), 998–1006 (1997) 6. Kalli, K., Jackson, D.A.: RR optical spectrum analyser with 20-KHz resolution. Opt. Lett. 17(15), 1090–1092 (1992) 7. Suzuki, S., Oda, K., Hibino, Y.: Integrated-optic double-RRs with a wide free spectral range of 100 GHz. J. Lightwave Tech. 13(8), 1766–1771 (1995) 8. Moslehi, B., Goodman, J., Tur, M., Sahw, H.: Fiber-optic lattice signal processing. Proc. IEEE 72(7), 909–930 (1984) 9. Mandal, S., Dasgupta, K., Ghosh, S.K., Basak, T.K.: A generalized apporach for modelling analysis of ring-resonator performance as optical filter, vol. 264, pp. 97–104. Elsevier (2006) 10. Dey, S., Mandal, S.: Modelling and analysis of quadruple optical RR performance as optical filter using vernier principle. In: Optics Communications, vol. 285, pp. 439-446. Elsevier (2012) 11. Dey, S.B., Mandal, S., Jana, N.N.: Enhancement of free spectral range using pentuple microresonator. Apld. Optics, 51(29) (2012) 12. Boeck, R., Shi, W., Chrostowski, L., F.Jaeger, N.A.: FSR-eliminated Vernier racetrack resonators using grating-assisted couplers. IEEE Photon. J. 5(5) (2013) 13. Dey, S.B., Mandal, S.: Wide free spectral range triple RR as optical filter. Optical Engg. SPIE 50 (2011) 14. Saeung, P., Yupapin, P.P.: Generalized analysis of multiple RR filters: modelling by using graphical approach. Optiks Elesevier 119, 465–472 (2008) 15. Troia, B., Passaro, V.N., Leonardis, F.D.: Investigation of wide-FSR SOI optical filters operating in C and L bands. Telfor J. 4(1), 37–42 (2012) 16. Rabus, D.G.: Integrated RRs: The Copendium, vol. XVI. Springer (2007)

Performance Analysis of Pentuple Micro-optical Asymmetric Ring …

205

17. Nagrath, I.J., Gopal, M.: Control System Engineering, vol. 5. New Age International Publishers (2007) 18. Addya, S., Dey, S.B.: Anti-symmetric quadruple RR based optical filter. Optik 126, 5865–5869 (2016) 19. Madsen, C.K., Zhao, J.H.: Optical filter design and analysis: a signal processing approach. Wiley (1999)

MPPT Control of Hydrokinetic Energy Conversion System Anuj Kumar Pandey, Irfan Ahmed, and Pooja Sharma

Abstract Hydropower is one of the best sources of power as flowing water is available in a sufficient amount in nature. The Hydrokinetic Energy Conversion System (HECS) uses kinetic energy of the running water to rotate the turbine blades and generate power. This energy conversion system has been designed, and the output power of this system is effectively controlled with the Perturbed and Observe technique of Maximum Power Point Tracking (MPPT) control. A Permanent Magnet Synchronous Machine (PMSM) is coupled with the turbine to generate electricity. The generator’s output power is then fed to a three-phase uncontrolled rectifier to convert it to DC from AC. A boost converter is utilized to control this DC power by controlling duty ratio. The system model has been designed and simulated. The validation of MPPT control strategy has been done with the help of simulated result of HECS. Keywords Hydrokinetic turbine · Boost converter · MPPT · PMSG

1 Introduction HECS uses kinetic energy of the water to run the turbine blades. As flowing water is available in abundant amount in nature in the forms of rivers streams, waterways and tidal currents [1]. As these resources are available in large number throughout the world, it can be one of the major renewable sources of energy to increase the current capacity of hydroelectric energy. As in wind energy, control of hydrokinetic energy conversion system is done using power electronics. An active bridge rectifier or a diode rectifier is utilized to convert the generated AC quantity into the DC quantity and a DC–DC converter A. K. Pandey (B) · I. Ahmed EED, NIT Durgapur, 713209 Durgapur, West Bengal, India e-mail: [email protected] P. Sharma IT, BIT Sindri, 828123 Dhanbad, Jharkhand, India © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2021 S. Choudhury et al. (eds.), Intelligent Communication, Control and Devices, Advances in Intelligent Systems and Computing 1341, https://doi.org/10.1007/978-981-16-1510-8_22

207

208

A. K. Pandey et al.

i.e. Boost converter is used to control the level of DC. In some applications where hydrokinetic energy conversion system is connected to grid, a three-phase inverter is used to connect to the grid. The use of MPPT technique in hydrokinetic turbine is quite similar to that of wind turbine [2]. In P & O technique of MPPT, perturbation is created in a control variable and output electrical power is observed. For wind turbine, the MPPT is achieved by changing the duty ratio of the DC–DC converter [3, 4]. In [5], the flow of water determines the rotational speed of the turbine thereby utilizing the Perturb and Observe MPPT technique in river-based hydrokinetic turbine. This paper depicts the model of three main elements of hydrokinetic energy conversion system (HECS), the PMSG and diode rectifier with boost converter. Kinetic energy of flowing water runs the hydrokinetic turbine blades. Generated AC power is converted to DC using diode bridge rectifier and this power is then given to the boost converter whose output is fed to battery or to the grid. The duty ratio of boost converter is used as control variable as perturbation in duty ratio of boost converter is created and the output electrical power is observed. This control scheme is mainly based on the P&O technique of the MPPT, where maximum power point will be obtained while changing the duty cycle.

2 Configuration of HECS The HECS [6] is shown in Fig. 1. Kinetic energy of flowing water rotates turbine blades. The PMSG and converter circuit consist of a synchronous generator to convert mechanical energy to the electrical form. Then a uncontrolled converter (AC/DC rectifier) changes from AC to DC, a DC–DC converter for control purpose and to store the energy battery banks are used. In this paper, we are using MATLAB/SIMULINK for dynamic modeling of the hydrokinetic energy conversion system. In stand-alone systems, battery banks are preferably used and in case of grid-connected system, a voltage source inverter is used to feed the generated voltage to the grid [7]. Fig. 1 Stand-alone HECS

MPPT Control of Hydrokinetic Energy Conversion System

209

3 Modeling of the HECS We now discuss the hydrokinetic energy conversion system, which consists of hydrokinetic turbine modeling and PMSG and power converter circuit model below.

3.1 Hydrokinetic Turbine Model Based on steady-state characteristics and power relations of hydrokinetic turbine, the MATLAB model is developed. For hydrokinetic turbine, [6] output power is given as Pm = 0.5ρπr 2 v 3 C p (λ, β)

(1)

where Pm = the turbine output power, ρ = water density, r = blade radius (in meters), v = water velocity (in m/s), C p = hydrokinetic turbine’s power coefficient, λ = tip speed ratio (TSR) and β = pitch angle. Therefore, if ρ, r and v are constant, [6] then Pm is proportional to C p , which is controlled by the value of TSR λ and pitch angle β. Pitch angle is considered unchanged in this paper. [8] A simplified relation between C p and λ when pitch angle is constant (β = 12°) is given as    λ−8  2 −1 16 0.57λ2 20 λ λ + 1.32 + 0.667 Cp = − 27 B C(λ + 0.5B)

(2)

where B is number of blades = 3, and C is ratio of C l to C d , which is 30. Figure 2 shows MATLAB model of hydrokinetic turbine based on hydrokinetic

Fig. 2 Hydrokinetic turbine MATLAB model

210

A. K. Pandey et al.

turbine analysis and power relations in Eqs. 1 and 2.

3.2 PMSG and Power Converter Circuit Model The schematic of the permanent magnet synchronous generator, three-phase rectifier, converter and the load is shown in Fig. 3. The PMSG is modeled as E (EMF), R (stator resistance) and L (stator inductance). R is neglected in this paper because it is usually small. As the generated power from pmsg is AC, a three-phase diode bridge rectifier is connected to convert this power into DC power. The control of this DC power is done using a boost converter. In variable speed operations in renewable energy applications, PMSG is preferred so it is adopted in this work. PMSG does not need an external supply, brushes and slip rings. [9]. Contrary to Induction machines, the power factor of the PMSG is higher and increased power density and these are suitable for low rotor speed having multipole structure without gearbox [10]. Figure 4 shows the MATLAB model of PMSG, three-phase diode rectifier, boost

Fig. 3 Circuit diagram of PMSG, converters and the load

Fig. 4 MATLAB model of PMSG, three-phase diode rectifier, boost converter and capacitive load

MPPT Control of Hydrokinetic Energy Conversion System

211

Fig. 5 Proposed control of Hydrokinetic turbine

converter and load. To implement the MPPT technique in the system, duty ratio of boost converter D is taken as variable. There will be only one duty ratio for each specific speed of the water to extract maximum power. So, to track utmost power point, tracking of D is done. As we are using a boost converter between rectifier and load, the equation for bus voltage is obtained as following 1 Vbus = . Vd 1− D

(3)

Substituting these values Pg can be expressed as [6] π(1 − D)Vbus Pg = √ 6Ln p ωm



 (Kωm )2 −

π(1 − D)Vbus √ 3 6

2 .

(4)

From Eq. 4, it is seen that controlling the duty ratio of the boost converter will finally control the power generated from the PMSG. Control Approach for MPPT. The main objective of this hydrokinetic energy conversion system is to maximize output power at different water velocity or different turbine speed. In Fig. 5, a boost converter is utilized to implement MPPT control in the hydrokinetic energy conversion system. In order to implement the MPPT control, small change in D of the boost converter is done and the output power is measured at respective duty cycles. As we have seen earlier (in Eq. 4) that for a fixed water speed, there will be only one duty cycle at which maximum power would be generated. In this wayfinding, the optimum duty cycle will help to track the maximum power point in system.

212

A. K. Pandey et al.

Fig. 6 Flowchart of the proposed MPPT control algorithm

4 Proposed MPPT Control Algorithm To extract maximum output (power) from system, a P&O technique of MPPT is shown as flow chart Fig. 6. Here, D = duty ratio and P = measured output power, and K is step change in D [5]. In P & O technique of MPPT, small change in D is done and output power is measured. This output is then compared with the previous output and if this power output is greater than previous then the duty cycle is updated. This process of updating duty cycle is repeated until duty cycle corresponding to maximum power output is found.

5 Results and Discussion The proposed MPPT technique is implemented on hydrokinetic energy conversion system and simulation of the system is done in MATLAB/Simulink R2016a. In Fig. 7, block diagram of whole system is shown, which consists of hydrokinetic turbine, which is coupled with PMSG and output power of PMSG is fed to the boost converter whose duty cycle is controlled for extracting the optimal operating power from the system. The simulation results of the whole system implemented with control strategy and different parameters of the system are given here. The control strategy is validated through simulation in MATLAB/Simulink.

MPPT Control of Hydrokinetic Energy Conversion System

213

Fig. 7 Block diagram of hydrokinetic system with MPPT

Fig. 8 Simulated result of Cp

Fig. 9 Generated voltage of PMSG

The simulation results for PMSG, Cp, turbine speed, torque, D, output current, voltage and power are shown in Figs. 8, 9, 10, 11, 12 and 13. The results of different parameters power coefficient, water speed, turbine speed and output power without MPPT are given in Table 1. In further analysis, all these parameters with MPPT control are shown in Table 2. The output power using MPPT is displayed in Fig. 14.

6 Discussion Comparing both the results from results of Tables 1 and 2, it is established that output power while implementing MPPT is much greater than when the system was without MPPT control. Also, the power coefficient in the system while implementing MPPT

214

Fig. 10 Current, voltage and power

Fig. 11 Speed, stator current and electromagnetic torque of PMSG

Fig. 12 Turbine speed

Fig. 13 Output electrical power

A. K. Pandey et al.

MPPT Control of Hydrokinetic Energy Conversion System Table 1 Steady state parameters of hydrokinetic system

Table 2 Values of the parameters of the hydrokinetic system with MPPT

215

Sl No.

Parameter

Value

1

Output power

244 W

2

Power coefficient

0.36

3

Water speed

2 m/sec

4

Turbine rotational speed

86 rad/sec

Sl No.

Parameter

Value

1

Maximum output Power

410 W

2

Power coefficient

0.38

3

Water speed

2 m/sec

4

Turbine rotational speed

95.62 rad/sec

Fig. 14 Simulation of the output power with MPPT control

in the system is improved. The output power without MPPT was 244 W and after implementing MPPT, the power comes out to be 410 W.

7 Conclusion Hydrokinetic system is simulated and the results have been presented and validated using MATLAB/Simulink. It is shown that the output power was improved when P&O technique of the MPPT was implemented on the system. The values have been verified and presented in this work. For every rotational speed of turbine, there is only one duty cycle at which output power is maximum, also, for tracking, maximum power duty ratio D is tracked. The control of duty cycle of boost converter for maximum power is equivalent to implementing MPPT in the system.

216

A. K. Pandey et al.

References 1. Khan, M.J., Bhuyan, G., Iqbal, M.T., et al.: Hydrokinetic energy conversion systems and assessment of horizontal and vertical axis turbines for river and tidal applications: a technology status review. Appl. Energy 86(10), 1823–1835 (2009) 2. Khan, M.J., Iqbal, M.T., Quiacoe, J.E.: River current energy conversion systems: progress, prospects and challenges. Renew. Sust. Energy Rev. 12(8), 2177–2193 (2008) 3. Koutroulis, E., Kalaitzakis, K.: Design of a maximum power tracking system for wind-energyconversion applications. IEEE Trans. Ind. Electron 53(2), 486–494 (2006) 4. Putri, R.I., Pujiantara, M., Priyadi, A., et al.: Maximum power extraction improvement using sensorless controller based on adaptive perturb and observe algorithm for PMSG wind turbine application. IET Electr. Power Appl. 12(4), 455–462 (2018) 5. Hauck, M., Rumeau, A., Bratcu, A.I., et al.: Identification and control of a river-current-turbine generator—application to a full-scale prototype. IEEE Trans. Sustain. Energy 9(3), 1365–1374 (2018) 6. Zhou, H.: Maximum power point tracking control of hydrokinetic turbine and low-speed highthrust permanent magnet generator design. Missouri University of Science and Technology Rolla (2012) 7. Murthy, S.S., Ahuja, R.K.: A novel solid state voltage controller of three phase self excited induction generator for decentralized power generation. In: International Conference on Power Control and Embedded Systems (2010) 8. Manwell, J.F., McGowan, J.G., Rogers, A.L.: Wind Energy Explained. Wiley (2002) 9. Ashourianjozdani, M.H., Lopes, L.A.C., Pillay, P.: Power control strategy for fixed-pitch PMSG- based hydrokinetic turbine. In: Proceedings of 2016 IEEE International Conference on Power Electronics, Drives and Energy Systems (PEDES), Trivandrum, India, December, pp. 1–6 (2016) 10. Melfi, M.J., Evon, S., McElveen, R.: Induction versus permanent magnet motors. IEEE Ind. Appl. Mag. 15(6), 28–35 (2009)

Using Adaptive Reduced Power Subframes to Mitigate Inter-Cell Nosiness in Heterogeneous Networks Arun Kumar Singh , Vikas Pandey, and Ranjan Mishra

Abstract With the remarkable impact and fast growth of the mobile networks, the mobile base stations have been increased too, especially in the high population areas. These base stations will be overloaded by users, for that reason the small cells (like Picocells) were introduced. However, the inter-cell nosiness will be high in this type of Heterogeneous networks. There are many solutions to mitigate this nosiness like the Inter-Cell Nosiness Coordination (ICNC), and then the further enhanced ICNC where the almost blank subframes are used to give priority to the (victim users). But it could be a waste of bandwidth due to the unused subframes. For that reason, in this paper, we proposed an adaptive reduced power subframe that reduces its power rate according to the user’s SINR in order to get a better throughput and to mitigate the inter-cell nosiness. When the user is far from the cell, the case will be considered as an edge user and will get a higher priority to be served first. The results show that the throughput of all users in the Macrocells and Picocell will be improved when applying the proposed scheme. Keywords Inter-cell nosiness · Almost blank subframes · Heterogeneous network

1 Introduction The Heterogeneous network in the Long-Term Evolution (LTE) and LTE- Advanced (LTE-A) is considered as one of the solutions that were introduced by the Third Generation Partnership Project (3GPP) to increase the network coverage and increase its A. K. Singh Saudi Electronic University, Riyadh, Saudi Arabia e-mail: [email protected]; [email protected] V. Pandey Babu Banarasi Das University, Lucknow, India e-mail: [email protected] R. Mishra (B) University of Petroleum and Energy Studies, Dehradun, India e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2021 S. Choudhury et al. (eds.), Intelligent Communication, Control and Devices, Advances in Intelligent Systems and Computing 1341, https://doi.org/10.1007/978-981-16-1510-8_23

217

218

A. K. Singh et al.

capacity. However, because of the co-channel and multi-cells in such networks, the Inter-Cell Nosiness (ICN) can be considered as an issue. Generally, the user and the cell throughput of the whole network will be affected negatively, and this negative impact will affect the edge user mostly. The main component of the Heterogeneous network can be listed as Macrocell, Picocell, Femtocell, and the User Equipment (UE) which can also be classified according to the serving cell, Picocell user (PUE), Macrocell user (MUE), Femtocell user (FUE). Moreover, the users can also be classified according to their location within the cell, so if the user is in the center of the cell coverage, the case will be considered as a normal user, but if the user is at the edge of the cell coverage, then the case will be considered as an edge user [1]. Figure 1 shows the main components of the Heterogeneous network. As can be seen in Fig. 1, the main Macrocell will add a nosiness to the Low-Power Cell users (LPCs), the green line is showing the serving cells’ connection while the red line showing the nosiness from the Macrocell. Usually, the UE will select the cell with a maximum Reference Signal Received Power (RSRP) among the surrounding cells. However, in some cases and because of the differences in the power levels in the Heterogeneous network, the UE may select the Macrocell as a serving cell because it has the highest power, and this leaves the small cells unused. This may cause nosiness to the other small cells in the network [2]. To solve this problem, the Cell Range Expansion (CRE) was introduced in order to get a balance to the network. The concept of the CRE is adding the Cell-Specific Offset (CSO) to the Reference Signal Received Power of the small cells to get a larger coverage area and to serve more users. But the cell expanded areas will still

Fig. 1 Components of Heterogeneous networks. (https://www.researchgate.net/figure/An-illust ration-of-heterogeneous-networks-with-untrusted-SBSs-The-macro-BSs-are-overlaid_fig6_3185 76543)

Using Adaptive Reduced Power Subframes …

219

have nosiness due to Macrocell high power. Figure 2 shows the CRE region (the gray area) [3]. From release 10, the Third Generation Partnership Project (3GPP) introduced the Time-Domain Inter-Cell Nosiness Coordination (TD-ICNC) schemes. In this proposal, the nosiness caused by the multi-channels will be mitigated and the power of the Macrocell will be lowered in certain subframes to let the other small cells use these subframes [2]. When using the TD-ICNC method, the transmission power of some subframes of the Macrocell within the resource blocks will be reduced to a minimum in the CRE, to reduce the nosiness to the victim UEs, and this type of subframe is called Almost Blank Subframes (ABS) [4–8]. Figure 3 shows these types of subframes. The Scheme works by muting the Macrocell in certain subframes namely (ABS) and gives the priority to other cells to use a protected subframe to avoid the nosiness that may be caused by the Macrocells to the other cells. Here the N is the normal

Fig. 2 Cell range expansion. (https://www.semanticscholar.org/paper/Interference-Managementwith-Cell-Selection-Using-Moon-Kim/5c64d6fd5710e4f3ead3c6b507e43b609af6493d)

Macrocell Macroc ell Users

AB S

Macroc ell Users

AB S

Macroc ell Users

AB S

Macroc ell Users

AB S

Macroc ell Users

AB S

Picocel l Users

PS F

Picocel l Users

PS F

Picocell Picocel l Users

PS F

Picocel l Users

PS F

Picocel l Users

PS F

Fig. 3 Subframe example showing the (ABS) in Heterogeneous network

220

A. K. Singh et al.

subframe, ABS is the Almost blank subframe in which the Macrocell gets muted or uses a very low power, and the PSF is the protected subframe. These ABS subframes are usually used only for common signals (control, page signal, broadcast signals) by the Macrocells [9, 10], and using these subframes along with CRE will improve the throughput of the cell-edge users as long as the Macrocell is silent or reduced power [7, 11]. In this paper, we proposed an adaptive ICNC that adaptively changes the percentage of power in the (new version) of the almost blank subframes based on the priority of the UE in the queue according to its signal to nosiness plus noise ratio and its distance from the serving cell. The rest of the paper is organized as follows: section II will show the related works, section III will show the proposed method, section IV will show the results, and finally, the conclusions will be introduced in section V.

2 Related Work in Icnc The idea of using CRE has attracted researchers to improve the Heterogeneous network. Many researchers change the CRE region depending on SINR like in [7, 12] and other researchers suggested to extend the region based on the analysis and update of the CSO as proposed [13]. Then the TD-ICNC was introduced by 3GPP to reduce the nosiness from multi-cells toward Picocells and Femtocells in the Heterogeneous networks by using the ABS [14], in which the Macrocell is abstained from transmission in this period of subframes because these subframes are protected to be used by cell-edge users who are using the same frequency band [15, 16]. To control and coordinate these subframes, a scheduler has to be used like the proposal that is introduced by [17]. The researchers used a decentralized resource management framework by using the dynamic FFR to coordinate the inter-cell nosiness between the UEs and give a priority to the edge users. The throughput of the low-power cells generally and the cell-edge specifically will be enhanced when using the ABS scheme as can be seen in [9, 15, 16]. TD-ICNC scheme has been chosen by [7] to be used along with the extended version of CRE, which divided the CRE region into inner and outer areas. The whole bandwidth has been shared with the Macrocells and the low-power cells (Picocells), and the author used the Reduced Power Subframe (RPS) along with the Almost Blank Subframe (ABS). This will keep an optimal performance for the Picocells while keeping the Heterogeneous network performance balanced. In release 11 LTE, the 3GPP proposed the new scheme which is the further enhanced ICNC [18, 19]. The scheme will reduce the power for some subframes for a specific node instead of blank it totally, such subframes called Low-Power ABS (LP-ABS) which will give a better efficiency to the network. The scheme has been used by [19], and it gives a higher priority to the UE in the center of the Macrocells. These schemes allow to decrease the power to 30 dB as mentioned in [20, 21] to give a better performance to the edge users in the Macrocells.

Using Adaptive Reduced Power Subframes …

221

The value of power level is not constant and can be variable which depends on many factors, like ABS power ratio, the value of the extended region in CRE, and the model of the channel. However, the value of 30dBm has always been used by the researcher even though it is not always the optimum value [21–23]. Because of the fact that the network performance can be changed according to the power level in the LP-ABS and the extended area of the CRE, the authors of [25] proposed an adaptive ICNC based on FFR. Also, the authors of [24] introduced a method where the small cells can organize themselves in the network to reduce the nosiness by suggesting a Dynamic FFR (DFFR) method. The authors of [2] also suggest a mechanism that changes the power of ABS dynamically to get a better Quality of Service (QoS). The authors used the traffic mode that supports multimedia communications only. Different Cell-Selection Offsets (CSOs) have been suggested by [25] to change the coverage of the CRE area. The CSO is changed adaptively in order to improve the performance of the Heterogeneous network. The authors in [26] suggested a new novel idea that changes the CRE area and the ABS dynamically based on the cell traffics of each cell in the Heterogeneous network [26]. From the above literature, it is obvious that CRE and the change of the power rate in the ABS and LP-ABS can increase the throughput of the cells, the users in the cells, and the edge users as well. From that, we suggest an adaptive power ABS that can be changed according to the distance between the user and the cell which will change the value of the SINR and give a higher priority to the edge user in the scheduler.

3 Proposed Work The user throughput calculation is based mainly on many factors, and one of these factors is the SINR, which is used to calculate the signal to nosiness ratio and take the noise into account [26]. Also, the distance plays a role in the SINR value that when the distance between the UE and serving cell is increased, the nosiness from other cells will be increase too. The proposed algorithm uses TD-ICNC with the adaptive rate of LP-ABS power. This power can be changed according to the distance from the serving cells in order to increase the throughput of the edge users in the Picocells. Figure 4 shows the proposed algorithm, where N is a normal subframe, ABS is the almost blank frame, and RP is the reduced power subframe where our proposed scheme will be changing the power adaptively. In order to control and manage the resource block effectively, the scheduler has to give different priorities for the users depending on some factors. The (victim users) who are the users that are located at the edge of the cell will have a higher priority, and the scheduler will give the priority to the user with less SINR to use the subframe of the Picocells during the ABS and RP subframes of the Macrocells, after that other

222

A. K. Singh et al.

Macrocell N

ABS

RP

RP

RP

N

N

N

N

N

N

N

N

N

N

Picocell N

N

N

N

N

Fig. 4 Proposed scheme subframes

users can connect and use the desired subframes. In this paper, we simulate the LTE-A HetGen network with a central eNodeB that has three Macrocells and two Picocells for each Macrocell which gives a total of 6 Picocells. As can be seen in Fig. 5, the X2 interface is used to connect these cells coordinating the nosiness and to share the resources between them. Different number of users are taken in different scenarios starting from 100, 200, 300, and 400, and these UEs are distributed randomly in the network and set to attach the cell with the higher Reference Signal Received Power. We considered the nosiness that came from the neighboring Macrocells. Thena bias of 3 dB is added for the CRE each time to check the network behavior and the edge user throughput along with the cell throughput. In this paper, we used a system-level simulation using a full buffer for traffic mode. Different values of CRE have been taken in the results with different power ratios for the reduced power subframes. When reducing the power of the RP subframes in the Macrocell, the throughput for the edge users and average throughput for the whole UE in the Picocell can be improved. The reason for this is that there will be more Fig. 5 Heterogeneous network scenario

XX

Using Adaptive Reduced Power Subframes …

223

Throughput (Mbps)

60 50 40 30 20 10 0 User 30

User 50 Without

End User With

Fig. 6 User throughput with different distances

power for the Pico UE to use the subframes and also will be given a higher priority in the scheduler. For that, we proposed a method to change the power of the users adaptively depending on the SINR, in which the power rate of the reduced power subframe can change accordingly. We can see the difference in the throughput for the UEs with and without the proposed scheme and it is clear that the throughput is improved when using the proposed scheme especially for the edge users. The Macrocells are considered as the coordinator for many Picocells and can share the resources between them. It can be clearly seen that the throughput of all users in the Macrocells and the edge users has been improved. The edge users get better throughput because the scheduler has set a priority for them. Finally, we take a different scenario to check the same user throughput with different distances from the cell, with and without using our proposed method, and the results can be seen in Fig. 6. As can be seen and because of the scheduler and the proposed scheme, the user that has a higher distance from the serving cell will get a higher priority. For that reason, we can see a better performance for the users on the edge.

4 Conclusions The enhanced ICNC can be considered as a good solution to solve the problem of ICN but it could have unused subframes in some cases. So, in this paper, we proposed an adaptive method that uses reduced power subframes which can change their power adaptively according to the user distance and consequently change the SINR. The edge users have the most benefit from this scheme as the scheduler focuses on those

224

A. K. Singh et al.

users. The results show that the average throughput of the users and edge users for both Macrocells and Picocells has improved and the throughput for the whole network as well.

References 1. Wannstrom, J., Mallinson, K.: Heterogeneous networks in LTE. 3GPP (2014) 2. Wang, Y.-C., Chen, S.-T.: Adaptive configuration of time-domain eICIC to support multimedia communications in LTE-A heterogeneous networks. In: 2017 IEEE 18th International Symposium on A World of Wireless, Mobile and Multimedia Networks (WoWMoM), pp. 1–6 (2017) 3. 3GPP-TS22.220: Service requirements for Home Node B and Home eNode B (Release 14), 3rd Generation Partnership Project-TSG Services and System Aspects (2017) 4. Lindbom, L., Love, R., Krishnamurthy, S., Yao, C., Miki, N., Chandrasekhar, V.: Enhanced inter-cell nosiness coordination for heterogeneous networks in LTE-advanced: a survey (2011). arXiv:1112.1344 5. Ye, Q., Rong, B., Chen, Y., Caramanis, C., Andrews, J.G.: Towards an optimal user association in heterogeneous cellular networks. In: 2012 IEEE Global Communications Conference (GLOBECOM), pp. 4143-4147 (2012) 6. Guvenc, I.: Capacity and fairness analysis of heterogeneous networks with range expansion and nosiness coordination. IEEE Commun. Lett. 15, 1084–1087 (2011) 7. Al-Aaloosi, A.B.A.: Inter-cell nosiness mitigation in LTE-advanced heterogeneous mobile networks. University of Salford (2017) 8. Samsung, G.T.R.W.: R1–111466 “Full Buffer Evaluation Results for CoMP Scenario 3 and 4” (2011) 9. Oh, J., Han, Y.: Cell selection for range expansion with almost blank subframe in heterogeneous networks. In: 2012 IEEE 23rd International Symposium on Personal, Indoor and Mobile Radio Communications-(PIMRC), pp. 653–657 (2012) 10. Akyildiz, I.F., Gutierrez-Estevez, D.M., Balakrishnan, R., Chavarria-Reyes, E.: LTE-advanced and the evolution to beyond 4G (B4G) systems. Phys. Commun. 10, 31–60 (2014) 11. Daeinabi, A., Sandrasegaran, K., Zhu, X.: Performance evaluation of cell selection techniques for picocells in LTE-advanced networks. In: 2013 10th International Conference on Electrical Engineering/Electronics, Computer, Telecommunications and Information Technology, pp. 1–6 (2013) 12. Naganuma, N., Nakazawa, S., Suyama, S., Okumura, Y., Otsuka, H.: Adaptive control CRE technique for eICIC in HetNet. In: 2016 Eighth International Conference on Ubiquitous and Future Networks (ICUFN), pp. 4–6 (2016) 13. Gu, X., Deng, X., Li, Q., Zhang, L., Li, W.: Capacity analysis and optimization in heterogeneous network with adaptive cell range control. Int. J. Antennas Propag. 2014, 215803 (2014) 14. 3GPP-TSG-RAN: TS 36.300, “Overall description (Release 14) (2017) 15. Sabella, D., Rapone, D., Fodrini, M., Cavdar, C., Olsson, M., Frenger, P., et al.: Energy management in mobile networks towards 5G. In: Energy Management in Wireless Cellular and Ad-hoc Networks, pp. 397–427. Springer (2016) 16. Pedersen, K.I., Wang, Y., Strzyz, S., Frederiksen, F.: Enhanced inter-cell nosiness coordination in co-channel multi-layer LTE-advanced networks. IEEE Wirel. Commun. 20, 120–127 (2013) 17. Mendrzik, R., Castillo, R.A.J., Bauch, G., Seidel, E.: Nosiness coordination-based downlink scheduling for heterogeneous LTE-A networks. In: 2016 IEEE Wireless Communications and Networking Conference, pp. 1–6 (2016) 18. Soret, B., Wang, H., Pedersen, K.I., Rosa, C.: Multicell cooperation for LTE-advanced heterogeneous network scenarios. IEEE Wirel. Commun. 20, 27–34 (2013)

Using Adaptive Reduced Power Subframes …

225

19. Merwaday, A., Mukherjee, S., Güvenç, I.: HetNet capacity with reduced power subframes. In: 2014 IEEE Wireless Communications and Networking Conference (WCNC), pp. 1380–1385 (2014) 20. R1-120023, G.: TSG-RAN WG1, Analysis of feasibility and standard impact of reduced power ABS (2012) 21. 3GPP R1–113635: TSG-RAN WG1 “Performance evaluation of FeICIC with zero and reduced power ABS” (2011) 22. R1-121489, G.: Discussion on the features and signalling support for non-zero transmit power ABS. In: 3rd Genereation Partnership Project-TSG-RAN1 (2012) 23. Hu, H., Weng, J., Zhang, J.: Coverage performance analysis of FeICIC low-power subframes. IEEE Trans. Wirel. Commun. 15, 5603–5614 (2016) 24. Hu, H., Weng, J., Zhang, J., Zhang, J., Wang, Y.: Modelling and analysis of reduced power subframes in two-tier femto HetNets. In: 2016 IEEE 83rd Vehicular Technology Conference (VTC Spring), pp. 1–5 (2016) 25. Saito, T., Adachi, F.: De-centralized adaptive 2-step inter-cell nosiness coordination in distributed MIMO. In: 2018 24th Asia-Pacific Conference on Communications (APCC), pp. 224–228 (2018) 26. Hassan, R.A., Idris, A., Adto, H., Ramadhan, M., Kassim, M.: Reduction of Inter-Cell nosiness in close proximity cell using dynamic fractional frequency reuse method. In: 2017 IEEE Conference on Systems, Process and Control (ICSPC), pp. 157–161 (2017)

Solution of Economic Load Dispatch Using Flower Pollination Algorithm Praveen Kumar Mahto, Kumari Sarwagya, Suman Ranjan, and Upendra Prasad

Abstract To offer dependable and continuous electric power to customers, power utilities feature a lot of monetary and industrial troubles in function, preparation and control of electric systems. Almost all of the power system optimization problems are intricate and posses nonlinear characteristics with intense equality and inequality limitations. Economic Load Dispatch (ELD) is an central aspect in the power system for the proper functioning of the generating units and for the scheduling of a power system that settles on the generation’s planning. To solve this kind of highly nonlinear optimization problem, Flower Pollination Algorithm (FPA) Technique is widely used. FPA is an evolutionary algorithm that has found application in a broad array of optimization conditions in recent times as well as it is well equipped to deal with optimization of functions containing several variables and having nonlinear characteristics. The proposed optimization strategy can deal with economic load dispatch problems including subject to constraints. Here, flower pollination algorithm technique is implemented for 6 units and 13 units. In comparison with the several previous methods, the presented solution method has been proved to do well in all three test cases. Conceiving the superiority of the results obtained using the FPA technique, it can be proposed as a potential candidate for the solution of the ELD problem. Keywords ELD · Valve point · Flower pollination algorithm

1 Introduction In the current world scenario, most of the equipment is electrical energy-based equipment. If we see all these, then we find daily used-based appliances like electric bulb, fan, television, laptop, mobile, refrigerator, etc., and all these appliances play a very important role in daily life. Another example is the manufacturing industry; nowadays, all of these types of industry operate on electric energy. In the transportation industry, all metro most of the train and some of the buses and auto rickshaw operate P. K. Mahto · K. Sarwagya (B) · S. Ranjan · U. Prasad Department of EE, BIT Sindri, Dhanbad, Jharkhand, India e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2021 S. Choudhury et al. (eds.), Intelligent Communication, Control and Devices, Advances in Intelligent Systems and Computing 1341, https://doi.org/10.1007/978-981-16-1510-8_24

227

228

P. K. Mahto et al.

on electric energy. To supply this huge amount of power demand, different types of power generation plants are currently working. Combination of different fuels, different power plants as well as load allocation to different fuels and different characteristic power s become important for minimizing the total fuel cost of operation. In the minimization process, many of the physical limitations of generating plant are continuously taken care. In optimum operation of power system, ELD problem which takes care the units constraints is the most important optimization problem [1, 2]. ELD is a vital aspect of the power system for the proper functioning of the generating units and for the scheduling of a power system that settles on the generation’s planning. Factors which affect the ELD problem such as valve-point loading characteristic, etc. are given in [3–5]. Due to above-mentioned cause, ELD becomes nonlinear and non-convex problem. Different types of techniques which are classical mathematical programming that used to solve ELD problem are heuristic programming methods [6–8]. Solutions obtained from these methods are still not up to the mark and there is large scope to improve these solutions to the optimum value. The effectiveness of this solution is not always sure [9]. In recent years, many number of nature-inspired optimization techniques have been used to get optimal solution, and example of such type of techniques are “Genetic Algorithm (GA)” [10] and its related algorithms. This technique has been implemented effectively to get the solution for complicated optimization problems, but there are some deficiencies found in the performance of GA in recent research. The performance of GA suffers a major setback in terms of global optimal solution due to premature convergence of characteristic curve [10]. Several other optimization techniques such as particle swarm optimization [11–14], differential evolution [15] optimization techniques, Tabu search optimization technique [16] as well as other heuristic techniques [17] are defined in past to determine the economic load dispatch problem. Due to its effortlessness and usefulness, PSO has paying attention to many researchers’ prospects. The natural phenomena which inspired the PSO are the flock of bird and schooling of fish. These are stretchy vigorous, populationbased techniques which are adopted by many people for solving ELD problems and various power system problems. Though, it is sluggish to converge. Also, there are contradictions between the processes of exploration and exploitation. But these two factors must be well balanced to obtain good optimization result. But in the case of Flower Pollination Algorithm (FPA), it has switch probability which is the only key parameter that turns the algorithm into comfort implementation and faster to reach optimal solution [18]. In 2012, FPA was first developed by Yang. Inspiration for this technique is taken from the pollination process of flowering plants; FPA practicability and competence performance for solving dissimilar types of conventional and contemporary optimization problems successfully and proficiently. It is proven that by Yang that Flower Pollination Algorithm is effortless, very competent and can do better than Modified PSO (MPSO), Improve PSO (ISPSO), Krill Herd Algorithm IV (KHA-IV) and Grey Wolf Optimization technique (GWO). Its convergence rate is fundamentally exponential. In the present paper, we apply FLA to find the best solution for the ELD problem of the power system. Here, FLA technique proficiency is tested for different numbers

Solution of Economic Load Dispatch …

229

of bus systems considering standard problems and ELD problems, and the result we get are compared with another nature-inspired technique result mentioned in recently worked literature with the intent of discovering basic major feature of the proposed technique. Flower pollination algorithm technique is implemented for 6 units and 13 units systems and the results are compared with the previously published results of MPSO, ISPSO, Krill Herd Algorithm IV (KHA-IV) and Grey Wolf Optimization technique (GWO) and it is found that the proposed FLA technique is superior than that of above mentioned optimization techniques.

2 Statement of Crisis The idea of the ELD problem is nonlinear in nature subjected to certain sets of nonlinear constraints. A.

Cost Function

The objective function of ELD is a quadratic function consisting of certain coefficients and can be defined as [1] Fi (Pi ) = ai .Pi2 + bi .Pi + ci

(1)

where ai , bi and ci are the coefficients of the quadratic equation, Pi is the total output power and F (Pi ) is the objective function. When considering the Valve-Point loading (VPL) effect, Eq. (1) gets modified as shown in Eq. (2) [1]     Fi (Pi ) = ai .Pi2 + bi .Pi + ci + abs ei × sin f i × Pimin − Pi

(2)

where ei and f i take care of VPL effect; m is the total number of generators engaged. B.

Different Limits

Certain constraints must be taken take when solving the economic load dispatch problem which are as follows. (1)

Total Power Limits

The total power demand is given by [1] m 

P − PD − PL = 0

(3)

|i=1

where the demand in power is represented as PD and PL is the transmission losses occurred. [2]

230

P. K. Mahto et al.

PL =

N  N 

Pi Bi j P j +

i=1 j=1

(2)

N 

B0i Pi + B00

(4)

i=1

Generator Constraints

The active and reactive power capacity of the generators must be kept within the maximum and minimum limits of the generator to avoid heating and instability issues. The constraints for active and reactive power are given in Eqs. (5) and (6), respectively [3]. Pi min ≤ Pi ≤ Pi max

(5)

Q i min ≤ Q i ≤ Q i max

(6)

where Pimin and Pimax are the smallest and peak active power of the i-th unit and Qimin and Qimax are the smallest and peak reactive power outputs of the i-th unit [8, 9]. (3)

Transformer Constraints

The transformer constraints are also strictly kept with the limits and can be represented as shown in Eq. (7). Ti min ≤ Ti ≤ Ti max

(7)

where Timin and Timax are the least and maximum limit of the tap-setting of the ith unit of the tap-changing transformer, and the tap-setting lies between 0.9 and 1.1. C.

Flower Pollination Algorithm.

Xin-she Yang proposed an advanced population-based optimization methodology which is acknowledged as Flower Pollination Algorithm [18]. Flower pollination behaviour is mimicked by that technique. FPA is a native intellectual procedure of coupling in plants along with carrying pollen by pollinators for insects and so on. According to biological evolution, survival of the fittest is the goal of flower pollination and the optimal breeding of vegetation in terms of quantity as well as fitness. Those phenomena can be considered as an optimization procedure of flower vegetation species. Flower pollination can be developed as an algorithm using the following four rules: 1. 2. 3.

Global pollination procedure is expressed biotic and cross-pollination, and pollen-transmitting pollinators can blow across length (Rule 1). Local pollination is represented by abiotic and self-pollination (Rule 2). Flower commitment can be taken as an identity to a breeding probability which is proportionate to the sameness of the involved two flowers (Rule 3).

Solution of Economic Load Dispatch …

4.

231

Knob possibility P2 [0 1] controlled switching of local pollination and global pollination (Rule 4).

Above given rule changes into the following mathematical equation. A mathematical equation that combines “universal pollination stride”(rule1) and “flower commitment” (rule 3) is represented as follows:   xit+1 = xit + γ L(λ)(g∗ − xit )xit+1 = xit + γ L(λ) g∗ − xit

(8)

where xit represents “pollen i” or the “solution vector xi ” for “iteration t”, and the best solution for this process is g∗ . For controlling the step size, the scaling factor is γ . Strength of the pollination corresponds by “Lévy flights-based step size” that is L(λ). Lévy distribution drawn how insects can take off over an elongated distance with various distance steps and it is given by the following equation   L = λr (λ) sin(π λ/2)/π ∗ 1/S 1+λ (S  S0  0)

(9)

r (λ) = Gamma function Validation of this distribution is for large steps s > 0. The combination of rule 2 and rule 3 can be represented as the following mathematical equation.   xit+1 = xit + ∈ x tj − xkt

(10)

Here xtj and xk are the pollen from different flowers of the same plant species. If the same species or same population gives xtj and xtk and  drawn from a uniform distribution in [0, 1], then it is equivalently known as a local random walk. At all scales, that is both local and global, flower pollination activities can occur. Before the proposed algorithm parameters are able to be adjusted, many times they have been performed. Size of population N = 50 is sufficient for economic load dispatch problems. (N iter ) = 2000 is considered as the sufficient number of iteration. P = 0.8 is switching probability. In the proposed paper, we are exploring the proficiency and success rate of flower pollination algorithm when we are implementing it to economic load dispatch problems which are many times confront in power plant station.

3 Proposed Algorithm of FPA STEP 1: STEP 2:

With arbitrary solutions, we “initialize a population of size m” where m is the number of flowers/pollen gametes. From starting population, we find the most fittest value that is the best solution

232

P. K. Mahto et al.

STEP 3:

STEP 4:

STEP 5: STEP 6: STEP 7: STEP 8:

Now we define switching probability and stopping condition. Switching probability p ∈ [0, 1] and stopping condition may be accuracy or fix number of generation/iteration. In this application, we consider the stopping condition as the maximum number of iteration. Set initial iteration count as t = 1. For all flowers in the population from 1 to m, perform global pollination using (6) and (7) if rand < p; or else perform local pollination using (8) and ∈. With the help of switching probability, we find a new solution for all flowers. Population is updated by new solution if they are better than the previous one. Current best solution g∗ is found among the flowers in the population. If the convergence criteria are satisfied (t equals tmax ), output the best solution for ELD problem; otherwise set t = t + 1, and repeat the steps 4–8.

4 Result and Discussion For verification of practicability of the implemented method, two different number of bus systems are taken into consideration. These are (1) (2)

Test Case I: 6-generating unit test system without VPL effect and with VPL effect. Test Case II: 13-generating unit test system without VPL effect and with VPL effect.

The FPA methodology has been functional to two special test systems, e.g. 6generating unit and 10-generating unit. The MATLAB programs of software which are written in MATLAB language and personal computer are used for execution that has 4 GB RAM, 2.5 GHz core and i-5 processor. A.

Test system 1

The unit characteristic data is provided in Table 1. The demand is 1263 MW in case of a system having a valve-point effect. Consequently, the results are shown in Table 3 and compared with the other previously defined methods (Table 2). The result shown in Table 3 clearly shows that the whole cost of fuel for 6-bus test system is 15276 $/hr and 14127 $/hr for the system without VPL effect and with VPL effect, respectively, which is superior than the previously published results for the same test bus system. The convergence characteristic using the proposed flower pollination algorithm for the 6-bus test case is represented in Fig. 1 where the optimization curve is converging at a faster rate resulting in a quick solution to the given objective function.

Solution of Economic Load Dispatch …

233

Table 1 Cost Coefficients of 6-Generating Unit Test System Without VPL Unit

ai

1

756.7988

bi

ci

38.539

0.15247

Pmin i

Pmax i

10

125

2

451.3251

46.1591

0.10587

10

150

3

1243.5311

38.3055

0.03546

35

210

4

1049.9977

40.3965

0.02803

35

225

5

1356.6592

38.2704

0.01799

125

315

6

1658.5696

36.3278

0.02111

130

325

Table 2 Cost coefficients of 6-generating system including VPL ci

ei

fi

Pmin i

Pmax i

7.0

240

300

0.035

100

50

0.0095

10.0

200

200

0.042

50

200

0.0090

8.5

220

200

0.042

80

300

4

0.0090

11.0

200

150

0.063

50

150

5

0.0080

10.5

220

150

0.063

50

200

6

0.0075

12.0

190

150

0.063

50

120

Unit

ai

bi

1

0.0070

2 3

Table 3 Comparison of 6-unit result with and without VPL effects Unit

Without VPL

With VPL

FPA (without VPL effect) FPA

ISPSO [19] MPSO

PSO

P1

447.299

459.5793

477.2

447.1874

447.4970

P2

170.9975

199.6776

184.95

173.5060

173.3221

P3

263.6131

300

226.88

260.9553

263.0594

P4

125.5987

52.3153

107.00

144.0583

139.0594

P5

172.2983

150.4736

165.00

163.2156

165.4761

P6

83.1936

100.9554

PD

1263

1263

1263.00

1263.00

1263.00

14127

15421.332

15441.00

15450.00

Fuel cost ($/hr) 15276

B.

102.00

86.2934

87.1280

Test System 2

The load requirement is 1800 MW. The unit characteristic data are specified in Table 4. The best solutions obtained from the FPA method are given in Table 5 and compared with previous results. The convergence characteristic of the FPA technique for the case is revealed in Fig. 2. The result shown in Table 5 clearly shows that the total fuel cost for 13-bus test system is 17943 $/hr and 15148 $/hr for the system without VPL effect and with

234

P. K. Mahto et al.

Fig. 1 Convergence curve of 6-unit test system

Table 4 Cost coefficients of 13-generating unit test system without VPL Pmin i

Pmax i

0.035

0

680

0.042

0

360

200

0.042

0

360

240

150

0.063

60

180

240

150

0.063

60

180

7.74

240

150

0.063

60

180

0.00324

7.74

240

150

0.063

60

180

0.00324

7.74

240

150

0.063

60

180

9

0.00324

7.74

240

150

0.063

60

180

10

0.00284

8.60

126

100

0.084

40

120

11

0.00284

8.60

126

100

0.084

40

120

12

0.00284

8.60

126

100

0.084

55

120

13

0.00284

8.60

126

100

0.084

55

120

Unit

ai

bi

ci

ei

fi

1

0.00028

8.10

550

300

2

0.00056

8.10

309

200

3

0.00056

8.10

307

4

0.00324

7.74

5

0.00324

7.74

6

0.00324

7 8

VPL effect, respectively, which is superior than the previously published results for the same test bus system. The convergence characteristic using the proposed flower pollination algorithm for the 13-bus test case is represented in Fig. 2 where the optimization curve is converging at a faster rate resulting in a quick solution to the given objective function.

Solution of Economic Load Dispatch …

235

Table 5 Cost coefficients of 13-generating unit test system with VPL Unit power output

FPA (without VPL)

FPA (with VPL)

MPSO (without VPL)

GWO (with VPL)

P1

535.6293

446.2678

425.098

807.1247

P2

230.1029

74.5201

182.5087

144.869

P3

251.5992

295.8428

133.5717

297.9434

P4

114.4931

60.8642

162.445

60

P5

102.253

60

153.9582

60

P6

85.9225

61.2199

113.9438

60

P7

96.5198

160.6061

133.8305

60

P8

96.224

164.9611

104.7926

60

P9

92.0435

111.0364

85.6033

60.0362

P10

40.9291

112.5069

66.7367

40

P11

40.0017

115.8002

60.8971

40.0267

P12

57.0269

81.9585

77.3235

55

P13

57.303

55

99.2915

55

PD

1800

1800

1800

1800

Fuel cost

17943

15148

19372.22

18442.59

Fig. 2 Convergence curve of 13-unit test system

236

P. K. Mahto et al.

5 Conclusion The proposed FPA are effectively applied to solve ELD problems with generator limitations in the form of linear equality and inequality limitation. Here, the success rate of FPA was tested for 6 units and 13 units. Results are compared with the previously reported literature works. After studying all two cases, it is found that the flower pollination algorithm has the capability to provide optimum result and enhanced convergence curve and efficacy as compared to other methods such as ISPSO, MPSO GWO and KH_IV. Thus, the proposed FPA method shows potential for solving intricated issues in power systems. In comparison with the several previous methods, the presented solution method has been proved to do well in all two test cases. Conceiving the superiority of the results obtained using the FPA technique, it can be proposed as a potential candidate for the solution of the ELD problem.

References 1. Amjady, N., Nasiri-Rad, H.: Economic dispatch using an efficient real-coded genetic algorithm. IET Gener. Transm. Distrib. 3(3), 266–278 (2009) 2. Elsayed, W.T., Hegazy, Y.G., Bendary, F.M., El- Bages, M.S.A.: Review on accuracy issues related to solving the non-convex economic dispatch problem. Electr. Pow. Syst. Res. 31, 325–332 (2016) 3. Vo, D.N., Ongsakul, W.: Economic dispatch with multiple fuel types by enhanced augmented Lagrange Hopfield network. Appl. Energy 91(1), 281–289 (2012) 4. Chaturvedi, K.T., Pandit, M., Srivastava, L.: Particle swarm optimization with crazy particles for nonconvex economic dispatch. Appl. Soft Comput. 9(3), 962–969 (2009) 5. Yang, L., Fraga, E.S., Papageorgiou, L.G.: Mathematical programming formulations for nonsmooth and non-convex electricity dispatch problems. Electr. Pow. Syst. Res. 28, 302–308 (2013) 6. Yang, H.T., Yang, P.C., Huang, C.L.: Evolutionary programming based economic dispatch for units with non- smooth fuel cost functions. IEEE Trans. Power Syst. 11(1), 112–118 (1996) 7. Park, J.H., Kim, Y.S., Eom, I.K., Lee, K.Y.: Economic load dispatch for piecewise quadratic cost function using Hopfield neural network. IEEE Trans. Power Syst. 8(3), 1030–1038 (1993) 8. Lee, K.Y., Sode-Yome, A., Park, J.H.: Adaptive Hopfield neural networks for economic load dispatch. IEEE Trans. Power Syst. 13(2), 519–526 (1998) 9. Zhan, J., Wu, Q.H., Guo, C., Zhou, X.: Economic dispatch with non-smooth objectives—part II: dimensional steepest decline method. IEEE Trans. Power Syst. 30(2), 722–733 (2015) 10. Ciornei, I., Kyriakides, E.A.: GA-API solution for the economic dispatch of generation in power system operation. IEEE Trans. Power Syst. 27(1), 233–242 (2012) 11. Eberhart, R, Kennedy, J.A.: New optimizer using particle swarm theory. In: Proceedings of the Sixth International Symposium on MICRO MACHINe and Human Science 1, 995.MHS’95, pp. 39–43 (1995) 12. Chaturvedi, K.T., Pandit, M., Srivastava, L.: Self-organizing hierarchical particle swarm optimization for nonconvex economic dispatch,. IEEE Trans. Power Syst. 23(3), 1079–1087 (2008) 13. Panigrahi, B.K., Pandi, V.R., Das, S.: Adaptive particle swarm optimization approach for static and dynamic economic load dispatch. Energ. Conver. Manage. 49(6), 1407–1415 (2008)

Solution of Economic Load Dispatch …

237

14. Sun, J., Palade, V., Wu, X.J., Fang, W., Wang, Z.: Solving the power economic dispatch problem with generator constraints by random drift particle swarm optimization. IEEE Trans. Ind. Inf. 10(1), 222–232 (2014) 15. Amjady, N., Sharifzadeh, H.: Solution of non- convex economic dispatch problem considering valve loading effect by a new modified differential evolution algorithm. Int. J. Electr. Power Energy Sys. 31, 893–903 (2010) 16. Lin, W.M., Cheng, F.S., Tsay, M.T.: An improved tabu search for economic dispatch with multiple minima. IEEE Trans. Power Syst. 17(1), 108–112 (2002) 17. Babu, G.S., Das, D.B., Patvardhan, C.: Real-parameter quantum evolutionary algorithm for economic load dispatch. IET Gener. Transm. Distrib. 2(1), 22–31 (2008) 18. Yang, X-S.: Flower pollination algorithm for global optimization. In: Unconventional computation and natural computation, pp. 240–249. Springer (2012) 19. Hardiansyah: A modified particle swarm optimization technique for economic load dispatch with valve-point effect. Int. J. Intelligent Syst. Appl. 07, 32–41 (2013)

Fault Detection in Overhead Lines Using IoT-Enabled Fault Passage Indicator Monika Yadav, Prayanshu Verma, and Devender Kumar Saini

Abstract The electrical distribution systems are broad networks and play a major role in electric power system. However, electricity power cut in particular area poses a massive threat to the power system as they lead to long interruptions of power which mean huge outage time and large economic losses. In this paper, the effect of power outages on the consumer end is discussed along with the solutions to make the distribution system reliable and efficient by using IoT-enabled Fault Passage Indicators (FPI) in the distribution lines which give visual indication of any fault that occurs on the overhead lines. The proposed framework is utilized to detect and identify the specific fault location with respect to phase in a very short time. In addition to this, the programming of the microcontroller has been used for the purpose of real-time monitoring. In this way, the consumers of electricity experience shorter outage times, and moreover, the operating lineman spends less time in searching the fault location and is able to isolate and repair the faulty part which thereby results in cost saving. Keywords Fault passage indicators · Distribution system · Fault location · IoT

1 Introduction Distribution systems are the most crucial link in the entire power sector network. The main role of an electricity distribution system is to meet the customer’s demand for power after accepting the mass electrical power from transmission or subtransmission substations. An important target in all the electrical power systems M. Yadav (B) · P. Verma · D. K. Saini University of Petroleum & Energy Studies, Dehradun, India e-mail: [email protected] P. Verma e-mail: [email protected] D. K. Saini e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2021 S. Choudhury et al. (eds.), Intelligent Communication, Control and Devices, Advances in Intelligent Systems and Computing 1341, https://doi.org/10.1007/978-981-16-1510-8_25

239

240

M. Yadav et al.

Fig. 1 Typical probability distribution per fault type

is to keep significant continuity of power but it is practically impossible to do so because of various faults that occur in distribution systems. The typical probability dispersion per fault type is mentioned in Fig. 1 [1]. For that, compliances of grid code are mandatory [2]. Some electrical faults remain for a limited timeframe and return back to the typical working stage and these faults are known as temporary faults. On the other hand, the second kind of electrical fault is permanent faults that will stay until the short circuit is recognized and eliminated. If the first kind of faults, temporary electrical faults, are not removed, in the end, they will definitely convert into permanent faults sometimes [3]. However, there are four sorts of faults that can happen in distribution frameworks: they are line to ground fault (L-G), line to line fault (L-L), double-line to ground fault (LL-G), and line to line to line to ground fault (LLL-G). Mostly, 70–80% of faults occur due to equipment failures, physical accidents, etc. Moreover, natural events also cause single line to ground or line to line to ground faults. In overhead lines, line to ground faults occur mostly because of lightning incited transient high voltage, and falling of trees brought about by wind is also a significant reason for faults [4]. In electrical distribution systems, fault detection is one of the main functions to reduce power outage time which results in less electricity economic losses and less power interruption at the consumer end. The complete cycle of electrical fault resolution contains a total of three phases: first, the grouping or detection of abnormal current and voltage, second, the location of occurrence of the fault just to have a quick solution for the issues that emerge in the electrical power system network, and at last third, the faulty section being cleared inside very short timeframe manageable to prevent harm to the healthy section of the complete network [5]. Figure 2 shows the installed FPI on overhead line that senses the faults [6]. There are various researches and techniques that have been developed to distinguish and identify the faults in distribution systems and they are categorized as: conventional techniques such as impedance-based methods, technique based on traveling wave, and (AI) artificial-intelligence techniques like Support Vector Device (SVD), Fuzzy logic, Artificial Neural Network (ANN), Matching methodology, and

Fault Detection in Overhead Lines Using IoT …

241

Fig. 2 FPIs on overhead line

Genetic Algorithm (GA). However, the process of detecting the location of fault utilizing intelligent techniques is very tedious since they require some training information for handling and are time-taking [3]. Apart from all these techniques, there were some methods used previously for the fault identification like first, “7-node frameworks”. In this method, a few experiments for 7-node frameworks are reenacted and to produce a set of base regulation for all the samples taken, the MRA (Multi-Resolution Analysis) of the current imprint got from fundamental substation feeder is utilized to compute the estimations of electrical fault levels [4]. The second and third methods were “Fuzzy Inference System” and “Adaptive Neuro fuzzy Inference System”. The distinction between FIS and ANFIS is that, with the FIS, just fixed enrollment works that are picked subjectively are used. Nonetheless, ANFIS enrollment capacities are adjusted to a verifiable information set. The modeling of FIS depends vigorously on the client’s translation of the connection between both the data input as well as output. On another side, ANFIS improves the cycle of the process by adjusting the output and input enrollment function to the relationship of an example set of input/output information [5]. In Fig. 3, different methods of fault location are shown [7]. In this paper, a fault recognition and location method utilizing IoT-enabled Fault Passage Indicators (FPI) sensors for distribution systems is proposed. With the help of IoT (Internet of things), real-time simulation can be achieved. Further, it contains a brief historical review of FPIs along with the merits and demerits which concludes that FPI is a cost-effective technique in order to increase the operational efficiency during fault management [8] and it also helps us to locate the faults with very bright LEDs and therefore sensibly reduces the downtimes and huge power outages at the consumer end.

2 Historical Review: Fault Passage Indicators A fault indicator is a device that can be situated on the overhead lines at some key areas on the distribution system that will give an indication if it detects any fault current passing in the line on which it is located. FPI has the capacity of differentiating the

242

M. Yadav et al.

Fig. 3 Categorization of earth fault location techniques

fault current from the load current owing through the feeder [9]. It is also known as Faulted Circuit Pointer (FCP), and the gadget is utilized in distribution networks as a method for naturally distinguishing and recognizing faults to lessen outage time. Commonly fault passage indicators sense magnetic field brought about by electric current flows through a conductor or link. Among them, few likewise use the estimation of the electric field brought about by voltage in the conductor. The first fault passage indicator came into the trade from Horstmann (Germany) in 1946. The E.O. Schweitzer Manufacturing Company (presently a division of Schweitzer Engineering Laboratories, Inc.) acquainted an item to the United States in 1948. The first FPIs were manually reset gadgets. Later fault passage indicators were consequently reset on a particular framework restoration or after a set timeframe. At the time of fault on a grounded framework, extra current passes through a conductor and induces a magnetic field, which is recognized up by the fault pointer and appears as a state change on LED or remote visual indication device. The severity of occurrence of a fault in distribution lines is shown in Table 1 [3] which represents Table 1 Severity of occurrence of fault

Type of faults

Percentage of fault occurrence (%)

Severity

L-G-F

70

Less severe

L-L-F

15

Less severe

LL-G-F

10

Less severe

LL-L-F

5

More severe

Fault Detection in Overhead Lines Using IoT …

243

that line to ground fault (L-G) is maximum.

3 Advantages of FPIs 1. 2. 3. 4.

Compact, lightweight development which helps in easy installation. Strong, basic clamping mechanisms fit different conductor sizes, ensuring quick, dependable installation. They lessen fault-finding time by 50%. They assist utilities with fast service of restoration.

4 Methodology 4.1 Proposed Method for Fault Location Identification In the non-automated distribution systems, operating line crews have been sent to the faulty feeder so that they can locate the faulty segment and segregate it from the system. In the case of automated distribution systems, the switchgear is opened and then shut by following some predefined calculations and computations as per the fault location results. In these two cases, the process can make superfluous wear on protective equipment attached with feeder because of the uncertainty of the fault location and results in an increase in the maintenance time and decrease in equipment’s life. So, Fault Passage Indicator (FPI) is a solution that assists operating line crews to locate the faulty section [10]. In order to make the proposed methodology clear, the block diagram is shown in Fig. 4. In this block diagram, the supply will be given to the transformer, the rectifier, and afterward the regulator. The microcontroller plays a significant function through which the task of sending the signal to the wi-fi module is accomplished. Communication between the devices also plays an important role, and various types of communication links are mentioned in the paper [11]. The study basically utilizes the concept of Ohm’s law, that is, when a low DC voltage is applied at the end of the feeder via series resistor arrangement (link lines), at that point the electric current will differ contingent on the fault location in the cable as resistance is corresponding to the separation. In the event that there is a short out (Line to Ground), the voltage across series resistors arrangement changes as per the resistance that changes with separation. In this method, there are three phases (R, Y, B) representing transmission lines up to 4 km and also used few LEDs for fault indication and constant checking of transmission lines. In addition to it, for the purpose of creating the fault, few sets of switches are used which are placed at 1, 2, 3, and 4 km transmission lines with the help of which fault execution takes place. Further, the programming of the microcontroller is done using Arduino IDE software so that the voltage where the fault has occurred

244

M. Yadav et al.

Fig. 4 Block diagram of the proposed method

will be sent to the microcontroller which will sense the current in the circuit cable, and thereby the analyzed data will be displayed to the LCD screen. Moreover, we created an SSID which is a separate wireless network and with the assistance of the ESP8266 wi-fi module, the complete data along with the fault location will be sent to an IoT platform or site for real-time simulation. Also, the merits and demerits of different methods used in fault location and detection are shown in Table 2 [7].

5 Simulation Results To demonstrate the effectiveness of the system in locating faults, the proposed model is being tested by performing a simulation using “iotgecko” platform such that the real-time simulation is achieved. For identifying the location of the faulty section, a few cases are covered. It is found that the data received in the display screen and in the IoT platform are the same. The above figures are demonstrating the prototype of the model along with the outcomes which we got in IoT simulation. In Fig. 5, the model is shown in which a screen is placed which is displaying all the three phases along with the distances. So, when there is any fault that occurs in any of the phases, at that point screen shows the corresponding distance of that respective phase, and moreover, the same location of the fault can likewise be seen in the IoT platform during simulation which is appeared in Fig. 5a for single phase, i.e., R phase-3 km. Similarly, for two phases and three phases simultaneously, the data can be seen in Figs. 6 and 7a.

Fault Detection in Overhead Lines Using IoT …

245

Table 2 Summary of the fundamental qualities of different methods Methods

Impedance based

Traveling wave Injection

IED/FPI

Major advantages

(a) Easy execution (b) Cost effective

Theoretical accuracy

(a) Verified in practice (b) Just current estimations needed along different feeders

(a) Verified functionality (b) High affectability (c) It is carried by smart grid development

Main drawbacks

(a) No persuading confirmation in compensated/repaid networks (b) Reliant on-line parameters evaluation

(a) Branches (b) Extremely less restricted involvement in distribution networks

(a) Transmitter is required (b) IED (Intelligent Electronic Device) is required along the feeders (c) Few practical limitations are present (d) Precise location of any fault must be found by different methods

(a) LEDs are required along the feeders (b) Precise location of any fault must be found by different methods (c) At least most executions need expensive estimations

Relevant in No compensated networks

Yes

Yes

Yes

Relevant in isolated neutral networks

Yes

Yes

No

Yes

Proved by field tests

Partially

Not in distribution networks

Yes

Yes

Featured (a) Measurement prerequisites precision is required (b) Network information

(a) High sampling rate (b) Costly detection device

(a) Injection gadget at the essential substation (b) Current sensors (c) Communication

(a) LEDs along the feeder (b) At least current sensors. (c) Communication

Outlook

Worth more inspections

Relevant in remunerated networks

Alongside the improvement of smart frameworks, these sorts of techniques will be progressively normal

After various methodologies, the prospect is inauspicious

246

M. Yadav et al.

Fig. 5 Prototype of proposed model. a Simulation of R phase

Fig. 6 Prototype of proposed model. a Simulation of YB Phase

Fig. 7 Prototype of proposed model. a Simulation of RYB Phase

6 Conclusion In this paper, the method of fault location and identification along with the real-time simulation is discussed. Over late many years, a ton of work and attempts have been placed into the advancement of fault location and identification as well. An ideal technique needs to be simply basic and profitable which is appropriate in different sorts of systems. Despite the fact that the faults in power distribution systems are

Fault Detection in Overhead Lines Using IoT …

247

very common, but identifying the location of the fault in such systems has either been tedious or time-consuming with conventional techniques. Today, fault passage indicators are very promising and accurate. It significantly reduces the outage time by finding the source that caused the fault. Moreover, the presented technique is based on the real-time simulation using IoT platform which results in less power interruption on the consumer end and also reduces power outage duration.

References 1. Zeljkovic, C., Mrsic, P., Lekic, D., Erceg, B., Matic, P., Zubic, S., Balcerek, P.: Performance assessment of fault locators and fault passage indicators in distribution networks by the nonsequential monte carlo simulation (2018) 2. Yadav, M., Saini, D.K., Pal, N.: Review and compliances of grid code with renewable energy (RE) integration. In: Proceedings of International Conference on Large Scale Grid Integration. Renewable Energy India, pp. 1–9 (2017) 3. Gururajapathy, S.S., Mokhlis, H., Illias, H.A.: Fault location and detection techniques in power distribution systems with distributed generation, vol. 74 (2017) 4. Thakre, M., Gawre, S.K., Mishra, M.K.: Distribution system faults classification and location based on Wavelet Transform (2013) 5. Babayomi, O.O., Oluseyi, P.O.: Intelligent fault diagnosis in a power distribution Network (2016) 6. Vidya Sagar, E.: Effect of Alternative Supply, Disconnecting Switches and Fault Passage Indicators on Reliability of Radial Distribution Feeder (2017) 7. Farughian, A., Kampulainen, L., Kauhaniemi, K.: Review of methodologies for earth fault indication and location in compensated and unearthed MV distribution networks (2017) 8. Bjerkan, E.: Efficient fault management using remote fault indicators, CIRED (2009) 9. Viplav Chaitanya Reddy, B.: Outage management system with fault passage indicator (2015) 10. Janjic, A., Velimirovic, L.: Integrated fault location and isolation strategy in distribution networks using Markov decision process (2020) 11. Yadav, M., Pal, N., Saini, D.K.: Microgrid control, storage, and communication strategies to enhance resiliency for survival of critical load. IEEE Access 8, 169047–169069 (2020)

Linear State Estimation for Power Systems Based on PMU Measurements Biswaranjan Mishra, S. S. Thakur, Pankaj Rai, and D. K. Tanti

Abstract A unique State Estimation (SE) technique is proposed here for power systems by optimal positioning of Phasor Measurement Units (PMU). The key idea of this method is to maintain the full observability of the power systems and also reduce installation costs. This technique estimates the states of the electric network in a single iteration only by utilizing a linear measurement model. Therefore SE quality is enhanced. IEEE 5, 14, 30, 57, and 118 bus test systems and 38 bus real systems of Damodar Valley Corporation (DVC) with various simulated operating situations have been investigated in this method. To make these above systems more realistic, a season load profile of the IEEE reliability test system (RTS)-96 bus is applied. The simulation results of the novel technique were compared with the famous Weighted Least Square (WLS). It shows the superiority of the technique proposed and its correctness for online implementation. Keywords Optimal placement · State estimation · Reliability test system · Weighted least square state estimation · Phasor measurement unit · Synchronized measurement

1 Introduction It is essential to collect accurate information on the states of a power system for a wellknown electric network at different loading situations for economical applications [1]. State Estimation (SE) is the method by which a statistical approach is used to estimate the best possible system states by minimizing or maximizing a chosen criterion.

B. Mishra (B) · P. Rai · D. K. Tanti Birsa Institute of Technology Sindri, Dhanbad, Jharkhand 828123, India e-mail: [email protected] S. S. Thakur National Institute of Technology Durgapur, Durgapur, West Bengal 713209, India © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2021 S. Choudhury et al. (eds.), Intelligent Communication, Control and Devices, Advances in Intelligent Systems and Computing 1341, https://doi.org/10.1007/978-981-16-1510-8_26

249

250

B. Mishra et al.

The SE preliminary concept in power systems was introduced in [2–4]. Hereafter, research works in SE have developed enormously, and the application has become famous in electric control centers. In power systems, SE algorithms were generally designed as over-determined non-linear equation systems. For the solution, Weighted Least Square (WLS) SE scheme is utilized in [5–7]. It has been revealed in energy control centers that it is challenging to retain measurement synchronism using the SCADA system. The investigator presented that synchronized voltage of positive sequence and current measurements were received against PMUs to overcome this in [8, 9], which could be utilized directly in power system SE. The researchers have made numerous attempts to optimally include PMUs due to their expensive infrastructure with full observability of the electric networks by various mathematical models [10–16]. In [17], an analysis of the placement of PMUs was carried out for enhancing hybrid SE using the conventional Gauss–Newton method. This technique uses both the measurements obtained from PMUs and SCADA systems. There are three primary necessities for power systems SE, such as convergence, observability, and performance, introduced to compare the impact of PMU placement. In a unified algorithm, both observability analysis and BD detection were presented in [18] for the best possible PMUs positioning in the SE of the electric network. In [19], a novel PMU-based robust SE method was developed for monitoring power systems in real-time at different operating conditions. To make it more robust and adaptive, weight assignment function based on the distance of large undesirable perturbation from the PMU measurements was used. It has been observed from the earlier literature survey in the area of SE as (i) the maximum number of the techniques need the calculation of Jacobin matrix continually to obtain the states of the system (ii) PMU usage not only eliminates the measurement errors, but also the large costs of installing and maintaining PMUs are prime issues (iii) Another difficulty is that the online computation of states utilizing the minimum number of measurements without compromising calculation using the aforementioned method is very complicated. A novel technique has been presented here to overcome all these lacunas. The innovation of the proposed approach is that it decreases the implementation expenses by optimally positioning PMUs, preserving the full system’s observability. In contrast, without any iterative method, it utilizes a simple measurement model to obtain a complex voltage of the bus.

2 Problem Formulation This manuscript utilizes a popular WLS SE technique to solve the SE issues. First, the basic model is presented, and in the next section, descriptions of the proposed technique are elaborated. SE based on WLS SE Method A power system consist of the following vectors-

Linear State Estimation for Power Systems …

251

z: Measurement vector (m dimension) x: State vector (n dimension). It can be portrayed as z = h(x) + μ Here, h(x): m-dimensional vector of power flow equations. μ: m-dimensional vector with noise.   E[μ] = 0; E μ · μT = Rr Here, the expectation operator is E[·] and ‘T’ denotes matrix transposition. ‘Rr ’ is defined as the covariance matrix considering measurement error. Mainly, the SE technique calculates the state vector. The index of performance is diminished when J (x) = [z − h(x)]T Rr−1 [z − h(x)] The solution of the iterative process obtained by minimizing (3) is x =



H T Rr−1 H

−1

H T Rr−1 z



where x = x p+1 − x p is known as correction vector and z = z − h(x p ) is mismatch vector. Here, p and H = ∂h(x)/∂ x are called iteration index and Jacobian matrix, respectively.

2.1 Proposed Method In this paper, to solve the SE issue, a unique SE method is presented. This new technique is represented in the following ways: (i) Optimal positioning of PMUs to minimize installation costs to retain full system observability (ii) PMU measurements used for optimal filtering. Optimal Deployment of PMUs A PMU is a measuring device. It measures the complex voltage of the bus and branch currents in a synchronized manner. So, the entire connected buses are observable by deploying PMUs on one bus. Also, the installation cost of the PMU is more. Accordingly, here PMUs are placed optimally to maintain the whole observability of the system and reduce placement costs. Here, PMUs are positioned at the buses with the highest visibility, enabling them to become observable physically connected with these buses. This method aims to decrease the quantity of PMUs necessary for full observability of power networks. So as to implement such a method, an L matrix is determined. There is a description

252

B. Mishra et al.

of the components of the L matrix, such as

L jk

⎧ ⎨ 1, if buses j and k are connected = 1, if j = k ⎩ 0, otherwise

Here, L is the connectivity matrix of the bus. The size of L is n ×n and n represents the number of buses. It is computed by transforming its elements into 1 or 0 forms from the bus admittance matrix (Y). The sum of the individual column of the visibility matrix L can be determined to construct a vector S (size 1 × n) to find out the utmost bus connectivity. The components of S are represented by Sj =

n

LTjk

k=1

The bus having maximum connectivity corresponds to the elements of S with the highest value. Therefore, a PMU is placed on this bus. After that, the placement of PMU updates the visibility array λ with size (1 × n). So, the elements of λ depending on observability buses.

λj =

1, if the bus j is observable 0, or else

The row and column corresponding to the connectivity matrix L are omitted after the placement of PMU to this bus. When all the visibility array elements are equal to one, then the entire system becomes completely observable. The cost function below is optimized by placing PMUs optimally as  = Min

n

xj

j=1

Here, X is a matrix (with size 1 × n), the elements of which are 1 if PMUs are deployed on the consequent buses, or else 0. So, by deploying PMUs optimally, the entire system becomes observable if L · XT ≥ [1]n×1 Here, X = [x 1 , x 2 , x n ]. Synchronized Phasor Measurements-Based Optimal Filtering For the computation of an accurate estimate of the complex voltage magnitude of the bus, a simple measurement model is suggested.

Linear State Estimation for Power Systems …

253

z =E·V+ε Here, z is the complex measurement matrix with m dimension obtained from PMUs. E is the coefficient vector having size (m × n). Depending on the form of measurement and connectivity of buses, the elements of which are either 1 or 0, or the line parameters function. V is a vector of complex bus voltage and ε a complex random variable vector with size m containing white noise properties where m represents measurement numbers. The statistical properties of ε is E(ε j ) = 0 ∀ j The mathematical representation of the correlation vector is as follows  E[ε j ·

εk∗ ]

=

0, if j = k (σ R2 j + σ I2j ), if j = k

Here, σ R2 j : variances of real components of ε j and σ I2j : variance of Imaginary components of ε j . The covariance of ε is represented as follows E[ε · ε∗ ] = Rcc T

∧

The expected magnitude of vector of the complex voltage V of the bus can be achieved by optimizing Min f (V) = [z − EV]T Rcc [z − EV] V

As is linear, optimization of gives ∧  −1 T −1 V = ET R−1 E Rcc · z cc E

3 Procedure for Computation The primary implementation of the computational process of the proposed method is presented below.

254

B. Mishra et al.

Fig. 1 Flow diagram for the deployment of PMUs optimally

3.1 Optimal Placement of PMUs Here, PMUs are placed optimally on an offline mode by utilizing the buses whose connectivity is maximum. The flow diagram is portrayed in Fig. 1.

3.2 Proposed SE with PMU Measurement The optimally deployed PMU measurements are utilized as measured in the previous step. The complex bus voltage of the system is estimated by using the measurements mentioned above. The flowchart of the novel SE scheme is illustrated in Fig. 2.

Linear State Estimation for Power Systems …

255

Fig. 2 Flow diagram for the novel SE technique

4 Simulation Results The simulated result obtained from the proposed has been inspected on IEEE 5, 14, 30, 57, and 118 bus systems. It is also investigated on realistic 38 bus of DVC system with various operating situations. To confirm the supremacy of the method, the results achieved from the novel method are compared with the WLS SE method with a similar operating environment. The results obtained from this method are compared with the WLS SE scheme with alike operating situations.

4.1 Simulation Description The simulation was accomplished by changing the load at every bus by considering the IEEE RTS-96 bus with winter seasonal load profile [20] of 51th week of Tuesday for a period of 24 h. Here, a jitter is portrayed through a normally distributed random fluctuation of 2% of 24 time samples trend component with zero mean and standard deviation. The power factor of the system was kept constant throughout the simulation, accompanied by the active counterpart with the reactive power. It was assumed that total load variation was distributed among the generators according to their factor

256

B. Mishra et al.

of participation. The complex voltages and complex line current measurements were taken from PMU deployed buses along with associated lines. Henceforth, to check the efficacy of the proposed algorithms under different operating conditions, sudden change of load has been simulated between 9th–11th, 2nd–4th, 3rd–5th, 11th–13th and 15th–17th, 10th–12th, 15th–17th, 19th–21th time steps of IEEE 14, 30, 57, and 118 test system, respectively. Similarly, sudden change of load is carried out at 6th–8th and 15th–17th-time samples of DVC 38 bus practical systems.

4.2 Discussion of Results The novel SE technique has been inspected among IEEE 5, 14, 30, 57, and 118 test systems along with real 38 bus DVC system comprehensively with different performing circumstances. The actual load profile of 51th week of the winter season of Tuesday of IEEE RTS-96 bus system is presented in Fig. 3. This load profile is applied on each IEEE test along with DVC 38 bus practical system.

Fig. 3 Actual load profile of IEEE RTS-96 bus system of Tuesday of 51th week of the winter season

Linear State Estimation for Power Systems …

257

Here, results are portrayed for the brevity of the aforementioned simulated conditions. In Table 1, the positions of deployment of PMUs optimally among all standard five IEEE test systems along with realistic 38 bus DVC system were provided. Thereby, the novel method competently minimizes the required PMU numbers for the power systems’ complete observability. The proposed method has so many exciting results with the aforementioned simulated circumstances. But, some results are presented here because of insufficient space. Here, load variation is carried out throughout the simulation by utilizing the IEEE RTS-96 bus system. Figures 4 and 5 show a comparison of bus voltage magnitude and real power at bus number 39 and 68 of IEEE 118 bus during sudden load variation. In both these figures, it is found that the results of the novel technique are superior over WLS SE. The voltage magnitudes and reactive line powers between buses 20–21 have been compared in Figs. 6 and 7 at bus 8 of DVC 38 bus practical systems’ sudden load variations. From these graphs, it is observed that the proposed method of SE outperforms WLS SE techniques. This also shows the suitability of the proposed method in a practical system over WLS SE techniques.

Fig. 4 Comparative result of bus voltage magnitude at bus 39 of IEEE 118 bus during sudden load variation

258

B. Mishra et al.

Fig. 5 Comparative result of real power at 68th bus of IEEE 118 bus during sudden load variation

5 Conclusions A novel method of linear-based SE has been portrayed here. This scheme utilizes the deployment of PMUs optimally to reduce the cost of installation. In addition, this technique utilizes a linear measurement model to provide SE solution in one step with no iterative process. This method reduces the computational burden and complexity because of the constant-coefficient matrix, E, rather than the Jacobian matrix. Furthermore, this also decreases matrices’ size due to the use of complex bus voltage measurements as a state variable. It dramatically reduces the computational burden. The effectiveness of the proposed method, utilizing measurements obtained from the position of PMUs optimally, was verified on IEEE 5, 14, 30, 57, and 118 bus test systems,and real 38 bus DVC system with different operating situations. The results portrayed here show that the proposed technique performance is better than the WLS SE method. Further, it is confirmed that the novel method is easy to implement and suitable for real applications.

Linear State Estimation for Power Systems …

259

Fig. 6 Comparative result of voltage magnitude at 8th bus of 38 bus of DVC system during sudden load variation

260

B. Mishra et al.

Fig. 7 Comparative result of reactive line powers between buses 20–21 of 38 bus of DVC system during sudden load variation

Table 1 Placement of PMUs optimally in five standard IEEE bus test and practical DVC systems Test systems

Numbers of PMU deployed Optimal PMU position

IEEE 5 bus

1

2

IEEE 14 bus

4

2; 6; 7; 9

IEEE 30 bus

10

2; 4; 6; 9; 10; 12; 15; 18; 25; 27

IEEE 57 bus

17

1; 4; 7; 9; 15; 20; 24; 25; 27; 32; 36; 38; 39; 41; 46; 50; 53

IEEE 118 bus 31

3; 5; 9; 12; 15; 17; 21; 23; 28; 30; 36; 40; 44; 46; 51; 54; 57; 62; 64; 68; 71; 75; 80; 85; 86; 91; 94; 101; 105; 110; 114

DVC 38 bus

1; 2; 5; 8; 10; 12; 14; 17; 19; 21; 26; 30; 32

13

Acknowledgements The authors gratefully acknowledge the NPIU, Govt. of India, for financial support of this research work under a collaborative research scheme (1-5728530491).

Linear State Estimation for Power Systems …

261

References 1. Durgaprasad, G., Thakur, S.S.: Robust dynamic state estimation of power systems based on M-estimation and realistic modeling of system dynamics. IEEE Trans. Power Syst. 13(4), 1331–1336 (1998) 2. Schweppe, F.C., Wildes, J.: Power system static-state estimation, Part I: Exact model. IEEE Trans. Power Appar. Syst. PAS-89(1), 120–125 (1970) 3. Schweppe, F.C., Rom, D.B.: Power system static-state estimation, Part II: Approximate model. IEEE Trans. Power Appar. Syst. PAS-89(1), 125–130 (1970) 4. Schweppe, F.C.: Power system static-state estimation, Part III: Implementation. IEEE Trans. Power Appar. Syst. PAS-89(1), 130–135 (1970) 5. Stagg, G.W., Dopazo, J.F., Klitin, O.A., Vanslyck, L.S.: Techniques for the real-time monitoring of power system operations. IEEE Trans. Power Appar. Syst. PAS-89(4), 545–555, (1970) 6. Schweppe, F.C., Handschin, E.J.: Static state estimation in electric power systems. Proc. IEEE 62(7), 972–982 (1974) 7. Wood, J., Wollenberg, B.F.: An introduction to state estimation in power systems. In: Power generation, operation and control, 2nd ed., pp. 453–512. Wiley (2003) 8. Phadke, A.G., Thorp, J.S., Karimi, K.J.: State estimlatjon with phasor measurements. IEEE Trans. Power Syst. 1(1), 233–238 (1986) 9. Phadke, A.G.: Synchronized phasor measurements in power systems. IEEE Comput. Appl. Power 6(2), 10–15 (1993) 10. Baldwin, T.L., Mili, L., Boisen, M.B., Adapa, R.: Power system observability with minimal phasor measurement placement. IEEE Trans. Power Syst. 8(2), 707–715 (1993) 11. Phadke, A.G.: Synchronized phasor measurements—a historical overview. In: IEEE/PES transmission and distribution conference and exhibition, vol. 1, pp. 476–479 (2002) 12. Nuqui, R.F., Phadke, A.G.: Phasor measurement unit placement techniques for complete and incomplete observability. IEEE Trans. Power Delivery 20(4), 2381–2388 (2005) 13. Dua, D., Dambhare, S., Gajbhiye, R.K., Soman, S.A.: Optimal multistage scheduling of PMU placement: an ILP approach. IEEE Trans. Power Delivery 23(4), 1812–1820 (2008) 14. Madani, V., et al.: PMU placement considerations—a roadmap for optimal PMU placement. In: 2011 IEEE/PES power systems conference and exposition, Phoenix, AZ, pp. 1–7 (2011) 15. Saha Roy, B.K., Sinha, A.K., Pradhan, A.K.: An optimal PMU placement technique for power system observability. Int. J. Electr. Power Energy Syst. 42(1), 71–77 (2012) 16. Dalawai, P.P., Abhyankar, A.R.: Placement of PMUs for complete and incomplete observability using search technique. In: 2013 Annual IEEE India conference (INDICON), Mumbai, pp. 1–5 (2013) 17. Li, X., Scaglione, A., Chang, T.H.: A framework for phasor measurement placement in hybrid state estimation via Gauss-Newton. IEEE Trans. Power Syst. 29(2), 824–832 (2014) 18. Gou, B., Kavasseri, R.G.: Unified PMU placement for observability and bad data detection in state estimation. IEEE Trans. Power Syst. 29(6), 2573–2580 (2014) 19. Zhao, J., et al.: Power system real-time monitoring by using PMU-based robust state estimation method. IEEE Trans. Smart Grid 7(1), 300–309 (2016) 20. Grigg, C., et al.: The IEEE reliability test system-1996. a report prepared by the reliability test system task force of the application of probability methods subcommittee. IEEE Trans. Power Syst. 14(3), 1010–1020 (1999)

Improvement of PID Controllers by Recurrent Fuzzy Neural Networks for Delta Robot Minh Le Thanh, Luong Hoai Thuong, Pham Thanh Tung, and Chi-Ngon Nguyen

Abstract The objective of this study is to control rotational angles of a 3degree freedom robot for tracking to the reference trajectories. That is a parallel robot, named Delta, with complex movements in shaping, processing with high efficiency, high load capacity, widely used in industry. The PID controller has been successfully developed for the Delta robot. However, when changing the robot’s parameters such as load, input coupling and friction, the PID controller is difficult to archive control criteria. This article proposes and tests a solution to improve the PID controller by combining it with a recurrent fuzzy neural network (RFNN) controller, so-called RFNN-PID controller. In the proposed solution, the PID controller plays the main role of controlling the Delta robot and the RFNN controller takes charge of a supplemental role to gain with the changes of control conditions. The RFNN-PID and PID controllers will be tested in the same conditions in MATLAB/Simulink. Simulations illustrate that the proposed controller is better than the traditional one, obtaining a response time of about 3.9 ± 0.1 (s) without steady-state error. Keywords Delta robot · PID · Recurrent fuzzy neural network · Identification

1 Introduction Parallel robot control is a topic that many researchers are still developing. With its flexible mechanism, the advantages of speed, force, and accuracy, delta robots have become popular and widely used in industry [1]. One of these studies is the PD and LQR controllers presented by Joao Fabian [2]. The main idea is to use embedded NI my RIO hardware programmed in LabView software to compare the two PD controllers and LQR robot trajectory controllers. The other is the nonlinear PID M. Le Thanh · L. H. Thuong · P. T. Tung Vinh Long University of Technical Education, Vinh Long, Vietnam C.-N. Nguyen (B) Can Tho University, Can Tho, Vietnam e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2021 S. Choudhury et al. (eds.), Intelligent Communication, Control and Devices, Advances in Intelligent Systems and Computing 1341, https://doi.org/10.1007/978-981-16-1510-8_27

263

264

M. Le Thanh et al.

shown by the HAN [3], which used a dynamic nonlinear function to reduce noise for better robot tracking. The third technique is using an intelligent PD controller to overcome the trap trajectory tracking for parallel robots [4]. Ahmed Chemori et al. have successful in applying a PID controller for a real robot with high acceleration up to 100 G [5]. The PID controllers have been successfully developed for 3–6 DOF robots [6– 9]. However, when changing robot’s parameters such as load, input coupling, and friction, the PID controllers are difficult to archive control criteria by their fixed parameters. So that, the neural networks and fuzzy logic are applied to improve controlling of the Delta robot [10]. The RFNN controllers and identifiers have been developed and applied for nonlinear systems [11–15]. During the control process, the RFNN identifier can estimate the object’s sensitivity, called Jacobian information. That information is used for online training the RFNN controller. Therefore, by using RFNN, the control technique is flexible to adjust its parameters online adapting to the control conditions. This paper aims to propose and test a solution to improve the traditional PID controller by combining it with an RFNN controller, so-called RFNN-PID controller. In this solution, the PID controller gives the main role of controlling the Delta robot and the RFNN controller keeps in charge of supplemental role to gain with the changes of control conditions. The RFNN-PID controller will be compared with the traditional PID controller in MATLAB/Simulink to determine their control criteria. The rest part of this paper focuses on the Delta robot model, RFNN identifier design, RFNN controller design, and simulation results.

2 Materials and Methods 2.1 Delta Robot Model Delta robot model is a system of rigid bodies connected by joints. The parallelogram mechanisms that connect the driving links to the mobile platform are modeled as homogeneous rods with universal and spherical joints at two ends. By using this model, the robot is seen as a system with eight bodies, including three legs that each leg having two links, one fixed plate connecting to the upper legs, and one mobile plate connecting to the lower legs. The geometry of the Delta robot is shown in Fig. 1 [16–18]. Where θi|i=1,2,3 are driving angles of actuated links—angles of upper legs; αi|i=1,2,3 are the passive angles that determine the position of the connecting rods; and P = [x p ,yp ,zp ]T is the displacement matrix of the center of the mobile plate; R and r are radiuses of upper and lower plates; L i|i=1,2 are upper and lower arm lengths. Hence, the position of the robot is determined by the generalized coordinates as q = [θ1 θ2 θ3 x P y P z P ]. The motion equation is established by using the Lagrange formula:

Improvement of PID Controllers by Recurrent Fuzzy …

265

Fig. 1 Geometry of Delta robot [18]

d dt



∂T ∂ q˙k



 ∂ fi ∂T = Qk − λi (k = 1, 2, . . . , m) ∂qk ∂qk i=1 r

−

(1)

where qk is the extrapolation coordinates; f i is the linking equations, Qk is the extrapolation force; and λi is the Lagrange factor. With this model, the vector of extrapolation coordinates θ ∈ R 6 and the number of associated equations is 3, so m = 6, and r = 3. We divide the forces acting on the robot into potential forces and forces without potential energy, the extrapolation force Qk is calculated as follow Qk = −

∂π np + Qk ∂qk

(2)

np

where Q k are extrapolation forces corresponding to forces that are not possible. Virtual power of extrapolation is not so δ A = τ1 δθ1 + τ2 δθ2 + τ3 δθ3 np

np

np

(3) np

So we have: Q 1 = τ1 , Q 2 = τ2 , Q 3 = τ3 , cases Q k = 0 with k = 4, 5, 6. Substituting kinetic expressions, potentials and equations into Eq. (1), we get the motion equation of the robot as the system. The differential equation—algebra is as follows:     1 2 ¨ I I y + m b l1 θ1 = gl1 m 1 + m b cos θ1 + τ1 − 2

266

M. Le Thanh et al.

  2λ1l1 sin θ1 (R − r ) − cos α1 sin θ1 x p − sin α1 sin θ1 y p − cos θ1 z p

(4)

   1 ¨ II y + θ2 = gl1 m 1 + m b cos θ2 + τ2 − 2   2λ1l1 sin θ2 (R − r ) − cos α2 sin θ2 x p − sin α2 sin θ2 y p − cos θ2 z p

(5)

   1 I I y + m b l12 θ¨3 = gl1 m 1 + m b cos θ3 + τ3 − 2   2λ3l1 sin θ3 (R − r ) − cos α3 sin θ3 x p − sin α3 sin θ3 y p − cos θ3 z p

(6)



m b l12











m p + 3m b x¨ p = −2λ1 cos α1 (R − r ) + l1 cos α1 cos θ1 − x p −     2λ2 cos α2 (R − r ) + l1 cos α2 cos θ2 − x p − 2λ3 cos α3 (R − r ) + l1 cos α3 cos θ3 − x p 





(7)



m p + 3m b y¨ p = −2λ1 sin α1 (R − r ) + l1 sin α1 cos θ1 − y p −     2λ2 sin α2 (R − r ) + l1 sin α2 cos θ2 − y p − 2λ3 sin α3 (R − r ) + l1 sin α3 cos θ3 − y p

(8)

      m p + 3m b z¨ p = − 3m b + m p g + 2λ1 z p + l1 sin θ1 +     2λ2 z p + l1 sin θ2 + 2λ3 z p + l1 sin θ3

(9)

2  l22 − cos α1 (R − r ) + l1 cos α1 cos θ1 − x p −  2  2 sin α1 (R − r ) + l1 sin α1 cos θ1 − y p − l1 sin θ1 + z p = 0

(10)

2  l22 − cos α2 (R − r ) + l1 cos α2 cos θ2 − x p −  2  2 sin α2 (R − r ) + l1 sin α2 cos θ2 − y p − l1 sin θ2 + z p = 0

(11)

2  l22 − cos α3 (R − r ) + l1 cos α3 cos θ3 − x p −  2  2 sin α3 (R − r ) + l1 sin α3 cos θ3 − y p − l1 sin θ3 + z p = 0

(12)

From the motion equations of the parallel robot (4) to (12), according to [1, 12, 18] we have the model of the construction object in Simulink as shown in Fig. 2.

2.2 RFNN-PID Controller Design 2.2.1

Delta Robot RFNN Identifier

The RFNN identifier is shown in Fig. 3 that consists of 4 layers [13–15, 19, 20] with a structure 2-10-25-1.

Improvement of PID Controllers by Recurrent Fuzzy …

267

Fig. 2 Delta robot model in MATLAB/Simulink

Fig. 3 Structure of the RFNN identifier

Layer 1: This is the input layer with 2 recurrent lines embedding the time series into the RFNN. The output of layer 1 is calculated by (13): Oi1 (k) = xi1 (k) + θi1 Oik (k − 1), i = 1, 2

(13)

With θi1 is the connection weight at the current time k; and xi1 (k)|i=1,2 include the current control signal and the past output of the response:  T xi1 (k) = x11 (k), x21 (k) = [u(k), y(k − 1)]T

(14)

Layer 2: This is the fuzzy layer that consists of (2 × 5) nodes, each node representing a related function of the Gaussian form with mean value mij and standard deviation σ ij , and defined as (15).

268

M. Le Thanh et al.



Oi2j (k)

O 1 (k) − m i j = exp − i  2 σi j

2

, i = 1, 2; j = 1, 2, . . . , 5

(15)

At each node on the fuzzy layer, there are 2 parameters that are automatically adjusted during the online training of the RFNN identifier, is mij and σ ij . Layer 3: This layer provides fuzzy inferences. Each node corresponds to a fuzzy rule. The qth fuzzy rule can be described as: Oq3 (k) =



2 Oiq (k)|i, j=1,2,...,5 i

(16)

i

Layer 4: This is the output layer defined as follows: Oi4 (k) =



wi4j O 3j (k)|i=1; j=1,2,...,25

(17)

j

where wi4j is the connecting weights from the 3rd layer to the 4th layer, is the weight representing the output impact level of the ith output corresponding to j rule of the RFNN: O14 (k) = ym (k) = fˆ[u(k), y(k − 1)]

(18)

The performance RFNN identifier training is based on a cost function in (19) [11]: E I (k) =

2 1 1 [y(k) − ym (k)]2 = y(k) − O I41 (k) 2 2

(19)

The RFNN’s weights will be adjusted according to (20) [11]:   ∂ E I (k) W I (k + 1) = W I (k) + W I (k) = W I (k) + η I − ∂ WI

(20)

where ηI is the learning rate; W I = [θ I , mI , σ I , wI ]T is the weight vector of RFNN updated during training the RFNN. The subscript I presents the RFNN identifier, called RFNNI. Given eI (k) = y(k) − ym (k) is the difference between plant’s output and RFNN’s output, then the gradient E I (·) with respect to W I is determined as (21): ∂ E I (k) ∂ ym (k) ∂ O 4 (k) = −e I (k) = −e I (k) I 1 ∂ WI ∂ WI ∂ WI

(21)

The weight of each RFNNI network layer is updated as follows [12, 19]: w 4I i j (k + 1) = w 4I i j (k) + ηwI I e I (k)O I3i

(22)

Improvement of PID Controllers by Recurrent Fuzzy …

269

2 O I1i j (k) − m I i j m I i j (k + 1) = m I i j (k) + ηmI I e I (k)w 4I ik O I3k (23)  2 σI i j k

2 2 O I1i j (k) − m I i j  σ I i j (k + 1) = σ I i j (k) + ησI I e I (k)w 4I ik O I3k (24)  3 σI i j k

(−2) O I1i j (k) − m I i j O I1i j (k − 1)  θ I1i (k + 1) = θ I1i (k) + ηθI I e I (k)w 4I ik O I3k  2 σI i j k (25) 

In addition, to estimate the output of the model ym (k), the RFNNI must also estimate Jacobian information ∂ y(k)/∂u(k), determined as follows [12, 20, 21]:

⎫ ⎧ 1 3 ∂ y(k)  4 ⎨ ∂ O I q (−2) O I i j (k) − m I i j ⎬ = wI i j  2 ⎩ ⎭ ∂u(k) ∂ O I2qs σI i j q s

2.2.2

(26)

RFNN-PID Controller

In this part, for controlling a robot arm, an RFNN controller (RFNNC) is combined with a traditional PID controller with its structure is illustrated in Fig. 4 [12–14]. In Fig. 4, the RFNNI is used to identify the model of the robot, based on y(k − 1) and u(k). That makes the RFNN model is simple and decreases the number of neurons. Through training, the RFNNI estimates the output trajectories of the delta robot by (18). The training performance criterion is used as (27) [11]:

Fig. 4 Control system based on RFNNs

270

M. Le Thanh et al.

E C (k) =

2 2 1 1 4 u r f nnc (k) − u(k) = u r f nnc (k) − OC1 (k) 2 2

(27)

where control signal u(k) is the sum of upid (k) and u r f nnc (k) that are the output of the PID controller and the output of the RFNNC, respectively. Using back-propagation (BP) algorithm, RFNNC’s weights will be updated by (28):   ∂ E C (k) WC (k + 1) = WC (k) + WC (k) = WC (k) + ηC − ∂ WC

(28)

In which, ηc is the learning rate, and W C = [θ C , mC ,σ C , wC ]T are updated during training. Subscript C presents the RFNNC. Given eC (k) = urfnn (k) − u(k), the gradient E C (·) with respect to W C is defined as (29): ∂u r f nnc (k) ∂ E C (k) ∂ O 4 (k) = −eC (k) = −eC (k) 1 ∂ WC ∂ WC ∂ WC

(29)

The weight of each RFNNC network layer is updated as follows [12, 20]: wC 4 4 3 wCi j (k + 1) = wCi j (k) + ηC u pid (k)eC (k)OCi

(30)

1 2 OCi − m (k) Ci j j 4 3 eC (k)wCik OCk (31) m Ci j (k + 1) = m Ci j (k) + ηCm C  2 σCi j k

2 1 2 OCi − m (k)  Ci j j 4 3 σCi j (k + 1) = σCi j (k) + ηCσC eC (k)wCik OCk (32)  3 σCi j k

1 1 OCi − m (−2) OCi (k)  Ci j j j (k − 1) 1 1 θC 4 3 eC (k)wCik OCk θCi (k + 1) = θCi (k) + η  2 σCi j k (33) 

3 Specifications and Simulation Results 3.1 Parameters of Delta Robot The robot’s parameters are described in Table 1 [12]. And the parameters of the RFNNC and the RFNNI are randomly initialized.

Improvement of PID Controllers by Recurrent Fuzzy … Table 1 Delta robot mechanical specifications

271

Symbol

Value

Unit

L1

0.3

m

L2

0.8

m

R

0.26

m

r

0.04

m

α1

0

rad/s

α2

2π/3

rad/s

α3

4π/3

rad/s

m1

1.2

kg

mb

0.2

kg

mp

0.75

kg

3.2 Simulation Results The MATLAB/Simulink control system for the Delta robot with the desired trajectory as an ellipse curve (34). xd = 0.19 sin(π t) + 0.3 yd = 0.12 cos(π t) + 0.2 zd = −0.8

(34)

The response of the traditional PID and the RFNN-PID controllers are presented in Figs. 5, 6, 7 and 8 including load changed. Simulation results indicate that the RFNN-PID controller is better than the PID controller, with the setting time is about 3.9 ± 0.1 s, and the steady-state error is eliminated. Comparing the circle and the ellipse trajectories, we have stable steady-state responses shown in Fig. 9. And the control criteria of the PID controller and RFNNPID controller are presented in Table 2. The results show that the proposed solution must be better than the classical one.

4 Conclusion This study proposes and tests a solution to improve the traditional PID controller by combining it with a recurrent fuzzy neural network controller, so-called RFNN-PID controller. In this solution, the PID controller plays the main role of controlling a MIMO nonlinear plant and the RFNNC takes charge of a supplemental role to gain with the changes of control conditions. The RFNN-PID controller is compared with the traditional PID controller on a 3-degree freedom Delta robot. Experiments show that the control criteria of the system with the RFNN-PID controller are improved

M. Le Thanh et al.

Thetas of Delta Robot (rad/s) theta1-Ref theta1RFNNC-PID theta1-PID

1

0

0

0.4 0.2 0 -0.2 -0.4 0

theta3-ref,pid,rfnnc-pid

theta2-ref,pid,rfnnc-pid

theta1-ref,pid,rfnnc-pid

272

2

4

6

8

10

12

14

18

20

theta2-Ref theta2RFNNC-PID theta2-PID

2

4

6

8

10

12

14

16

18

20

theta3-Ref theta3RFNNC-PID theta3-PID

1

0

16

0

2

4

6

8

10

12

14

16

18

20

Time(s)

Fig. 5 Comparison of traditional PID and RFNN-PID controller

better than the PID controller. The proposed algorithm is stable while changing the control conditions such as increasing load of the Delta robot. In further work, we will apply the RFNN-PID controller on a real Delta robot that under development.

error2

error1 PID and RFNNC-PID

Improvement of PID Controllers by Recurrent Fuzzy …

273

Errors of Theta (rad/s)

0.04

error1-RFNNC-PID error1-PID

0.02 0 -0.02 -0.04 0

0.02 0 -0.02 -0.04 -0.06 -0.08 0

2

4

6

8

10

12

14

18

20

error2-RFNNC-PID error2-PID

2

4

6

8

10

12

14

0.2

error3

16

16

18

20

error3-RFNNC-PID error3-PID

0

-0.2 0

2

4

6

8

10

12

14

16

18

20

Time(s)

Fig. 6 Errors of responses Fig. 7 Trajectory tracking of traditional PID and RFNN-PID controller

Trajectory of Delta Robot

0.4

0.35

0.3

Y(m)

0.25 Reference Trajectory PID-Trajectory PID-RFNNC-Trajectory

0.2

0.15

0.1

0.05 0.1

0.15

0.2

0.25

0.3

X(m)

0.35

0.4

0.45

0.5

274

M. Le Thanh et al.

Fig. 8 Responses when changing load from 4.71 to 7.05 kg

Trajectory of Delta Robot

0.4

0.35

0.3

Y(m)

0.25

0.2 Reference Trajectory PID-Trajectory PID-RFNNC-Trajectory

0.15

0.1

0.05 0.1

0.15

0.2

0.25

0.3

0.35

0.4

0.45

0.5

0.4

0.45

0.5

X(m)

Fig. 9 Trajectory changing comparison

Trajectory of Delta Robot

0.4 0.35 0.3

Y(m)

0.25 0.2 Reference Trajectory PID-Trajectory PID-RFNNC-Trajectory

0.15 0.1 0.05 0 0.1

0.15

0.2

0.25

0.3

0.35

X(m)

Table 2 System control quality standards Response

PID controller Rise time (s)

Overshoot (%)

RFNN-PID controller Settling time (s)

Rise time (s)

Overshoot (%)

Settling time (s)

Theta 1

3.134

1.873

5.6 ± 0.1

3.134

1.699

4.2 ± 0.1

Theta 2

2.838

1.882

6.9 ± 0.1

2.838

1.881

3.9 ± 0.1

Theta 3

2.698

1.859

4.2 ± 0.1

2.698

0.503

3.6 ± 0.1

Improvement of PID Controllers by Recurrent Fuzzy …

275

References 1. Merlet, J.: Parallel Robots. P.O. Box 17, 3300 AA Dordrecht. Kluwer Academic Publishers, The Netherlands (2000) 2. Fabian, J., Monterrey, C., Canahuire, R.: Trajectory tracking control of a 3 DOF delta robot: a PD and LQR comparison. In: 2016 IEEE XXIII International Congress on Electronics, Electrical Engineering and Computing (INTERCON), Piura, pp. 1–5 (2016) 3. Han, J.: From PID to active disturbance rejection control. IEEE Trans. Industr. Electron. 56(3), 900–906 (2009) 4. Hernández, J.M.E., Aguilar-Sierra, H., Aguilar-Mejía, O., Chemori, A., Arroyo-Núñez, J.: An intelligent compensation through B-spline neural network for a delta parallel robot. In: 6th International Conference on Control, Decision and Information Technologies (CoDIT), Paris, France, pp. 361–366 (2019) 5. Chemori, A., Natal, G.S., Pierrot, F.: Control of parallel robots: towards very high accelerations. In: SSD: Systems, Signals and Devices, Mar 2013, Hammamet, Tunisia, p. 8, ffirmm-00809514 (2013) 6. Anoop, R., Achu, K.: Control technique for parallel manipulator using PID. Inter. J. Eng. Res. Technol. 5(7), 56–59 (2016) 7. Zhang, Y.-X., Cong, S., Shang, W.-W., Li, Z.-X., Jiang, S.-L.: Modeling, identification and control of a redundant planar 2-DOF parallel manipulator. Inter. J. Control, Autom. Syst. 5(5), 559–569 (2007) 8. Saied, H., Chemori, A., El Rafei, M., Francis, C., Pierrot, F.: From non-model-based to modelbased control of PKMS: a comparative study. In: Proceedings of 1st International Congress for the Advancement of Mechanism, Machine, Robotics and Mechatronics Sciences, pp. 50–64, Beirut, Lebanon (2017) 9. Su, Y.X., Duan, B.Y., Zheng, C.H.: Nonlinear PID control of a six-DOF parallel manipulator. IEEE Proc. Control Theory Appl. 151(1), 95–102 (2004) 10. Widhiada, W., Nindhia, T.G.T., Budiarsa, N.: Robust control for the motion five fingered robot gripper. Inter. J. Mecha. Eng. Robot. Res. 4(3), 226–232 (2015) 11. Liu, J.: Radial Basis Function (RBF) Neural Network Control for Mechanical Systems: Design, Analysis and Matlab Simulation, pp. 55–69. Springer-Verlag, Berlin, Heidelberg (2013) 12. Thanh, L.M., Thuong, L.H., Loc, P.T., Nguyen, C.-N.: Delta robot control using single neuron PID algorithms based on recurrent fuzzy neural network identifiers. Inter. J. Mech. Eng. Robot. Res. 9(10) (2020) 13. Slama, S., Errachdi, A., Benrejeb, M.: Adaptive PID controller based on neural networks for MIMO nonlinear systems. J. Theor. Appl. Inf. Tech. 97(2), 361–371 (2019) 14. Sun, W., Wang, Y.: A recurrent fuzzy neural network based adaptive control and its application on robotic tracking control. Neural Inf. Process. Lett. Rev. 5(1) (2004) 15. Hasanpour, H., Beni, M.H., Askari, M.: Adaptive PID control based on RBF NN for quadrotor. Int. Res. J. Appl. Basic Sci. 11(2), 177–186 (2017) 16. Abdelaziz, O.T., Maged, S.A., Awad, M.I.: Towards dynamic task/posture control of a 4 DOF humanoid robotic arm. Int. J. Mech. Eng. Robot. Res. 9(1), 99–105 (2020) 17. Elayaraja, D., Ramakrishnan, R., Udhayakumar, M., Ramabalan, S.: Intelligent control of mobile robot using C++. Int. J. Mech. Eng. Robot. Res. 3(2), 429–434 (2014) 18. Dung, N.D.: Reverse dynamics of parallel delta space robots. Ph.D. Dissertation on Mechanical Engineering: 9.52.01.01, Vietnam National Library, code:629.892/Ð455L (2018) 19. Slama, S., Errachdi, A., Benrejeb, M.: Neural adaptive PID and neural indirect adaptive control switch controller for nonlinear MIMO systems. Math. Probl. Eng. 2019 (2019) 20. Lee, C.H., Teng, C.C.: Identification and control of dynamic systems using recurrent fuzzy neural networks. IEEE Trans. Fuzzy Syst. 8(4), 349–366 (2000) 21. Wei, S., Lujin, Z., Jinhai, Z., Siyi, M.: Adaptive control based on neural network. In: You, K. (ed.) Adaptive Control. InTech (2009)

Vision-Based Measurement of Leaf Dimensions and Area Using a Smartphone Chanh-Nghiem Nguyen, Dang-Khoa Thach, Quoc-Thang Phan, and Chi-Ngon Nguyen

Abstract Many vision-based techniques have been proposed and handheld devices have been commercially available for leaf area measurement. However, they are not cost-effective nor simple enough for use. As an effort to tackle this problem, this study proposed an easy vision-based approach toward measuring the leaf dimensions and area using a smartphone, a cost-effective and available device. Acceptable measurement accuracies were achieved from experiments with different leaf shapes and sizes in terms of both error and error percentage. The leaf area had the largest error percentage of 2.3%; however, its largest mean error percentage was only 0.6 ± 0.5%. The smallest error was associated with the leaf width with the largest mean error and mear error percentage of 0.61 ± 0.35 mm and 1.5 ± 1.0%, respectively. It is thus promising to apply this simple approach for screening the dimensions and area of many leaves with different shapes and sizes. Keywords Leaf · Dimension · Area · Smartphone

1 Introduction As a major exporter of agricultural products, agriculture is one of the most important economic sectors in Vietnam. In 2017, shiso leaves (Perilla frutescens var. crispa) were exported to Japan with the premium price of 0.03 USD per leaf [1]. To satisfy the requirements for export, leaves must go through rigorous inspection. However, leaf inspecting and grading are still manually performed [2], which is prone to humanbased errors and might lead to many quality issues or the dissatisfaction of consumers. To tackle these problems and promote an automatic process of leaf grading based on its uniformity and quality, non-invasive approaches have been developed. By applying linear regression, the area of the leaf could be estimated from its dimensions (e.g., length, width) [3]. This cost-effective method was successfully adopted C.-N. Nguyen (B) · D.-K. Thach · Q.-T. Phan · C.-N. Nguyen Department of Automation Technology, College of Engineering Technology, Can Tho University, Can Tho, Vietnam e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2021 S. Choudhury et al. (eds.), Intelligent Communication, Control and Devices, Advances in Intelligent Systems and Computing 1341, https://doi.org/10.1007/978-981-16-1510-8_28

277

278

C.-N. Nguyen et al.

for many kinds of leaves, such as chestnut, hazelnut, fava beans, ginger, maize, onion, pepper, and cacao [3]. However, such approaches are laborious because they require sufficient measurements for corresponding leaf morphologies. More processing techniques using computer vision have been proposed [3–6]. Some downsides of these studies were that clear and image distortion was not considered, and they were limited to certain types and species of leaves. Moreover, the camera had to be fixed in a complicated setup. Although simple and comfortable-use handheld and industrial devices are commercially available [7, 8], they are costly and somewhat complicated. To promote a comfortable and easy measurement of leaves, this study aims to propose a simple approach to measuring the dimensions (i.e., width, height, and perimeter) and area of a leaf using a smartphone. This approach is very cost-effective and suitable for personal use to monitor the development of the leaf by screening its dimensions and area.

2 Material and Methods The conceptual figure of measuring the dimensions and area of a target leaf using a smartphone is illustrated in Fig. 1. A plastic plate is used to hold an A4 paper that had printed square mesh with a size of 130 mm × 130 mm. The printed square mesh area was adequate to encompass our leaf targets of interest in this preliminary study of using a smartphone for leaf dimension and area measurement. The target

Fig. 1 Conceptual figure of leaf dimension measuring using a smartphone

Vision-Based Measurement of Leaf Dimensions and Area …

279

leaf should be put at the center of the square mesh. Then, a transparent plastic was put on the target leaf to push the leaf flat before capturing its image for an accurate measurement. The image of the target leaf was captured at a distance from 15 to 30 cm, which is a comfortable distance for an operator to hold the plastic plate with one hand and hold the smartphone to capture the image with the other hand.

3 Leaf Dimension Measurement Let I0 be the original image of the leaf whose dimensions will be determined. Let R O I0 be the desired image region in the image I0 within which the square mesh can be found. A leaf object should be within R O I0 and its dimensions will be determined in the following steps: (1) Square mesh extraction, (2) Geometric distortion correction, and (3) Leaf dimension measurement.

3.1 Square Mesh Extraction The purpose of square mesh extraction is to obtain the desired image region R O I0 in the image I0 within which the square mesh can be found. To reduce the computational cost, the desired image region R O I0 is determined from a resized version of the original image I0 . The square mesh region extraction is illustrated in Fig. 2 and is obtained as follows.

3.1.1

Square Mesh Detection from Resized Image

Let I1 be the resized gray image of the original image I0 with the scale factor of 1/4. The square mesh region R O I1 in the image I1 is obtained as follow. • Apply average filter on I1 ; • Apply Canny edge detection on the filtered image to have the resulting image as shown in Fig. 2b; • Find the contour of the square mesh based on the Canny edge detection result. A contour C is said to be of the square mesh if its area aC (in pixel number) satisfy the following criterion: amin < aC < amax

(1)

where amin and amax are the minimum and maximum area of the square mesh contour in pixel number; • Obtain the square mesh region R O I1 from the bounding box of its contour.

280

C.-N. Nguyen et al.

(a)

(b)

(c)

(d)

Fig. 2 a Original gray image; b Result of Canny edge detection; c Contour of square mesh; d Square mesh region

3.1.2

Determining Square Mesh Region in Original Image

Let (x1 , y1 ) and (x2 , y2 ), respectively, be the coordinates of the top left and bottom right pixels of the square mesh R O I1 extracted from the resized image I1 . The coordinates of the top left and bottom right pixels of the corresponding square mesh R O I0 extracted from the original image I0 are, respectively, calculated as

Vision-Based Measurement of Leaf Dimensions and Area …

281

  x1 = x1 − x ∗ 4

(2)

  y1 = y1 − y ∗ 4

(3)

  x2 = x2 + x ∗ 4

(4)

  y2 = y2 + y ∗ 4

(5)

where x and y, respectively, are compensated horizontal and vertical distances to preserve the four outermost corners of the square mesh so that they can be detected for geometric distortion correction. In this study, x and y were set equals to 5 pixels to yield the resulting square mesh region R O I0 as illustrated in Fig. 2d.

3.2 Geometric Distortion Correction Digital images generally suffer from different distortion types, which can be classified into lens distortion and perspective distortion [9]. The main focus of this study is to correct perspective distortion so that the area of a square grid in the square mesh can be used to yield the dimensions of the desired leaf object.

3.2.1

Square Mesh Corners Detection

The corners of the square mesh in the image region R O I0 can be obtained by the Harris corner detector, a very common corner detection operator that is commonly used in computer vision algorithms. However, it may yield more than one solution for one corner in the image. To obtain the four outermost corners of the square mesh precisely, Shi-Tomasi corner detector [10], an improved version of the Harris corner detector, was applied in this study. Before corner detection, contrast enhancement [11] had been performed on the image region R O I0 to improve corner detection results.

3.2.2

Perspective Transform

In this study, the square mesh should be corrected from perspective distortion to obtain the precise ratio of pixel numbers and the actual area of a squared image area. Therefore, the idea is to convert the square mesh into a squared region by the perspective transformation. A planar perspective transformation is a linear mapping on homogeneous vectors, which is represented by a 3 × 3 homogeneous matrix H as

282

C.-N. Nguyen et al.

x = Hx

(6)

⎤ ⎡ ⎤⎡ ⎤ h 11 h 12 h 13 x x ⎣ y  ⎦ = ⎣ h 21 h 22 h 23 ⎦⎣ y ⎦ s h 31 h 32 h 33 s

(7)

or ⎡

The degree of freedom for the perspective transform is 8 [12] so H can be determined by matching a set of the four outermost corner points (A, B, C, D) of the square mesh in R O I0 with their corresponding points A , B  , C  , D  which define a squared image region as shown in Fig. 3. The coordinates of these points are calculated as x A = x D = min(x A , x D )

(8)

y A = y B = min(y A , y B )

(9)

x B = xC = x A + w

(10)

yC = y D = y A + w

(11)

w = max(x B − x A , xC − x D , y D − y A , yC − y B )

(12)

where

After H is obtained, all pixels in the region R O I0 can be computed from the perspective transform H.

Fig. 3 Illustration of perspective transformation

Vision-Based Measurement of Leaf Dimensions and Area …

283

3.3 Leaf Dimension Measurement The actual area of the square mesh printed on an A4 paper attached on the surface of the leaf holder was Asquare_mesh = 1302 = 16900 mm2 .

(13)

According to (8)–(12), the square mesh after geometric distortion correction has an area asquare_mesh = w 2 .

(14)

From (13) and (14), the area of a pixel on this distortion-corrected square mesh is calculated as Apixel =

Asquare_mesh (mm2 /pixel). asquare_mesh

(15)

The pixel size is thus dpixel =



Apixel (mm).

(16)

With the knowledge of the pixel size in each image capture, the contour of the leaf object is then calculated from the distortion-corrected square mesh so that the leaf dimensions can be determined.

3.3.1

Leaf Area

Let aleaf be the image area of the leaf object, calculated as the total number of pixels inside the contour Cleaf of the leaf. The area of the leaf is calculated by Aleaf = aleaf ∗ Apixel (mm2 ).

3.3.2

(17)

Leaf Perimeter

The perimeter of the leaf object in the image can be defined as the length of the leaf’s contour Cleaf in pixel number pleaf = length(Cleaf ) (number of pixels). Therefore, the perimeter of the leaf can be calculated as

(18)

284

C.-N. Nguyen et al.

Pleaf = pleaf ∗ dpixel (mm).

3.3.3

(19)

Leaf Width and Length

To determine the 2D dimensions of the leaf accurately, it is necessary to determine the location of the leaf base and apex in the distortion-free square-mesh image area. As the leaf is placed in the holder in an arbitrary orientation, it is important to determine the locations of either the midrib or the leaf base and apex. Principal Component Analysis (PCA) is an orthogonal linear transformation that transforms the data into a new coordinate system such that the greatest variance can be obtained by the projection of the data on the first coordinate, i.e., the first principle component [13]. Because large variance can be obtained by projecting the data on symmetry axes, PCA has been widely used to find object symmetry for 3D alignment [14] or for studying the object geometric morphometrics [15]. In the case of 2-dimensional pixel data, PCA seeks to rotate the two original axes X and Y so that the new axis X lies along the direction of maximum variation in the data [13]. Despite the differences in the extent of bilateral symmetry among the leaves of the same plant species, many plants have leaves that exhibit bilateral symmetry [16]. Therefore, by applying PCA on the edge pixels of the leaf, the orientation of the bilateral symmetry axis, which normally aligns with the leaf midrib is obtained. The leaf contour can be rotated so that the leaf midrib is approximately horizontal (Fig. 4). Let W bb and H bb be, respectively, the width and height of the bounding box drawn from the rotated leaf contour. The width of the leaf object in the image can be defined as the height of the rotated bounding box in pixel number x

Fig. 4 Rotation of concave leaf’s contour

hleaf y

rotated leaf’s contour

wleaf

leaf apex

Δ

vertical search limit for leaf base

rotated bounding box leaf base

leaf centroid

Vision-Based Measurement of Leaf Dimensions and Area …

285

wleaf = Hbb (number of pixels).

(20)

Therefore, the actual width of the leaf can be calculated as Wleaf = wleaf ∗ dpixel (mm).

(21)

Let pleftmost (xleftmost , yleftmost ) and prightmost (xrightmost , yrightmost ) be the leftmost point and rightmost point of the rotated leaf contour, respectively. Let define pbase (xbase , ybase ) and pcentroid (xcentroid , ycentroid ) the leaf base point and the centroid of the rotated leaf contour, respectively. It is apparent that the leftmost point of the contour is the leaf base if the contour is convex (Fig. 5). In general, the leaf is not perfectly bilateral symmetry. To determine the leaf base location for both convex and concave cases of the rotated leaf contour, a vertical search range  = (ylow , yhigh ) is defined from the height of its bounding box where the search limits are defined as ylow = ycentroid − 0.05 ∗ Hbb

(22)

yhigh = ycentroid + 0.05 ∗ Hbb

(23)

The leaf base point can be determined as pbase =

p(x p , y p ) ∈ C , if local maximum of x p exists otherwise pleftmost ,

(24)

where C is the left part of the rotated contour limited by the vertical search range , and p is the local maximum point along the x-direction. With the knowledge of the leaf base and leaf apex, the height of the leaf object in the image can be calculated as h leaf =



xrightmost − xbase

2

2  + yrightmost − ybase (number of pixels).

(25)

Fig. 5 Leaf’s contour of convex type

leaf base

leaf apex

286

C.-N. Nguyen et al.

Fig. 6 Conceptual figure of accuracy assessment

The actual height of the leaf can be calculated as Hleaf = h leaf ∗ dpixel (mm).

4 Accuracy Assessment It is obvious that the accuracy of leaf dimension measurements largely depends on the geometric distortion correction in the proposed approach. To verify the measurement results, the same leaf objects were scanned by HP Scanjet G4010 at an acceptably high resolution of 1200 dpi (Fig. 6). Then the same procedure for leaf dimension measurement in which geometric distortion correction was excluded was applied to these scanned images. The measurement results on these scanned images were used for comparison with those performed by the proposed approach using a smartphone that captured the leaf images on the plastic holder and measured the leaf dimensions.

5 Experimental Results To evaluate the efficiency of the proposed approach for leaf dimension measurement, 28 leaf objects of different shapes and sizes were selected. The statistics of these leaves are listed in Table 1. A few of them are shown in Fig. 7. Each leaf object was captured three times with affordable capture distances from about 10 to 30 cm (Fig. 8) by an Android smartphone on which an application had been developed and installed to perform all measurements of leaf dimensions (Fig. 9). All leaf objects were also scanned by HP Scanjet G010 at 1200 dpi to provide distortion-free images for accuracy assessment.

Vision-Based Measurement of Leaf Dimensions and Area … Table 1 Statistics of leaf dataset

Description

Min/Mean/Max

Width

20/44/70 (mm)

Height

40/80/119 (mm)

Area

1225/2430/4845 (mm2 )

287

Fig. 7 Scanned images of leaves of different shapes and sizes

Fig. 8 Images captured at three different distances

Let m and t be the leaf dimensions calculated, respectively, from the smartphonecaptured image and the scanned image of the same leaf object. The measurement error is calculated as error = |m − t|.

(27)

The percentage error of a measurement is calculated as percentage_error =

|m − t| 100%. t

(28)

288

C.-N. Nguyen et al.

Fig. 9 Screenshots of Android application: a Preview of a selected image and b display of measurement results

b

a

To understand the effect of capture distance on the measurements, the statistics of measurement errors from images captured at the closest and farthest distance were also calculated (Table 2). Besides, the statistics of measurement error in percentages were calculated and shown in Table 3 to provide better insights into the relationship of the measurement error concerning the leaf size. The statistics shown in Tables 2 and 3 revealed that capturing the leaf image at a closer distance might lead to a smaller error for most cases. Among the measured Table 2 Statistics of measurement error Measurement error

Width (mm)

Height (mm)

Perimeter (mm)

Area (mm2 )

Image sets

Minimum

All images

Closest capture distance

Farthest capture distance

0.09

0.11

0.19 0.61 ± 0.35

Mean ± σ

0.59 ± 0.35

0.55 ± 0.35

Maximum

1.37

1.27

1.31

Minimum

0.48

0.58

0.55 1.54 ± 0.71

Mean ± σ

1.42 ± 0.67

1.24 ± 0.60

Maximum

3.29

2.93

3.29

Minimum

0.04

0.05

0.05

Mean ± σ

3.05 ± 1.93

2.30 ± 1.77

3.72 ± 1.87

Maximum

7.08

6.67

7.08

Minimum

0.11

0.47

0.11

Mean ± σ

12.99 ± 9.88

15.22 ± 9.50

12.99 ± 9.88

Maximum

43.14

34.72

31.55

Vision-Based Measurement of Leaf Dimensions and Area …

289

Table 3 Statistics of measurement percentage error Percentage error (%)

Width

Height

Perimeter

Area

Image sets All images

Closest capture distance

Farthest capture distance

Minimum

0.2

0.2

0.4

Mean ± σ

1.4 ± 0.9

1.3 ± 0.9

1.5 ± 1.0

Maximum

4.7

4.1

4.7

Minimum

0.7

0.7

1.0

Mean ± σ

1.8 ± 0.6

1.6 ± 0.5

2.0 ± 0.6

Maximum

3.5

2.9

3.5

Minimum

0.0

0.0

0.0

Mean ± σ

1.5 ± 1.0

1.1 ± 0.9

1.8 ± 1.0

Maximum

4.4

3.3

4.4

Minimum

0.0

0.0

0.0

Mean ± σ

0.6 ± 0.5

0.7 ± 0.5

0.6 ± 0.5

Maximum

2.3

1.9

2.0

dimensions (i.e., width, height, and perimeter), the leaf width had the smallest mean error of only 0.61 ± 0.35 mm. The perimeter had the largest mean error of 3.72 ± 1.87 mm with the mean percentage error of 1.8 ± 1.0%. In terms of percentage error, the leaf height had the largest value of 2.0 ± 0.6%. Although the area experienced the largest error (43.24 mm2 ), its mean percentage error was only 0.6 ± 0.5% with a maximum of 2.3%. Therefore, the proposed approach for measuring leaf dimensions and area had acceptable accuracies and was suitable for many screening purposes.

6 Conclusion To facilitate the measurement of leaf dimensions and area, a smartphone-based measurement approach was proposed. Experiments were carried out with various image capture distances, leaf shapes, and sizes to evaluate the proposed approach. The experimental results showed the largest error was associated with the leaf area. However, the largest and mean percentage error was only 2.3% and 0.6 ± 0.5%, respectively. For leaf dimensions, the perimeter had the largest mean error of 3.72 ± 1.87 mm, which corresponded to the mean percentage error of only 1.8 ± 1.0%. Therefore, the proposed smartphone-based approach was acceptable for a screening of the dimensions and area of many leaves with different shapes and sizes. Further improvement on the measurement accuracy might be achieved with an initial camera calibration process and vibration reduction for higher image quality.

290

C.-N. Nguyen et al.

References 1. vtv.vn: Vietnamese perilla exported to Japan (2017). https://english.vtv.vn/news/vietnameseperilla-exported-to-japan-20170712152526544.htm. Accessed 02 October 2020 2. Tuyen, N., Tu, A.: Sisho’s farm (2017). https://vnexpress.net/trang-trai-tia-to-700-dong-motla-xuat-di-nhat-3614332.html. Accessed 20 October 2020 3. Zhang, W.: Digital image processing method for estimating leaf length and width tested using kiwifruit leaves (Actinidia chinensis Planch). PLoS ONE 15(7), 1–14 (2020). https://doi.org/ 10.1371/journal.pone.0235499 4. Easlon, H.M., Bloom, A.J.: Easy leaf area: automated digital image analysis for rapid and accurate measurement of leaf area. Appl. Plant Sci. 2(7), 1400033 (2014). https://doi.org/10. 3732/apps.1400033 5. Tech, A.R.B., da Silva, A.L.C., Meira, L.A., de Oliveira, M.E., Pereira, L.E.T.: Methods of image acquisition and software development for leaf area measurements in pastures. Comput. Electron. Agric. 153(February), 278–284 (2018). https://doi.org/10.1016/j.compag. 2018.08.025 6. Mahanti, N.K., Chakraborty, S.K., Babu, V.B.: Non-destructive estimation of spinach leaf area: image processing and artificial neural network based approach. Curr. J. Appl. Sci. Technol. 39(16), 146–153 (2020). https://doi.org/10.9734/CJAST/2020/v39i1630746 7. CI-203 Wand Leaf Area Meter. https://ictinternational.com/products/ci-203/ci-203-wand-leafarea-meter/ 8. LI-COR: LI-3100C Area Meter. https://www.licor.com/env/products/leaf_area/LI-3100C/. Accessed 02 October 2020 9. Li, X., Zhang, B., Sander, P.V., Liao, J.: Blind geometric distortion correction on images through deep learning. In: Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition, June 2019, pp. 4850–4859 (2019). https://doi.org/10.1109/CVPR. 2019.00499 10. Shi, J., Tomasi, C.: Good features to track. In: Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition, pp. 593–600 (1994). https://doi.org/ 10.1109/CVPR.1994.323794 11. Kim, Y.T.: Contrast enhancement using brightness preserving bi-histogram equalization. IEEE Trans. Consum. Electron. 43(1), 1–8 (1997). https://doi.org/10.1109/30.580378 12. Tekalp, A.M.: Digital video processing, no. January 1995 (2017) 13. Wallisch, P., Lusignan, M., Benayoun, M., Baker, T., Dickey, A., Hatsopoulos, N.: MATLAB for Neuroscientists: An Introduction to Scientific Computing in MATLAB (2013) 14. Chaouch, M., Verroust-Blondet, A.: Alignment of 3D models. Graph. Models 71(2), 63–76 (2009). https://doi.org/10.1016/j.gmod.2008.12.006 15. Savriama, Y.: A Step-by-step guide for geometric morphometrics of floral symmetry. Front. Plant Sci. 9 (2018). https://doi.org/10.3389/fpls.2018.01433 16. Shi, P., Zheng, X., Ratkowsky, D.A., Li, Y., Wang, P., Cheng, L.: A simple method for measuring the bilateral symmetry of leaves. Symmetry (Basel) 10(4), 1–10 (2018). https://doi.org/10.3390/ sym10040118

Long-Range Radio and Vision Node Based Waste Management System Shaik Vaseem Akram, Rajesh Singh, and Anita Gehlot

Abstract Long Range (LoRa) radio is low power and long-range communication protocol that meets the requirement of the internet of things. In the context of waste management, automation and communication are required to be enhanced for a sustainable and effective waste management system. In this study, we are proposing the implementation of long-range radio and vision node for real-time monitoring of the waste management system. Vision-based sensor mote comprises of long-range radio, level measurement sensor, weighing senor and Pi camera for communicates the sensory information to the data transmission unit via long-range radio. The data transmission unit transmits the data to the cloud server via internet connectivity. The sensory data and visuals data logs in the cloud server. The authorities are able to the real-time monitoring the visuals of the bin, this will assist the authorities to implement an effective waste management system. Keywords Long range (LoRa) radio · Vision node · Waste management · Cloud server

1 Introduction Waste management is an organizational unit that is established for managing waste from its generation to the final disposal. In the context of waste management, it is a challenging one for municipal authorities to monitor the garbage bins manually, it is time-consuming, require a large number of workforce and also transport cost rises due to repeated visit to the bins. Waste management is one of the crucial areas where the infrastructure is needed to improvise for the effective implementation of smart cities. Internet of Things (IoT) is a promising technology that connects the bins and also monitors bins through internet connectivity from anywhere [1]. The implementation of the Internet of Things (IoT) is one of the possibilities for the implementation of smart cities. As the Internet of Things (IoT) connects the physical S. V. Akram (B) · R. Singh · A. Gehlot School of Electronics and Electrical Engineering, Lovely Professional University, Punjab, India e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2021 S. Choudhury et al. (eds.), Intelligent Communication, Control and Devices, Advances in Intelligent Systems and Computing 1341, https://doi.org/10.1007/978-981-16-1510-8_29

291

292

S. V. Akram et al.

devices to the internet, here the physical devices are low power constrained devices because these devices work on the battery power supply. So in order to the meet the requirement of IoT, we need low power, long-range, low data rate, and costeffectiveness for transmission of the data from the connecting the physical device to the virtual world [2]. LPWAN (Low Power Wide Area Network) is an tailor-made wireless communication technologies for the IoT [3, 4]. Long Range (LoRa) is one of the LPWAN wireless technologies that transmits the data for 5 km range in the urban area and 10 km in rural area [2]. LoRa radio works on chirp spread spectrum modulation technique and it utilized the 868 MHz band in the Europe region and 433 MHz band in the Asia region [5]. In this study, we are embedding the LoRa (LoRa) radio at the end devices (sensor mote) for transmitting the status of bins to long-range. LoRa (Long Range) based end devices will be deployed in the bins for transmitting the information of sensing the level of the garbage and also senses the amount of the garbage in the bins. This end device communicate the sensory information to the data transmission unit, this act as bridge between the end device and cloud server. A vision node is also embedding in our system for authenticating the operation of the prototype in visual mode. Vision node captures the visuals of images frequently for clarifying whether the garbage in the bins is cleared properly are not. Vision node is an additional add for effective implementation of smart waste management in terms of security and authentication. In order to implement the vision node, we need a specific controller unit that can process the capture visuals and transmits. Raspberry pi 3 controller is the controller unit that is capable of capturing the visuals and transmit them to the cloud server via wireless communication protocol [5]. In this study we are embedding the raspberry Pi 3 controller as an end device, that is able to process the visuals through the pi camera and also transmits the sensory information to the cloud server via long range radio communication. The cloud server receives the visuals and sensory information of the bin through the data transmission unit via internet connectivity. The data transmission unit is the integration of long-range radio and wi-fi modem, it receives the data from the end device, i.e., vision node via long-range radio and communicates the data to the cloud server over the internet protocol through Wi-Fi modem. The graphical user interface of the cloud server presents the real-time visuals and sensory information of the many bins assists the authorities to implement the waste management in realtime scenario. The visuals and sensory data available in the cloud server can be used for analyzing of data assists the authorities to check the pattern of bins. This will conclude the authorities to increase and decrease the bins in a particular location.

2 Prior Art RFID (Radio-Frequency Identification) technology is initially technology that is implemented for monitoring of bins and trucks. GPRS and GPS communication technologies are incorporated in the bins and trucks for communicating the status and location of the bin [6]. With the assistance of cloud server and Internet of Things

Long-Range Radio and Vision Node Based Waste Management System

293

(IoT), a real-time monitoring system is implementing in the waste management system for effective communication of overflow of the bins. Cloud-based smart garbage bins are implemented in this study for exchanging the information of the level of waste in the bin to the cloud server via the internet [7]. A prototype and algorithm are implemented for executing the automated solid waste management using an integrated sensing system. The integrated sensing system is mounting on the lid works for sensing the level of the waste bin [8]. A smart solid waste collection system is implementing for the collecting the waste in the city of turkey. This collection system is integrating RFID (Radio-frequency Identification) and GSM (Global System for Mobile) technologies have been integrating with the bins and trucks for effective collection of the waste [9]. GIS (Geographic Information System) mapping technology is implementing for the allocation of the solid waste bins for effective implementation of the waste collection system. Vehicle Route Problem (VRP) is used for optimizing the routes of the trucks for productive waste collection with minimum transportation cost [10, 11]. Things speak cloud server and Arduino controller unitbased internet of things system is implemented for monitoring the garbage in the bin and also for the monitoring the pungent gas that may evolve from the garbage. The things speak server connects to the municipal authority for collecting the waste on time to maintain the hygienic environment in urban [12, 13]. A green IoT (Internet of Things) waste management system is proposing for combating the waste and emission of greenhouse gases. Raspberry Pi 3 is controller unit, ultrasonic sensor is integrating with it for communicating the status of the bin over the internet through the Wi-Fi module. The Wi-Fi module is inbuilt in the Raspberry Pi 3 controller for transmitting the information to the cloud server by connecting with the internet [14]. Long-Range (LoRa) communication-based nodes are implementing for communicating the information of the garbage bins with low power consumption over long range distance. LoRa-based gateway covers the long distance and also reduces the infrastructure cost and also lessen the burden on the workforce [15, 16].

3 Proposed Architecture Long Range Radio is an advanced, low power and long-range wireless communication that is embedding in the end devices and gateway. We are promising a long-range radio and vision node based smart waste management system and the architecture of the waste management system is shown in Fig. 1. ‘n number of vision-based sensor mote will be deployed in the bins for collecting the sensory information and also for capturing the visuals of the bins for authentication and security. Vision-based sensor mote is embedding with long-range (LoRa) radio communication protocol for overcoming the communication challenge in the waste management system for reliable and long-range data transmission between the sensor mote and cloud server. Long-Range (LoRa) radio activates for the transmission of the data, the remaining time it will be sleep mode. This feature enables the LoRa for low power consumption at the sensor mote and this will enhance the battery performance of the sensor

294

S. V. Akram et al.

Vision based Sensor Mote ‘1’

Data transmission unit

Vision based Sensor Mote ‘2’

Internet Long range communication

Vision based Sensor Mote ‘n’

Cloud based waste management Server

Fig. 1 Architecture of vision node-based waste management system

mote. Long Range (Radio) is a transceiver unit, it communicates the data of the bins to the data transmission unit. This data transmission unit is also embedding with long-range radio, it receives the data and controller unit activates the Wi-Fi modem to transmit the data to the cloud server via internet connectivity. Sensory and visual data of the bins are logged in the cloud server, the data is represented in the graphical user interface (GUI). This GUI delivers the real-time data of the different bins at a single point will enhance the authorities to utilize the resources in smart and effective manner. Vision-based sensor mote is the primary unit that enables to monitor the bins wirelessly from any location. Figure 2 presents the components that exist in the vision-based sensor mote. Raspberry pi 3 is main controller unit of the mote, that processes the visuals and sensory data effectively. Raspberry Pi 3 controller is system on chip (SoC) unit, it consists of 64-bit quad-core processor. This controller unit comprises of 40 GPIO (General-purpose input/output) that enables to connect the physical devices to the controller unit. Inbuilt HDMI port and wi-fi modem, camera serial interface facilities exist in this controller unit [17]. LoRa is a low-powered wireless communication for long-distance. LoRa uses chirp spread spectrum modulation for modulating the signals in ISM (Industrial, Scientific and Medical). Chirp spread spectrum modulation broadcasts a narrow band signal over broader channel bandwidth. Long Range (LoRa) radio is also a bi-directional communication that is capable to transmit and receive the data over a long distance. LoRa operates in the spreading factor (SF) in between 7 and 12. SF 7 delivers the data in a short interval of time and SF 12 takes the long time interval

Long-Range Radio and Vision Node Based Waste Management System

295

Display unt Long Range radio

Pi camera Raspberry pi 3 controller

Level measurement sensor

Weighing sensor Battery power supply

Fig. 2 Vision based sensor mote

to deliver the same amount of the data [18]. HC SR 04 ultrasonic sensor is used as a level measurement sensor for measuring the distance to an object by sending ultrasonic waves. This ultrasonic sensor is a four-pin module, whose pin names are Vcc, trigger, echo, and ground. Power the ultrasonic sensor using a regulated + 5 V through the Vcc and ground. Two I/O pins are trigger and echo, and they are connecting to the I/O pins of the microcontroller. This ultrasonic sensor is useful to measure the distance with the range of 2–400 cm [19]. The ultrasonic sensor senses the level of the garbage of the bin and when it is about to reach the threshold level, it intimates the controller unit. Load cell sensor is used as weighing sensor, that weighs the amount of the garbage present in the bins. Load cell sensor is capable of converting the force into an electronic signal. Load cell sensor works on the wheat stone bridge circuit and generate electronic signal may be voltage change or current change. The range of load cell sensors is up to 110 kg [20]. Load cell sensor activates when the threshold level of the bin is reached, this sensor senses the weight of the garbage in the bin and intimates the controller unit. The data transmission unit is the integration of long rang radio and Wi-Fi modem, that act as a bridge between the sensor mote and cloud server. Figure 3 illustrates the components present in the data transmission unit. Long Range (LoRa) radio is a transceiver unit, that receives the sensory and visual information of the bins. The controller unit activates the wi-fi modem to transmit the sensory and visual data to the cloud server over internet protocol. The display unit presents the data in digital form.

296

S. V. Akram et al.

Long Range radio

Display Unit

Controller unit

Wi-Fi module

Battery power supply

Fig. 3 Data transmission unit

4 Hardware Specification Raspberry Pi 3 Controller: The Raspberry Pi 3 controller is a chip (SoC) device that comes with the Broadcom BCM2837 64 bit Quad Core Processor and is integrated with 40 pins. Of the 40 pins, 26 GPIO (general purpose input/output) pins are used to power physical devices using physical computing. This controller has built-in 802.11n Wi-Fi capabilities for transmitting data over the internet and is powered by a 5 V voltage USB port. With the support of the HDMI port, the PC-interfaced raspberry pi and USB ports are useful for connecting the camera. HC SR 04 Ultrasonic Sensor: Ultrasonic sensor based on ultrasonic waves. Generally, this sensor is used to determine the distance to an object. This sensor measures the distance to the object by calculating the time taken between the transmitter and the obtained ultrasonic pulses. Ultrasonic sensor is capable of measuring distances between 2 and 400 cm. Long Range Radio (LoRa): LoRa is a long-distance low-powered wireless networking. LoRa uses chirp-spread frequency modulation to modulate signals in ISM (Industrial, Science and Medical). Chirp frequency modulation transmits a narrow band signal over a larger bandwidth channel. LoRa uses the 868 MHz ISM band in Europe, 915 MHz in North America and 433 MHz in Asia. LoRa has a wide range of communications up to 10–40 km in rural areas and 1–5 km in urban areas. Power supply of 3.3 V is required for powering the LoRa. Load Cell Sensor: A load cell is a transducer that transforms force into measurable electrical output. The electrical output can be changed in the voltage and current. The load cell sensor works on the principle of wheat stone bridge circuit. The voltage supply is range in between 2.6 and 5.5 V.

Long-Range Radio and Vision Node Based Waste Management System

297

Wi-Fi Module: ESP 8266 Wi-Fi module that consists of an inbuilt TCP/IP stack for connecting to the Wi-Fi network. ESP 8266 wi-fi module is 32-bit type microcontroller that comprises 16 GPIO pins and Voltage of 3.3 V. UART (Universal asynchronous receiver-transmitter) and I2 S (Inter-IC Sound) are present in the ESP 8266 module.

5 Conclusion Smart waste management system is a necessary area to be enhanced for the implementation of smart cities. In this study, we present the implementation of IoT based waste management system with the assistance of long-range radio and wi-fi modem. A vision-based sensor mote will be embedded in the bins for real-time sensing and visual capturing of the bins and communicating to the cloud server via wireless communication. Raspberry Pi 3 controller and LoRa radio are embedding in the vision-based sensor mote for real-time capturing and send it to over the long-range wireless communication. The proposed system will enable to real-time capturing of the bins over long range distance with reliable and cost-effective communication protocol. Cloud server will enable to visualize the sensory and visuals data of the bins in the GUI, this will assist the authorities to enhance the implementation of waste management system with less infrastructure and operation cost.

References 1. Marques, P., et al.: An IoT-based smart cities infrastructure architecture applied to a waste management scenario. Ad Hoc Netw. 87, 200–208 (2019). https://doi.org/10.1016/j.adhoc. 2018.12.009 2. Mekki, K., Bajic, E., Chaxel, F., Meyer, F.: A comparative study of LPWAN technologies for large-scale IoT deployment. ICT Express 5(1), 1–7 (2019). https://doi.org/10.1016/j.icte.2017. 12.005 3. Patel, D., Won, M.: Experimental study on low power wide area networks (LPWAN) for mobile internet of things. In: IEEE Vehicular Technology Conference, 2017-June, May 2017. https:// doi.org/10.1109/VTCSpring.2017.8108501 4. Raza, U., Kulkarni, P., Sooriyabandara, M.: Low power wide area networks: an overview. IEEE Commun. Surv. Tutorials 19(2), 855–873 (2017). https://doi.org/10.1109/COMST.2017. 2652320 5. Marais, J.M., Malekian, R., Abu-Mahfouz, A.M.: LoRa and LoRaWAN testbeds: a review. In: 2017 IEEE AFRICON: Science, Technology and Innovation for Africa, AFRICON 2017, pp. 1496–1501 (2017). https://doi.org/10.1109/AFRCON.2017.8095703 6. Hannan, M.A., Arebey, M., Begum, R.A., Basri, H.: Radio Frequency Identification (RFID) and communication technologies for solid waste bin and truck monitoring system. Waste Manag. 31(12), 2406–2413 (2011). https://doi.org/10.1016/j.wasman.2011.07.022 7. Hong, I., Park, S., Lee, B., Lee, J., Jeong, D., Park, S.: IoT-based smart garbage system for efficient food waste management. Sci. World J. 2014 (2014). https://doi.org/10.1155/2014/ 646953

298

S. V. Akram et al.

8. Al Mamun, M.A., Hannan, M.A., Hussain, A., Basri, H.: Integrated sensing systems and algorithms for solid waste bin state management automation. IEEE Sens. J. 15(1), 561–567 (2015). https://doi.org/10.1109/JSEN.2014.2351452 9. Cavdar, K., Koroglu, M., Akyildiz, B.: Design and implementation of a smart solid waste collection system. Int. J. Environ. Sci. Technol. 13(6), 1553–1562 (2016). https://doi.org/10. 1007/s13762-016-0993-4 10. Khan, D., Samadder, S.R.: Allocation of solid waste collection bins and route optimisation using geographical information system: a case study of Dhanbad City, India. Waste Manag. Res. 34(7), 666–676 (2016). https://doi.org/10.1177/0734242X16649679 11. Chaudhary, S., Nidhi, C., Rawal, N.R.: Gis-based model for optimal collection and transportation system for solid waste in Allahabad city. Adv. Intell. Syst. Comput. 814, 45–65 (2019). https://doi.org/10.1007/978-981-13-1501-5_5 12. Misra, D., Das, G., Chakrabortty, T., Das, D.: An IoT-based waste management system monitored by cloud. J. Mater. Cycles Waste Manag. 20(3), 1574–1582 (2018). https://doi.org/10. 1007/s10163-018-0720-y.S.Indexed 13. Yerraboina, S., Kumar, N.M., Parimala, K.S., Jyothi, N.A.: Monitoring the smart garbage bin filling status: an IoT application. Int. J. Civil Eng. Technol. 9(6), 373–381 (2018) 14. Maksimovic, M.: Leveraging internet of things to revolutionize waste management. Int. J. Agric. Environ. Inf. Syst. 9(4), 1–13 (2018). https://doi.org/10.4018/IJAEIS.2018100101 15. Lozano, Á., Caridad, J., De Paz, J.F., González, G.V., Bajo, J.: Smart waste collection system with low consumption LoRaWAN nodes and route optimization. Sensors (Switzerland) 18(5), 1–24 (2018). https://doi.org/10.3390/s18051465 16. Sheng, T.J., et al.: An internet of things based smart waste management system using LoRa and tensorflow deep learning model. IEEE Access 8, 148793–148811 (2020). https://doi.org/ 10.1109/ACCESS.2020.3016255 17. Raspberry Pi 3: Specs, benchmarks and testing—The MagPi magazine. https://magpi.raspbe rrypi.org/articles/raspberry-pi-3-specs-benchmarks. Accessed 13 October 2020 18. Home page | LoRa Alliance® . https://lora-alliance.org/. Accessed 13 October 2020 19. Ultrasonic Module—Wiki. https://wiki.sunfounder.cc/index.php?title=Ultrasonic_Module Accessed 13 October 2020 20. load cell sensor specifications—Google Search. https://www.google.com/search?sxsrf=ALe Kk00qq30kVM-OHSi85AwIFHZqzin-mA%3A1602576559613&ei=r2CFXBJcb99QPZn YaQDw&q=load+cell+sensor+specifications&oq=load+cell+sensor+sp&gs_lcp=CgZwc3 ktYWIQAxgAMgUIABDJAzIGCAAQFhAeOggIABAWEAoQHlDRKlinMGDWOmgAcA F4AIABjwaIAd0IkgEHMC4yLjYtMZgBAKABAaoBB2d3cy13aXrAAQE&sclient=psy-ab. Accessed 13 October 2020

Predicting the COVID-19 Cases in India Arpit Jain, Abhinav Sharma, and S. Nitisha Bharathi

Abstract COVID-19 has challenged the scientific community to its core and researchers, government and healthcare workers are working persistently to fight the pandemic. Data driven prediction for disease spread is a vital information which assists policymakers to plan their moves against the pandemic. In this paper, we have compared the performance of the two most important time-series forecasting methods: Prophet and ARIMA for predicting the COVID-19 cases in India, considering the effect of disease spread imposed by unlocking regulation of the government. The performance parameters comparison shows that predictions obtained from Prophet are more accurate and may assist the government agencies and health care professionals for effective planning in post lockdown. The prediction model also includes the constraints and factors exhibited by the phased lockdown and unlocked strategies followed by the government of India, the prediction models designed in the initial days of April–July 2020 did not give the accurate predictions for September– October as there had been major changes in the public movement-related policies by the government. The current prediction model holds true as the public movement in India is almost back to normal (excluding the operation of educational institutes of India). Keywords COVID-19 · Prophet · ARIMA · Machine learning · Time-series forecasting

A. Jain · A. Sharma (B) Department of Electrical and Electronics Engineering, University of Petroleum and Energy Studies, Dehradun, India e-mail: [email protected] A. Jain e-mail: [email protected] S. Nitisha Bharathi Department of Applied Mathematics and Computational Sciences, PSG College of Technology, Coimbatore, India e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2021 S. Choudhury et al. (eds.), Intelligent Communication, Control and Devices, Advances in Intelligent Systems and Computing 1341, https://doi.org/10.1007/978-981-16-1510-8_30

299

300

A. Jain et al.

1 Introduction The COVID-19 has hit the globe at a colossal scale. With worldwide reported cases of 4.35 million [1], and declared as a pandemic by World Health Organization, it has led to severe impact on humanity. Being a highly contagious disease, it has put up global health services to its severe challenges [2]. This is high time for the research community around the world to gather up and provide their contribution in fighting the crisis. The healthcare professionals are battling at the frontend by contributing relentlessly, with the biological and healthcare researchers being deeply involved in vaccine/treatment development. This pandemic has united the entire mankind for developing innovative solutions to fight the pandemic. There are numerous ways in which modern technologies can assist mankind in the fight against the pandemic, some of them are: population screening, contact tracing, robots assisting healthcare professionals, minimizing human contact with autonomous systems, speeding up the drug discovery, efficient diagnostics, Chabot for various sectors to name a few. The role of robotic agents in healthcare has been discussed in Yang et al. [3] and the adoption of a robot assisted surgery procedure is given in Kimming et al. [4]. The role of contact tracing and its significance in controlling COVID-19 cases has been studied by Hellewell et al. [5], Marcel et al. [6]. Various countries are fighting to minimize the losses due to the outbreak, however, a common trait is enforcing lockdown, which has become the main defence mechanism for India as well [7]. At the time of writing, India had more than 8,00,000 [1] active cases, amidst a population of 1.3 Billion [8]. This has presented an interesting case of the epidemic fight to the world when the anticipated value of R0 for COVID19 cases lies between 2 and 3 [9]. However, a major sacrifice made to flatten the curve is the Indian economy, which has suffered due to continuous lockdown from March 23, 2020. India reported its first case of COVID-19 on January 30, 2020, with infection origin from China [10]. Tracking the first 50 cases of COVID-19 in India gives us a big picture of state-wise spread and the steps endured by the government to handle the epidemic and their efficiency [11]. On January 30, India reported its first 3 COVID-19 cases in the state of Kerala. This was followed by Delhi, Telangana, Tamil Nadu. The situation became more critical when the state of Rajasthan reported 17 cases (16 Italian tourists along with their 1 Indian driver). On March 5, the national capital region of India (NCR) reported cases in Ghaziabad and Gurgaon. Kerala reported the youngest patient on March 9, a 3-year-old girl which had a travel history to Italy. On March 10, Maharashtra’s Pune reported 2 cases, the first two cases for the state. The first case of third level transmission was discovered in Meerut, a person who came in contact with 6 patients in Agra, who contracted the infection from a patient in Delhi, who had a travel history to Italy. As of June 1, the number of cases has increased by a huge factor. This paper presents a demographic-based detailed analysis of the COVID-19 spread in India. The study is further added by the prediction of the outbreak along with a comparison of various predictive models. The manuscript layout is as follows: Sect. 2 presents the analysis of COVID-19 cases

Predicting the COVID-19 Cases in India

301

in India as of June 1, 2020. Section 3 depicts the prediction of COVID-19 cases and is followed by the conclusion.

1.1 Impact of AI in COVID-19 The word Artificial Intelligence was first coined in the year 1956, at the Dartmouth conference. Since its inception, the technology has experienced exponential growth in different sectors of our day-to-day life. Internet of Things (IoT) is another booming technology which basically collects, sends and acts on data that it collects from the environment. Artificial Intelligence and IoT part of industrial revolution 4.0 have jointly revolutionized health, agriculture and automation industry. In March 2020, smart hospitals were built in Wuhan, China, for COVID-19 patients which use AI and IoT. Hospitals use smart IoT devices for contactless health monitoring of patients and robots for cleaning and sanitizing different areas of hospitals. Advanced data analytics is making a significant contribution in understanding the disease spread and equipping the world government leaders and various officials with deep insight into the trends. Artificial Intelligence or Machine learning algorithms are useful in predicting the spread and with accurate predictions the government can decide the future actions needed to contain the disease. There has been a lot of work in by worldwide programming community, for instance, GitHub Inc. has more than 1 million codes with more than 50,000 active repositories related to COVID-19 or coronavirus. AI technology is useful for fast drug development and treatment [12]. In the starting phase of COVID-19 pandemic, the medical fraternity in China diagnosed the virus using computed tomography (CT) and X-ray images due to the limitation of testing kits. In [13–15] the author proposed deep learning neural network model for COVID-19 diagnosis and concluded the approach can also be used for diagnosing other chest diseases like tuberculosis and pneumonia. AI assisted intelligent humanoid robots can be used to reduce the human contact and spread of COVID-19. In Italy, Tommy named robots have been used for measuring blood pressure, oxygen saturation and temperature of the patients [16]. Robots also find applications in disinfecting and sterilizing public places, COVID-19 testing, food and medicine delivery, as well as entertaining patients in hospitals and quarantine centres, thereby reducing the workload of doctors and nurses. In near future considering this pandemic, robots, self-driving cars will play a crucial role in the tourism industry [17]. Swarm intelligence controlled drones can find applications to monitor the movement of people in a denser city under lockdown and also can be used for sanitizing the affected areas. Natural Language Processing (NLP) based Chatbots can be built to handle queries and provide assistance related to COVID-19 in hospitals, police control rooms and other public sectors, directly and indirectly, involved in this pandemic. WHO and center for disease control and prevention (CDC) have already incorporated Chatbots in their websites to provide information about the spread and symptoms of COVID-19 to millions of users across the world.

302

A. Jain et al.

Tracking of COVID-19 patients and the people who came in contact with infected people in the recent past is a critical step to control the spread of this virus. The passengers that recently arrived in a city or country can be given IoT-based wrist bands which continuously track their movement and monitor that they are following the quarantine rules. IoT enabled smart devices can be used for monitoring the patient’s health and uploading the information on the cloud so that data can be analyzed remotely in a control room and preventive measures can be taken. The main cause of the spread of COVID-19 is physical contact, IoT enabled smart home gadgets like smart cameras, smart security systems, smart speakers and smart air conditioners can prevent humans to come in direct contact with vulnerable surfaces. Smart devices can also be used in hospitals and offices to avoid direct physical contact. In some cases, it is the Contact Volume Editor that checks all the pdfs. In such cases, the authors are not involved in the checking phase.

2 Demographics-Based Analysis Preparation India marked its entry in the list of countries worst hit by COVID-19 on May 25th, 2020 [18]. Figure 1 represents the data for countries reporting a maximum number of COVID-19 cases. The details of COVID-19 cases in India can be seen in Fig. 2. The mortality rate of COVID-19 patients in India is around 1% and reported to be around 4% worldwide. To get a detailed analysis of Indian states affected by the COVID-19,

Fig. 1 Countries with maximum COVID-19 reported cases

Predicting the COVID-19 Cases in India

303

Fig. 2 COVID-19 cases in India

we will compare and analyze the number of cases for ten states reporting the highest number of COVID-19 cases as depicted in Table 1. These states make up more than 80% of the total reported cases in India. Table 1 States having the highest number of COVID-19 reported patients S. no

State

Confirmed cases

Recovered

Recovery rate (per 100)

Active cases

Deaths

Death rate (per 100)

1

Maharashtra

1,384,446

1,088,322

97.32

296,124

37,091

2.68

2

Andhra Pradesh

693,484

629,211

99.16

64,273

5828

0.84

3

Karnataka

601,767

485,268

98.52

116,499

8883

1.48

4

Tamil Nadu

597,602

541,819

98.41

55,783

9520

1.59

5

Uttar Pradesh

399,082

342,415

98.55

56,667

5784

1.45

6

Delhi

279,715

247,446

98.08

32,269

5361

1.92

7

West Bengal

257,049

225,759

98.07

31,290

4958

1.93

8

Odisha

222,734

185,700

99.59

37,034

912

0.41

9

Kerala

196,106

128,224

99.60

67,882

776

0.40

10

Bihar

193,600

163,407

99.41

30,193

1135

0.59

304

A. Jain et al.

Fig. 3 Moving average for states reporting maximum number of COVID-19 cases in India

These states make up 75.5% (4,825,585 of 6,391,960 as of Oct. 1, 2020) of total cases in India. The average death rate exhibited by these states is around 1.3%. Figure 3 depicts the 5-day moving average for these states. The moving average gives us long-term trend for time series. In this case, it can be thought of as a parameter which depicts the surge in the trend of daily cases and it is clearly visible that daily reported cases are increasing, which is a big concern for healthcare officials and the government. Investigating further the reasons behind the surge of cases in these states, the following parameters were primarily responsible: 1 2. 3.

Exposure to international tourists, diplomats or working professionals. Population density in the state. Non-adherence to governments lockdown strictness.

India is a country with a population of 1,387,297,452 (1.3 Billion) which is equivalent to 17.7% of the world’s total population, despite having a mere 2.4% of the world’s land. With a total land area of 2,973,190 Km2 , the population density in India approximates 464 per Km2 . This puts up India already in a critical zone with respect to COVID-19 contamination which can only be controlled by social distancing. Wasdani and Prasad [19] discussed the reality of social distancing for the urban poor in India. The spread of COVID-19 will highly be affected by the population-related

Predicting the COVID-19 Cases in India

305

Table 2 Indian states reporting maximum COVID-19 cases and corresponding population S.

Region

Confirmed Deaths Cured

no 2286

Population area KM2 density /KM2 (2011 census)

1

Maharashtra 67,655

29,329 112,374,333

307,713

2

Tamil Nadu

22,333

173

12,757

7,214,703

130,058

555

3

Delhi

19,844

473

8478

16,787,941

1484

11,297

4

Gujarat

16,779

1038

9919

60,439,692

196,024

308

5

Rajasthan

8831

194

5927

68,548,437

342,239

201

6

Madhya Pradesh

8089

350

4842

72,626,809

308,245

236

7

Uttar Pradesh

7823

213

4709 199,812,341

240,928

828

8

West Bengal

5501

317

2157

91,276,115

88,752

1029

9

Bihar

3815

21

1710 104,099,452

94,163

1102

10

Andhra Pradesh

3679

62

2349

162,968

303

49,577,103

365

parameters. Table 2 depicts the population of these states as per census 2011 and the population density as per the official figures. One of the biggest challenges the government and healthcare professionals will face is the fewer number of health centres for an enormous population.

3 COVID-19 Predictions for India Predicting COVID-19 figures for the Indian population using the most accurate scientific method is very crucial. Artificial intelligence and cloud-based prediction models can be proven beneficial in combating the prediction figures for COVID-19 as accurate datasets are available [1, 20]. There already has been significant research in predicting the COVID-19 cases. Tiwari et al. [21] used the dataset from China to analyze the spread and infer details for the Indian scenario, which seems to be a legitimate method for early prediction models as the Indian dataset in the initial days were not promising enough for prediction. The authors had predicted a number of 100,000 COVID-19 cases in India by April 22, 2020. However, the actual cases reported till April 22, 2020, is 21,370 [22]. This clearly depicts the inaccuracy of the model. Tuli et al. [23] developed a machine learning model to fit a Generalized Inverse Weibull (GIW) Distribution model [24] described as below f (x) = k · γ · β · α β · x −1−β · exp(−γ

 α β x

306 Table 3 Phase wise lockdown enforced due to COVID-19 in India

A. Jain et al. Lockdown

Start date

End date

Duration

Lockdown 1.0 25th March 2020 14th April 2020 21 days Lockdown 2.0 15th April 2020

3rd May 2020

19 days

Lockdown 3.0 4th May 2020

17th May 2020

14 days

Lockdown 4.0 18th May 2020

31st May 2020

14 days

Lockdown 5.0 1st June 2020

30th June 2020

30 days

where α, β, γ are model parameters. The authors used an iterative weight-based machine learning algorithm for curve fitting and predicting the expected number of cases. The authors claimed cases to be at a peak during April 2020. The total cases in India were predicted to be 409,418, and a worldwide total number of cases to 6,734,075. Tomar and Gupta [10] used LSTM architecture (Long Short Term Memory) based RNN (Recurrent Neural Network) for prediction. Authors trained the model using data till April 4, 2020, and predicted the results up to April 9, 2020, with an error range of ±24%. For efficient and accurate time-series forecasting, there have been enormous research in the healthcare industry [25], financial industry [26], telecom industry [27], environment engineering [28]. In this article, we will predict the anticipated cases utilizing Prophet and ARIMA prediction forecasting methods.

3.1 Lockdown and Unlock Timeline in India On 22nd March, India observed its first ‘Janta Curfew’ which is a voluntary lockdown suggested by the Indian Prime Minister. After which the lockdown 1.0 was implemented from 25th March 2020. The timeline for lockdown is given in Table 3. As of, May 17, 2020, the nationwide lockdown in India has been loosened off with a new set of rules and our study reveals the impact of lifting lockdown on the outbreak of COVID-19 in India. The migrant workers throughout the country are now travelling to their native states. On June 1, 2020, there have been further relaxations in lockdown. Due to growing financial strain on the Indian economy [10, 11] and the adverse conditions of migrant workers the Indian government announced unlock [12]. Table 4 depicts the timeline for unlock.

3.2 Prophet Prediction Model The prophet model is a time-series forecasting method based on additive non-linear trends, which cater to seasonality (daily, weekly, and yearly) and holiday effects. For predicting the COVID-19 cases, utilization of these trends is beneficial as the movement of the public depends on these factors and the lockdown and unlock are converted to holiday effects. The prophet forecasting model can be summarized as

Predicting the COVID-19 Cases in India Table 4 Phase wise unlock enforced due to COVID-19 in India

307

Unlock

Start date

End date

Duration

Unlock 1.0

1th June 2020

30th June 2020

30 days

Unlock 2.0

1st July 2020

31st July 2020

31 days

Unlock 3.0

1st August 2020 31st August 2020 31 days

Unlock 4.0

1st September 2020

30th September 2020

30 days

Unlock 5.0

1st October 2020

31st October 2020

31 days

f (t) = g(t) + s(t) + h(t) + t where g(t) represents the trend function, s(t) represents periodic changes, h(t) represents the effects of holidays, and  t represents the error term not accommodated by the model. Trend function: The trend function is a non-linear saturating function that can be described by the following logistic function: g(t) =

C 1 + exp(−k(t − m))

where C is the carrying capacity, k is the growth rate, t is the observation point and m is an offset parameter. The trend function will consist of the contagious parameters and disease spread. Seasonality: The cyclic patterns are described by seasonality. For COVID-19 prediction, these parameters can be taught on weekdays, where there is very less restriction on people’s movement and weekends when a relaxed lockdown is imposed from Friday 19:00 h to Monday 07:00 h. The seasonality function comprises of Fourier series based function:     N   2π nt 2π nt an cos + bn sin s(t) = P P n=1 where P is the regular period, a and b are observed seasonality and N is the number of observations. Holidays and events: These do not follow a periodic pattern and are relatively large changes in times-series. In our case, the relaxation in public movement and unlock parameters will result in an event and will generate a spike in the number of cases. The event function is depicted by the parameter Ki that corresponds to the change in the forecast h(t) = Z (t)ki where, Z (t) = [1(t ∈ D1 ), . . . , 1(t ∈ D L )] Using the prophet model, the forecast of COVID-19 cases is depicted in Fig. 4. The RMSE and R2 for the prediction model using Prophet is 9872.98 and 0.98.

308

A. Jain et al.

Fig. 4 COVID-19 cases prediction with Prophet

3.3 ARIMA Prediction Model Autoregressive integrated moving average models an extension to autoregressive moving average (ARIMA) method and is a good algorithm for analyzing and predicting non-stationary time-series data [25]. ARIMA is the integration of autoregressive (AR), and Moving Average (MA) terms, with p autoregressive terms and q moving average terms. The ARIMA model is depicted by

y = c + t +

p  i=1

∅i yt−1 +

q 

θi t−1

i=1

where t is assumed as an independent and homogeneously distributed random variable, c represents the system constant variable, y is the value of the variable in the temporal moment t. The prediction using the ARIMA model is depicted in Fig. 5. The model exhibited an RMSE and R2 of 8007.15 and 0.93.

4 Conclusion This paper provides the prediction of COVID-19 cases during the unlock period when the government of India has given guidelines for various activities such as mass movement, tourism, migration, tour and travels, etc. In this article, two time-series

Predicting the COVID-19 Cases in India

309

Fig. 5 COVID-19 cases prediction with ARIMA

prediction models: Prophet and ARIMA are explored for COVID-19 case prediction in the post lockdown period. The predictions of ARIMA were close to the actual cases reported. The effect of preventive measures taken by the Indian government has been observed and it clearly indicates a reduction in the spread of the virus. The possibility of herd immunity in near future is not feasible and in order to suppress the number of cases, a strategic lockdown is recommended or the number of cases in India will rise at an unprecedented rate.

References 1. Dong, E., Du H., Gardner, L.: An interactive web-based dashboard to track COVID-19 in real time. Lancet Infect. Diesases 533–534 (2020) 2. Wang, L.-S., Wang, Y.-R., Ye, D.-W., Liu, Q.-Q.: A review of the 2019 Novel Coronavirus (COVID-19) based on current evidence. Int. J. Antimicrob. Agents (2020) 3. Yang, G.-Z., Nelson, B.J., Murphy, R.R., Choset, H., Christense, H., Collins, S.H., Dario, P., Goldberg, K., Ikuta, K., Jacobstein, N., Kragic, D., Taylor, R.H., McNutt, M.: Combating COVID-19—The role of robotics in managing public health and infectious diseases. Sci. Robot. 1–3 (2020) 4. Kimming, R., Verheijen, R.H., Rudnicki, M.: Robot assisted surgery during the COVID-19 pandemic, especially for gynecological cancer: a statement of the Society of European robotic gynaecological surgery (SERGS). J. Gynecol. Oncol. 31(3), 1–7 (2020) 5. Hellewell, J., Abbott, S., Gimma, A., Bosse, N.I., Jarvis, C.I., Russell, T.W., Munday, J.D., Kucharski, A.J., Edmunds, W.J.: Feasibility of controlling COVID-19 outbreaks by isolation of cases and contacts. Lancet Global Health 8(4), 488–496 (2020)

310

A. Jain et al.

6. Marcel, S., Christian, A.L., Richard, N., Silvia, S., Emma, H., Jacques, F., Marcel, Z., Gabriela, S., Manuel, B., Annelies, W.-S., Isabella, E., Matthias, E., Nicola, L.: COVID-19 epidemic in Switzerland: on the importance of testing, contact tracing and isolation. Swiss Med. Weekly 11–12 (2020) 7. Nair, A.G., Gandhi, R.A., Natarajan, S.: Effect of COVID-19 related lockdown on ophthalmic practice and patient care in India: results of a survey. Indian J. Ophthalmol. 68(5), 725–730 (2020) 8. Department of Economic and Social Affairs: Revision of World Population Prospects 2019. United Nations, Department of Economic and Social Affairs (2019) 9. D’Arienzo, M., Coniglio, A.: Assessment of the SARS-CoV-2 basic reproduction number, R0 , based on the early phase of COVID-19 outbreak in Italy. Biosaf. Health 1–3 (2020) 10. Tomar, A., Gupta, N.: Prediction for the spread of COVID-19 in India and effectiveness of preventive measures. Sci. Total Environ. 728 (2020) 11. Rawat, M.: Coronavirus in India: tracking country’s first 50 COVID-19 cases; what numbers tell, 12 March 2020. https://www.indiatoday.in/india/story/coronavirus-in-india-tracking-cou ntry-s-first-50-covid-19-cases-what-numbers-tell-1654468-2020-03-12. Accessed 1 June 2020 12. Vaishya, R., Javaid, M., Haleem Khan, I., Haleem, A.: Artificial intelligence (AI) applications for COVID-19 pandemic. Diabetes Metab. Syndr. Clin. Res. Rev. 14(4), 337–339 (2020) 13. Ozturk, T., Talo, M., Yildirim, E.A., Baloglu, U.B., Yildirim, O., Acharya, R.U.: Automated detection of COVID-19 cases using deep neural networks with X-ray images. Comput. Biol. Med. 121 (2020) 14. Panwar, H., Gupta, P., Siddiqui, M.K., Menendez, R.M., Singh, V.: Application of deep learning for fast detection of COVID-19 in X-rays using nCOVnet. Chaos Solitons Fractals 138 (2020) 15. Ucar, F., Korkmaz, D.: COVIDiagnosis-Net: deep Bayes-Squeezenet based diagnostic of the coronavirus disease 2019 (COVID-19) from X-ray images. Med. Hypotheses 140 (2020) 16. Romero, M.E.: Tommy the robot nurse helps Italian doctors care for COVID-19 patients. The World (2020) 17. Zeng, Z., Chen, P.-J., Lew, A.A.: From high-touch to high-tech: COVID-19 drives robotics adoption. Tour. Geogr. 22(3), 724–734 (2020) 18. Desk, E.W.: Indian Express COVID-19 Tracker: India’s state-wise and global cases, deaths and recoveries. The Indian Express, New Delhi (2020) 19. Wasdani, K.P., Prasad, A.: The impossibility of social distancing among the urban poor: the case of an Indian slum in the times of COVID-19. Local Environ. Int. J. Justice Sustain. 25(5), 414–418 (2020) 20. Ritchie, H.: Coronavirus source data. Our World in Data (2020). https://ourworldindata.org/ coronavirus-source-data. Accessed 2020 21. Tiwari, S., Kumar, S., Guleria, K.: Outbreak trends of coronavirus disease–2019 in India: a prediction. Disaster Med. Publ. Health Prep. 1–6 (2020) 22. Worldometer: India Coronavirus – worldometer. Worldometer (2020). https://www.worldomet ers.info/coronavirus/country/india/. Accessed 5 June 2020 23. Tuli, S., Tuli, S., Tuli, R., Gill, S.S.: Predicting the growth and trend of COVID-19 pandemic using machine learning and cloud computing. Internet of Things 11 (2020) 24. Gusmao, F.R.D., Ortega, E.M., Cordeiro, G.M.: The generalized inverse weibull distribution. Stat. Pap. 52(3), 591–619 (2011) 25. Duarte, D., Faerman, J.: Comparison of Time Series Prediction of Healthcare Emergency Department Indicators with ARIMA and Prophet. AIRCC Publishing Corporation, Dubai (2019) 26. Weytjens, H., Lohmann, E., Kleinsteuber, M.: Cash flow prediction: MLP and LSTM compared to ARIMA and Prophet. Electron. Commer. Res. (2019)

Predicting the COVID-19 Cases in India

311

27. Pinho, A., Costa, R., Silva, H., Furtado, P.: Comparing time series prediction approaches for telecom analysis. In: ITISE: International Conference on Time Series and Forecasting: Theory and Applications of Time Series Analysis. Springer, Spain (2019) 28. Hrnjica, B., Bonacci, O.: Lake level prediction using feed forward and recurrent neural networks. Water Resour. Manag. 33, 2471–2484 (2019)

Experimental Evaluation of the Performance of Diaphragm Spring Clutch of a Four-Stroke Multi-cylinder Petrol Engine Under Dry Friction Conditions Shriram Dravid, Jitendra Yadav, Santosh Kumar Kurre, Adesh Kumar, and Roushan Kumar Abstract It is a known fact that the clutch is a flexible machine element in vehicles and its performance is considerably defined by the friction characteristics of the surfaces. The poor performance of the clutch is unfavourable to the passenger comfort and detrimental to good driving conditions. Fluctuations in engine shaft and load shaft speed may result in non-smooth driving and fatigue failure of components. The present work is intended to investigate the motion characteristics of the engine driveline, i.e. the engine shaft and the load shaft under constant throttle position. An experimental study is performed to evaluate the performance of the diaphragm spring clutch of a four-stroke multi-cylinder petrol engine i.e. Maruti 800. Non-uniform motion of the vehicle’s driveline components under different operating conditions like gear position, throttle and load is studied and the fluctuations at the engine side and load side are recorded to establish the correlation between them for different operating conditions. It is observed that the load shaft speed does not follow the engine shaft speed curve in an identical fashion. This clearly indicates that the clutch is operating continuously in slipping mode. The study proposes to increase the stiffness of the springs for forth gear operations. Keywords Dry clutch · Clutch slip · Lockup · Driveline · Vibration

S. Dravid Mechanial Engineering Department, SD Bansal College, Indore, MP, India J. Yadav · S. K. Kurre (B) · R. Kumar Department of Mechanical Engineering, University of Petroleum and Energy Studies, Dehradun 248007, India e-mail: [email protected] A. Kumar Department of Electrical and Electronics Engineering, University of Petroleum and Energy Studies, Dehradun 248007, India © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2021 S. Choudhury et al. (eds.), Intelligent Communication, Control and Devices, Advances in Intelligent Systems and Computing 1341, https://doi.org/10.1007/978-981-16-1510-8_31

313

314

S. Dravid et al.

1 Introduction In present days, more focus studies are performed to achieve desirable comfort in driving and excellent engagement performance of transmission system, in cars and light motor vehicles. Driver and passenger comfort can be enhanced by reducing gear rattle and body boom noises transmitted into the passenger compartment. The development of automobile engines has been along the lines of constantly increased power, so that the speed of the vehicle increases, which created the problem of vibration. To reduce vibration, flexible engine mountings were introduced. With the introduction of flexible mountings, clutch chatter became a major problem. At the time of disengagement of the clutch, the engine is forced forward due to flexible mountings. As the clutch engagement is started, the engine has a tendency to move backwards to its original position. Continuous engagement and disengagement causes clutch chatter. Clutch chatter also starts occurring when the clutch gets old, resulting in fatigue, wear, and discomfort in driving conditions. The transmission efficiency of the torque converter is enhanced by suppressing engine torque fluctuation to be transmitted to a drive train [1]. Clutch slip occurs when the rotation speed of the output shaft is more than 3% lower than that of the input shaft [2]. The development of clutch systems for passenger cars continues to put more and more focus on the torsional damper design [3]. Automated manual transmission is used in modern vehicle [4]. Bostwick and Szadkowski [5], have reported that, the friction clutch produces judder during the staring engagement of the vehicle. Due to judder, the driving comfort and smooth starting hampers, it also harms the drive train. However, judder is experienced as vibration and noise of the drive train of vehicles. Brassart et al. [6] have analyzed the dynamic behaviour of a tractor driveline with respect to torsional vibrations to improve transmission quality. Steinel [7] reported that the vibration and noise of the axle and power transmission component may be reduced by selecting the appropriate clutch. Duque et al. [8] reported that the torsional vibration of the engine is transferred to the driveline which creates body boom and rattle noise. The torsional damper can be integrated with clutch disc to lower the torsional vibration. Vandenplas et al. [3] studied the driveline torsional vibration of vehicles with the automatic gearbox and found an increase in the fuel consumption. They also observed that the efficiency of the engine was improved if the torque converter disengaged early and bypassed by a lock-up clutch. Doyle, Jr. and Lynn Faulkner [9] have studied the Torsional vibrations of the hydraulic system, which consists of a diesel engine, propeller shaft, spline shaft with clearance, mechanical transmission clutch engagement, universal joints, load and hydraulic pump. Higher torsional vibrations caused in spline shaft, propeller and U-joint failures. In all of these cases, the study was confined to the engagement phase of the clutch only, and it was shown that the engine shaft and load shaft attain a common speed after the engagement phase is over. Mizumoto et al. [10] have studied slip control of the lock up clutch using an adaptive output feedback system. Adachi et al. [11] have observed that the controlled torque capacity of the first clutch make the fast and smooth and fast acceleration. is achieved by controlling the torque capacity of

Experimental Evaluation of the Performance of Diaphragm Spring Clutch …

315

the first clutch using the stroke control system. Mishra et al. [12] have studied the inertia phase control of the clutch using non-linear control theory and design a model. Behrooz Mashadi et al. [13] found that the dry clutch effectively controls the vehicle’s longitudinal vibration of the driveline. Jiang et al. [14] studied the launch control method of the twin clutch transmission vehicle and found that the model predictive control system is effective to control the slip of the clutch and reduces the time of the launching process. Hu et al. [15] designed an observer-based feedback controller to control the slip of automatic transmission. Naruse [16] evaluated the slip loss of the torque converter by using the minimum slip lockup clutch control system and found the improved fuel economy of the vehicle. Yadav et al. [17, 18] also reflected on the stick–slip mode of operation of clutches and its dependency on various system parameters. Yadav et al. [19] suggested the circumvention of stick–slip from the system. The study of these fluctuations is the subject matter of the present work. In the present work, percentage fluctuations are investigated for different operating conditions of driveline components, like gear position, throttle and load.

2 Experimental Setup An experimental setup was developed, around the engine and the driveline of multi cylinder petrol engine of a four wheeler vehicle to transfer the speed of engine and load shaft. The engine used for this purpose was a laboratory-based engine of a four wheeler vehicle (Maruti 800) mounted on a base frame and its wheels have been removed. The wheel shaft of the driveline is supported on brackets fastened onto the base frame. To extend the wheel shaft, we used the wheel studs and tightened a shaft to carry circular discs to provide inertia equivalent load of the vehicle and also the rotary inertia corresponding to the wheels. A round disc is connected to the engine pulley and a small pin is welded to the round disc in the centre to transfer the motion of the engine to the encoder through a flexible PVC coupling. The arrangement for connecting the shaft encoder to the load shaft is shown in Fig. 1. The shaft encoder is mounted on the counter shaft, as shown. The connection between the counter shaft and the encoder shaft is through a rigid PVC tube, in order to accommodate any misalignment. For measurement of load torque, an arrangement is made with the help of angles welded together to form a C shape and a plate is tightened at the lower end of the channel to form a rectangle. A scale and a dial gauge are connected with the help of nut bolts and hangers. This is shown in Fig. 2.

3 Experimental Results As described in the previous section, the experimental set up was used to obtain the speeds of the engine shaft and the load shaft with the clutch engaged and engine

316

S. Dravid et al.

Fig. 1 Load Shaft and Encoder connecting arrangement

running under constant throttle and constant load condition. The observations that were taken for the fourth gear are presented here for different speeds. The version in speeds of engine and load shaft is shown in Fig. 3, 4, 5, and 6. Table 1 shows Statistical Data of 4th Gear at 1.2 kg-m Load Torque. It is palpable from Figs. 3, 4, 5, and 6 that the load shaft speed curve does not follow the engine shaft speed curve in an identical fashion. This clearly indicates that the clutch is operating continuously in slipping mode, i.e. the engine side plates of the clutch do not hold the load side plates tightly enough. It is also noted that the fluctuations in load shaft speed are certainly higher as compared to the fluctuations in the engine shaft speed for lower and medium speed range. However, as soon as speed increases further, then it is seen that for higher speed range, fluctuations of engine shaft increase.

4 Conclusion An important outcome of this work is that there is no direct correlation between the fluctuations of engine shaft and load shaft speeds. Although, provisions are made by the manufacturers to keep these fluctuations as minimum as possible, the constraints

Experimental Evaluation of the Performance of Diaphragm Spring Clutch …

Fig. 2 Measuring Arrangement with Load Shaft and Brake Drum

Fig. 3 Engine and load shaft speeds (4th gear 1.2 kg-m load torque, 1570 RPM)

317

318

Fig. 4 Engine and load shaft speeds (4th gear 1.2 kg-m load torque, 2100 RPM)

Fig. 5 Engine and load shaft speeds (4th gear 1.2 kg-m load torque, 2320 RPM)

S. Dravid et al.

Experimental Evaluation of the Performance of Diaphragm Spring Clutch …

319

Fig. 6 Comparison of engine and load shaft speed fluctuations (4th gear 1.2 kg-m load torque)

Table 1 Statistical data of 4th gear at 1.2 kg-m load torque Engine RPM

Load RPM

Engine fluctuations

Load fluctuations

Percentage slip (%)

1570

1564

20.02

43.72

0.38

2100

2072

34.06

62.54

1.31

2320

2273

63.66

53.25

2.02

imposed by systems sometimes limit the optimum solution. The fluctuation at the two sides of the drive leads to the poor performance of the clutch. The poor performance of the clutch is unfavourable to the passenger comfort and detrimental to good driving conditions. One important observation is that, most of the time the engine runs in fourth gear, therefore, control of slip in fourth gear is important. This may be achieved by providing a stiffer spring in case of the fourth gear. The present work proposes the guideline to deal with the performance of clutch to develop it further.

References 1. Osawa, M., Hibino, R., Yamada, M., Kobo, K., Kobiki, Y.: Application of H ∞ control design to slip control system for torque converter clutch. IFAC Proc. 28, 157–162 (1995). https://doi. org/10.1016/s1474-6670(17)45689-8 2. Noh, D., Kim, S., Kim, Y., Jang, J.: Lubricant flow analysis for effective lubrication of tractor forward/reverse clutch. Heliyon. 3, e00295 (2017). https://doi.org/10.1016/j.heliyon. 2017.e00295

320

S. Dravid et al.

3. Vandenplas, B., Gotoh, K., Dutre, S.: Predictive Analysis for Engine/Driveline Torsional Vibration in Vehicle Conditions using Large Scale Multi Body Model (2003) 4. Sharifzadeh, M., Pisaturo, M., Farnam, A., Senatore, A.: Joint structure for the real-time estimation and control of automotive dry clutch engagement. IFAC-PapersOnLine. 51, 1062–1067 (2018). https://doi.org/10.1016/j.ifacol.2018.09.053 5. Bostwick, C.C., Szadkowski, A.: Self-excited vibrations during engagements of dry friction clutches. SAE Trans. 689–701 (1998) 6. Drouin, B., Goupillon, J.-F., Goupillon, J.-F., Brassart, F., Brassart, F., Gublin, F.: Dynamic modeling of the transmission line of an agricultural tractor. SAE Trans. 189–200 (1991) 7. Steinel, K.: Clutch tuning to optimize noise and vibration behavior in trucks and buses (2000) 8. Duque, E.L., Lemes, D.V., Galvani, S.A., Nigro, F.E.B.: Analyzing the torsional vibration of engines in dynamometer previewing the impacts in clutch disc calibration (2004) 9. Doyle, G.R., Faulkner, L.L.: Torsional vibrations in a mechanical drive (1982) 10. Mizumoto, I., Takagi, T., Yamanaka, K.: Adaptive output feedback control system design for slip control of a lock-up clutch time-delay system. IFAC (2013). https://doi.org/10.3182/201 30703-3-FR-4038.00067 11. Adachi, K., Ashizawa, H., Nomura, S., Ochi, Y.: Development of a clutch control system for a hybrid electric vehicle with one motor and two clutches. IFAC (2011). https://doi.org/10.3182/ 20110828-6-IT-1002.02671 12. Mishra, K.D., Srinivasan, K.: Robust nonlinear control of inertia phase in clutch-to-clutch shifts. IFAC-PapersOnLine. 28, 277–284 (2015). https://doi.org/10.1016/j.ifacol.2015.10.040 13. Mashadi, B., Badrykoohi, M.: Driveline oscillation control by using a dry clutch system. Appl. Math. Model. 39, 6471–6490 (2015). https://doi.org/10.1016/j.apm.2015.01.061 14. Jiang, Z., Liu, Q., Dong, S., Chen, H.: Launch coordination control based on twin-clutch torque distribution for DCT vehicle. IFAC-PapersOnLine. 51, 904–909 (2018). https://doi.org/ 10.1016/j.ifacol.2018.10.086 15. Hu, Y., Tian, L., Gao, B., Chen, H.: Design of observer-based output feedback clutch slip controller for automatic transmission. IFAC (2011). https://doi.org/10.3182/20110828-6-IT1002.03270 16. Naruse, T.: The tribology of a minimum-slip lock-up clutch-control system. Tribol. Int. 27, 25–30 (1994). https://doi.org/10.1016/0301-679X(94)90059-0 17. Yadav, J., Agnihotri, G.: Modeling and simulation of the dynamic response of a generic mechanical linkage for control application under the consideration of nonlinearities imposed by friction. In: Proceedings of the International Conference on Nano-electronics, Circuits & Communication Systems. pp. 319–330 (2017) 18. Yadav, J., Agnihotri, G.: Proposed Critical Damping for a Spring Mass System to Avoid Stick Slip. J. Inst. Eng. Ser. C. 96, 331–335 (2015) 19. Yadav, J., Agnihotri, G.: Circumvention of Friction-Induced Stick-Slip Vibration by Modeling and Simulation. In: Intelligent Communication, Control and Devices. pp. 1687–1695. Springer (2018)

Cognitive Radio-Based Spectrum Allocation Method for Next Generation Cellular Networks Rajarao Manda and B. V. Sudhakar Reddy

Abstract The dynamic spectrum allocation is one of the important methods to meet high spectral efficiency and high data rates in wireless communication systems. This dynamic radio spectrum allocation is becoming a challenging task for present and next generation wireless communication systems due to the lack of available radio resources, time-varying channel conditions, and high-speed data requirements of broadband multimedia users. In this paper, we propose an intelligent Inter-Cell Interference Coordination (ICIC) scheme, which we call as Cognitive Cell-level Resource Allocation (CCRA) scheme for LTE downlink. This scheme allows a base station (e-Node B (eNB)) to use the neighbouring eNB’s physical resource blocks (PRBs) opportunistically. This scheme follows the SFR ICIC method which enhances the cell-edge user SINR by varying the power over the spectrum. In this scheme, the e-Node B autonomously takes the decision on which spectrum to be used within the cell for cell-edge users, so that Inter-Cell Interference (ICI) will not occur to its neighbouring cells. This increases the cell-edge user service rate while maintaining the ICI levels in neighbouring cells at a prescribed value. It is observed that the proposed scheme is outperforming the previous schemes. Keywords Long Term Evolution (LTE) · Orthogonal Frequency Division Multiple Access (OFDMA) · Cognitive radio · SINR · ICI · PRB · ENB · Inter-Cell Interference Coordination (ICIC)

R. Manda (B) Dept. of Electrical and Electronics, UPES, Dehradun, India e-mail: [email protected] B. V. Sudhakar Reddy Dept. of Electronic and Communication Engineering, RGUKT, Nuzvid, India © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2021 S. Choudhury et al. (eds.), Intelligent Communication, Control and Devices, Advances in Intelligent Systems and Computing 1341, https://doi.org/10.1007/978-981-16-1510-8_32

321

322

R. Manda and B. V. Sudhakar Reddy

1 Introduction The 3GPP Long Term Evolution (LTE) and LTE-Advanced are the standards to provide Broadband Wireless Access to the users. These standards are targeted to support 100 Mbps (1Gbps) in downlink and 50 Mbps (500 Mbps) in uplink with a scalable spectral band of 20 to 100 MHz channel bandwidth (by carrier aggregation) with multiple input multiple output (MIMO)–OFDMA technology [1–3]. Due to the continuous increase in the number of mobile users, spectrum allocation is one of the foremost concerns in future generation wireless networks to meet the higher system capacity/cell throughput. From the last few decades, frequency reuse is one of the best concepts used in wireless networks to increase system capacity. In LTE networks, the frequency reuse one proposed in multi cell scenario based on technical requirements and targets. By which an unacceptable Inter-Cell Interference (ICI) will occur, especially at cell edge regions. This ICI causes degradation in cell-edge user signal to interference noise ratio (SINR). Because of this, per user throughput depends on the position of the cell user. To resolve the problem of ICI at the cell edge, users various frequency reuse schemes have been proposed. One is called Reuse Partitioning [4–6], in which the total available bandwidth of the system will be divided into several disjoint parts and each part can be used at different frequency reuse and these schemes achieve different SINRs. In this case, based on the user location (near or far from the base station) and SINR, the spectrum can be allocated from one of the parts which suit well to that user. In LTE, the main issue is cell edge user performance, so according to Reuse Partitioning, the available bandwidth can be divided into two disjoint parts and one can be used at reuse one and the other can be used at reuse three. Fractional Frequency Reuse (FFR) [7–9] and Soft Frequency Reuse (SFR) [10, 11] are the schemes which are recently proposed. FFR is almost similar to the reuse partitioning scheme except the power ratio of transmit power over frequency reuse one to the transmit power over frequency reuse three is less than 1, where in the latter scheme, it is equal to one. So, the special case of FFR for which the power ratio is equal to one is the reuse partitioning scheme. The SFR scheme, which follows reuse three scheme, in this case each cell will get one-third of available bandwidth with high transmit power and the remaining bandwidth in each cell will be used at low transmit power. Therefore, each cell can use an orthogonal or disjoint sub band at high transmit power. Users nearer to the base station, called cell center users (CCUs), will get high SINR as compared to edge users, so these low transmit power sub band can be used to allocate for the transmissions of CCUs. Then the cell edge users (CEUs) can be allocated to high transmit power sub band which is being used at reuse three. In this scheme, each cell will use the entire bandwidth with varying power levels. However, these schemes are static and unable to adapt to the traffic load that leads to spectrum underutilization in one cell and spectrum scarcity arises in one cell. Due to their inability to adapt to the varying traffic loads, there is a scope to adapt the basic SFR and FFR schemes. However, for the case of LTE, the central entity is not present among the eNBs, which is in the previous cellular generations such as BSC in GSM and RNC in HSPA standards. Therefore, to adapt

Cognitive Radio-Based Spectrum Allocation Method …

323

these schemes according to the traffic load, a decentralized scheme is needed. Here the decentralized means the scheme can run in each eNB with the exchange of some signalling information among them. However, this inter-eNB signaling increases the backhaul communication cost and scheduling delay. Therefore, a more autonomous scheme is needed in order to redistribute the resources among the base stations according to the traffic loads. So, the solution for the addressed problem can be, use SFR or FFR as a base-line spectrum allocation method or an ICIC (inter-cellinterference coordination) to increase cell-edge user SINR and use Cognitive radio principles on top of it to mitigate the Spectrum scarcity and Spectrum underutilization problems. The system analysis and problem formulation are given in Sect. 2. The proposed scheme for LTE downlink which uses the CR principle to achieve the higher throughput is given in Sect. 3. It also discusses the performance evaluation and comparison of the proposed scheme. Section 5 concludes the proposed method.

2 System Analysis and Problem Formulation In the cellular networks, it is a usual case that one cell may be highly loaded for some time and adjacent cells may be lightly loaded for the same time. This is due to the random locations of users and they can be concentrated anywhere in the cellular service area that results in “Spectrum scarcity” problem for one cell and Spectrum underutilization for another cell. As we have seen for LTE, the SFR scheme and FFR scheme can be a better option to increase both cell throughput and cell-edge user throughput. As its name “Cognitive Radio” suggests, it should work intelligently and autonomously without disrupting the ongoing transmissions of other radios. It would use the spectrum of others for whom it is originally assigned when the spectrum is vacant. It vacates the spectrum that in use for its transmissions when the primary users of the spectrum suddenly come up and start the transmissions.

2.1 Spectrum Allocation in LTE System Spectrum allocation is the main concern in Frequency Division Multiplexing (FDM) based cellular system. It distributes the available spectrum for cellular system among the adjacent cells so that their transmissions will not interfere. The interference will occur when the same frequency spectrum/channel is used in multiple cells, this is called frequency reuse, and the amount of the interference depends upon the distance between the cells. The number of adjacent cells, whose allocated frequencies are different, is called as cluster size and it depends upon the frequency reuse distance which is the distance between two cells to which the same frequency is allocated. Since the OFDM signal is generated by Fast Fourier transform, each base station

324

R. Manda and B. V. Sudhakar Reddy

will process the entire bandwidth allocated to LTE system for the transmission or reception. Frequency reuse can be adopted in LTE by nullifying some subcarriers’ power by giving zero signals at the same indexed inputs of IFFT block. Frequency reuse can be quantified in form of cluster size, for example, the cluster size 3 for a cellular coverage area means frequency reuse three. The frequency reuse factor is reciprocal of frequency reuse.

2.2 Inter-Cell Interference By using the same frequency channel in multiple cells, one cell’s transmission will interfere with other cell transmissions at the receiver. The receiver receives the signal which is destined to it and also from other co-channel cells which use the same frequency. The latter part of the received signal interferes in decoding of the former which is required, and it results in interference. This interference is called as CoChannel Interference (CCI) or Inter-Cell Interference (ICI). However, the effect of ICI will be considered in terms of Signal to Interference Ratio (SIR). It is the ratio of received power of Signal of Interest (SOI) from own base station (own cell user) at the user position (base station) and the total received interference power from other cells as given in Equation Pi SIRi =  Ncc Pn n=1 n = i where Pi and N cc are the received signal power from ith base station and a number of co-channel cells, respectively. Pn is the received signal power from the co-channel cell base station.

3 Prosed Method Based on Cognitive Spectrum Allocation In this proposed scheme, eNB will command its associated mobile terminals who have lesser path gains to the neighbouring eNBs to measure the interference power from the neighbouring eNBs through sensing and estimate maximum interference free transmit power (MIFTP) (since the location of an adjacent base station is fixed and known, the MIFTP can be estimated) in each group of 12 contiguous subcarriers (equal to the PRB length in the frequency domain) of all useful subcarriers. In return, users will send their estimated MIFTP over each PRB in the frequency domain to the base station. Then the base station will fuse this collection of information to find the appropriate transmission power over each PRB by which a prescribed ICI can

Cognitive Radio-Based Spectrum Allocation Method …

325

be achieved. By doing this, eNB can make a resource chart or map, in which, each PRB will have the transmission power level that is declared by it with the fusion of collected MIFTP information over all PRBs. In this negotiated resource chart, if a PRB’s transmit power level is low, which means that in the adjacent cell that PRB is used at a high transmit power level. Similarly, if a PRB’s transmit power level in the resource chart is high, it means that the corresponding PRB in the adjacent base station cell is used at a low transmit power level. Once the resource chart is obtained by a base station, the base station can use it for user level resource allocation purpose in its own cell until any request comes from other adjacent base stations. In the worst case, the obtained resource chart may have a few number (or zero) of PRBs with high transmission power level. But if, with this resource chart, the base station cannot serve enough number of its cell edge users, then the base station can send a request to the adjacent base station through X2 interface to decrease the power level over some number of PRBs. The estimated MIFTP value over each PRB can be quantized into only two values, ‘low’ and ‘high’. So, we will get the resource chart with PRBs at low or high transmission power levels. Then this scheme will be converted into an opportunistic version of Soft Frequency Reuse (SFR) as if an adjacent cell’s base station is not using high transmission power over some PRBs, and the negotiated resource chart after the fusion process in the current cell will have those PRBs at high transmission level, then the current base station can use those PRBs for cell edge users. An eNB can get any number of PRBs over which it can use high transmit power through sensing and negotiation process. However, the total transmit power of eNBs is fixed (20 (40) Watts for 5 (10) MHz LTE) so this total transmit power restricts the transmit power over these high transmit power PRBs. Let us denote N high as the number of PRBs over which the eNB can use high transmit power, N low is the number of PRBs used at low transmit power and the respective power levels over these PRBs as Phigh and Plow . Since the sum of transmit power over the total number (NPRB) of PRBs is equal to the total transmit power (Ptot ) of the eNB. Ptot = Nhigh · Phigh + Nlow · Plow

(1)

If we choose the power ratio as between high transmit power and low transmit power over PRBs then from Eq. (1) we can find Phigh and Plow as Phigh =

Ptot Nhigh + γ · Nlow

(2)

Plow = γ

Ptot Nhigh + γ · Nlow

(3)

If an eNB can use a high transmit power over all the PRBs, then in Eq. (2) N low will be equal to zero and Phigh will be Ptot NPRB which is equal to the transmit power over PRBs when frequency reuse one is used. So, our scheme is converted

326

R. Manda and B. V. Sudhakar Reddy

to reuse one scheme when the neighbouring base stations are under loaded. Therefore, our proposed scheme can achieve reuse one scheme cell throughput when the neighbouring eNBs are under loaded that is when ICI is at a low level. It may so happen that the negotiated resource chart is inappropriate to serve cell edge users for a given required edge user throughput and the other base stations are not responding to the sent request to decrease power level over some resource blocks (this may be happening in the case where the other adjacent cells also have a higher number of cell edge users), then a fairness feature may be introduced such that each cell’s base station satisfy a required level of its cell edge user throughput. For that purpose, we propose to put a limit on the number of PRBs for each cell to use at high power level. If the estimated resource chart through sensing has only a few number of high transmission power level PRBs which cannot serve the edge users’ arrival traffic for a given required cell edge user throughput, then it counts the number of PRBs over which transmission power can be high and if that number is less than the threshold number of PRBs over which transmission power can be high, it sends the request to other adjacent base stations to decrease the power level over a given number of PRBs. when a base station receives that request, it decreases the transmission power over those PRBs exceeding the threshold number. It takes no action if the number of high power users is within the threshold. Once the resource chart is prepared as per the proposed scheme based on the received CQI, users will be categorized into two sets: one is cell centre user set and another is cell edge user set as a low CQI user can be presumed as a cell edge user and a high CQI user can be presumed as the cell centre user. Then allocate high transmission power level PRBs to cell edge users if present, otherwise allocate to cell centre users if required.

4 Performance Analysis and Simulation Result The performance is evaluated for both the SFR scheme and the proposed CCRA scheme in two scenarios. In the first scenario, a 27-cell structure is considered and the performance is evaluated at the centre cluster of 3 adjacent sectors as shown in Fig. 1. The remaining sectors act as ICI sources to these sectors. In this scenario, a given number of edge users are fixed in each sector of the cluster. Since our CCRA scheme gains the advantage of non-uniform traffic distributions. The number of users per cell is fixed to 24 but the number of cell-edge users is different. The simulation parameters for this first scenario are given in Table 1. In each simulation run, users are uniformly distributed and if a user is in the outer one-third of the cell region, it is considered as cell-edge user. According to the power levels used over PRBs in the neighbouring cells, we determine how many PRBs we can use on high transmit power and this power is determined from Eq. (2). After negotiating the power levels, the CQI is estimated on each PRB for each user as per their received SINR. According to this calculated CQI, throughput is calculated. For example, if the calculated CQI or SINR is 1 dB then QPSK and rate 14 for the 10%

Cognitive Radio-Based Spectrum Allocation Method …

327

Fig. 1 coverage area for 27 cells Table 1 Simulation parameters

Parameter

Value

Carrier frequency

1.9 GHz

Cell layout

27 cells, 3 sectors per cell

Bandwidth per PRB

180 kHz

Number of PRBs

24

Number of reserved PRBs for cell edge

8

Scheduling

Round Robin

Thermal noise PSD

−174 dBm/Hz

UE noise figure

7 dB

Cell radius

1 km

Area of the cell edge region

1/3 rd of cell area

Path loss model

COST231-Hata

Mobile terminal height

1.6 m

eNB height

32 m

Shadowing standard deviation

8 dB

eNB total transmitting power

43 dB

Γ

0.4

Number of active users per sector

24

Traffic model

Full buffer

CQI reporting resolution

1 dB

328

R. Manda and B. V. Sudhakar Reddy

of BER is chosen then the throughput is calculated per Scheduling Block (SB = two consecutive PRBs in time) as ThSB = 12.14 · n · rate(1 − BER)

(4)

where n is the number of bits that can be modulated one symbol that is log2 (M), M is the constellation size. For QPSK n = 2 the considered LTE channel bandwidth to evaluate the performance of the CCRA is 5 MHz which results in total 25 PRBs per sector site. For the sake of simplicity, only 24 PRBs are allowed for the transmission of user data. For comparison purposes the cell edge frequency reuse three is taken, then with the SFR scheme, each sector will get 8 number of PRBs with transmit power large as compared to the transmit power of remain number (16) of PRBs. It is assumed that the time domain scheduler will always select 24 users for frequency domain scheduling in each TTI in other words always 24 (which is equal to the total number of PRBs per sector site) number of users will be present in each sector site. Since our scheme is based upon the opportunistic usage of the other cells’ high power PRBs and it converges to the traditional SFR when the cell edge load is uniform. This is because, in uniform loads, each cell uses the same number of high transmit power PRBs as per their threshold number of PRBs and that threshold is chosen as 8 PRBs which is the number of PRB at which high transmit power can be used in the SFR scheme. To check the performance, a non-uniform cell edge load is considered in the three cells but the number of users per cell is the same, that is, 24. It is evaluated that the average throughput per user in the system when the sector #1 of centre cluster has 16 users at the edge region (8 users at cell centre region), sector #2 and sector #3 has 4 users at the cell edge region (20 users in the cell centre region) for both SFR and CCRA. The resulted CDF plot of the average user throughput in the total three cells is shown in Fig. 2 from which we can observe that with our proposed scheme the overall average user throughput is increased from the conventional SFR scheme. Here we can see that our proposed scheme for this considered traffic load achieves approximately 45 Kbps improvement over the SFR scheme. Figure 3 shows the sum Fig. 2 Average user throughput

Cognitive Radio-Based Spectrum Allocation Method …

329

Fig. 3 Sum of the sector throughput for different number of edge users in each sector

of the considered sectors throughput for a different number of edge users in each sector. Here the total number of users per sector is the same, that is, 24 and the total number of edge users in the three sectors equal to 24. The real axis parameter n in figure is related to the number of edge users in the sectors as: Number of edge users in sector #1 is 24 − 2 (n − 1). Number of edge users in sector #2 is (n − 1). Number of edge users in sector #3 is (n − 1). If we vary n from 1 to 9, the number of edge users in sector #1 changes from 24 to 8 that is when n is equal to 9 each of the sectors will have the same number of edge users that is equal to 8. For the n = 1 situation, the sector #1 has 24 edge users and zero edge users in sectors #1 and #3. Therefore, sectors #2 and #3 which are neighbours to sector #1 uses low power over the PRBs, thereby according to our scheme, sector #1 can use high transmit power over the PRBs to serve all its 24 edge users whereby increased sector throughput. Figure 3 shows all possible values of n, i.e. wherever the number of edge users is unequal in the sectors, the proposed scheme is achieving higher throughput. One can also observe that for n = 9, where each sector uses the same number of PRBs at high transmit power, our scheme here is converted to the SFR scheme and results in the same throughput as SFR. For the second scenario for which the proposed scheme is again evaluated, in this case, 57 sectors in the simulated service area, one-third of them are fully loaded and the remaining sectors are half loaded. These one-third sectors are randomly chosen among the 57 sectors for each simulation run. Then the user throughputs and sector throughputs are evaluated as per the evaluation method given in that section for both the SFR scheme and CCRA scheme. The resulted average sector throughput CDF curves are shown in Fig. 4. It is observed from this figure that our proposed CCRA scheme is outperforming the SFR scheme and it is giving approximately 130 Kbps improvement in the sector throughput among these highly loaded sectors, now the percentage of edge users is varied from 33 to 100%. The effect on the average sector

330

R. Manda and B. V. Sudhakar Reddy

Fig. 4 Average sector throughput CDF curves

throughput and the edge user blocking probability is shown in Fig. 5 and Fig. 6, respectively. It is observed that the proposed CCRA scheme is outperforming the SFR scheme for all the considered edge user distributions. The improvement in the sector throughput with the CCRA scheme is high as the percentage of edge users increases. This is because, with the increase in the number of edge users in the cell, SFR cannot serve those increased edge traffic, whereas our proposed scheme which uses the high transmit power PRBs of the neighbouring lightly loaded cells can serve

Fig. 5 Average sector throughput vs percentage of edge users

Cognitive Radio-Based Spectrum Allocation Method …

331

Fig. 6 Edge user blocking probability vs percentage of edge users

the exceeded edge user traffic. This result in more traffic is served and so more sector throughput is achieved. Since eNBs are not connected to a centralized body (such as RNC in WCDMA) in LTE, it is difficult to allocate resources at Cell-level as per traffic load. In our proposed scheme, a cell’s base station can attain the resources autonomously by measuring the interference from the base station of adjacent cells and estimating MIFTP. Through this scheme, a highly loaded cell can get sufficient resources from its lightly loaded neighbour cells without going to co-ordinate with them. With the proposed scheme, the temporal variations in the arrival traffic load could not create ‘spectrum scarcity’ and ‘spectrum underutilization’ problems. The proposed scheme allows the high transmission power level PRBs opportunistically by a base station so increased spectral efficiency can be obtained. The problem with SFR or FFR that, as the ratio of cell edge users and cell centre user’s increases, then the overall throughput will go down because the allocated bandwidth for cell edge users is fixed. But with our proposed scheme, that we have shown in results, the overall throughput is high for our scheme as compared to the SFR scheme. And also, the service rates for cells can be increased by our proposed scheme. The simulation parameters given in Table 1 are considered while doing the simulations.

332

R. Manda and B. V. Sudhakar Reddy

5 Conclusion The performances of the ICIC schemes such as SFR and FFR are studied and analyzed. Then it is concluded that both the schemes are performing well with different trade-offs between cell throughput and cell-edge throughput. However, due to time-varying traffic loads in each cell, there was a need for adaptive versions of these schemes. Due to the absence of RNC in the LTE, it required only decentralized adaptive schemes. On the other hand, it is observed that Cognitive Radio (CR) principles can take decisions autonomously on the spectrum usage without causing interference to the others. In order to adopt decentralized spectrum allocation methods for LTE, CR can be a better possible option to incorporate into the LTE spectrum allocation methods. It is observed that this scheme provides increased sector throughput as compared to the SFR scheme when the traffic is unevenly distributed among the cells. It is seen that the proposed scheme improved the sector throughput by 130 Kbps in comparison to the SFR scheme. In the proposed scheme for uplink, each cell-edge user communicates the eNBs cell IDs in its interfering zone to the serving eNB then eNB allocates such PRB to the user which is not in use at those eNBs. By doing this, the neighbouring cell’s allocated PRBs can be used opportunistically whereby the edge user blocking rates can be reduced. It is observed that the proposed scheme reduced the edge user blocking probability drastically as compared to the FFR scheme and also it increased the edge user spectral efficiency by approximately 22%.

References 1. 3GPP TR 25.913, Requirements for Evolved UTRA (EUTRA) and Evolved UTRAN (EUTRAN), v.8.0.0 (December 2008) 2. 3gpp tr 25.892v6.0.0, 3GPP; technical speciation group radio access network: feasibility study for orthogonal frequency division multiplexing (OFDM) for utran enhancement (release 6) (2004) 3. Martin-Sacristan, D., Monserrat, J. F., Cabrejas-Penuelas, J., Calabuig, D., Garrigas, S., Cardona, N.: On the way towards fourth-generation Mobile: 3GPP LTE and LTE-Advanced. EURASIP J. Wirel. Commun. Netw. 1, 1 (2009). https://doi.org/10.1155/2009/354089 4. Halpern, S.W.: Reuse partitioning in cellular systems. In: 33rd IEEE Vehicular Technology Conference, Toronto, ON, Canada, pp. 322–327 (1983). https://doi.org/10.1109/VTC.1983. 1623156 5. Chong, P.H.J., Leung, C.: Capacity improvement in cellular systems with reuse partitioning. J. Commun. Netw. 3(3), 1–8 (2001). https://doi.org/10.1109/JCN.2001.6596800 6. Chen, S.L., Chong, P.H.J.: Dynamic channel assignment with flexible reuse partitioning in cellular systems. In: 2004 IEEE International Conference on Communications (IEEE Cat. No.04CH37577), Paris, France, vol. 7, pp. 4275–4279 (2004). https://doi.org/10.1109/ICC. 2004.1313354 7. Lei, H., Zhang, L., Zhang, X., Yang, D.: A novel multi-cell OFDMA system structure using fractional frequency reuse. In: 2007 IEEE 18th International Symposium on Personal, Indoor and Mobile Radio Communications, Athens, pp. 1–5 (2007). https://doi.org/10.1109/PIMRC. 2007.4394228

Cognitive Radio-Based Spectrum Allocation Method …

333

8. Bilios, D., Bouras, C., Kokkinos, V., Papazois, A., Tseliou, G.: Optimization of fractional frequency reuse in long term evolution networks. In: 2012 IEEE Wireless Communications and Networking Conference (WCNC), Shanghai, pp. 1853–1857 (2012). https://doi.org/10. 1109/WCNC.2012.6214087 9. Saquib, N., Hossain, E., Kim, D.I.: Fractional frequency reuse for interference management in LTE-advanced hetnets. IEEE Wirel. Commun. 20(2), 113–122 (2013). https://doi.org/10.1109/ MWC.2013.6507402 10. Yu, Y., Dutkiewicz, E., Huang, X., Mueck, M., Fang, G.: Performance analysis of soft frequency reuse for inter-cell interference coordination in LTE networks. In: 2010 10th International Symposium on Communications and Information Technologies, Tokyo, pp. 504–509 (2010). https://doi.org/10.1109/ISCIT.2010.5665044 11. Giambene, G., Yahiya, T.A.: LTE planning for soft frequency reuse. In: 2013 IFIP Wireless Days (WD), Valencia, pp. 1–7 (2013). https://doi.org/10.1109/WD.2013.6686468

Investigation of Analog Parameters and Miller Capacitance Affecting the Circuit Performance of Double Gate Tunnel Field Effect Transistors Deepak Kumar, Shiromani Balmukund Rahi, and Piyush Kuchhal

Abstract TCAD Simulations for 30 nm double gate tunnel field effect transistor (DGTFET) reports steeper subthreshold swing, SS ~ 15 mV/dec, I ON ~ 10–4 A/μm, and low off-state current I OFF ~ 10−15 A/μm as desirable parameters for low voltage applications. The unity gain frequency (f T ) increases with V gs and maximizes at 5.2 × 1011 Hz for V gs = V ds = 0.7 V. It is investigated that the gain-bandwidth product (GBP) also increase with Vgs and maximized at 2.63 × 1011 Hz for V ds = 0.7 V at V gs = 0.6 V. Transconductance frequency product (TFP) increases initially with V gs (0–0.7 V) and maximizes at 4.46 × 1011 Hz/V for V ds = 0.7 V. Higher value of V ds results in better response time of the DGTFETs, i.e., increasing V ds from 0.1 to 0.8 V, the transit time (t r ) of the electron decreases from 4 to 0.1 ps resulting faster switching operation. Transient performance of DGTFETs reports that at supply voltage (V DD ) = 0.7 V, increasing the load capacitance (C L , 10–200 pF) the total delay increases from 0.18 to 1.9 ns. It is also noticed that the % peak voltage overshoot (% V p ) decreases from 42.8 to 2.14% due to decrease in computed values of miller capacitance (CMIL ) from 11.27 to 4.32 fF. Maintaining C L = 15 fF, increasing V DD reports significant variation in voltage peak overshoot from 35 to 26.25% and total delay also decreases from 8 to 0.2 ns for V DD = 0.1–0.8 V. Keywords Band-to-band tunneling (BTBT) · Analog · Transient · TFET · Miller capacitance · Verilog A model · Symica D D. Kumar (B) Department of Electrical and Electronics Engineering, University of Petroleum and Energy Studies, Dehradun 248007, Uttarakhand, India e-mail: [email protected] S. B. Rahi Department of Electrical Engineering, Indian Institute of Technology Kanpur, Kanpur 208016, India e-mail: [email protected] P. Kuchhal Department of Applied Sciences, University of Petroleum and Energy Studies, 248007 Dehradun, India e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2021 S. Choudhury et al. (eds.), Intelligent Communication, Control and Devices, Advances in Intelligent Systems and Computing 1341, https://doi.org/10.1007/978-981-16-1510-8_33

335

336

D. Kumar et al.

1 Introduction The transistor density is continuously increasing with scaling conventional— MOSFETs which results in huge power dissipation inside integrated chips (ICs). The aggressive scaling of supply voltage (V DD ) for the CMOS devices which further reduces the dynamic power dissipation (~CL × VDD 2 ) and same is low-power applications for advance technology nodes [1–4]. The threshold voltage (V th ) is needed to be scaled down proportionally as the supply voltage (V DD ) along with dimensional scaling to achieve sufficient on current (I ON ) in order to maintain satisfactory circuit and device performance. But Scaling down V th further increases off-state current (I OFF ) significantly, hence increases static leakage power dissipation (~V DD × I OFF ) irrespective of scaled down V DD and these impacts can be understood by the limitation of subthreshold swing (SS ~ 60 mV/decade) in MOSFETs due to conventional thermionic emission of electrons from source to channel. Therefore, it becomes essential to explore other structures of (FETs), those that have the potential to function at low supply voltage (V DD ), hence minimize the switching power. To meet these expectations, a new semiconductor device architecture based on quantum transport mechanism (QTM), called Tunneling Field Effect Transistor (TFET) is studied nowadays [5–16]. TFETs can be considered as a potential device structure for ultralow power applications due to their band-to-band (B2B) tunneling transport phenomenon [17–25]. Furthermore, the TFETs possess high-κ dielectric for deposition over the gate dimensions enhances the tunneling Electric field strength across junction formed across the body and the channel. Equation 1 shows that SS (∂V GS /∂logI DS ) of the TFET consist of different parameter as compared to a MOSFET [26–30]. 

1 d Veff E + b dE SS = ln ln(10) + Veff d Vgs E 2 d Vgs

−1 (1)

The Eq. 1 shows that SS is not limited by kT/q; can achieve a steep SS for lower gate voltages as compared to MOSFETs. The electric field (E) across the sourcechannel junction is another key parameter to reduce the SS (< SS ~ 60 mV/dec) which further tends to reduce the subthreshold leakage power dissipation (E L ) in circuits as given by Eq. 2 [31, 32]. Due to its B2BT electron tunneling injection mechanism, a TFET has the potential to achieve lower SS particularly for low-power applications and have ability to cross the barrier of scaling limitations imposed by MOSFETs [30, 33–38]. 2 · 10 E L ∝ VDD

−VDD SS

(2)

A 30 nm gate length double gate tunnel field effect transistor (DGFET) is simulated to achieve I–V and C–V characteristics in the Sentaurus 2D simulator. The features of the system are further used to evaluate the analog and transient (digital) output of DGTFETs by using Symica DE [39–45] to integrate a Verilog A model-based look-up table. The suitability of DGTFETs for analog applications is also studied by

Investigation of Analog Parameters and Miller Capacitance …

337

comparing the analog output at different supply voltages (V DD ) of Double Gate (DG) n-channel TFETs. The analog parameters such as transconductance (gm ), gain bandwidth product (GBP), transit time (t r ), transconductance frequency product (TFP), and unity gain cut off frequency f T [46] are also reported as Si DGTFET performance. In DG TFET, the source and drain region’s charges are coupled due to B2BT process leads to significant increase in Cgd induces of miller capacitance at drain’s node. The higher miller capacitance (C MIL ) may have an impact on the peak voltage overshoots during transient performance [47, 48]. In order to analyze the effect of V DD and load capacitance (C L ) on transient parameters and miller capacitance, digital simulations of DGTFETs based inverters are also reported.

2 Setup of Device Simulations for DGTFET A DGTFET device structure is shown in Fig. 1 with the dimensions of the gate length L g = 30 nm, body thickness t si = 7.0 nm, gate oxide thickness t ox = 1.0 nm; body dielectric εsi = 11.8 εo , oxide dielectric εox = 21.0 εo (HfO2 ). For NTFET, the source is uniformly doped with Boron (B) of 1.0 × 1020 cm−3 , doping of Phosphorus (P) of 1.0 × 1020 cm−3 is done at the drain region and B of 1.0 × 1016 cm−3 is doped in the channel region, respectively. Similarly for PTFET, the source is uniformly doped with P of 1.0 × 1020 cm−3 , doping of B of 1.0 × 1020 cm−3 is done at the drain region and B of 1.0 × 1016 cm−3 is doped in the channel region, respectively. TCAD simulations are performed in order to carry out the electrical characteristic (I–V /C–V ) by considering dc and ac signal at 5 MHz) analysis. Non local B2BT model, Fermi statistics, SRH recombination, Poisson’s, and continuity equations are coded in the simulation script to achieve device simulation [49]. Simulations also considered gate metal work function and the width of the device of 4.2 eV and 1.0 μm, respectively. Fig. 1 Structure of DGTFET used for TCAD simulation Source

gate oxide

tox

Channel

Drain

gate oxide

Lg

tsi

Metal contact

338

D. Kumar et al.

3 Results and Discussion 3.1 Analysis of I–V/C-V Characteristics of DGFET Here in this structure, the intensity of electric field below the gate dielectric is enhanced by applying HfO2 as a gate dielectric which further results in enhancing the tunneling probability (T p) of charge carriers as given by Eq. 3, and the same can be visualized from Fig. 2c; where λ is screen length, φ is the tunneling barrier potential, E g is energy band gap of silicon and m* is the electron’s effective mass [50]. 

 √  1.5 2m ∗ 4λ  . Eg Tp ∼ exp − .  3 h φ + E g

(3)

To understand the electrical characteristics of N type DGTFET, it is important to analyze its energy band diagram for the investigation of charge transport process. Figure 2a depicts the transfer characteristics (I ds –V gs ). The insignificant variations in I–V characteristics are noticed for V ds (0.1–0.7 V) at V gs (0.1 < V gs < 0.3 V). It is also observed that I DS increases with increasing V ds for higher values of V gs . For V gs = 0.3–0.7 V, a noticeable upward shift in the transfer characteristics is noticed for V ds between 0.1 and 0.3 V and after a further increase in V ds (>0.3 V), Ids shows saturation behavior. Figure 2a depicts that DG TFET possesses very low off current (I OFF ~ 5.0 × 10–14 A/μm) which shows the capability to reduce power consumption (~V DD × I OFF ) in standby mode for electronic circuits [51]. Simulation results indicate that this structure exhibits a steep subthreshold swing (SS ~ 15 mV/dec), higher on-state current (I ON ~ 0.1 mA/μm), and higher switching current ratio (I ON /I OFF ~ 5 × 1010 ).

Fig. 2 a Transfer characteristics (I ds –V gs ) of DG NTFET, b Output characteristics (I ds –V ds ) of DG NTFET, c the energy band diagram of a simulated DG NTFET at V ds of 0.7 V showing tunnel barrier modulation

Investigation of Analog Parameters and Miller Capacitance …

339

The current conduction phenomenon can be understood from the TCAD simulations of the energy band diagram as shown in Fig. 2c, which shows the off state (V gs = 0.0 V, I ds ~ 5 × 10–14 A/μm) and the on-state (V gs = 0.2 V, I ds ~ 10–7 A/μm). Output characteristics of NDGTFET are shown in Fig. 2b which shows the saturation point for I ds at V ds > 0.4 V and for V ds > 0.7 V for V gs 0.5 and 0.7 V and higher output impedance can be also predicted for these V ds ranges. In the TFETs, the pinch is shifting to higher values of V ds for higher values of V gs as depicted in Fig. 2b. It is clearly seen that the conduction band of the channel region is not overlapping with the valence band of the source (P+) region (0.0 V < V gs < 0.2 V) results in a wide tunneling barrier across the source-channel region as shown in Fig. 2c. In this situation, practical device current is very low and at this point, the device current is known as off-current (I OFF ). A sharp energy band bending is noticed at the source–channel interface for V gs changes from 0.2 to 0.7 V, which further lowers the tunneling barrier across the source-channel interface, hence increases the T p so that electrons get transported from the valence band of the source to the conduction band of the channel. The transportation of carriers takes place under the effect of B2BT. For a particular drain voltage, the higher the V gs (higher will be the electric field), the lower will be the tunneling barrier across the source-channel region and higher will be the electron tunneling probability resulting in higher tunneling drain current, Ids (see Eq. 3 and Fig. 2c). The C–V characteristics for NDGTFET show that the total gate capacitance (C gg ) is closely followed by drain capacitance (C gd ) for all the values of V gs as shown in Fig. 3 for V ds ranges, 0.1–0.7 V. This can be understood as charges of source-drain regions are directly coupled and governed by B2BT charge transport mechanism in Tunnel FET [23]. Source side tunnel barrier promotes the lower values of C gs (see Figs. 2c and 3) because there are insufficient minority carriers provided by the source, as the source junction is under reverse bias [52–54]. C gd increases for V gs due to reduction in tunneling barrier. Increasing the V ds , a small amount of voltage drop occurs across drain to channel region resulting in lower values of C gd as compared to C gs . Higher the V ds , higher will be the voltage drop across the drain interface results lower values of C gd as clearly shown in Fig. 3. Higher drain voltage results in higher threshold voltage which further reduces energy barrier across the drain side (shifting of the conduction band of drain towards higher energy levels). Further, the C–V curves are utilized to investigate and analyze the analog behavior of TFET and the same has reported, additionally, the transient simulations are carried out to analyze the impact of V DD and C L over the miller capacitance (C MIL ) in order to investigate the performance of DGTFET inverter.

3.2 Analysis of Analog Performance of DGTFET As for better amplification action, the FET device should be highly sensitive to the input signal variations and this feature of any amplifier is characterized by a small signal parameter called transconductance (gm = ∂I DS /∂V GS ) which determines the

340

D. Kumar et al.

Fig. 3 C–V Characteristics of TFET at various drain to source voltage (V ds )

gain of the device required for designing of the analog amplifiers. The steep SS of DG TFET leads to higher values of gm are achieved. Figure 4 shows the gm for different values of V ds (0.1–0.7 V). It is also seen from Table 1 that increasing the Vds the ratio (r = gm2 /gm1 ) is also increases significantly from 2.33 to 7 for V ds variations (0.1–0.7 V), respectively. From Fig. 4, it is observed that at the lowest value of V ds = 0.1 V, the gm is not showing the sensitive behavior for V gs , i.e., gm increases only 2.33 times (from 0.75 × 10–4 (s) to 1.75 × 10–4 (s)) for V gs 0.4 V to 0.7 V, respectively, as listed in Table 1. It can be seen from table gm is more sensitive to V gs for higher values of V ds such that r = 6 at V ds = 0.5 V and r = 7 at V ds = 0.7 V, respectively, and this property is suitable for amplification action for FET devices in order to achieve the higher voltage gain. Another critical parameter for amplification action is short circuit unity gain frequency (f T ) and transit time (t r ) which play a significant role in the analysis of frequency response and carrier transportation time, respectively. High frequency performance of analog circuits is explained in terms of unity gain frequency (f T ) as given by Eq. 4. Figure 5a shows the variation of cut off frequency (f T ) with V gs of

Investigation of Analog Parameters and Miller Capacitance …

341

Fig. 4 Impact of V ds voltage on gm for DGFET performance Table 1 Listing the Impact of Vds on the gm V ds (V)

gm1 (S) at V gs = 0.4 V)

gm2 (S) at (V gs = 0.7 V)

r = gm2 /gm1

0.1

0.75 ×

1.75 ×

2.33

0.3

2 × 10–4

7 × 10–4

3.5

0.5

2 × 10–4

1.2 × 10–3

6

0.7

2×

1.4 ×

7

10–4

10–4

10–4

10–3

Fig. 5 a Impact of drain to source voltage on cut off frequency (f T ) and b impact of drain to source voltage transit time (t r )

342

D. Kumar et al.

DGTFET device for various drain to source voltages. fT =

gm   2π Cgs + Cgd

(4)

1 20 · π · f T

(5)

tr =

As given in Eq. 4, the combined effect of different gm and capacitances (C gs , C gd ) can be observed at cut off frequency (f T ). As seen from Fig. 5a, due to the higher gm value, f T initially increases with increasing V GS and reaches its peak value at 0.1 × 1011 Hz to 5.2 × 1011 Hz for V ds = 0.1 V to 0.7 V, respectively. Afterward, C gd for rising V gs increases at a faster pace than gm. For low-power applications, the greater value of fT makes DG TFET attractive. The frequency of unit gain (f T ) is also capable of investigating charge transport time in the FET and can be measured in terms of transit time (t r ). The t r is the time spent in transporting the carriers from the source to drain region and Eq. (5) has provided the same. Figure 5b demonstrates the transit time variance (t r ) with varying gate to source voltage (V gs ) for the different drain to source voltage to investigate the effect of drain bias. As the inversion layer is substantially increased as the carriers now travel along a shorter path through the inversion layer and this effect can be clearly depicted from Fig. 5b showing t r starts reducing with increasing V gs . For a particular value of V gs (0 < V gs < 0.75 V), the lower values of transit time (4– 0.1 ps) are noticed at the values of V ds = 0.1–0.7 V at V gs = 0.75 V, respectively, as clearly depicted from Fig. 5b. Higher drain voltage results in a strong electric field in the vicinity of the drain and channel region which efficiently swept out the electrons from channel to drain. The tr also characterizes the switching ability of the FET devices and for fast switching operations, lower values of the transit time are preferred. Hence, it can be estimated as better switching ability of TFETs at higher drain voltages. One of the major characteristics of the amplifiers for satisfactory operation is higher bandwidth. GBP is a parameter which shows how much the amplifier is capable to amplify the low frequency message signal and high frequency message signals. Figure 6a, b shows the impact of V ds versus V gs on GBP and TFP parameter, respectively, as given by Eq. 6 and 7. GBP =

gm ; 20 · Cgd

(6)

TFP =

gm ∗ fT Ids

(7)

It is another crucial investigation of DGFET performance for higher frequency application. From Fig. 6, it is clearly observed that GBP increases initially with Vgs because of increasing behavior of gm and achieves maximum values of 1.98 × 109 Hz, 1.14 × 1010 Hz, 1 × 1011 Hz, and 2.63 × 1011 Hz for different values of V ds

Investigation of Analog Parameters and Miller Capacitance …

343

Fig. 6 a Impact of V ds on GBP; b impact of V ds on TFP

= 0.1 V, 0.3 V, 0.5 V, and 0.7 V, respectively. TFP is also following the same trend such that it saturates at its maximum value of 4.31 × 1010 Hz/V, 3.42 × 1011 Hz/V, 3.81 × 1011 Hz/V, and 4.46 × 1012 Hz/V for different values of V ds for 0.1 V– 0.7 V, respectively (see Fig. 6b). The driving point behind the GBP variations is the combined effect of gm and C gd , i.e., for a particular range of V gs , gm is dominated by C gd and afterward C gd starts dominating over gm (see Figs. 2 and 4) resulting in peak values in GBP. It is observed that as V ds increases, the peaks of the GBP curves start shifting toward the higher values of V gs ; GBP peak at V gs = 0.49 V for V ds = 0.5 V and GBP peak V gs = 0.6 V for V ds = 0.7 V as depicted from Fig. 6a. TFP for V ds = 0.7 V starts falling for V gs > 0.6 V due to falling in fT (see Fig. 5a).

3.3 Transient Performance of TFET to Analyze the Impact on Miller Capacitance (CMIL ) In this section, the impact of supply voltage (V DD ) and load capacitance (C L ) on the Miller capacitance (C MIL ) which is formed by B2BT tunneling has been investigated. Several parameters for digital applications are computed and listed for inverter based on DG TFET. To carry out the transient analysis, a 4 ns time period digital pulse (V in ) is applied at the input of DG TFET inverter and output digital waveform is measured at node V out . Figure 7 has shown the Vout for various C L in order to evaluate the inverter performance in terms of rise time (T r ), fall time (T f ), delay (t), peak overshoot (V p ), % peak overshoot, miller capacitance (C MIL ). It is clearly seen from Table 2 and Fig. 7 that the % peak overshoots are diminishing from 42.8 to 2.14% and total delay increasing (0.18–1.9 ns) with increasing the CL (10–200 fF). This behavior can be explained by taking the impact of miller capacitance as given by Eqs (8 and 9). The extent of the voltage peak overshoot can be estimated from the

344

D. Kumar et al.

Fig. 7 Input and output waveforms of double gate TFET inverter for various capacitive load (C L )

Table 2 Impact of load capacitance (C L ) on the transient performance of DGTFET V DD 0.7 V

C L (fF)

Rise time T r (ns)

Fall time T f (ns) Delay (ns)

V p (V)

1

10

2

50

3

(%V p )

C MIL (fF)

0.2

0.16

0.18

0.60

0.55

0.575

0.30

42.80

11.27

0.08

11.42

100

1.2

1.0

6.45

1.1

0.04

5.71

6.06

4

150

1.4

5

200

2

1.5

1.45

0.025

3.57

5.55

1.8

1.9

0.015

2.14

4.32

following equations based on the law of charge conservation (Eqs. 6 and 7) [55, 56]. CMIL · (VM − VDD ) + CL · VM = VDD · (CMIL + CL )

(8)

Vp = VM − VDD = CMIL · VDD /(CMIL + CL )

(9)

where C MIL denotes the Miller capacitance across drain and gate nodes consisting of both comprising inverter structure with N DGTFET and P DGTFET, C L represents the load capacitance connected externally, maximum output voltage (V out ) represented by V M , V p is the output peak overshoot voltage, and V DD is the supply voltage. In silicon-based TFETs, these equations clearly demonstrate the effect of C MIL on V p . Increasing load capacitance (C L ) raises the time constant (RC ), and thus increases the drain node’s charge and discharge time in accordance with Eq. 10. This

Investigation of Analog Parameters and Miller Capacitance …

345

effect can be validated in Fig. 7 and the parameters are shown in Table 2, showing that the delay increases with the CL .   V = Vo 1 − e−t/RC L

(10)

At the same instant, it is also observed that voltage peak overshoot occurs in the transient performance of the TFET inverter for each capacitance load. Figure 7 shows the diminishing voltage peak overshoots with an increasing load capacitance. The % peak overshoots are directly proportional to C MIL . Table 2 has listed the values of C MIL for various C L . The charge conservation equations (Eqs. 8 and 9) are applied to compute the value of C MIL (listed in Table 2) and it is found that the miller capacitance (C MIL ) decreases (11.27–4.32 fF) with increasing C L (10–200 fF), hence results in diminishing % peak overshoots. From the detailed discussion, it can be understood that to diminish the % peak overshoots, the higher values of load capacitance need to be considered so that impact of C MIL gets diminished in FTETs. Figure 8 shows the impact of V DD on the Miller Capacitance (C MIL ). The transient performance of DGTFET inverter is listed in Table 3 has listed its performance parameters. Upon increasing the V DD , the delay of the inverter decreases, and shows saturation in delay for V DD from 0.6 to 0.8 V. Increasing V DD results in higher drain current, but for higher drain voltages, the drain current gets saturated as shown in Fig. 2a. From Table 3, it is clearly seen that the delay has decreased by 50% (from 0.8 to 0.4 ns) for VDD from 0.5 to 0.6 V, 37.5% (from 0.4 to 0.25 ns) for V DD from 0.6 to 0.7 V, and by 20% (0.25–0.2 ns) for V DD from 0.7 to 0.8 V, respectively. It is also observed that a 50% decrease in delay for the VDD from 0.5 to 0.6 V and only

Fig. 8 Input and output waveforms of double gate TFET inverter at various drain to source voltage (V ds )

346

D. Kumar et al.

Table 3 Impact of V DD on the transient performance of DGTFET CL = 15 fF

V DD (V DD = V in )

Rise time T r (ns)

Fall time T f (ns)

Delay (ns)

V p (V)

(%V p )

C MIL (fF)

1

0.5

0.8

0.8

0.8

0.175

35

8

2

0.6

0.4

0.4

0.4

0.2

33

7.5

3

0.7

0.25

0.25

0.25

0.22

31.4

6.88

4

0.8

0.2

0.2

0.2

0.21

26.25

5.35

20% decrease in delay for V DD from 0.7 to 0.8 V due to saturation of drain current at higher drain voltages. An opposite trend is observed for the % peak overshoot; % V p does not show significant variation for V DD from 0.5 to 0.6 V, it only changes from 35 to 33%. However, for the V DD range from 0.7 to 0.8 V, there is a significant variation in % V p are noticed during the simulations from 31.4 to 26.25%. Simulations results also investigated that miller capacitance (C MIL ) continuously decreasing from 8 to 5.35 fF for increasing the V DD from 0.5 to 0.8 V and this impact is also noticed in the % peak voltage overshoots from 35 to 26.25% which further results in better delay performance of TFET inverter (see Fig. 8 and Table 3).

4 Conclusion Simulations results investigate that the quantum band to band tunneling (B2BT) charge transport process is responsible for current conduction in DGTFETs producing steep subthreshold swing (SS ~ 15 mV/dec) and high I ON /I OFF ratio ~1011 with the presence of very low leakage current (~5 × 10–14 A/μm), and thus can be considered for low-power applications. Implementation of I–V /C–V data sets in the form look table coded with verilog-A model can be considered as one of the effective way to analyze circuit behavior of DGTFETs device structure. Simulation results for GBP, TFP, transit time (t r ), gm , f T investigate that the DGTFET device structure can be considered for analog applications. Findings from transient analysis describe the existence of miller capacitance (C MIL ) present at gate-drain node of TFET inverter, which is affected by voltage (V DD ) and load capacitance (C L ). The combined study of analog and transient parameters of TFET device structure describes the suitability of TFET for integrated circuits applications.

References 1. Baravelli, E., Gnani, E., Gnudi, A., Reggiani, S., Baccarani, G.: TFET inverters with n-/pdevices on the same technology platform for low-voltage/low-power applications. IEEE Trans. Electron Devices 61(2), 473–478 (2014)

Investigation of Analog Parameters and Miller Capacitance …

347

2. Nikonov, D.E., Young, I.A.: Overview of beyond-CMOS devices and a uniform methodology for their benchmarking. Proc. IEEE 101(12), 2498–2533 (2013) 3. Nikonov, D.E., Young, I.A.: Benchmarking of beyond-CMOS exploratory devices for logic integrated circuits. IEEE J. Explor. Solid-State Comput. Devices Circuits 1, 3–11 (2015) 4. Pan, C., Naeemi, A.: An expanded benchmarking of beyond-CMOS devices based on Boolean and neuromorphic representative circuits. IEEE J. Explor. Solid-State Comput. Devices Circuits 3, 101–110 (2017) 5. Lu, H., Paletti, P., Li, W., Fay, P., Ytterdal, T., Seabaugh, A.: Tunnel FET analog benchmarking and circuit design. IEEE J. Explor. Solid-State Comput. Devices Circuits 4(1), 19–25 (2018) 6. Guenifi, N., Rahi, S.B., Ghodbane, T.: Rigorous study of double gate tunneling field effect transistor structure based on silicon. Mater. Focus. 7(6), 866–872 (2018) 7. Kumar, D.: Performance evaluation of double gate tunnel FET based chain of inverters and 6-T SRAM cell. Eng. Res. Express. 1(2), 025055 (2019) 8. Chen, S., Liu, H., Wang, S., Li, W., Wang, X., Zhao, L.: Analog/RF performance of T-shape gate dual-source tunnel field-effect transistor. Nanoscale Res. Lett. 13(1), 321 (2018) 9. Der Agopian, P.G., Martino, J.A., Vandooren, A., Rooyackers, R., Simoen, E., Thean, A., Claeys, C.: Study of line-TFET analog performance comparing with other TFET and MOSFET architectures. Solid-State Electron. 128, 43–47 (2017) 10. Baravelli, E., Gnani, E., Gnudi, A., Reggiani, S., Baccarani, G.: TFET inverters with n-/pdevices on the same technology platform for low-voltage/low-power applications. IEEE Trans. Electron Devices 61(2), 473–478 (2014) 11. Khatami, Y., Banerjee, K.: Steep subthreshold slope n-and p-type tunnel-FET devices for lowpower and energy-efficient digital circuits. IEEE Trans. Electron Devices 56(11), 2752–2761 (2009) 12. Zhuge, J., Verhulst, A.S., Vandenberghe, W.G., Dehaene, W., Huang, R., Wang, Y., Groeseneken, G.: Digital-circuit analysis of short-gate tunnel FETs for low-voltage applications. Semicond. Sci. Technol. 26(8), 085001 (2011) 13. Bizindavyi, J., Verhulst, A.S., Verreck, D., Sorée, B., Groeseneken, G.: Large Variation in Temperature Dependence of Band-to-Band Tunneling Current in Tunnel Devices. IEEE Electron Device Lett. 40(11), 1864–1867 (2019) 14. Wu, P., Appenzeller, J.: Reconfigurable black phosphorus vertical tunneling field-effect transistor with record high on-currents. IEEE Electron Device Lett. 40(6), 981–984 (2019) 15. Verhulst, A.S., Vandenberghe, W.G., Maex, K., De Gendt, S., Heyns, M.M. and Groeseneken, G.: Complementary silicon-based heterostructure tunnel-FETs with high tunnel rates. IEEE Electron Device Lett. 29(12), 398–1401 (2008) 16. Kim, S.W., Choi, W.Y., Sun, M.C., Kim, H.W. and Park, B.G.: Design guideline of Si-based L-shaped tunneling field-effect transistors. Jpn. J. Appl. Phys., 51(6S), 06FE09 (2012) 17. Mookerjea, S., Datta, S.: Comparative study of Si, Ge and InAs based steep subthreshold slope tunnel transistors for 0.25 V supply voltage logic applications. In: 2008 Device Research Conference, pp. 47–48. IEEE (2008) 18. Gandhi, R., Chen, Z., Singh, N., Banerjee, K., Lee, S.: Vertical Si-Nanowire $ n $-Type Tunneling FETs With Low Subthreshold Swing ($\leq\hbox {50}\\hbox mV/decade $) at Room Temperature. IEEE Electron Device Lett. 32(4), 437–439 (2011) 19. Asthana, P.K., Goswami, Y., Basak, S., Rahi, S.B., Ghosh, B.: Improved performance of a junctionless tunnel field effect transistor with a Si and SiGe heterostructure for ultra low power applications. RSC Adv. 5(60), 48779–48785 (2015) 20. Luo, Z., Wang, H., An, N., Zhu, Z.: A tunnel dielectric-based tunnel FET. IEEE Electron Device Lett. 36(9), 966–968 (2015) 21. Rahi, S.B., Bahniman, G.: High-k Double Gate Junctionless Tunnel FET with Tunable Bandgap. RSC Adv. 5(67), 54544–54550 (2015) (Impact factor: 3.049). https://doi.org/10. 1039/C5RA06954H 22. Rahi, S.B., Asthana, P. Gupta, S.: Heterogate junctionless tunnel field-effect transistor: future of low-power devices. J. Comput. Electron. 16(1), 30–38 (2017) (Impact factor: 1.637). https:// doi.org/10.1007/s10825-016-0936-9

348

D. Kumar et al.

23. Wang, X., Tang, Z., Cao, L., Li, J., Liu, Y.: Gate Field plate structure for subthreshold swing improvement of Si line-tunneling FETs. IEEE Access 7, 100675–100683 (2019) 24. Lu, H., Paletti, P., Li, W., Fay, P., Ytterdal, T. and Seabaugh, A.: Tunnel FET analog benchmarking and circuit design. IEEE J. Explor. Solid-State Comput. Devices Circuits 4(1), 19–25 (2018) 25. Strangio, S., Settino, F., Palestri, P., Lanuzza, M., Crupi, F., Esseni, D., Selmi, L.: Digital and analog TFET circuits: design and benchmark. Solid-State Electron. 146, 50–65 (2018) 26. Kim, M.S., Liu, H., Li, X., Datta, S., Narayanan, V.: A steep-slope tunnel FET based SAR analog-to-digital converter. IEEE Trans. Electron Devices 61(11), 3661–3667 (2014) 27. Elnaggar, M., Shaker, A., Fedawy, M.: Modified hetero-gate-dielectric TFET for improved analog and digital performance. In: 2018 13th International Conference on Computer Engineering and Systems (ICCES), pp. 683–687. IEEE (2018) 28. Imenabadi, R.M., Saremi, M., Vandenberghe, W.G.: A novel PNPN-like Z-shaped tunnel fieldeffect transistor with improved ambipolar behavior and RF performance. IEEE Trans. Electron Devices 64(11), 4752–4758 (2017) 29. Asra, R., Shrivastava, M., Murali, K.V., Pandey, R.K., Gossner, H., Rao, V.R.: A tunnel FET for $ V_ {DD} $ scaling below 0.6 V with a CMOS-comparable performance. IEEE Trans. Electron Devicesv 58(7), 1855–1863 (2011) 30. Datta, S., Liu, H., Narayanan, V.: Tunnel FET technology: a reliability perspective. Microelectron. Reliab. 54(5), 861–874 (2014) 31. Choi, W.Y., Park, B.G., Lee, J.D., Liu, T.J.K.: Tunneling field-effect transistors (TFETs) with subthreshold swing (SS) less than 60 mV/dec. IEEE Electron Device Lett. 28(8), 743–745 (2007). https://doi.org/10.1109/LED.2007.901273 32. Ionescu, A.M., Riel, H.: Tunnel field-effect transistor as energy-efficient electronic switches. Nature 479(7373), 329–337 (2011). https://doi.org/10.1038/nature10679 33. Mookerjea, S., Krishnan, R., Datta, S., Narayanan, V.: Effective capacitance and drive current for tunnel FET (TFET) CV/I estimation. IEEE Trans. Electron Devices 56(9), 2092–2098 (2009) 34. Nirschl, T., Wang, P.F., Weber, C., Sedlmeir, J., Heinrich, R., Kakoschke, R., Schrufer, K., Holz, J., Pacha, C., Schulz, T., Ostermayr, M.: The tunneling field effect transistor (TFET) as an add-on for ultra-low-voltage analog and digital processes. In: IEDM Technical Digest. IEEE International Electron Devices Meeting, pp. 195–198. IEEE (2004) 35. Kumar, D., Jain, P.: Performance of dual metal-double gate tunnel field effect transistor with different dielectrics. In: Proceeding of International Conference on Intelligent Communication, Control and Devices 2017, pp. 927–933. Springer, Singapore 36. Mallik, A., Chattopadhyay, A.: Drain-dependence of tunnel field-effect transistor characteristics: the role of the channel. IEEE Trans. Electron Devices 58(12), 4250–4257 (2011) 37. Xing, H.G., Zhou, G., Li, M., Lu, Y., Li, R., Wistey, M., Fay, P., Jena, D., Seabaugh, A.: Tunnel FETs with tunneling normal to the gate. In: 2013 Third Berkeley Symposium on Energy Efficient Electronic Systems (E3S), pp. 1–1. IEEE (2013) 38. Sedighi, B., Hu, X.S., Liu, H., Nahas, J.J., Niemier, M.: Analog circuit design using tunnelFETs. IEEE Trans. Circuits Syst. I Regul. Pap. 62(1), 39–48 (2014) 39. Narang R, Saxena M, Gupta RS, Gupta M. Device and circuit level performance comparison of tunnel FET architectures and impact of heterogeneous gate dielectric. JSTS: J. Semicond. Technol. Sci. 13(3), 22436 (2013) 40. Saripalli, V., Datta, S., Narayanan, V., Kulkarni, J.P.: Variation-tolerant ultra low-power heterojunction tunnel FET SRAM design. In: Proceedings of the 2011 IEEE/ACM International Symposium on Nanoscale Architectures, 2011 Jun 8, pp. 45–52. IEEE Computer Society (2011) 41. Lee, Y, Kim, D., Cai, J., Lauer, I., Chang, L., Koester, S.J., Blaauw, D., Sylvester, D.: Lowpower circuit analysis and designbased on heterojunction tunneling transistors (HETTs). IEEE Trans. Very Large Scale Integr. (VLSI) Syst. 21(9), 1632–1643 (2013) 42. Avci, U.E., Morris, D.H., Hasan, S., Kotlyar, R., Kim, R., Rios, R., Nikonov, D.E., Young, I.A.: Energy efficiency comparison of nanowire heterojunction TFET and Si MOSFET at Lg

Investigation of Analog Parameters and Miller Capacitance …

43.

44.

45.

46. 47.

48. 49.

50.

51.

52.

53.

54.

55.

56.

349

= 13nm, including P-TFET and variation considerations. In: 2013 IEEE International Electron Devices Meeting, 2013 Dec 9, pp. 33–4. IEEE (2013) Trivedi, A.R., Carlo, S., Mukhopadhyay, S.: Exploring tunnel-FET for ultra low power analog applications: a case study on operational transconductance amplifier. In: Proceedings of the 50th Annual Design Automation Conference, 2013 May 29, p. 109. ACM (2013) Raghav, N., Bansal, M.: Analysis of power efficient 6-T SRAM cell with performance measurements. In: 2017 International Conference on Innovations in Control, Communication and Information Systems (ICICCI), 2017 Aug 12, pp. 1–4. IEEE (2017) Palomo, F.R., Fernández-Martínez, P., Mogollón, J.M., Hidalgo, S., Aguirre, M.A., Flores, D., López-Calle, I., de Agapito, J.A.: Simulation of femtosecond pulsed laser effects on MOS electronics using TCAD Sentaurus customized models. Int. J. Numer. Model. Electron. Networks Devices Fields 23(4–5), 379–399 (2010) Anand, S., Amin, S.I., Sarin, R.K.: Analog performance investigation of dual electrode based doping-less tunnel FET. J. Comput. Electron. 15(1), 94–103 (2016) Yang, Y., Tong, X., Yang, L.T., Guo, P.F., Fan, L., Yeo, Y.C.: Tunneling field-effect transistor: Capacitance components and modeling. IEEE Electron Device Lett. 31(7), 752–754 (2010). https://doi.org/10.1109/LED.2010.2047240 Mookerjea, S., Krishnan, R., Datta, S., Narayanan, V.: On enhanced Miller capacitance effect in interband tunnel transistors. IEEE Electron Device Lett. Biswas, A., Dan, S.S., Le Royer, C., Grabinski, W., Ionescu, A.M.: TCAD simulation of SOI TFETs and calibration of non-local band-to-band tunneling model. Microelectron. Eng. 1(98), 334–337 (2012) Knoch J.: Optimizing tunnel FET performance-Impact of device structure, transistor dimensions and choice of material. In: 2009 International Symposium on VLSI Technology, Systems, and Applications, 2009 Apr 27, pp. 45–46. IEEE (2009) Singh, K.S., Kumar, S., Nigam, K., Tikkiwal, V.A.: Tunnel field effect transistor for ultra low power applications: a review. In: 2019 International Conference on Signal Processing and Communication (ICSC), 2019 Mar 7, pp. 286–291. IEEE (2019) Wang, P.Y., Tsui, B.Y.: Investigation into gate-to-source capacitance induced by highly efficient band-to-band tunneling in p-channel Ge epitaxial tunnel layer tunnel FET. IEEE Trans. Electron Devices 63(4), 1788–1790 (2016) Jain, P., Kumar, D.: Drive current boosting using pocket implant near to the strained SiGe/Si source with single-metal/dual-metal double-gate tunnel field-effect transistor. In: Proceeding of International Conference on Intelligent Communication, Control and Devices, pp. 943–950. Springer, Singapore (2017) Kumar, D., Jain, P.: Double gate tunnel field effect transistor with extended source structure and impact ionization enhanced current. Inintell. Commun., Control. Devices, pp. 973–980. Springer, Singapore (2018) Mookerjea, S., Krishnan, R., Datta, S., Narayanan, V.: Effective capacitance and drive current for tunnel FET (TFET) CV/I estimation. IEEE Trans. Electron Devices 56(9), 2092–2098 (2009) Shoji, M.: CMOS digital circuit technology. In: Englewood Cliffs, ch. 4, pp. 189–190. PrenticeHall, NJ (1988)

Systematic Framework for Early Fire Detection and Smart Evacuation Using loRa—A Long-Range and Low-Power Communication Protocol Gajanand Birajdar, Rajesh Singh, and Anita Gehlot

Abstract With rapid growth in urbanization more number of people are attracting towards urban areas. Due to which density of human being in urban area is increasing day by day. To fulfill the demand of leaving area, people developing high rise buildings. With this there is increase in chances of mishaps that may occur due to earthquake, fire accident etc. These mishaps may lead dangerous to human beings as well as infrastructure. Fire is an unexpected event and it occurs very often but once it occurs it affects great manner to human and his infrastructure. Early detection of fire may assist to avoid unnecessary losses occurring fire mishaps. In the view of addressing this issue we are proposing hybrid architecture for building detection and smart evacuation system using LoRa and IoT. Keywords LoRa · IoT · Fire detection · Smart evacuation · Fire detection architecture · Building fire

1 Introduction With rapid growth of Industry and technology, the life of human being is becoming smart and comfortable. In recent days, fire accidents are affecting the lives of human beings and their infrastructure. In some cases, fire is more dangerous if it spreads to the surrounding and it becomes uncontrollable.Avoiding fire spread and diminishing fire sources is the key challenge. Installing a fire alarm system will address early detection of a fire event and avoid forthcoming losses. Most of the fire alarm systems are based on wireless sensor network due to their popularity in various aspects such as industrial automation, target tracing, environmental change monitoring, etc. In WSN-based system, sensors are the key elements which are mainly used to monitor as well as collect data from the target. But, most of the time these sensors may fail due to power limitation, as WSN-based network consumes more power. Also, in some scenarios, single sensor-based fire alarming G. Birajdar (B) · R. Singh · A. Gehlot School of Electronics & Electricals, Lovely Professional University, Punjab, India e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2021 S. Choudhury et al. (eds.), Intelligent Communication, Control and Devices, Advances in Intelligent Systems and Computing 1341, https://doi.org/10.1007/978-981-16-1510-8_34

351

352

G. Birajdar et al.

system may not be efficient, hence it is better to have a multi-sensor system for more reliability. Today, smart cities and smart homes are equipped with a smart system to deal with various situations. Hazardous events such as fire accidents also need to be addressed in a smart way. At present, the buildings which are equipped with fire alarm systems are capable to provide predefined exit path with static direction in case of fire accident. Let us assume the scenario where the fire is spreading in the same direction through which people are moving to exit. In such scenario, the existing fire alarm system may not be reliable and secure to follow a predefined exit path. Also, it is difficult to find out the direction of fire flow and people density stuck inside the building. If the information about fire accident is conveyed to the fire department or concerned authority with exact fire location and real-time monitoring, it will help them to assist and expedite the fire mitigation and people evacuation. Thus, to address above mentioned issues, we are proposing novel hybrid architecture based on LoRa and IoT. The contribution of the paper is as follows: • Proposing a multisensory model for early fire detection. • Vision node incorporated to monitor real-time fire spread and people density inside the building • Providing real-time exit path using intelligent and dynamic exit path display. • LoRa is used for low and long-range communication to share real-time data on cloud. The remaining paper is organized in the following sections. The background work is mentioned in Sect. 2. Section 3 discusses the proposed architecture and detailed design along with a flowchart of work. Details of each part are elaborated in Sect. 4. Finally, the article is concluded in Sect. 5.

2 Background Work LoRa is Long Range, low-power wireless, low data rate communication protocol for building an IoT network developed by Cycleo of Grenoble, France. LoRa uses license-free sub-gigahertz radio frequency bands. It provides long-range transmissions (1–15 km) with low-power consumption [1]. It enables the communication between remote sensors and gateways connected to the network Server and Application Servers. Multi-sensor-based fire alarming system provides reliable fire detection [2]. It is important to provide a safe and optimal exit path for reducing fire casualties and it is discussed in [3] using an improved Dijkstra algorithm. Locating the spreading fire is more important than localizing people inside the building during a fire accident and it is addressed in [4]. Lightweight Communication and Marshallingbased real-time fire monitoring and visualization system is presented and discussed in [5]. Open system architecture with intelligent environment monitoring and evacuation guidance system presented in [6]. Finding people count with more accuracy in case of abnormal scenario is of utmost importance and it is implemented using an

Systematic Framework for Early Fire Detection and Smart Evacuation Using loRa …

353

adaptive background model in [7]. Raspberry pi along with OpenCV is used to detect human heads and to provide a count of humans in [8]. Mobile terminal integrated with GIS technology is used to alarm emergency and to provide navigation path to exit the building in case of fire emergency [9].

3 System Architecture The fire is an unexpected event that occurs rarely, but once it occurs, its consequences are very disastrous. It affects life human as well as his property. To address fire accidents, we have proposed a systematic architecture constituted by end nodes with fire detection capability, controlling and actuating nodes, co-coordinators, LoRA node, and finally user interface. Figure 2 shows the proposed architecture based on LoRA network for fire and safety in the building during an emergency. It is an integration of various layers like end devices, LoRa Gateway, Network layer, and Application layer (Fig. 1). End nodes in the proposed system are deployed to collect environmental changes inside the building. The environmental parameter could be temperature, humidity, gas detection, etc. These parameters can be detected with sensors such as temperature sensors, humidity sensors, gas detectors, etc. As soon as the detected values cross the threshold level, we can understand there may be chances of fire accident, and further, it can be confirmed with the vision node. End nodes also include actuators

Vision Node 1

2.4 GHz RF communication channel

End device 1 with LoRA & RF modem End device N with LoRA & RF modem End device 1 with LoRA & RF modem

Evacuation path display controller 1

Safety operation controller1 LoRa communication channel

Application server UG87 LoRaWAN Gateway Network server

2.4 GHz RF communication channel

Evacuation path display controller2

Safety operation controller2

End device N with LoRA & RF modem

3G/4G/Etherent Bachhaul Connectivity 2.4 GHz RF communication channel

End device N with LoRA & RF modem

Evacuation path display controller N

Safety operation controllerN

End device N with LoRA & RF modem Video communication channel

Video communication channel to cloud server

Video stream controller and Local Server1

Vision Node 1

Video communication channel

Video stream controller and Cloud Server

Video stream controller and Local Server N

Fig. 1 Architecture of building fire detection and smart evacuation using LoRa and IoT

354

G. Birajdar et al. Start 1

Initialize camera and Read Fire and people count

Read all Sensor data from End Node

F

If Smoke reading= Fire?

F

If gas reading= Fire?

1

1 T

F

If heat reading= Fire?

F

1

1 T

T

T

Evacuation Display

Fire flow detected?

Send data to Cocoordinator Node

Fire Alarm and Sprikler

LoRa Gateway

Sending data to Cloud

Count People Desnity

Real Time monitoring

End

Fig. 2 Architecture of Building fire detection and smart evacuation using LoRa and IoT

such as sprinkler system, dynamic exit path display, alarming system, etc., which helps to alert people inside the building and guide them to exit safely. LoRa Gateway is responsible for collecting data from end devices and sending them over network to cloud. It has the capability to integrate data from thousands of end devices in the building or residential area within a range of few kilometers with low-power consumption. Network server is responsible for network management functions such as traffic management, frame routing, network administration, data monitoring, etc. It sends data to the application layer over Wi-Fi, Ethernet or Cellular network. It acts as a mediator between Lora gateway and End user application. Application Server deals with interpreting data received from LoRa based end devices through network server. After collecting data, further we can apply intelligent techniques such as Artificial intelligence, Machine learning to solve business problems and showcase the data in an effective manner to the end user. The end user can view the data either on the mobile-based application or web-based application. Detail flowchart of the proposed architecture is shown in Fig. 2. It starts with reading all sensory data connected to end node and also initializing vision node. Then check for threshold value of sensory data for an actual fire event. After confirming the fire event, information is shared with the respective coordinator node. This coordinator node is equipped with intelligence and it can take further decisions such as actuating sprinkler system, blowing alarm, etc. These can be placed across each floor. Then LoRa gateway will collect data from all co-ordinators and sends data to the cloud environment for further course of action. We can place a single

Systematic Framework for Early Fire Detection and Smart Evacuation Using loRa …

355

LoRa gateway for each building as it has low-power and long-range communication capability.

4 Details of Components In this section, components are elaborated in detail with their parts and uses in architecture. Figure 3 shows components of end devices with LoRa and RF modem, meant to access data from the real environment. It is integrated with many sensors such as smoke, gas, flame, heat, etc., and controlling units such as microcontroller / microprocessor-based system such as Arduino, Raspberry Pi, ESP8266, etc. In the current situation, fire alarming systems equipped inside buildings are capable of providing a predefined exit path for evacuees. To provide a secure and reliable exit path display of exit direction must be intelligent and dynamic as per situation. Evacuation LED display used to display safe route during fire emergency situation. It will guide the people to safe exit by displaying appropriate direction as shown in Fig. 4 Vision node made of Raspberry Pi and camera with audio recorder is shown in Fig. 5. A properly designed vision node will help to monitor real-time conditions of the environment during fire accidents [10]. It works seamlessly even in power failure as it operates on batteries. It also helps to visualize the density of people to be evacuated. Sprinkler System provided here in Fig. 6 is to control fire and to assist in evacuation. It is fully automatic and discharges water when a fire event is detected and it is controlled by programmable control. If the building is equipped with a proper fire detection system and a well-designed sprinkler system, then we can control fire and save human lives and property. A fire sprinkler system is an active type of fire protection system that discharges water in the event of a fire.

Power supply Circuit

Gas Flame Heat

Signal conditioning and optimization

Smoke

Other Fig. 3 Components of end devices with LoRa & RF modem

Customised LoRa and 2.4 GHz contoller board

356

G. Birajdar et al.

Fig. 4 Components of the evacuation display system

Customised 2.4 GHz RF contoller board with evacuation planning LED matrix driver

Power supply Circuit

LED display system

Fig. 5 Components of vision and information recorder device

Audio recorder

Camera

Raspberry Pi controller

Power Supply circuit

Fig. 6 Components of safety operation controller

Sprinkler

Power supply Circuit

Driver circuit for sprinkler Customised 2.4 GHz RF contoller board for safety opeartion

Systematic Framework for Early Fire Detection and Smart Evacuation Using loRa …

357

Fig. 7 Components of LoRa Gateway

Customised LoRa controller board

Hooter driver circuit

Power supply Circuit

Hooter

LoRa Gateway as shown in Fig. 7 is intended to provide long-range communication capability to the existing fire alarm system. It acts as a mediator between the end device and the network server. It will collect data from end devices and send it to the cloud network for further course of action.

5 Conclusion and Discussion Fire occurs in building due to negligence and short circuits. They cause great loss to residential community and unrecoverable damages to their property. Consequently, fire events must be detected early to prevent such type of mishaps to be happening. The main purpose of this article is to present novel architecture based on LoRa and IoT with smart fire detection and intelligent evacuation system. It meets requirements such as reliable early fire detection using the multisensory system, vision node for real-time fire monitoring and people counting, integrating with the cloud through low-power and long-range communication technology, etc. With the proposed system architecture, we can ensure early fire detection and safe evacuation. Also, the system will alert the fire department authority about the fire accident and send information through user interface. In future, we are intended to implement and analyze the feasibility of this architecture through customized hardware.

References 1. Kolobe, L., Sigweni, B., Lebekwe, C.K.: Systematic literature survey: applications of LoRa communication. Int. J. Electr. Comput. Eng. 10(3), 3176–3183 (2020) 2. Saeed, F., Paul, A., Rehman, A., Hong, W.H., Seo, H.: IoT-Based intelligent modeling of smart home environment for fire prevention and safety. J. Sens. Actuator Networks 7(1) (2018). https://doi.org/10.3390/jsan7010011

358

G. Birajdar et al.

3. Zheng, W., Cai, Q.: Implementation of intelligent fire evacuation route based on Internet of things. In: Proc. 2015 IEEE Adv. Inf. Technol. Electron. Autom. Control Conf. IAEAC 2015, vol. 1, pp. 934–938 (2016). https://doi.org/10.1109/IAEAC.2015.7428693 4. Majumder, S., O’Neil, S., Kennedy, R.: Smart apparatus for fire evacuation—an IoT based fire emergency monitoring and evacuation system. In: 2017 IEEE MIT Undergrad. Res. Technol. Conf. URTC 2017, vol. 2018-Janua, pp. 1–4 (2018). https://doi.org/10.1109/URTC.2017.828 4186 5. Beata, P.A., Jeffers, A.E., Kamat, V.R.: Real-time fire monitoring and visualization for the post-ignition fire state in a building. Fire Technol. 54(4), 995–1027 (2018). https://doi.org/10. 1007/s10694-018-0723-1 6. Yu, K.-M. et al.: Construct an intelligent evacuation guidance system with open system architecture 155(Acaai), 214–217 (2018). https://doi.org/10.2991/acaai-18.2018.50 7. Singh, C., Sohani, M.: Estimation of crowd density by counting objects. In: Proc. - Int. Conf. Trends Electron. Informatics, ICEI 2017, vol. 2018-Janua, pp. 235–238 (2018). https://doi.org/ 10.1109/ICOEI.2017.8300923 8. Syed Ameer Abbas, S., Anitha, M., Vinitha Jaini, X.: Realization of multiple human head detection and direction movement using Raspberry Pi. In: Proc. 2017 Int. Conf. Wirel. Commun. Signal Process. Networking, WiSPNET 2017, vol. 2018-Janua, pp. 1160–1164 (2018).https:// doi.org/10.1109/WiSPNET.2017.8299946 9. Huixian, Shaoping, Z.: Design of intelligent fire-fighting system of a big public building emergency evacuation navigation system for mobile terminal. In: 13th Int. Conf. Comput. Sci. Educ. ICCSE 2018, no. Iccse, pp. 90–95 (2018). https://doi.org/10.1109/ICCSE.2018.8468780 10. Gajjar, V., Khandhediya, Y., Gurnani, A.: Human detection and tracking for video surveillance: a cognitive science approach. In: Proc. - 2017 IEEE Int. Conf. Comput. Vis. Work. ICCVW 2017, vol. 2018-Janua, pp. 2805–2809 (2017)

The Development and Implementation of Competency-Based ITS for Learners with Learning Disability Akanksha Bisht and Neelu Jyothi Ahuja

Abstract An Intelligent Tutoring System provides customized and immediate feedback or instruction to the learner, without any intervention from a physical educator. Intelligent tutoring systems tend to solve the problem faced by learners by giving learning material or instruction, but it does not enforce learners to complete the learning content or achieve mastery in a particular skill or competency. For focusing on learner’s skill development and one skill at a time, competency-based learning is used. Competency-based learning allows a learner to access another competency only after he/she achieves mastery in the first one. This paper discusses implementing competency-based learning in an intelligent tutoring system for learning disabled learners. Keywords Intelligent tutoring system · Competency-based learning · Learning disability · Dyslexia · Dysgraphia · Dyscalculia

1 Introduction The basic objective of the teaching–learning process is to help the learner in achieving learning objectives by assisting whenever they face a problem in understanding a concept or an idea. In the classroom environment, one for all policy is used as it is not for an educator to consider the learning needs of all learners and provide assistance accordingly. Today, many learning environments have been developed for supporting the learning activities of learners. In these systems, the learner is provided support in the form of adaptive interfaces, path personalization, and competency-based content [1]. Such systems provide appropriate instructions, learning content, and feedback to the learner without the intervention of an educator, they simplify the work of A. Bisht (B) · N. J. Ahuja University of Petroleum and Energy Sudies, Dehradun, India e-mail: [email protected] N. J. Ahuja e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2021 S. Choudhury et al. (eds.), Intelligent Communication, Control and Devices, Advances in Intelligent Systems and Computing 1341, https://doi.org/10.1007/978-981-16-1510-8_35

359

360

A. Bisht and N. J. Ahuja

educators. Therefore, the development of effective learning systems is an important focus in the world of digitized learning [2]. In a traditional classroom environment, educators apply several theories while assisting learners to achieve predefined learning objectives. However, competencybased learning (CBL) emphasizes the mastery of specific and measurable skills during the various stages of the learning process. Learner focuses on a single competency (skill) at a time and can move to another competency only if he/she has mastered the current competency. Another important aspect of CBL is that a learner can skip a learning module completely if he/she shows mastery in that module [3]. CBL is an outcome-based instructional method that was originated in 1972. The main objective of CBL is to assist students in achieving established competency standards. The assessment and learning process of CBL considers that there is a vast difference among the learners and uses criteria-based assessment to achieve a higher level of self-learning objectives, instead of comparing with other students. CBL focuses on skill-based learning and provides unlimited time to students who participate in the learning process [4]. For evaluation purposes, tests are widely accepted as a tool because of their ease of deployment and generality. In the 80s, Computer Administered Tests emerged, which now have been evolved to the present Computerized Adaptive Tests (CAT). A CAT is used for evaluating a learner and according to the learner’s level of knowledge or test performance, it updates the student model. The result of CAT plays an important role in determining the competency of a learner [5]. Intelligent tutoring systems (ITS) are the computer-based instructional system that offers customized learning content and feedback to support the learner’s learning process. In ITS, inferences about the learner’s level are made and based on that the learning content is provided. For providing a more personalized environment to the learner, CBL can be used. In this paper, the implementation of CBL in ITS is presented to provide personalized learning content to the learner. For determining the competency of the learner, he/she has to take a pre-test. The result of the pre-test reveals the learning problem or a combination of learning problems that he/she has. Based on this result, a unique Instruction Plan (IP) is provided to each learner in the learner’s preferred learning style.

2 Literature Review In [6], authors have discussed a trialogue-based ITS that uses the learning competency of peer agents to investigate learner engagement, reading, attitude, and system selfefficacy while learning English. As a result of their study, they found out that using peer agents in ITS enhances learner engagement and improves learning outcomes. They found out that the peer agent with low competency agent increases the engagement time of the learner, whereas high or medium competency agent motivates the learner to achieve higher academic goals.

The Development and Implementation of Competency-Based ITS …

361

In [7], the authors examined the effects of using pedagogical agents as learning companions on learner’s attitudes, learning, and self-efficacy. The agent can belong to two competencies—low or high, and interaction type—proactive or responsive. The test was performed on 72 students of computer course. Treatments were randomly assigned to the students that were of the category low responsive, highly responsive, low proactive, and high proactive. They found out that students assigned high competency in both interaction types, displayed a more positive attitude, and achieved higher scores. However, students, who were assigned low competency, displayed enhanced self-efficacy belief while doing learning tasks. The authors concluded that interaction and competency should be assigned according to the motivational goals and desired to learn. In [8], the authors discussed a Technical Entrepreneurial Vocational Education and Training (TEVET) system and its challenges while providing technical education in Malawi. Competency-based education (CBE) aims at the skill and responsibility development in the students. However, CBE cannot develop the sense of responsibility on its own, rather it requires the correct mindset and attitude of both students and instructors. Authors recommend the following factors for making CBE successful: • Change the attitude and mindset of students and instructors. • Provide relevant and up-to-date learning material on time. In [9], the authors presented a new algorithm for an adaptive e-learning environment called competency-based guided-learning algorithm (CBGLA). For developing this algorithm, the authors used a mathematical derivation of CBL and computational process analysis. The proposed algorithm generates a personalized learning path for learners without involving the physical instructor. For implementing adaptive learning, the system provided continuous feedback, correction, and assessment. This experiment showed that under the guidance of CGBL, learners learned more effectively and were satisfied with the interface design and learning path mechanism of the system. In [10], the authors discussed the importance of computerized adaptive tests (CAT) in the diagnosis process while implementing competency education in ITS. To develop a CAT, the question selection process plays an important role so the questions must be chosen in a way that best contributes to learner assessment. The authors introduced a new question selection algorithm that enhances the accuracy level of the diagnosis based on fuzzy linguistic information. In [11], the authors discussed the role of tutors in e-learning environments. The tutor provides assistance to the learner and follows up their learning process. It is possible that sometimes requests made by learners may not match the competency and skills of the tutor. So collaboration of tutors is required for providing effective monitoring and improving the skills of the tutor. The authors presented a new technique for grouping tutors called K-complementarity. The system used this technique formed a group of tutors and was tested at Algerian University. The results showed that each group solved 80% of the tasks given to them that were acceptable.

362

A. Bisht and N. J. Ahuja

In [12], the authors presented an approach based on probability distributions for student modeling. For diagnosis purposes, an adaptive test was provided where learners can give partial credit instead of correct or incorrect. This test accurately evaluates the learner’s knowledge in several concepts and provides competency of the learner as result. In [13], authors reviewed research works conducted in India in the field of learning disability. The authors highlighted the role of the special educator in the education of LD learners and the impact of using technology in their education. This study pointed out that government policies and researchers are doing noticeable work and may deliver promising results in the future. In [14], authors presented a handwriting evaluation framework to analyze handwritten documents for dysgraphia profiling. For identification of dysgraphia, authors used the Structured Similarity Index Measure algorithm.

3 Proposed Module In this paper, at the onset, the philosophy and principles behind the implementation of CBL in ITS, have been discussed. In the proposed architecture, the target group comprises students with learning disabilities under the 4–10 age group. Here some competencies of three learning disabilities dyslexia, dysgraphia, and dyscalculia are taken as target problems. Dyslexia involves difficulty in reading due to a problem in identifying the sound of letters and decoding how sound is related to these letters. So the competencies related to dyslexia, taken in the proposed architecture are literacy skills, phonological awareness, reading fluency, and random naming/ decoding. Dysgraphia affects fine motor skills and handwriting. A learner with dysgraphia presents a problem with spelling, spacing, and putting thoughts on paper. The proposed architecture deals with sentence and word expression, visual-spatial response, motor skill, spelling, and legibility. Learners with dyscalculia present problems in number-related concepts, basic calculation, and learning how to manipulate numbers. In the current scope, focused competencies are counting, basic calculation, and reasoning. In the proposed architecture, a pre-test containing 19 questions is included for checking the level of the learner so that appropriate content can be provided to the learner according to his/her need. Table 1 presents the problematic skills associated with pre-test questions. The proposed architecture offers two levels for the learner—grade 1, grade 2. Grade 1 is the primary level and grade 2 is the higher level. Whenever a fresh learner logs into the system, he/she is directed to grade1. If an educator recommends grade 2 for a particular learner, then only the learner can access grade 2. The proposed ITS contains three phases: 1. 2. 3.

Pretest Learning process Posttest

The Development and Implementation of Competency-Based ITS … Table 1 Problem Skills associated with pre-test questions

Disability type

Problematic skills

Dyslexia

Literacy skills

363

Phonological awareness Reading fluency Random naming/ decoding Dysgraphia

Sentence and word formation Visual-spatial response Motor skill Spelling Legibility

Dyscalculia

Counting Basic calculation Reasoning

3.1 Pretest After entering the system, learners are provided a pre-test. Both grade 1 and grade 2 pre-tests comprise 19 questions each. Pretest questions are mapped to the type of disability, problematic skill, cognitive strength, and learning style. The target disabilities are dyslexia, dysgraphia, and dyscalculia, and problematic skills are mentioned in Table 1. Cognitive strengths taken in the proposed ITS are—memory and attention, audio-visual perception, and rational/ logical thinking. Learning styles taken in the proposed ITS are—video-based, story-based, and game-based learning style. Based on the score achieved in the pre-test, a learner profile is created. Learner profile comprises the type of disability, degree of disability, competencies that leaner need to improve (problematic skills), and learning style.

3.1.1

Type of Disability

The pre-test score is calculated individually for each disability type. If a learner scores less in a particular disability type, then it is concluded that the learner has that particular type of disability. The initial competency of a learner is determined based on the pre-test score. DL category consist of 7 questions: 1, 2, 3, 11, 14, 15, and 19; DG category consist of 6 questions: 4, 5, 10, 13, 16, and 18; and DC category consist of 6 questions: 6, 7, 8, 9, 12, and 17. If a learner scores 0–6 marks in the DL category, he/she is considered dyslexic, and if a learner scores 7 marks then the learner is categorized as not dyslexic. If a learner scores 0–5 marks in the DG category, he/she is categorized as DG, and if a learner scores 6 marks then he is considered as not DG. If a learner scores 0–6 marks in the DC category, he/she is

364

A. Bisht and N. J. Ahuja

Table 2 Subcategory of competency of learner based on pre-test score in DL

S. no

Score achieved

Degree

1

0

Extreme

2

1–3

High

3

4–5 (dysgraphia, dyscalculia)/6 (dyslexia)

Low

4

6 (dysgraphia, dyscalculia)/7 (dyslexia)

Nil

considered as DC, and if a learner scores 6 marks then the learner is categorized as not DC.

3.1.2

Degree of Disability and Problematic Skill

The degree of disability is calculated based on the score obtained through cumulative scoring from the response of questions asked in the pre-test. Each question belongs to a problematic skill. If a learner scores 0 marks in any disability category then his/her degree is considered as extreme. If he/she scores 1–3 marks then the degree is high, if the learner score 4–5 (in case of dysgraphia and dyscalculia)/ 6 (dyslexia) marks, then the degree is low. If he/she scores 6 (in case of dysgraphia and dyscalculia)/ 7 marks (dyslexia), then the learner does not have that particular disability. Problematic skill is determined based on wrong answers given by the learner in the presented. Target problematic skills are listed in Table 1. Table 2 presents the scoring pattern.

3.1.3

Cognitive Strength and Learning Style

Questions 1, 2, 6, 13, 15, and 17 are mapped to cognitive strength: memory and attention (MA), questions 3, 9, 10, 14, 18, and 19 are mapped to cognitive strength: audio-visual perception (AV), and questions 4, 5, 7, 8, 12, and 16 are mapped to cognitive strength: rational/logical thinking (RL). The scores achieved in the subcategory MA, AV, and RL are put in a stack in such a way that the bottom of the stack contains the least marks scored in one of the categories, and the top of the stack contains a subcategory in which learner scored highest marks. The cognitive strength is connected with learning styles. The learner who has memory and attention as cognitive strength, will get the content in presentation mode. The learner who has audio-visual perception as cognitive strength, will get content in video mode. The learner who has logical and rational processing as cognitive strength, will get content in form of serious games. So the learner is presented content in the learning style that is mapped to the conitive strength mentioned at the top of the stack. Figure 1 represents the learner profile that is determined after pre-test.

The Development and Implementation of Competency-Based ITS …

365

Fig. 1 Learner Profile

3.2 Learning Process Based on the learner profile, competencies of the learner that need to be improved is determined. After the competency is determined, a unique IP is allocated to each learner in the learning style suggested by the pre-test (top of the stack). This IP contains a combination of learning content based on the learner’s problem. According to CBL, learner works on one competency at a time. So from this IP, a single content piece is presented to the learner to strengthen that particular competency. A learner can go to the next competency only when he/she has mastered the current competency. So after the unit learning content is completed by the learner, an interaction quiz is presented to the learner. This interaction quiz helps in knowing if the learner can understand the provided content and whether the learning style of the content is suitable for the learner. If the learner presents a positive response in the interaction

366

A. Bisht and N. J. Ahuja

Fig. 2 Instruction plan delivered to the learner

quiz, then the learning process is continued. If the learner shows a negative response in the interaction quiz, it can be concluded that the learner is not able to understand the provided content in the given learning style, so the learning style of content is changed and the same learning content is provided to the learner in the learning style that is next to the top in the stack. When a learner completes the interaction quiz (shows mastery in current competency) then only he/she is allowed to access the next content piece. Figure 2 represents the IP delivered to the learner.

3.2.1

Grade 1

If a learner achieves a cent percent score in pre-test or other words shows mastery in all competencies, then it is concluded that learner has higher competency than assigned so the competency of the learner is upgraded and grade 2 pre-test is presented to the

The Development and Implementation of Competency-Based ITS …

367

Fig. 3 Changing competency of grade 1 learner on achieving a perfect score in the pre-test

Fig. 4 Changing competency of grade 2 learner based on achieving an extremely low score in the pre-test

learner, to determine the new competency level. The new competency of the learner is decided based on the score achieved in the grade 2 pre-test. Figure 3 represents the flow of competency change, as implemented by the proposed ITS.

3.2.2

Grade 2

The learner is provided learning content based on his/her problem. If a learner is found to have an extreme degree in all the three submodules, then there may be a chance that the learner can belong to grade 1, so, for better assessment, grade 1 pre-test is presented to the learner. So based on the grade 1 pre-test, the competency of this learner is determined. Figure 4 presents learner migration between grades depending upon performance in the pre-test.

3.3 Post-test After the learning process, a post-test is presented to the learner. The post-test contains questions related to dyslexia, dysgraphia, and dyscalculia. The difference in pre-test score and interaction quiz score and pre-test score and post-test score is calculated

368

A. Bisht and N. J. Ahuja

using fuzzy logic. The degree of difference is termed as learning gain. Based on the performance of the learner, learning gain is categorized into three categories: positive, negative, and neutral. The fuzzy rules used in the proposed work are mentioned in Table 3: Figure 5 represents screenshot of the final profile details of the learner after posttest. Table 3 Fuzzy rules Rule

Input Interactive_score_diff

Post_score_diff

Degree Learning_gain

Rule 1

>0

>0

Positive

Rule 2