Dynamics in Logistics: Proceedings of the 8th International Conference LDIC 2022, Bremen, Germany (Lecture Notes in Logistics) 3031053583, 9783031053580
143 7 21MB
English Pages [502]
Preface
Contents
Supply Chain Management
The Linkage Between Macro Logistics Capabilities and Micro Firm Performance Towards Framework Development for Supply Chain Performance Measurement
1 Introduction
2 Macro Higher-Order Capabilities and Firm Capabilities Towards Performance Optimization: A Dynamic Capability View
3 Approach
3.1 Framework for Linking Macro-micro View Towards Supply Chain Performance Measurement
3.2 Qualitative Integrative Reviews
4 Findings and Discussion
4.1 Linkages Between Macro Logistics Higher-Order Capabilities and Micro Firm Performance Objectives and Capabilities
4.2 Future Research on the Development of Macro Capability-Micro Performance Framework
5 Conclusion and Limitations
References
Integrating Regional Food Manufacturers into Grocery Retail Supply Chains in Germany
1 Introduction
2 Regionality and Retailing
2.1 Regional Food
2.2 Grocery Retail
2.3 Trends in Grocery Retail
3 Developing a Frame of Reference for Regional Supply Chain Concepts for Regional Food Manufacturers
3.1 General Structure of Grocery Retail Supply Chains
3.2 Integration Challenges
3.3 Final Model of Alternatives
3.4 Attributes and Indicators for Decision Making
4 AHP-Methodology
5 Findings
6 Conclusions
References
Enhancement of Crowd Logistics Model in an E-Commerce Scenario Using Blockchain-Based Decentralized Application
1 Introduction
2 Literature Review
2.1 Logistics
2.2 Crowd Logistics
2.3 Blockchain Technology
2.4 Research Gap Analysis
3 Proposed Model
3.1 Conceptual Model
3.2 The Detailed Solution
4 Model Capabilities
4.1 Encryption Mechanism
4.2 Model Transactions
5 Conclusion
References
Synthesising COVID-19 Related Research from a Logistics and Supply Chain Perspective
1 Introduction
2 Methodological Approach
3 Results
3.1 Sample Characterisation
3.2 Comparative Analysis
3.3 Covid-19 from an OPS/OR Point of View
4 Research Avenues for Future OPS/OR Research and Conclusions
4.1 Rethinking Supply Chain Strategies and Structures
4.2 Supply Chain Risk Management
4.3 Circular Supply Chains
4.4 Diversifying Logistics Services
4.5 Logistics for Social Welfare
5 Conclusion
References
Challenges and Approaches of Non-pharmaceutical Interventions for Airport Operations During Pandemic Situations
1 Motivation
2 Problem Analysis
3 Requirements
4 Current Approaches
5 Proposed Approach
6 Further Use Cases
7 Limitations, Conclusion and Outlook
References
Identifying Common Elements Within Supply Chain Resilience and Sustainability - An Exploratory Study Based on Bibliographic Analysis
1 Introduction
2 Literature Review
3 Research Methodology
4 Results
4.1 The Four Connecting Elements in SSCM and SCRES Literature
4.2 Capability of Transparency and Big Data Analytics
4.3 Main Path of SCRES and SSCM Based on Capability Research
5 Discussion
6 Conclusion
References
The Impact of Blockchain on Supply Chain Resilience
1 Introduction
2 Conceptual Framework
3 Methodology
3.1 Data Collection Method and Sample Description
3.2 Data Analysis Method
4 Findings
5 Discussion
6 Conclusion
References
Towards Supply Chain Resilience in Mining Industry: A Literature Analysis
1 Introduction
2 Characterizing the Mining Industry
2.1 Mining Industry in the Mineral Supply Chain
2.2 Mineral Deposit Main Source of Mining Industry
3 Scope and Methodology
4 Conceptualizing Supply Chain Resilience in Mining
4.1 Defining SCRes in Mining
4.2 Operationalizing SCRes Elements in Mining
5 Conclusion
References
Data Quality in Social Media Analytics for Operations and Supply Chain Performance Management
1 Introduction
2 OSC Performance and Unstructured Social Media Data
3 Data Quality Issues for OSCPM with Social Media Data
4 Methodology
5 Findings
5.1 Quantitative Analysis
5.2 Qualitative Analysis: Accuracy, Relevance and Believability
6 Discussion
7 Conclusion
References
Managing Interfaces Between Smart Factories and Digital Supply Chains
1 Introduction
2 Methodology
3 Related Work
3.1 Interface Challenges and Management
3.2 Design Principles of a Smart Factory
3.3 Interfaces in Adaptive Environments
4 Managing Interfaces Between Smart Factories and Digital Supply Chains
4.1 Definition
4.2 Control
4.3 Communication
5 Discussion and Conclusion
References
Maritime Logistics and Port Operations
Container Flow Generation for Maritime Container Terminals
1 Introduction
2 State of the Art
3 Conceptual Model
4 Visual Validation
5 Discussion and Outlook
References
Simulation-Based Port Storage Dimensioning to Mitigate Operational Instability
1 Introduction
2 Literature Review
3 Methodological Approach
3.1 Description of the Developed Simulation Model
3.2 Specifics Regarding the Simulation Experiments
4 Results
4.1 Effect of Storage Shortage on the Operational Instability of a Port Terminal
4.2 Approach to Dimension Port Storage Coping with Operation Instability
4.3 Discussion of Results
5 Conclusion
References
Integration of Renewable Energies at Maritime Container Terminals
1 Introduction
2 Problem Statement
2.1 Energy Demand of Quay Cranes as Main Energy Consumers at Maritime Container Terminals
2.2 Forecast of Energy Supply Through Renewable Resources
3 Simulation Procedure
3.1 Simulation Model Description
3.2 Variation of Input Parameters
4 Results and Discussion
4.1 Results for Throughput Times
4.2 Energy Consumption Analysis
5 Conclusion and Outlook
5.1 Conclusion
5.2 Outlook
References
An Appraisal of the Northern European LNG Bunker Ship Fleet
1 Introduction
1.1 LNG as Ship Fuel and Associated Supply Infrastructure
1.2 Research Gap
1.3 Methodological Approach
2 Supply Infrastructure for LNG Bunker Ships and Demand for LNG as Ship Fuel – State of Play and Future Developments
3 Operational Profile of the Northern European LNG Bunker Ship Fleet
3.1 Composition of the Fleet
3.2 Type of Operations
4 Conclusion
References
Investigating the Requirements of Automated Vehicles for Port-internal Logistics of Containers
1 Automated Container Transportation in Port Logistics
2 State of Research
3 Methodology
4 Results
5 Conclusion
References
Unmanned Vessels and the Law
1 Introduction
1.1 Current Legal Situation
1.2 Technology and Current Research Projects
2 UMV as ‘Ship’ and ‘Vessel’ Within the Meaning of International Maritime Law
2.1 Rights Granted to Ships and Vessels
2.2 Definition in International Maritime Law
2.3 Definition in National Maritime Law
3 Requirement of Safe Manning
3.1 Regulatory Framework
3.2 Status of Opinion
4 Look-Out Duty
4.1 Statutory Framework
4.2 Technical Solutions
4.3 Status of Opinion
5 Conclusion
References
Low Emission Choices in Freight Transport: Comparing Land and Short Sea Shipping Alternatives
1 Introduction
2 Methodology
3 Results
3.1 Conditions, Factors, and Variables
4 Discussion
5 Summary
References
Digital Twin Features for the Intelligent Container
1 Introduction
1.1 The Intelligent Container
1.2 Digital Twins in the Food Supply Chain
2 Making Models Updateable
2.1 State-Space Description
2.2 Complexity of Biological Systems
3 Software Structure of Digital Twins
3.1 Software Components
3.2 Converting Simulation Models into Event-Driven Services
3.3 From the Intelligent Container to a General Digital Twin
4 Summary
References
Transportation Networks and Vehicle Routing
A New Lower Bound for the Static Dial-a-Ride Problem with Ride and Waiting Time Minimization
1 Introduction
2 Literature Review
3 Problem Description
4 Mixed-Integer Program
4.1 Model Formulation
4.2 Calculation of an Initial Lower Bound
4.3 Bounding of Variables
5 Computational Study
5.1 Instances
5.2 Discussion of Results
6 Conclusion
References
An Auction-Based Multi-Agent System for the Pickup and Delivery Problem with Autonomous Vehicles and Alternative Locations
1 Introduction
2 Decentralized Generalized Pickup and Delivery Problem with Preferences
2.1 Problem Motivation
2.2 Local Problem Definitions
2.3 Local and Global Perspective
3 Multi-Agent System
3.1 General Approach
3.2 Order Agent
3.3 Vehicle Agent
4 Single-Agent System
5 Computational Study
5.1 Problem Instances
5.2 MAS vs. SAS
5.3 Discussion
6 Conclusions and Future Research
References
Strategic Bidding in Decentralized Collaborative Vehicle Routing
1 Introduction
2 Related Work
3 Auction Approaches
3.1 First-Price Auctions
3.2 Second-Price Auctions
4 Computational Study
4.1 Strategic Behaviour in a First-Price Auction System
4.2 Strategic Behaviour in a Second-Price Auction System
5 Conclusions
References
Artificial Intelligence in Urban Last Mile Logistics - Status Quo, Potentials and Key Challenges
1 Introduction
2 Methodology
2.1 Data Collection
2.2 Data Analysis
3 Findings and Discussion
3.1 AI Use Cases for Urban Last Mile Logistics
3.2 Potentials of AI for Urban Last Mile Logistics
3.3 Challenges of AI Implementation in Urban Last Mile Logistics
3.4 Central Findings and Implications
4 Conclusion
5 Limitations
References
An Inter-organizational Digital Platform for Efficient Container Transportation
1 Introduction
2 Digital Exchange of Information in Container Transportation Processes
3 Support of Container Transportation Processes by an Inter-organizational Digital Platform
3.1 Quotation and Selection of Carriers
3.2 Container Slot Booking
3.3 Preparation and Issue of Shipping Documents
3.4 Tracking and Tracing of Orders
4 Current State and Evaluation of the Platform
5 Conclusion and Outlook
References
SRP: A Sustainable Dynamic Ridesharing Platform Utilizing Blockchain Technology
1 Introduction
1.1 Research Problem and Relevance
1.2 Aim of the Study
2 Literature Review
2.1 Ridesharing
2.2 Blockchain Technology in Ridesharing Services
2.3 Research Gap
3 Proposed Solution
3.1 Encryption Mechanism
3.2 Transactions and the Flow of the Process
3.3 Framework Capabilities
4 Research Highlight
5 Conclusion
References
Yard Management: Identification and Evaluation of Critical Sub-processes with AHP
1 Introduction
2 Yard Management and Methodological Approach
3 AHP Prioritization
4 Discussion
5 Conclusion and Outlook
References
A Cargo Throughput Capacity Quantization Estimation Using Semi-Markov Jump System Filter Within Partial State Delay
1 Introduction
2 Problem Formulation
3 Method
4 Simulation Results
5 Conclusion
References
Applications of Pixel Oriented Mobility Modelling in Transport & Logistics
1 Introduction
2 Background
2.1 Macroscopic Models
2.2 Microscopic Models
3 Method Elicitation
4 Feasible Application Areas in Transport & Logistics
4.1 Vehicular Traffic
4.2 Port Operations
4.3 Freight Movements
4.4 Warehouse Operations
4.5 Airports, Commercial Aviation and Urban Air Mobility
4.6 Sensors and Internet of Things
5 Conclusion
References
Choosing the Right Technique for the Right Restriction – A Domain-Specific Approach for Enforcing Search-Space Restrictions in Evolutionary Algorithms
1 Introduction
2 Background
3 Methodology
3.1 Noise-sensitive Constraint Handling
3.2 Configurable Constraint Handling
4 Implementation
5 Evaluation
5.1 Setup of Evaluation
5.2 Results
5.3 Discussion
6 Conclusion and Further Research
References
Production Planning and Scheduling
Managing Complexity in Variant-Oriented Manufacturing: A System Dynamics Approach
1 Introduction
2 Literature Review
2.1 Complexity Measurement Approaches
2.2 Overview Complexity Metrics
2.3 System Dynamics Approaches
3 Methodology
4 Simulation and Analysis of the Model
5 Results
6 Conclusion
References
Dynamic Lot Size Optimization with Reinforcement Learning
1 Introduction
2 Challenges in Batch Size Planning
3 Methodology for the Calculation of a Lot Size
4 Dynamic Adaptation with Reinforcement Learning
5 Scenario and Realization
6 Observation and Action Space
7 Evaluation of the Different Policies
8 Implementation in the ERP System
9 Outlook
References
Integrated Simulation-Based Optimization Approach for Production Scheduling: A Use Case Application in a Machining Process
1 Introduction
2 Related Work and Contribution
3 Data-Driven Simulation-Based Optimization
3.1 Simulation-Based Optimization Approach
4 Use Case
4.1 Experiment Description
4.2 Results and Analysis
5 Conclusion and Outlook
References
Scheduling Workforce in Decentrally Controlled Production Systems: A Literature Review
1 Introduction
2 Background Information of Investigation
2.1 Workforce Scheduling Dilemma in Industry 4.0
2.2 Systematic Literature Review Process
3 State of the Art
3.1 Selected Relevant Trends in the Last Years
3.2 Classification of Contributions
3.3 Some Gained Insights
4 Conclusion
References
Maintenance 4.0: A Literature Review and SWOT Analysis
1 Introduction
2 Theoretical Background
2.1 Maintenance 4.0
3 Research Methodology
4 Results and Discussion
4.1 Bibliometric Analysis
4.2 Content Analysis
4.3 SWOT Analysis
5 Conclusion
References
Using Supervised Learning to Predict Process Steps for Process Planning of Third-Party Logistics
1 Introduction
1.1 Motivation
1.2 Problem Description
1.3 Research Approach
2 Methods
2.1 Data Basis
2.2 Sequence Analysis with N-Gram Analysis
2.3 Classification with Decision Trees
3 Results
3.1 Results of N-Gram-Sequence Analysis
3.2 Results of Decision Tree Classification
4 Conclusion
References
From Linear to Circular Packaging: Enablers and Challenges in the Fashion Industry
1 Introduction
2 Methodology
3 Discussion
3.1 Challenges for Circular Packaging Models
3.2 Enablers for Circular Packaging Models
4 Conclusion
5 Managerial Implications, Further Research and Limitations
5.1 Managerial Implications
5.2 Implications for Further Research
5.3 Limitations
References
Socio-technical Systems
Assessing Driver Fatigue During Urban Traffic Congestion Using ECG Method
1 Introduction
2 Method of Assessing Driver Fatigue
3 Results
3.1 Experiment Description
3.2 Data Processing
4 Conclusions
References
Technology Review for Guiding Persons in Airports and Other Hubs
1 Introduction
2 State of the Art
2.1 Human Recognition
2.2 Routing Components
2.3 Human Computer Interaction (HCI)
3 Selection of Technologies
4 Conclusion and Outlook
References
The Impact of the COVID-19 Pandemic on E-commerce Consumers’ Pro-environmental Behavior
1 Introduction
2 Hypotheses Development
3 Research Design
4 Results
4.1 Sample Description and Frequency Distributions
4.2 Methodology
4.3 Analysis and Results
5 Discussion
6 Conclusion
References
Developing a Serious Game for Intelligent Transportation Systems
1 Introduction
1.1 Problem Background/Motivation
1.2 Research Question and Methodological Approach
2 Gamification or Serious Game?
2.1 Gamification
2.2 Serious Games
2.3 What is Logistics Empire?
3 Logistics Empire – The Game
3.1 Phases of Game Development
3.2 ITS for Road Freight Transport
3.3 Game Design
4 Discussion and Conclusions
References
Author Index
Recommend Papers
![Dynamics in Logistics: Second International Conference, LDIC 2009, Bremen, Germany, August 2009, Proceedings [1 ed.]
3642119956, 9783642119958](https://ebin.pub/img/200x200/dynamics-in-logistics-second-international-conference-ldic-2009-bremen-germany-august-2009-proceedings-1nbsped-3642119956-9783642119958.jpg)
![Dynamics in Logistics: First International Conference, LDIC 2007, Bremen, Germany, August 2007, Proceedings [1 ed.]
9783540768616, 3540768610](https://ebin.pub/img/200x200/dynamics-in-logistics-first-international-conference-ldic-2007-bremen-germany-august-2007-proceedings-1nbsped-9783540768616-3540768610.jpg)
![Dynamics in Logistics : Twenty-Five Years of Interdisciplinary Logistics Research in Bremen, Germany [1 ed.]
9783030886622, 9783030886615](https://ebin.pub/img/200x200/dynamics-in-logistics-twenty-five-years-of-interdisciplinary-logistics-research-in-bremen-germany-1nbsped-9783030886622-9783030886615.jpg)
- Author / Uploaded
- Michael Freitag
- Aseem Kinra
- Herbert Kotzab
- Nicole Megow
File loading please wait...
Citation preview
Lecture Notes in Logistics Series Editors: Uwe Clausen · Michael ten Hompel · Robert de Souza
Michael Freitag Aseem Kinra Herbert Kotzab Nicole Megow Editors
Dynamics in Logistics Proceedings of the 8th International Conference LDIC 2022, Bremen, Germany
Lecture Notes in Logistics Series Editors Uwe Clausen, Fraunhofer Institute for Material Flow and Logistics IML, Dortmund, Germany Michael ten Hompel, Fraunhofer Institute for Material Flow and Logistics IML, Dortmund, Germany Robert de Souza, The Logistics Institute - Asia Pacific, National University of Singapore, Singapore, Singapore
Lecture Notes in Logistics (LNL) is a book series that reports the latest research and developments in Logistics, comprising: • • • • • • • • • • • • • • • • • • •
supply chain management transportation logistics intralogistics production logistics distribution systems inventory management operations management logistics network design factory planning material flow systems physical internet warehouse management systems maritime logistics aviation logistics multimodal transport reverse logistics waste disposal logistics storage systems logistics IT
LNL publishes authored conference proceedings, contributed volumes and authored monographs that present cutting-edge research information as well as new perspectives on classical fields, while maintaining Springer’s high standards of excellence, the content is peer reviewed. Also considered for publication are lecture notes and other related material of exceptionally high quality and interest. The subject matter should be original and timely, reporting the latest research and developments in all areas of logistics. The target audience of LNL consists of advanced level students, researchers, as well as industry professionals working at the forefront of their fields. Much like Springer’s other Lecture Notes series, LNL will be distributed through Springer’s print and electronic publishing channels. ** Indexed by Scopus (2021) **
More information about this series at https://link.springer.com/bookseries/11220
Michael Freitag Aseem Kinra Herbert Kotzab Nicole Megow •
•
•
Editors
Dynamics in Logistics Proceedings of the 8th International Conference LDIC 2022, Bremen, Germany
123
Editors Michael Freitag Production Systems and Logistics University of Bremen Bremen, Germany
Aseem Kinra Global Supply Chain Management University of Bremen Bremen, Germany
Herbert Kotzab Logistics Management University of Bremen Bremen, Germany
Nicole Megow Combinatorial Optimization and Logistics University of Bremen Bremen, Germany
ISSN 2194-8917 ISSN 2194-8925 (electronic) Lecture Notes in Logistics ISBN 978-3-031-05358-0 ISBN 978-3-031-05359-7 (eBook) https://doi.org/10.1007/978-3-031-05359-7 © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 This work is subject to copyright. All rights are solely and exclusively licensed by the Publisher, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed. The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. The publisher, the authors and the editors are safe to assume that the advice and information in this book are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the editors give a warranty, expressed or implied, with respect to the material contained herein or for any errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. This Springer imprint is published by the registered company Springer Nature Switzerland AG The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
Preface
Since 2007, the biennial International Conferences on Dynamics in Logistics (LDIC) offers researchers and practitioners from logistics, operations research, and engineering as well as from mathematics and computer science an opportunity to meet and discuss the latest developments in logistics and related domains. From February 23–25, 2022, the eighth LDIC was held online and hosted by the University of Bremen. This time, the conference featured the “International Diginomics-LogDynamics PhD Field Day” as a satellite event as well as a virtual tour through the LogDynamics Lab of the Bremen Research Cluster for Dynamics in Logistics (LogDynamics). Similar to its seven predecessors, the LogDynamics Research Cluster organized this conference in cooperation with the BIBA—Bremer Institut für Produktion und Logistik, which is an engineering research institute affiliated to the University of Bremen. The LDIC 2022 comprised of empirical, theoretical, methodological, and practice-oriented contributions addressing the modeling, planning, optimization and control of processes in supply chains, logistic networks, production systems, and material flow systems and facilities. LDIC 2022 provided a forum for the discussion of advances in that matter. The conference program considered three invited keynote speeches, 54 talks, and 39 scientific papers selected by a double-blind reviewing process. All selected papers are arranged within these LDIC 2022 proceedings. By this, the proceedings give an interdisciplinary outline on the state of the art of research in dynamics in logistics as well as identify challenges and solutions for logistics today and tomorrow. The volume is organized into the following main areas: • • • • •
Supply chain management, Maritime logistics and port operations, Transportation networks and vehicle routing, Production planning and scheduling, Socio-technical systems.
v
vi
Preface
There are many people whom we have to thank for their help in one or the other way. For pleasant and fruitful collaboration, we are grateful to the members of the international program committee: Julia Arlinghaus, Magdeburg (Germany); Till Becker, Emden/Leer (Germany); Valeria Belvedere, Milan (Italy); Tobias Buer, Maskat (Oman); Sergey Dashkovskiy, Würzburg (Germany); Malte Fliedner, Hamburg (Germany); Enzo M. Frazzon, Florianópolis (Brazil); Jens Heger, Lüneburg (Germany); Michael Henke, Dortmund (Germany); Soondo Hong, Pusan (Korea); Alexander Hübner, München (Germany); Dmitry Ivanov, Berlin (Germany); Hamid Reza Karimi, Milano (Italy); Matthias Klumpp, Essen (Germany); Rene de Koster, Rotterdam (Netherlands); Anne Lange, Luxemburg (Luxemburg); Frank Meisel, Kiel (Germany); Antônio G.N. Novaes, Campinas (Brazil); Jörn Schönberger, Dresden (Germany); Elen Twrdy, Ljubljana (Slovenia); Thorsten Wuest, Morgantown, WV (USA). Carrying the burden of countless reviewing hours, we wish to thank our Bremen LogDynamics colleagues Frank Arendt, Michael Beetz, Matthias Burwinkel, Gralf-Peter Calliess, Rolf Drechsler, Hans-Dietrich Haasis, Otthein Herzog, Hans-Jörg Kreowski, Walter Lang, Michael Lawo, Burkhard Lemper, Daniel Schmand, Klaus-Dieter Thoben, Yilmaz Uygun, Hendro Wicaksono as well as numerous other colleagues from all over the world for their help in the selection process. We are also very grateful to Aleksandra Himstedt, Ingrid Rügge, Matthias Burwinkel, Nicolas Kassel, Angelika Gühr, and Yasmin Sakhr for their support in the local organization and the technical assistance during the virtual conference. Moreover, we would like to acknowledge the financial support by the BIBA. Finally, we appreciate the excellent cooperation with Springer, which continuously supported us regarding the proceedings of all LDIC conferences. February 2022
Michael Freitag Aseem Kinra Herbert Kotzab Nicole Megow
Contents
Supply Chain Management The Linkage Between Macro Logistics Capabilities and Micro Firm Performance Towards Framework Development for Supply Chain Performance Measurement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Debarshee Bhardwaj and Aseem Kinra
3
Integrating Regional Food Manufacturers into Grocery Retail Supply Chains in Germany . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Janik Serkowsky, Herbert Kotzab, and Julia Fischer
15
Enhancement of Crowd Logistics Model in an E-Commerce Scenario Using Blockchain-Based Decentralized Application . . . . . . . . . Karthikeyan Navendan, Hendro Wicaksono, and Omid Fatahi Valilai
26
Synthesising COVID-19 Related Research from a Logistics and Supply Chain Perspective . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Işık Özge Yumurtacı Hüseyinoğlu, Ilja Bäumler, and Herbert Kotzab
38
Challenges and Approaches of Non-pharmaceutical Interventions for Airport Operations During Pandemic Situations . . . . . . . . . . . . . . . Ann-Kathrin Rohde, Birte Pupkes, Rafael Mortensen Ernits, Dennis Keiser, Michael Lütjen, and Michael Freitag Identifying Common Elements Within Supply Chain Resilience and Sustainability - An Exploratory Study Based on Bibliographic Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Piotr Warmbier and Aseem Kinra The Impact of Blockchain on Supply Chain Resilience . . . . . . . . . . . . . Anna Kolmykova
52
65 82
vii
viii
Contents
Towards Supply Chain Resilience in Mining Industry: A Literature Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Raúl Castillo-Villagra and Klaus-Dieter Thoben
92
Data Quality in Social Media Analytics for Operations and Supply Chain Performance Management . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104 Fabian Siekmann, Aseem Kinra, and Herbert Kotzab Managing Interfaces Between Smart Factories and Digital Supply Chains . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117 Bennet Zander, Kerstin Lange, and Hans-Dietrich Haasis
Maritime Logistics and Port Operations Container Flow Generation for Maritime Container Terminals . . . . . . . 133 Marvin Kastner, Ole Grasse, and Carlos Jahn Simulation-Based Port Storage Dimensioning to Mitigate Operational Instability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144 Yuri Triska and Enzo Morosini Frazzon Integration of Renewable Energies at Maritime Container Terminals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156 Felix Schütze, Anne Kathrina Schwientek, Ole Grasse, and Carlos Jahn An Appraisal of the Northern European LNG Bunker Ship Fleet . . . . . 168 Antje Roß and Kerstin Lange Investigating the Requirements of Automated Vehicles for Port-internal Logistics of Containers . . . . . . . . . . . . . . . . . . . . . . . . 179 Hendrik Rose, Ann-Kathrin Lange, Johannes Hinckeldeyn, Carlos Jahn, and Jochen Kreutzfeldt Unmanned Vessels and the Law . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 191 Eva Ricarda Lange Low Emission Choices in Freight Transport: Comparing Land and Short Sea Shipping Alternatives . . . . . . . . . . . . . . . . . . . . . . . . . . . 204 Esa Hämäläinen, Tommi Inkinen, and Eunice O. Olaniyi Digital Twin Features for the Intelligent Container . . . . . . . . . . . . . . . . 217 Reiner Jedermann, Walter Lang, Martin Geyer, and Pramod Mahajan
Transportation Networks and Vehicle Routing A New Lower Bound for the Static Dial-a-Ride Problem with Ride and Waiting Time Minimization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 231 Christian Pfeiffer and Arne Schulz
Contents
ix
An Auction-Based Multi-Agent System for the Pickup and Delivery Problem with Autonomous Vehicles and Alternative Locations . . . . . . . 244 Johan Los, Frederik Schulte, Matthijs T. J. Spaan, and Rudy R. Negenborn Strategic Bidding in Decentralized Collaborative Vehicle Routing . . . . . 261 Johan Los, Frederik Schulte, Matthijs T. J. Spaan, and Rudy R. Negenborn Artificial Intelligence in Urban Last Mile Logistics - Status Quo, Potentials and Key Challenges . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 275 Maximilian Engelhardt, Stephan Seeck, and Ben Geier An Inter-organizational Digital Platform for Efficient Container Transportation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 290 Michael Teucke, Eike Broda, and Michael Freitag SRP: A Sustainable Dynamic Ridesharing Platform Utilizing Blockchain Technology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 301 Roshaali Khaturia, Hendro Wicaksono, and Omid Fatahi Valilai Yard Management: Identification and Evaluation of Critical Sub-processes with AHP . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 314 Alina Müller, Thomas Keuschen, and Matthias Klumpp A Cargo Throughput Capacity Quantization Estimation Using Semi-Markov Jump System Filter Within Partial State Delay . . . . . . . . 326 Bingxuan Ren, Tangwen Yin, Hamid Reza Karimi, and Shan Fu Applications of Pixel Oriented Mobility Modelling in Transport & Logistics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 337 H. Niles Perera and H. Y. Ranjit Perera Choosing the Right Technique for the Right Restriction – A DomainSpecific Approach for Enforcing Search-Space Restrictions in Evolutionary Algorithms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 349 Christina Plump, Bernhard J. Berger, and Rolf Drechsler
Production Planning and Scheduling Managing Complexity in Variant-Oriented Manufacturing: A System Dynamics Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 363 Phillip Kießner and H. Niles Perera Dynamic Lot Size Optimization with Reinforcement Learning . . . . . . . . 376 Thomas Voss, Christopher Bode, and Jens Heger
x
Contents
Integrated Simulation-Based Optimization Approach for Production Scheduling: A Use Case Application in a Machining Process . . . . . . . . . 386 Ícaro Romolo Sousa Agostino, Mauricio Randolfo Flores da Silva, Enzo Morosini Frazzon, and Luciana Amaral Stradioto Neto Scheduling Workforce in Decentrally Controlled Production Systems: A Literature Review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 396 Julia Schwemmer, Mathias Kühn, Michael Völker, and Thorsten Schmidt Maintenance 4.0: A Literature Review and SWOT Analysis . . . . . . . . . 409 Danilo Ribamar Sá Ribeiro, Lúcio Galvão Mendes, Fernando Antônio Forcellini, and Enzo Morosini Frazzon Using Supervised Learning to Predict Process Steps for Process Planning of Third-Party Logistics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 423 Marius Veigt, Lennart Steinbacher, and Michael Freitag From Linear to Circular Packaging: Enablers and Challenges in the Fashion Industry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 435 Sarah Pfoser, Katharina Herman, Andrea Massimiani, Patrick Brandtner, and Oliver Schauer
Socio-technical Systems Assessing Driver Fatigue During Urban Traffic Congestion Using ECG Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 449 Nizami Gyulyev, Andrii Galkin, Tibor Schlosser, Silvia Capayova, and Oleksii Lobashov Technology Review for Guiding Persons in Airports and Other Hubs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 462 Axel Börold, Eike Broda, Nicolas Jathe, Dirk Schweers, Tobias Sprodowski, Waldemar Zeitler, and Michael Freitag The Impact of the COVID-19 Pandemic on E-commerce Consumers’ Pro-environmental Behavior . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 474 Simona Koleva and Stanislav Chankov Developing a Serious Game for Intelligent Transportation Systems . . . . 486 Ilja Bäumler, Moritz Elfers, Okan Dogtas, Fynn Gresens, and Sercan Eyigün Author Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 499
Supply Chain Management
The Linkage Between Macro Logistics Capabilities and Micro Firm Performance Towards Framework Development for Supply Chain Performance Measurement Debarshee Bhardwaj(B)
and Aseem Kinra
Professorship for Global Supply Chain Management, Universität Bremen, Bremen, Germany {bhardwaj,kinra}@uni-bremen.de
Abstract. Technological advancements, increased globalization, political, social, and environmental concerns are changing the world we live in. Firms applying a global strategic management approach to enter a new country have to dynamically assess and align their performance objectives and micro capabilities with the changing macro logistics higher-order capabilities and resources at the regional or country level, and vice versa. Higher-order capabilities are present in a specific location and emerge over time as a result of systematic interactions across firms and institutions, whereas micro capabilities and performance objectives are firmwide. Previous research on logistics and supply chain performance measurement hasn’t precisely addressed this macro-micro perspective with a global strategic management view. Furthermore, this aspect associating macro and micro capabilities and performance has been only projected with the help of case study-based empirical investigation within the previous research. To investigate the magnitude of this association, this research uses an integrative literature review strategy that includes a qualitative content analysis of the most important/cited supply chain performance measurement literature. With the help of a macro-micro supply chain performance measurement framework adapted from the supply chain performance measurement literature, the findings bring out the link between various macro logistics higher-order resources and capability elements (input) such as collaboration, knowledge sharing between logistics partners, transportation mode and energy efficiency and the strategic firm performance objectives and dynamic micro capabilities (output) such as sustainability and process innovation. It also extends framework propositions to guide future supply chain performance measurement research with a particular emphasis on global supply chain decision-making. Keywords: Supply chain performance measurement · Higher-order capabilities · Macro logistics capabilities · Micro firm performance
1 Introduction The rapid globalization of developing economy multinational companies (MNCs), as well as their ambition to enter and seek strategic resources in foreign markets, provides an © The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 M. Freitag et al. (Eds.): LDIC 2022, LNLO, pp. 3–14, 2022. https://doi.org/10.1007/978-3-031-05359-7_1
4
D. Bhardwaj and A. Kinra
opportunity to other corporations to learn about the capabilities of these businesses and offer advice to other MNCs beginning on this transformative path. MNCs from emerging economies invest abroad not just to obtain resources, but also to contribute their own experience and knowledge base. Taking a strategic management view, firms entering a new country must be dynamic in determining which firm performance objectives or dynamic micro/firm capabilities may be bolstered by the higher-order macro logistics capabilities and resources, and vice versa. This combination of macro higher-order capabilities, such as a collective invention or knowledge-sharing in R&D networks or government environmental and technological advancement policies, and dynamic micro firm capabilities, which are determined by the knowledge residing in individual firms on how to produce, distribute, or market goods, always determines how effectively a firm performs. Rather than companies per se, here macro higher-order capabilities are characterized by regional advantage, shared capabilities that develop from recurring interactions among geographically confined enterprises and the external infrastructural, institutional, and technological elements, and the business environment in which the company works. Towards this notion, macro logistics higher-order capabilities developed within the country like infrastructure, institutional and other external environmental regularities drive the micro-level firm´s strategic direction, optimizing the firm´s performance objectives such as cost or responsiveness (Barney et al. 2021). Wiengarten et al. (2014) highlight the positive effect of the logistics capabilities of a country on the firm´s integration and operations. A good case in point for country logistics performance assessment is the LPI (Logistics Performance Index) used by the World Bank since the year 2007 also points towards the same direction. This notion taking a macro-micro perspective has been also reposited within the resource-based view and dynamic capability theory. Research constraining supply chain has failed to show this linkage and hence, this aspect needs to be investigated even more. Studies by Hasegan et al. (2018); Irfani et al. (2019) have explicated the methodological boundaries of this linkage with the help of empirical investigation but have not fully extrapolated the theoretical boundary. A holistic theoretical footing is yet to be nurtured in terms of elucidating which macro logistics higher-order capabilities support organizations in enhancing their firm capabilities, which in turn evolve towards performance targets optimization. Adapting a macro-micro perspective framework developed by Rouse and Putterill (2003), this paper aims to explore the extent of this notion within the supply chain performance measurement literature. The research also proposes theoretical notions towards this to guide future supply chain performance measurement research and emerging economy organizations towards leveraging logistics higher-order capabilities for optimizing firm performance with an acute implication on global strategic management and supply chain decision making. The following research questions are developed: RQ1: How are macro logistics higher-order capabilities linked to micro firm capabilities and performance in supply chain performance measurement research? RQ2: What can be learned towards developing a supply chain performance measurement framework that integrates macro capabilities and micro firm performance?
The Linkage Between Macro Logistics Capabilities and Micro Firm Performance
5
The structure of the paper follows as such: the following section elucidates the theoretical foothold of macro higher-order capabilities and dynamic firm capabilities and performance considering dynamic capability view. Next, the approach of the research integrating macro-micro perspective conceptualization is broadened within the supply chain performance measurement literature where an established framework by Rouse and Putterill (2003) is defined and reformed towards attaining the main objective of the paper. Following that, the methodological overview is presented, which explains the content analysis strategy to developing the categories that would answer the research questions. Finally, the results and discussion section connects and elucidates the many macro logistics higher-order capabilities and micro firm capabilities and performance objectives, which are backed up by certain framework propositions.
2 Macro Higher-Order Capabilities and Firm Capabilities Towards Performance Optimization: A Dynamic Capability View A large account of literature developing the linkage between macro or higher-order capabilities and firm capabilities towards sustained performance and competitive advantage has been invariably associated with resource-based theory and dynamic capability theory (Akhtar et al. 2020; Foss 1996). This notion has been postulated towards the ability to integrate, build, and reconfigure internal and external resources, capabilities and competencies to attain a superior competitive advantage and firm performance (Teece et al. 1997). These capabilities and resources are either embedded within the organization and its practices (micro) or reside in the region (macro) emerging from systematic firm interactions or external institutional factors to address rapidly changing contextual environments. The idea of macro or higher-order capabilities (Barney and Clark 2007, Foss 1996; Akhtar et al. 2020) develops from macro and meso perspectives where geographical boundedness is emphasized. This helps to explain why a firm’s origin in terms of location or country has such a strong influence on its global performance and gives comparative and competitive advantages to firms. Higher-order capabilities are non-proprietary and intangible assets that are shared across a group of enterprises and are developed from external institutional support. Examples of higher-order capabilities may include, for example, standards, knowledge-sharing in R&D networks, collective invention, shared behavioral norms, government policies, quality indicators, legal requirements, technology and economic capability, market requirements, economic capabilities (Foss 1996; Akhtar et al. 2020). Firm or dynamic micro capabilities advances the heterogeneity in what firms can accomplish and how well they can do it in terms of developing capabilities. Excess physical, human, and organizational resources are a common occurrence in firms. To some extent, this is a matter of indivisibility. More importantly, it’s a matter of learning and experience impacts, which are a result of the firm’s normal operations - notably within the management team. The significant learning impacts are shown in abilities, which fundamentally differentiate firms and codetermine their opportunity sets or the range of profitable actions that the company can recognize and exploit. It encompasses
6
D. Bhardwaj and A. Kinra
human dynamic capabilities (e.g., human resource knowledge), physical dynamic (e.g., businesses’ geographical position and specialized equipment), and organizational, strong internal corporate governance and control, all of which may be used to implement valueenhancing corporate strategies. This paper examines micro-level dynamic capability both at the individual and organizational levels, as an extension of the Resource-based View (RBV) theory. Looking at micro-level dynamic capabilities in isolation has the disadvantage of not explaining why certain businesses maintain a competitive edge in dynamic contexts while others do not, despite having identical resources. Understanding how micro-and macro-level capabilities interact to achieve higher organizational performance necessitates addressing both at the same time. The whole notion linking macro logistics higher-order capabilities and micro firm capabilities and performance can be assimilated within a chain process within the strategic decision-making sphere. Towards this, macro capabilities influence the decisions of the firms in terms of what internal competencies firms want to enhance. This impacts the firm decision towards enhancing their performance goals and creating competitiveness (McIvor 2013). For eg, if in a country, the government stability and support are quite high, then firms have to decide whether they would want to invest in the collaboration capabilities which can enhance the firm goal such as sustainability and resilience. Also, within the strategic management literature, location capabilities require that international firms build and reconfigure competencies through understanding the host country’s environment and performing actions to leverage it (Pe’er et al. 2008). These capabilities can be a source of advantage and as a consequence a determinant of survival for internationalizing firms. Similarly, the link between decision-making effectiveness and organizational performance is further supported by studies on strategic decision-making. For example, companies with comprehensive and precise information on the anticipated link between choices and outcomes can increase strategic decision effectiveness and strategic decision-making indicates favorable future firm performance.
3 Approach 3.1 Framework for Linking Macro-micro View Towards Supply Chain Performance Measurement Coming back to RQ1, it is possible to conceptualize that there is a strong link between the value, rarity, inimitability of macro capabilities and firm performance and the relationship is mediated by the strategic decision-making effectiveness of the firm. This macro-micro view has been typically established through the lens of a superior logistics performance assessment structure. Superior logistics performance manifests resourceful cross-border operations enabled by countries in terms of efficient and consistent mobility (Roy and Schoenherr 2020). From Kinra et al. (2020), it can be postulated that there is a direct influence of the macro-level country´s logistics capabilities on the micro firm´s performance and its managerial decision-making countenance. For example, logistics infrastructural factors such as ports and transportation mode quality or institutional factors such as customs balance the firm´s strategic performance objectives responsiveness.
The Linkage Between Macro Logistics Capabilities and Micro Firm Performance
7
Rouse and Putterill (2003) have provided a framework for linking macro and micro capabilities and performance objectives and outcomes within the supply chain performance measurement literature. The framework (Fig. 1) extended by (Rouse and Putterill 2003 and Bititci et al. (2000) have the following aspects: Aspect 1: It is not static and evaluation of different dimensions takes place flowing through time in different levels and changing environments. Aspect 2: It shows a linkage as to how the macro capabilities and resource requirements (processes drivers) lead towards the desired strategic performance directions (performance objectives). Aspect 3: It can distinguish between measurements of improvement and measures of control where the planning and evaluation of control processes take place.
Fig. 1. Macro-micro organizational view within supply performance measurement. Adapted from Rouse and Putterill (2003).
Within the literature on country logistics performance assessment, performance measurement has also been suggested in the form of a linked structure connecting the cause (macro resources and capabilities) and result (micro firm performance objectives) determinants (e.g. Bookbinder and Tan 2003; Bowersox et al. 2003), taking a macro-micro organizational view. Anchoring on the RQ1 and RQ2, the study adopts this theoretical framework to see how the macro logistics capabilities can be connected to micro firm performance and provide some propositions towards macro capability-micro performance framework development for future research in supply chain performance measurement. 3.2 Qualitative Integrative Reviews The study adopts a qualitative integrative review (Cooper1989) as similar to Ojo (2019). For which a content analysis research approach is used for the systematic qualitative
8
D. Bhardwaj and A. Kinra
description of the evident content of the literature in supply chain performance measurement (Staudt et al. 2015). In convention with the design of integrative research review, the review is based on the following stages: (a) defining the topic and research questions that will lead the integrative research review (b) determination of a data-gathering strategy and the selection of numerous channels to avoid coverage bias (c) data evaluation and selection, including deciding which data to include in the review using selection criteria (d) Analysis and interpretation of the reviewed literature, including source statistics, a number of retrievals, and lastly reviewed literature (e) presenting the findings. Integrative reviews have many benefits within research. But the main motivation to apply this method has been mainly to evaluate the strength or the extent of the scientific evidence, between related areas of work and identify the learnings and direction for future research. Figure 2 illustrates the process map for the overall research design including research databases, search terms, inclusion and exclusion criteria and the utilized approach for data analysis.
Fig. 2. Process map
As it can be illustrated from Fig. 2, for answering RQ1 and RQ2, a literature review is conducted on the 50 most cited and connected papers overall and then a content analysis approach to applied to classify the 42 articles according to seven major categories: (i) Boundary of the research (ii) nature of paper: empirical or normative (iii) extent of macromicro view (iv) Macro logistics higher-order capabilities (v) Micro firm performance objectives and capabilities (iv) Content segment.
The Linkage Between Macro Logistics Capabilities and Micro Firm Performance
9
The paper shows the classification according to the different literature in terms of the various macro-level capabilities factors developing to the micro firm performance objectives. Citations have long been the primary parameter used by academia to assess the impact or significance of research (Aksnes 2006). Hence, only the most cited and connected papers are included as per bibliographic analysis on histcite application based on the metric LCS (local citation score), which is provided to the highest cited articles within a specific search sample. As journals are widely used in academia for gathering information, the conference proceedings papers, master and doctoral thesis and unpublished working papers are excluded. Finally, to enhance the quality of the review, the 35 best journals evaluated from the Harzing list are considered.
4 Findings and Discussion 4.1 Linkages Between Macro Logistics Higher-Order Capabilities and Micro Firm Performance Objectives and Capabilities In reference to the RQ1, Table 1 elucidates the extent of the macro-micro view in supply chain performance measurement fulfilling the different aspects (page 5) within the literature. Within the table, a categorization of the content in terms of which macro logistics higher-order capabilities are linked to micro firm capabilities and performance objectives is anticipated. After reviewing the 42 short-listed papers 8 relevant linkages were identified, which were then described using the content analysis approach. Table 1. Macro capabilities and micro performance objectives linkages from supply chain performance measurement literature. Boundary
References
Green supply Mollenkopf chain and et al. (2010) sustainability (3)
Nature of paper
Macro logistics higher-order capabilities and resources
Normative Collaboration, knowledge sharing between logistics partners
Hassini et al. Normative Transportation (2012) and mode and energy Empirical efficiency
Micro firm performance objectives and capabilities
Extent of macro-micro view in supply chain performance measurement
Sustainable competitive advantage
Medium: The time aspect has not been addressed
Environmental sustainability, time and cost, competitive advantage
High: All three aspects fulfilled (continued)
10
D. Bhardwaj and A. Kinra Table 1. (continued)
Boundary
References
Nature of paper
Macro logistics higher-order capabilities and resources
Micro firm performance objectives and capabilities
Extent of macro-micro view in supply chain performance measurement
Chiou et al. (2011)
Empirical
Collective learning capabilities and technological advancement within logistics services
Process innovation, green innovation
High: All three aspects fulfilled
Supply chain integration (1)
Nyaga et al. (2010)
Empirical
Firm collaboration
Satisfaction (Trust and commitment)
High: All three aspects fulfilled
Supply chain strategy (2)
Gunasekaran Empirical et al. (2004)
Operational level Strategic level and tactical level outcomes Gunasekaran Normative strategies of firms et al. (2004)
Low: Doesn´t clearly define what are the capabilities or what are the control elements
Humanitarian Beamon & supply Balcik, chain (1) (2008)
Normative Annual cost and recovery empirical (resources)
Response time Medium: The (Outcome) time aspect has not been addressed
Humanitarian Beamon supply (1999) chain (1)
Normative Cost-effectiveness Customer High: All (resources) responsiveness three (Outcome) aspects fulfilled
As it can be seen from Table 1, only 8 articles within the search, which provide evidence towards this linkage were identified. However, out of these 8, 6 of them are showing a high extent of this view as they incorporated at least two out of three aspects for the macro-micro performance view (Refer to aspects, Page5). This evidently shows that this aspect is present within the logistic and supply chain performance measurement literature but the pattern has not been accounted for enough. The different process drivers (macro logistics capabilities) and firm performance objectives and capabilities have been categorized within the table.
The Linkage Between Macro Logistics Capabilities and Micro Firm Performance
11
4.2 Future Research on the Development of Macro Capability-Micro Performance Framework A prominent account of this view has been grasped within the domain of green supply chain and sustainability with three articles (Chiou et al. 2011; Hassini et al. 2012; Mollenkopf et al. 2010) followed by the discipline of supply chain strategy and integration. Various collaborative capabilities such as cooperation, coordination, collective learning and information technology within logistics services of the firms have been considered as the input macro or higher-order capabilities elements for green supply chain performance improvement. This directed the rising amount of research in the adoption of collaborative capabilities by the organizational stakeholder for attaining sustainable performance objectives within the supply chain. This is specifically in line with Wilkinson et al. (2001) demonstrating the responsibilities played by governments and companies in achieving sustainability, with a focus on operations management competencies and human resource management. Towards this, several organizational goals and strategic objectives have also been accounted such as green innovation or sustainable environmental competitive advantage. Network theory which takes inter-organizational collaboration into account at a higher/global level has been suggested for investigation, even more for firms adapting sustainability objectives towards a green and lean global supply chain. The boundary of supply chain integration (Nyaga et al. 2010) has also shown this trait of inter-organizational collaboration as the main capability factor. In the light of answering RQ2 and taking all these notions into account, the following propositions can be formulated for future research towards framework development for supply chain performance measurement integrating logistics macro capabilities and micro firm performance: Proposition 1: Macro inter-organizational collaborative capabilities such as cooperation, coordination, collective information technology and learning should be considered as main capability factors for enhancing a firm´s sustainability performance towards global supply chain strategic decision making. Proposition 2: Logistics and supply chain performance measurement for designing a global supply chain network must relate to optimizing multiple organizational performance goals and strategies. Proposition 3: Network theory which involves inter-firm and inter-functional collaboration as macro logistics capability between multiple actors should be investigated for firms adapting to green and lean global supply chain involved in a continuously emerging and evolving dynamic environment. Extending to this notion of global supply chain network, Teece (2019) suggest that the difference in the growth and performance of the nation is directly affected by the capabilities within firms, their collaboration and countries as a whole. According to the business insolvency outlook (2020), countries such as India and Canada have the highest organizational insolvencies and bankruptcies in the year 2020. One of the major reasons for this is a lack of capability or resilience in terms of innovation, asset orchestration, market creation, knowledge sharing and resource adaptation within the industries.
12
D. Bhardwaj and A. Kinra
Considering a macro logistics higher-order capability, Hassini et al. (2012) have developed a framework for sustainable supply chain management and performance measures within a global context. It demonstrated that the macro logistics capability such as the transportation mode and energy efficiency have a high impact on environmental sustainability, timeliness and cost performance objectives in a global distribution network. Value proposition has been taken as the control measure within the study as there is always a trade-off between cost and customer acceptance within sustainable supply chain management. Towards this, two propositions are developed: Proposition 4: Firms within the same supply chain having conflicting strategies in a global distribution network faces difficulties to align their capabilities. Proposition 5: Assessment of supply chain performance within a global distribution network not only relates to time optimization as a performance objective but is also dependent on coordination between the different roles of the supply chain players. A significant aspect was also shown by Beamon (1999); Beamon and Balcik (2008) within the manufacturing and humanitarian supply chain performance measurement. Framework involving categorization and transformation of performance metrics from resource level to output level has been demonstrated. The time and environmental uncertainty constraints have also been projected in the literature, with flexibility being considered and adopted as a control measure, such as volume or response flexibility, which pertains to the evaluation of a supply chain responding to change in an uncertain environment. Taking this into account, this is the final proposition: Proposition 6: Flexibility should be taken as a very important control measure for improved organizational supply chain performance in an uncertain and evolving environment.
5 Conclusion and Limitations Answering the first research question, the study projected the firms’ behavioral propensity toward collaborative capabilities in the development of a sustainable and green supply chain. Decision-makers gain a broad understanding of macro logistics capability factors such as government regulations, infrastructure, and political conditions, as well as inter-organizational collaborative capabilities, and will no longer be able to ignore their impact when dealing with strategic global and cross-national activities. After that, the paper was able to profile six propositions that contribute to supply chain operations and performance measurement literature for framework development that elaborates the context of understanding logistics macro capabilities while also linking it to specific decision-making aspects at the micro firm level. However, the study lacks an expert-based validation of the different propositions and the linkages provided within the study. Moreover, the study is only limited to a literature review based on a limited critical number of studies which is not enough for implications within such a huge research area, supply chain performance measurement. In the future, this needs to be investigated even more and a pattern based on the various research discipline within logistics and supply chain management needs to be developed.
The Linkage Between Macro Logistics Capabilities and Micro Firm Performance
13
References 1. Akhtar, P., Ullah, S., Amin, S.H., Kabra, G., Shaw, S.: Dynamic capabilities and environmental sustainability for emerging economies’ multinational enterprises. Int. Stud. Manag. Organ. 50(1), 27–42 (2020) 2. Aksnes, D.W.: Citation rates and perceptions of scientific contribution. J. Am. Soc. Inform. Sci. Technol. 57(2), 169–185 (2006) 3. Barney, J.B., Clark, D.N.: Resource-based theory.: In Resource Based Theory: Creating and Sustaining Competitive Advantage. Oxford University Press on Demand (2007) 4. Barney, J.B., Ketchen, D.J., Wright, M.: Resource-Based theory and the value creation framework. J. Manag. 47(7), 1936–1955 (2021) 5. Beamon, B.M.: Measuring supply chain performance. Int. J. Oper. Prod. Manag. 19(3), 275– 292 (1999) 6. Beamon, B.M., Balcik, B.: Performance measurement in humanitarian relief chains. Int. J. Public Sect. Manag. 21(1), 4–25 (2008) 7. Bititci, U.S., Turner, T., Begemann, C.: Dynamics of performance measurement systems. Int. J. Oper. Prod. Manag. 20(6), 692–704 (2000) 8. Bleady, A., Ali, A.H., Ibrahim, S.B.: Dynamic capabilities theory: pinning down a shifting concept. Acad. Acc. Financ. Stud. J. 22(2), 1–16 (2018) 9. Bookbinder, J.H., Tan, C.S.: Comparison of Asian and European logistics systems. Int. J. Phys. Distrib. Logist. Manag. 33(1), 36–58 (2003) 10. Bowersox, D.J., Calantone, R.J., Rodrigues, A.M.: Estimation of global logistics expenditures using neural networks. J. Bus. Logist. 24(2), 21–36 (2003) 11. Chiou, T.Y., Chan, H.K., Lettice, F., Chung, S.H.: The influence of greening the suppliers and green innovation on environmental performance and competitive advantage in Taiwan. Transp. Res. Part E: Logistic. Transp. Rev. 47(6), 822–836 (2011) 12. Cooper, H.M.: Integrating research: A guide for literature reviews. Sage Publications, Inc (1989) 13. Dörnhöfer, M., Schröder, F., Günthner, W.A.: Logistics performance measurement system for the automotive industry. Logist. Res. 9(1), 1–26 (2016) 14. Foss, N.J.: Higher-order industrial capabilities and competitive advantage. J. Ind. Stud. 3(1), 1–20 (1996) 15. Gunasekaran, A., Patel, C., McGaughey, R.E.: A framework for supply chain performance measurement. Int. J. Prod. Econ. 87(3), 333–347 (2004) 16. Hasegan, M.F., Nudurupati, S.S., Childe, S.J.: Predicting performance – a dynamic capability view. Int. J. Oper. Prod. Manag. 38(11), 2192–2213 (2018) 17. Hassini, E., Surti, C., Searcy, C.: A literature review and a case study of sustainable supply chains with a focus on metrics. Int. J. Prod. Econ. 140(1), 69–82 (2012) 18. Irfani, D.P., Wibisono, D., Basri, M.H.: Design of a logistics performance management system based on the system dynamics model. Meas. Bus. Excell. 23(3), 269–291 (2019) 19. Kerr, S.: Accounting, budgeting and control systems in their organizational context: comments by the discussant. Acc. Organ. Soc. 8(2–3), 171–174 (1983) 20. Kinra, A., Hald, K.S., Mukkamala, R.R., Vatrapu, R.: An unstructured big data approach for country logistics performance assessment in global supply chains. Int. J. Oper. Prod. Manag. 40(4), 439–458 (2020) 21. McIvor, R.: Understanding the manufacturing location decision: the case for the transaction cost and capability perspectives. J. Supply Chain Manag. 49(2), 23–26 (2013) 22. Mollenkopf, D., Stolze, H., Tate, W.L., Ueltschy, M.: Green, lean, and global supply chains. Int. J. Phys. Distrib. Logist. Manag. 40(1–2), 14–41 (2010)
14
D. Bhardwaj and A. Kinra
23. Nudurupati, S.S., Garengo, P., Bititci, U.S.: Impact of the changing business environment on performance measurement and management practices. Int. J. Prod. Econ. 232, 1–15 (2021) 24. Nyaga, G.N., Whipple, J.M., Lynch, D.F.: Examining supply chain relationships: do buyer and supplier perspectives on collaborative relationships differ? J. Oper. Manag. 28(2), 101–114 (2010) 25. Ojo, T.K.: Quality of public transport service: an integrative review and research agenda. Transp. Lett. 11(2), 104–116 (2019) 26. Pe’er, A., Vertinsky, I., King, A.: Who enters, where and why? The influence of capabilities and initial resource endowments on the location choices of de novo enterprises. Strategic Organ. 6(2), 119–149 (2008) 27. Rouse, P., Putterill, M.: An integral framework for performance measurement. Manag. Decis. 41(8), 791–805 (2003) 28. Roy, V., Schoenherr, T.: Implications of Sectoral Logistical Capabilities for Export Competitiveness: A Public Policy Perspective for Interventions in the Logistics Sector. IEEE Transactions on Engineering Management, 1–14 (2020) 29. Staudt, F.H., Alpan, G., Di Mascolo, M., Rodriguez, C.M.T.: Warehouse performance measurement: A literature review. Int. J. Prod. Res. 53(18), 5524–5544 (2015) 30. Teece, D.J., Pisano, G., Shuen, A.: Dynamic capabilities and strategic management. Strateg. Manag. J. 18(7), 509–533 (1997) 31. Teece, D.J.: A capability theory of the firm: an economics and (strategic) management perspective. N. Z. Econ. Pap. 53(1), 1–43 (2019) 32. Wiengarten, F., Pagell, M., Ahmed, M.U., Gimenez, C.: Do a country’s logistical capabilities moderate the external integration performance relationship? J. Oper. Manag. 32(1–2), 51–63 (2014) 33. Wilkinson, A., Hill, M., Gollan, P.: The sustainability debate. Int. J. Oper. Prod. Manag. 21(12), 1492–1502 (2001)
Integrating Regional Food Manufacturers into Grocery Retail Supply Chains in Germany Janik Serkowsky1 , Herbert Kotzab2(B)
, and Julia Fischer3
1 Convivo, Weyhe, Germany 2 University of Bremen, Bremen, Germany
[email protected] 3 Brandenburgische Technische Universität Cottbus-Senftenberg, Cottbus, Germany
Abstract. This paper examines the integration of regional food manufacturers into the supply chains of large grocery retailers. It discusses the current integration problems and challenges of those supply chain concepts which are presently used in the food retail sector. In this paper, we assessed direct delivery, milkrun and intermediate stations with regard to their suitability for integrating regional food manufacturers into the structures of large grocery retailers and used an Analytical Hierarchy Process (AHP) approach for the evaluation. The results of the AHP show that milkruns with an intermediate station are best suited for this type of supplier integration. Keywords: Regional food manufacturer · Grocery retail · Milkrun · AHP
1 Introduction More than a 37 million consumers in Germany prefer to purchase regional products [1]. When it comes to the major reasons for buying regional food, German consumers indicate the support of regional economy, the better freshness of products, the positive effects for the environment as well as lower emissions [2–4]. Also, supermarket chains offer more and more regional products in their shelves, whereby the two largest German grocery retailers (in terms of sales volume) have an advantage as they operate their store networks with local independent retailers as opposed to those retailers with a centrally controlled store network [5, 6]. The Bavarian EdekaSouth-Group increased the sales of their regional private label by nearly 20% within one year [7]. Besides the potential positive effects of shorter supply chains and thus an improved environmental footprint, a higher share of regional products in the grocer retailers’ product assortment brings some challenges to their efficiency-driven high-volume purchasing strategies. Regional food producers though face a change in their distribution systems, which were traditionally organized as direct yard sales where consumers have to come to the yard in order to purchase the items. As consumers may not be willing to drive to the yards, producers could either deliver to the consumers’ homes or integrate with large grocery retailers. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 M. Freitag et al. (Eds.): LDIC 2022, LNLO, pp. 15–25, 2022. https://doi.org/10.1007/978-3-031-05359-7_2
16
J. Serkowsky et al.
In this paper we are interested to see which conceptual concept large grocery retailers can use to integrate regional food manufactures in their supply chain. We approach this question by examining the identified concepts and assess their supply chain integration suitability with an Analytical Hierarchy Process (AHP) approach. In order to solve our research problem, we first define the term ‘regionality’ and discuss further the challenges of regional supply chain integration. Afterwards we analyze current supplier structures of large grocery retailers (in terms of sales volume) and assess the attributes and indicators that are applicable for our evaluation and comparison of supply chain integration concepts.
2 Regionality and Retailing 2.1 Regional Food There is no common understanding on regional food as it is difficult to define regionality. According to Hausladen [8], the definition of the term ‘region’ includes criteria regarding geography, economy, culture or climate in order to come up with a unified spatial understanding of a region. Ermann [9] extends this by subjective aspects by using relations and/or the identification with an area. When it comes to regional food, the information that the product is from a certain region is crucial as Wegmann [10] points out. Besides the origin of the product, consumers also want to know how the respective region is defined [10]. Therefore, the European Union [11] refers to two seals of approval, the protected designation of origin and the protected geographical indication. However, there are additional seals, e.g. the regional window (Regionalfenster) in Germany, that shows from where the products come from, where it was processed as well as the share of regional ingredients [12]. For the purpose of this paper, we follow Buxel [13] and define regional food as food that stems from a radius of up to 50 km for the place of production. 2.2 Grocery Retail Retailing as such can be defined from a functional as well as institutional perspective [14]. Retailing from a functional perspective refers to the procurement and distribution of final products without further processing and focuses on the value-adding processes only. In case of defining retailing from an institutional point of view, it focuses on the firm that executes these functions [15]. Following an institutional understanding, different forms of retailers exist, depending whether they operate with physical stores, hybrid store formats or without stores (stationary, ambulant, non-stationary). When it comes to grocery retailers, stationary retail formats such as supermarkets, discount stores or hypermarkets are still dominating the sector [16]. The more than 34,000 food retailers in Germany represent a sales volume of nearly 200 billion Euro. The largest players in this sector are Edeka with a market share of more than 25% followed by Rewe-Group, Schwarz Group and Aldi-Group. These structures remained stable for the last 10 years [17].
Integrating Regional Food Manufacturers into Grocery Retail Supply Chains
17
2.3 Trends in Grocery Retail Single-Households. The development of the German population shows an aging of society, where more and more people tend to live in urban environments as well as in single households. Due to time pressure, there is less time for preparing meals, thus ready-made products, restaurant visits or food delivery services are becoming more popular [18]. Convenience Products. As a consequence, the market share of convenience products which are easy to prepare and to serve increases. Luetticken [19] differs hereby between five stages between basic stage (e.g. fruit or vegetables) to ready-to-eat (e.g. ready-made salad). Food Markets. These are either fixed or mobile locations that offer at small space a comprehensive choice of fresh-prepared food (e.g. food trucks) especially for the lunch break in city-office areas [20]. Such markets are also found as shopping islands in stationary grocery retail stores such as the bakery station or a sushi bar [18]. Importance of Regionality When Buying Food. The share of regional foods of grocery retail sales is around 20 per cent, which is larger than the share of organic food [21]. Wegmann [22] indicates the increasing importance of regionality for consumers who are also willing to pay more for regional food. However, this importance depends on the food category.
3 Developing a Frame of Reference for Regional Supply Chain Concepts for Regional Food Manufacturers In this section, we present the supply chain concepts as well as the attributes and indicators with which these concepts will be assessed. 3.1 General Structure of Grocery Retail Supply Chains Grocery Retail Supply Chain Structure. Nietsche and Fiegel [23] indicate that grocery retailers depend on efficient and reliable supply chains. These retailers operate a network of physical stores and distribution facilities and purchase their products mainly in bulk from manufacturers and ship them to their distribution facilities. From there, products are further delivered to the stores. The large assortments lead to various product requirements that need to be considered, grocery retailers apply a combination of different sourcing, delivery, and storage concepts. Sourcing. Grocery retailers use global as well as local sourcing strategies [24, 25]. The overall goal with global sourcing is to procure products on an international level in order to achieve low purchase prices due to economies of scale. Local sourcing reduces transportation costs, however purchasing prices may be higher as purchasing volumes are lower. Depending on the number of suppliers from which a certain product is procured, we can differ between single and multi-sourcing [25].
18
J. Serkowsky et al.
Delivery. Direct delivery is chosen if transport capacity can be fully exploited and if grocery retailers do not have problems to handle the large incoming flows. In case of smaller transport volumes, consolidation can be achieved by utilizing milk runs [27]. Storage. Hereby we can differ between central and regional distribution center networks or cross docking facilities, depending whether products are intermediately stored and later distributed or only processed for break-bulk and/or consolidation processes and immediately transported to the stores [27].
3.2 Integration Challenges A large grocery retailer who wants to increase its share of regional food is facing challenges in sourcing as well as the regional food manufacturer in distribution [23]. Furthermore, there are challenges for both parties when integrating their information flow systems. Sourcing. The required quantity of products may not be sourced from one supplier, but from many, who need to guarantee a high product quality on a continuous level [28]. Thereby, seasonality issues may negatively impact this goal, so that there might be times where there are no products offered at all [23]. Another issue refers to purchasing cycles and delivery times that need to be due in time so that products are made available in the stores [28]. The German Rewe Group has therefore implemented a special regional supplier collaboration program [29]. Finally, traceability is also an important issue to be considered as there is a legal requirement to provide information about the origin to authorities [30]. Distribution. The distribution of regional food from a manufacturer to a retail store can either be executed by the manufacturer or by a logistics service provider. This depends more or less on the distributed quantities with which transport capacity can be fully utilized. This also depends on other product requirements such as cooling that needs special vehicles [31]. This goes also in line with storage requirements depending on product type which may hinder a consolidated transport and/or storage [23]. Information Flow. It is necessary to provide transparency in regards of product origin and supplier structures also in order to fulfill traceability issues [27]. In addition, there are integration aspects when linking the various IT systems of large grocery retailers with small regional food manufacturers that can negatively impact the compatibility of the involved system structures [27].
3.3 Final Model of Alternatives Based on the previously presented notions as well as an analysis of literature that documents some empirical evidence on how regional food manufacturers organize their supply chains, following Fig. 1 summarizes the supply chain concepts that will be further evaluated on their integration suitability with the supply chains for large grocery retailers.
Integrating Regional Food Manufacturers into Grocery Retail Supply Chains
19
Fig. 1. Final model of supply chain concepts
The practical case evidence showed a focus on a multisourcing strategy where the product quantities of many regional food manufacturers are consolidated in order to meet the desired demand. Thus, we excluded the single sourcing strategy as the quantities of one regional food manufacturer may not be sufficient to meet the complete demand of a retailer for this particular product. Even though regional food manufacturers execute their supply chain operations in many different ways, the basic structure can be condensed to direct deliveries, milkruns and intermediate stations in combination with either direct delivery or milkrun. For the purpose of this paper, we made no difference whether the intermediate station is operated by the food manufacturer or by the retailer. 3.4 Attributes and Indicators for Decision Making The different possibilities will be assessed by using a mix of following attributes which refer to economic, technical as well as logistical indicators [32–34]: • Economic indicators: o Low current costs as additional costs cannot be easily transferred to end-user prices o Low investment costs as the existing logistics systems shall be easily adapted to include regional products • Technical indicators: o Reliable traceability to assure the regional character of products and their origin o Consideration of storage conditions to deal with the different storage conditions of the products • Logistical indicators: o Provision of sufficient quantity of regional products as a retailer want to offer these in all stores within a certain area
20
J. Serkowsky et al.
o Adherence to delivery times to guarantee frictionless logistics processes as well as full shelves o Efficient management of transport distances and routes in order to cope with small supplier structures These attributes and indicators will be used to compare the previously presented supply chain concepts.
4 AHP-Methodology We consider the integration of regional food manufactures into large grocery retail supply chains as a complex decision problem as the assessment of the various requirements refers to qualitative characteristic. Thus, a method is required that is able to cope with these qualitative and subjective evaluation issues. Consequently, we use the AHP as this approach is applicable for a multi-criteria decision problem, where the problem and its decision criteria are divided into sub problems and sub criteria so that a superordinate group can only be influenced by a subgroup (see Saaty 1990 or Goepel 2018). According to Vargas and Saaty (2012), our decision model is divided as outlined in Fig. 2. The decision problem at hand refers to the selection of a suitable supply chain concept that is able to integrate regional food manufacturers with the supply chains of large grocery retailers. The chosen decision criteria and respective sub criteria refer to the identified decision indicators, which were presented in the previous Sect. 3.3. The chosen alternatives include supply chain concepts as outlined in Fig. 1. Following the notions of Saaty (1994), we made pairwise comparisons of the relative evaluation for each criterion. Then the pairwise comparison results were aggregated to priorities and calculated a total priority result in order to identify the integration suitability of the selected supply chain concepts. We used the AHP Online System (AHP-OS) to perform the AHP (Goepel 2018) with secondary data.
Fig. 2. Hierarchical structure of the decision problem
Integrating Regional Food Manufacturers into Grocery Retail Supply Chains
21
5 Findings Figure 3 summarizes the results of our AHP-assessment and shows the priority results as well as the evaluation of comparing the alternatives.
Fig. 3. Results of the AHP-evaluation
The dominance of logistical and technical criteria is explained by the fact that a supply chain concept has to fulfil first its main purpose properly and has secondly to correspond with some legal requirements. On the sub criteria level it is evident that storage conditions are more important than delivery times and respective transport routes. Storage conditions have an impact on the product quality, thus this requirement is more important than traceability. Finally, current costs are more important than investment costs as the interest payments for leveraged investments belong also to current costs. Taking these conditions, the alternatives were further pairwise evaluated. With a total priority of nearly 31%, the intermediate station with milkruns appears to be the best integration alternative. Within this concept, retailers can expect the provision of the necessary quantities of products as well as guaranteeing a high adherence to delivery times. Furthermore, the concepts allow better storage conditions than the other concepts. Slightly behind ranks the concept of intermediate stations with direct delivery, which offers even better adherences to delivery times and storage conditions than intermediate stations with milkruns. It appears that intermediate stations allow necessary consolidation advantages in order to exploit given transport and storage capacities.
6 Conclusions Offering an increasing share of regional products in grocery retailing requires an extensive collaboration between large food retailers and regional food manufacturers. Supply chain integration plays thereby an important role and this paper shows on a conceptual level, which supply chain concepts can be used to realize a successful integration.
22
J. Serkowsky et al.
Due to the low quantities that individual local food producers are offering as well as due to the supply chain structures large grocery retailers are executing, there is a need for concepts that allow consolidation. This allows to keep quantity as well as delivery time promises. Consequently, our findings identified intermediate stations with milkruns as well as direct delivery as favorable solutions. As our findings are based on a more or less conceptual understanding, we suggest an empirical validation of our results by confronting them with grocery retail reality. This refers to a validation of our understanding of regionality, which is limited to an area of 50 km around a location. Recent studies show grocery retailers indicate to source their regional products within a range of 30 km of a supermarket store location, however further analyses identified ranges of more than 100 km on average. We also assumed the up- as well as downstream logistics processes of a retailer to be organized either as direct delivery or milkrun. We suggest confronting our suggestion with crossdocking concepts operated by retailers as well as manufacturers. We also recommend local food manufacturers to work more closely together in order to consolidate not only product quantities but also supplier power against large retailers as well as to examine the possibilities of outsourcing logistics processes to logistics service providers. From a societal perspective, we suggest increasing the integration of regional food manufacturers into the supply chains of large grocery retailers. Food grocery retailers fulfill with their assortment function and choices an important gate-keeper function for more sustainable consumer behavior. They also are able to change set up of existing supply chains towards more climate-friendly supply chains with short distances, lower emissions and positive consequences for the regional economy. As such, these retailers should take their chance using their power of forcing a steering effect in these two areas.
References 1. Leopold, S.: Wie sich das Verbraucherverhalten ändert [How consumer behaviour is changing]. https://www.agrarheute.com/land-leben/verbraucherverhalten-veraendert-572840. Accessed 03 July 2021 2. Umweltbundesamt: “Grüne” Produkte: Marktzahlen [‘Green’ products: market numbers]. https://www.umweltbundesamt.de/daten/private-haushalte-konsum/konsum-produkte/ gruene-produkte-marktzahlen#umsatz-mit-grunen-produkten. Accessed 21 July 2021 3. Zukunftsinstitut. https://onlineshop.zukunftsinstitut.de/shop/food-report-2020/. Accessed 30 June 2021 4. Deter, A.: EDEKA Südbayern setzt auf regionale Produkte heimischer Bauern [EDEKA Südbayern embraces regional products of local farmers]. https://www.topagrar.com/sue dplus/news/edeka-suedbayern-baut-seine-position-als-fuehrender-lebensmittelhaendler-inder-region-aus-12624373.html. Accessed 30 June 2021 5. Kavak-Küstner, Y.: Regionales Sourcing als logistisches Erfolgsmodell [regional sourcing as logistical success tool], https://www.bvl.de/blog/regionales-sourcing-als-logistisches-erfolg smodell. Accessed 03 Apr 2020 6. BR24: Supermärkte setzen verstärkt auf regionale Lebensmittel [Supermarkets focus on regional food]. https://www.br.de/nachrichten/wirtschaft/supermaerkte-setzen-verstaerktauf-regionale-lebensmittel,SaHxfWf. Accessed 30 Jun 2021
Integrating Regional Food Manufacturers into Grocery Retail Supply Chains
23
7. Hausladen, H.: Regionales Marketing: ein Marketing-Management-Ansatz für kleinräumige Kooperationsprojekte zur Erzielung regionaler Wettbewerbsvorteile, Marketing der Agrar- und Ernährungswirtschaft [regional marketing: a marketing management approach for regional cooperation projects to gain regional sustainable advantages, marketing of agriand food industry], Wissenschaftsverlag Vauk, Kiel (2001) 8. Ermann, U.: Regionalprodukte: Vernetzungen und Grenzziehungen bei der Regionalisierung von Nahrungsmitteln [regional products: linkages and border lines of regionlisation of food]. Franz Steiner, Stuttgart (2005) 9. Wegmann, C.: Lebensmittelmarketing: Produktinnovationen – Produktgestaltung – Werbung – Vertrieb [food marketing: product innovations – product design – promotion – sales], Gabler, Wiesbaden GmbH (2020) 10. Der Rat der Europäischen Union: Verordnung (EG) Nr. 510/2006 des Rates vom 20. März 2006 zum Schutz von geografischen Angaben und Ursprungsbezeichnungen für Agrarerzeugnisse und Lebensmittel [COUNCIL REGULATION (EC) No 510/2006, of 20 March 2006 on the protection of geographical indications and designations of origin for agricultural products and foodstuffs. https://eur-lex.europa.eu/legal-content/DE/ALL/?uri=CELEX%3A3200 6R0510. Accessed13 July 2021 11. Verbraucherzentrale: Regionale Lebensmittel [regional food]. https://www.verbraucherzent rale.de/wissen/lebensmittel/kennzeichnung-und-inhaltsstoffe/regionale-lebensmittel-11403. Accessed 22 Apr 2020 12. Buxel, H.: Schmeckt die Region? Verbrauchererwartungen im Hinblick auf regionale Lebensmittel. Leitfaden für zielgruppenspezifische Kommunikations- und Marketingkonzepte für Erzeuger und Vermarkter regionaler Lebensmittel [does the region tastes good? Consumer expectations in regards to regional food. A guide for target group specific communication and marketing concepts for food producers and marketers of regional food], https://www.geschmackstage.de/uploads/content_article/attachment/79/Fol der_Schmeckt_die_Region_2017_IT.pdf. Accessed 13 May 2020 13. Gittenberger, E.: Betriebsformenwahl älterer KonsumentInnen [store selection of elderly consumers]. Peter Lang, Frankfurt a.M (2018) 14. Kotzab, H.: Handel im Spannungsfeld von marketing, distribution und kooperation [retail in the tension field of marketing, distribution and cooperation]. In: Holzmüller, H., Schuh, A. (eds.): Innovationen im sektoralen Marketing: Festschrift zum 60. Geburtstag von Fritz Scheuch [Innovation in sectoral marketing. Commemorative for the 60th birthday of Fritz Scheuch]. pp. 53–69, Springer, Heidelberg (2005) 15. Nielsen: Nielsen Consumers Deutschland. Verbraucher – Handel – Werbung [Nielsen Consumers Germany. Consumers – Retail – Advertising]. The Nielsen Company, Frankfurt (2021) 16. BE: BVE-Jahresbericht 2019 [BE annual report 2019], https://www.bve-online.de/presse/inf othek/publikationen-jahresbericht/bve-jahresbericht-ernaehrungsindustrie-2019. Accessed 22 Apr 2020 17. KPMG: Trends im Handel 2025 [trends in retailing 2025], https://einzelhandel.de/images/pre sse/Studie_Trends_Handel_2025.pdf. Accessed 28 Apr 2020 18. Luetticken, I.: Convenience-Produkte – Lebensmittel mit dem Zusatznutzen Bequemlichkeit [convenience products – food with the additional utitility of convenience]. https://www.dlr.rlp. de/Internet/global/themen.nsf/0/78559B3A80661F56C12570E6002EFAFA?. Accessed 06 Aug 2020 19. Nieswandt, R.: Streetfood & Märkte in Köln [Street food and markets in Cologne]. https://blog. koelntourismus.de/kulinarische-entdeckungen/streetfood-und-maerkte-in-koeln/. Accessed 06 Aug 2020 20. Handelsverband: HDE-Zahlenspiegel 2020. https://einzelhandel.de/publikationen-hde/zah lenspiegel. Accessed 08 Aug 2020
24
J. Serkowsky et al.
21. BMEL: Deutschland, wie es isst – der BMEL-Ernährungsrport 2021 [Germany, as it eats – the BMEL nutrition report 2021]. https://www.bmel.de/DE/themen/ernaehrung/ernaehrungsrepo rt2021.html. Accessed 24 July 2021 22. Wegmann, C.: Regionalität von Lebensmitteln aus Marketingsicht [regionality of food from a marketing perspective]. J. für Verbraucherschutz und Lebensmittelsicherheit 10(1), 57–63 (2015) 23. Nitsche, B., Figiel, A.: Zukunftstrends in der Lebensmittellogistik – Herausforderungen und Lösungsimpulse [Future trends in food logistics, challenges and solutions] https://doi.org/10. 14279/DEPOSITONCE-5122 (2016) 24. Fröhlich-Glantschnig, E.: SWOT-Analyse – Möglichkeiten einer strategischen Beschaffungsanalyse [SWOT analysis – possibilities of a strategic sourcing analysis]. In: FröhlichGlantschnig, E. (ed.) Marketing im Perspektivenwechsel. Festschrift für Udo Koppelmann [Perspective changes in marketing. Commemorative for Udo Koppelmann], 1–16. Springer, Berlin, Heidelberg (2005). https://doi.org/10.1007/3-540-28560-1_1 25. Wannenwetsch, H.: Integrierte Materialwirtschaft und Logistik: Beschaffung, Logistik, Materialwirtschaft und Produktion, 2, überarbeitete Auflage [integrated material management and logistics: sourcing, logistics, material management and production, 2nd revised edition]. Springer, Berlin (2004) 26. Hessenberger, M.: Innovative Logistik Versorgungsstrategien, Standortkonzepte, Steuerungselemente [innovative logistics, sourcing strategies, location concepts, controlling elements]. Gabler Verlag, Wiesbaden (1997) 27. Chopra, S., Meindl, P.: Supply chain management: Strategy, planning, and operation. 7th edition. Always learning. Pearson, Boston MA (2019) 28. Vogt, L., Schütz, K., Mergenthaler, M.: Regionalisierung von Lieferketten in der Ernährungswirtschaft – Herausforderungen und Ansatzpunkte. [regionalisation of supply chains in food industry – challenges and starting points]. Forschungsberichte des Fachbereichs Agrarwirtschaft Soest. Nr. 39 (2016) 29. Anonymous: Wunsch nach Regionalität im Handel [Desire towards regionality in grocery retail]. https://www.biopress.de/de/inhalte/details/7474/wunsch-nach-regionalitaet-imhandel.html. Accessed 08 Jun 2020 30. Der Rat der Europäischen Union: Verordnung (EG) Nr. 178/2002 des Europäischen Parlaments und des Rates vom 28. Januar 2002 zur Festlegung der allgemeinen Grundsätze und Anforderungen des Lebensmittelrechts, zur Errichtung der Europäischen Behörde für Lebensmittelsicherheit und zur Festlegung von Verfahren zur Lebensmittelsicherheit [Regulation (EC) No 178/2002 of the European Parliament and of the Council of 28 January 2002 laying down the general principles and requirements of food law, establishing the European Food Safety Authority and laying down procedures in matters of food safety]. https://eur-lex. europa.eu/legal-content/DE/ALL/?uri=celex:32002R0178. Accessed 08 Aug 2020 31. Reindl, L.: Logistik für nachhaltig erzeugte Lebensmittel in der Region Freiburg - Überblick und Optimierungsvorschläge [logistics for sustainable produced food in the region Freiburg overview and optimisation suggestions]. Forschungsgesellschaft Die Agronauten e.V. (2012) 32. Disselkamp, M.: Lieferantenrating: Instrumente, Kriterien, Checklisten [supplier rating: tools, criteria, check lists]. Gabler, Wiesbaden (2013) 33. Koether, R.: Distributionslogistik: effiziente Absicherung der Lieferfähigkeit, 2, aktualisierte und überarbeitete Auflage [distribution logistics: efficiently ensuring delivery ability, 2nd updated and revised edition]. Springer-Gabler, Wiesbaden (2014) 34. Breckner, C.: E-Innovationsmanagement in der Automobilindustrie (Technologiemanagement, Innovation und Beratung) [E-innvation management in the automotive industry – technology management, innovation and consulting]. Eul Verlag, Köln (2004) 35. Saaty, T.: How to make a decision: The analytic hierarchy process. Eur. J. Oper. Res. 48(1), 9–26 (1990)
Integrating Regional Food Manufacturers into Grocery Retail Supply Chains
25
36. Goepel, K.: Implementation of an online software tool for the analytic hierarchy process (AHP-OS). Int. J. Anal. Hierarchy Process 10(3), 469–487 (2018) 37. Vargas, L.G., Saaty, T.: Models, Methods, Concepts & Applications of The Analytic Hierarchy Process. Springer, US, Boston, MA (2012) 38. Saaty, T.: Fundamentals of Decision Making and Priority Theory with the Analytic Hierarchy Process, 1st edn. Analytic hierarchy process series. RWS Publications, Pittsburgh, PA (1994) 39. Anonymous: ZDF-Doku offenbart Erstaunliches: Was “regional” bei Edeka und Rewe wirklich bedeutet [ZDF-documentary shows amazing: what does “regional” really mean for Edeka and Rewe]. https://www.watson.de/leben/nachhaltigkeit/807117365-was-regionalbei-edeka-und-rewe-wirklich-bedeutet-zdf-doku-offenbart-erstaunliches. Accessed 14 July 2021
Enhancement of Crowd Logistics Model in an E-Commerce Scenario Using Blockchain-Based Decentralized Application Karthikeyan Navendan, Hendro Wicaksono , and Omid Fatahi Valilai(B) Jacobs University Bremen, Campus Ring 1, 28759 Bremen, Germany {K.Navendan,H.Wicaksono,O.FatahiValilai}@Jacobs-University.de
Abstract. Globalization and developments in technology have contributed to the growth of numerous industries around the globe which are creating major impacts on today’s supply chains. The supply chain has drastically changed under the open-ended influence of globalization. The need for different types of mobilities is increasing due to urbanization and population growth, rapid development in the E-commerce industry, and the growing expectation of customers. Crowd logistics is one of these techniques that is gaining rapid attention in the logistics industries and many start-ups have started using this method in their business models. This paper has investigated the crowd logistics and the challenges like user trust, data safety and security, security of the financial transactions for both the customer and the crowd and tracking service quality. Using the Blockchain technology, an e-commerce crowd logistics conceptual model is proposed. Moreover, through different scenarios the capabilities of the proposed model besides the detailed flow of the business processes have been discussed. The conceptual model of a Blockchain-based crowd logistics has used the functionalities of Blockchain such as the smart contracts and the DApps and especially has increased the flexibility of the crowd logistics system. Keywords: Crowd logistics · Blockchain technology · Supply chain management
1 Introduction Today’s supply chains have been affected by the globalization and the introduced technologies [1]. In the past decade, the global industrial sector is experiencing a transition from the industrial economy to a digital and networked economy [2]. With the advancement of information technology, new electronic commerce business models have been emerging and the authors [2] states that businesses and the markets are no longer confined to a particular geographical area. They are linked together, forming a network of markets. Over the past decade, the consumption pattern due to electronic commerce growth has changed remarkably, and to handle this situation new logistics business models and different types of distribution systems have emerged [3]. There is a tremendous requirement to deliver the products at the right time, at the right place, at the right quantity, and © The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 M. Freitag et al. (Eds.): LDIC 2022, LNLO, pp. 26–37, 2022. https://doi.org/10.1007/978-3-031-05359-7_3
Enhancement of Crowd Logistics Model in an E-Commerce Scenario
27
to the right person is of utmost importance and is a very challenging task for logistics corporations [4]. The need for different types of mobilities is increasing due to urbanization and population growth, rapid development in the E-commerce industry, and the growing expectation of customers. To cope with these developments, new logistic methods are being implemented and tested. Crowd logistics is one of these techniques that is gaining rapid attention in the logistics industries and many start-ups have started using this method in their process. Though crowd logistics or peer-to-peer logistics have lots of advantages; this technique also possesses many challenges. According to [5] logistics is the involvement of people or a group of people for the delivery of goods or services, and crowd logistics or peer to peer logistics is said to be an involvement of unknown individuals; for example, the crowd, delivers products to crowdsourcer. The major challenge in crowd logistics is the trust between the crowdsourcers and the crowdsources. It is also mentioned [6] that the stress in CL (Crowd Logistics) or P2P (Peer to Peer) logistics is majorly due to unknown crowd identity. The crowdsourcers believe that sharing their private data with the unknown crowd is considered to be a threat [5]. The backbone for crowd logistics is the use of ICTs. But the use of ICTs has data safety, security, and privacy concerns; for example, the location-based services (LBS) that are used for connecting the receiver and deliverer [5]. The CL platforms take care of the trust part by collecting the crowd’s personal details such as the license, insurance, proof of dependable vehicles, etc. [7]. The main challenge of any crowdsourcing platform is to incentivize and engage a large enough crowd who is willing to participate and also capable of adding value to the system [8]. Discussions related to incentives for the crowds are less discussed and this is essential for the motivation of the crowd to perform well. It also emphasized that the security of the financial transactions to develop trust with the crowd is so crucial [7]. Crowd logistics require a large database of individuals, or critical mass to operate successfully. But the major challenges in involving this mass of actors is user trust, data safety and security, security of the financial transactions for both the customer and the crowd, tracking service quality - how to ensure service quality and managing incentives and compensation to keep the crowd motivated [8, 9]. To find proper solution for above challenges, Blockchain technologies seem to be effective solutions [10, 11]. Few countries are considering the application of Blockchain technologies in the public services. For example, e-Residencies have been issued by the country Estonia which enables their people to identify themselves online securely and also use them for starting location-independent businesses all over the world [12]. Estonia proves to be one of the progressive and technologically advanced countries which use Blockchain technologies to keep their citizen’s data safe and secure [13]. The application of Blockchain technologies for logistics and supply chain management have grown in terms of proposal of a Blockchain-based prototypes for tracking the historical data of vehicle tracking in China [14]. This application of Blockchain in this area helps in enhancing the data transparency between different players in the used car trade; overcoming the information asymmetry problem in this industry. The proposed prototype named BCVehis based on Blockchain technology helps in tracking the vehicle history. The authors also state that the Blockchain could be a disruptive technology in
28
K. Navendan et al.
the automotive sector; where the various business data relating to the vehicles could be stored and exchanged in a secure way, the vehicles unused capacity could be utilized by ridesharing and crowd logistics, and the vehicle data held by different organizations across the automotive industry could be linked to provide a transparent history of the vehicle lifecycle. This paper has investigated related literary works for the background of logistics and also the Blockchain technology and its capabilities. By discussing the research gap and using the capabilities of Blockchain technology, an e-commerce crowd logistics conceptual model is proposed. Further, a detailed flow of the business model is discussed.
2 Literature Review 2.1 Logistics Logistics has been an important function and has been used for over a long time from the building of the pyramids till today for the effective flow of materials. The author also highlights that Battles have been won and lost throughout history due to logistical strengths and skills or the lack thereof. Logistics has gained importance since the 1980s when the industries identified the potential assistance and benefits that logistics could offer the businesses. Business organizations have only recently begun to appreciate the critical role that logistics management can play in achieving a competitive edge [15]. Due to the increase in customer demand and the importance to fulfil the demand and also achieving a competitive advantage over many businesses and competitors, the companies started investing more money and time on logistics activities [5]. The drastic development was visualized in the logistics sector and the major contributor for such a development was the information and communication technology (ICT) [1, 16, 17]. ICT has become essential for managing the flow of materials, information, and monetary value among the supply chain entities by integrating, synchronizing, and making the information more visible among the partners and making the supply chain more responsive [18]. ICT in the logistics sector can be majorly used for transferring of information, planning the route and planning the mode of transportation, picking up the goods and delivering the goods, track and trace option and much more [19]. Due to the emergence of ICTs in the logistics sector, third party logistics and fourth-party logistics emerged, and in addition to these concepts, a new logistic activity emerged. This activity was called Crowd Logistics. 2.2 Crowd Logistics Crowdsourced Delivery or Crowd Logistics is one of these new possibilities. This approach comprises the utilization of private crowds to support delivery operations [9]. The crowd logistics concept was derived from the crowdsourcing technique where companies outsource the logistics process [5]. Crowd logistics are also termed as crowd shipping, crowdsourced delivery, cargo hitching. Crowd logistics is considered to be a realistic concept since it encourages passengers to transport goods for others using their additional carrying capacity on cars, bikes, buses, and aircraft. Crowd logistics makes use of
Enhancement of Crowd Logistics Model in an E-Commerce Scenario
29
information technology to distribute logistics resources more efficiently. Crowd logistics platforms are critical to crowd logistics as they are the enablers for the business. The crowd logistics platforms are said to be the backbone for the logistic process since crowd logistics platforms provide the ability for the distance routing, tracking, and tracing of goods, giving feedback for the delivery and for the payment of the services [20]. The main competitive advantages of crowd logistics in comparison with traditional logistics model, can be interpreted as the resource matching for enabling access to shared logistics resource pools which increase the flexibility for operations management and also affects highly the related risks inside the logistics planning [20]. Specially with the capability of accessing the shared logistics resources the unused fleet capacity in one firm can be shared with other logistics demanders which highly improve the sustainability within and among the logistics firms [6]. 2.3 Blockchain Technology The Blockchain technology at first came into fame during the early 2009 along with the bitcoin cryptocurrency [21]. Blockchain is a digital database or distributed ledger technology with cryptography and hashing algorithm. Blockchain is said to have a chain of blocks and each block in the Blockchain consist of data entered by the users and verified by the verifiers or miners. Each and every data in the Blockchain is public and can be visible to every user within the network, anyone can check, add and copy data but it is impossible to change it, making it an immutable technology [22]. It is transparent and quick, and it eliminates the need for a trusted third-party entity [23]. The application of Blockchains such as the smart contracts is a solution that is responsible to overcome the trusted third party involvement [24]. The concept of smart contracts is used in this study to manage the deposits and other payment transactions between the seller, buyer and the crowd (courier) with the help of the Decentralized applications (DApp) using SCs. Smart Contracts. Smart contracts are programmed contracts that run in the blockchain network. Users have the ability to create a contract by doing a transaction [22]. The authors [25] states that the smart contracts are chain codes and the chain codes are used by the peers in the nodes to execute a transaction and also to modify the ledger. The author mentions that the smart contracts guide in verifying, and negotiating enforcement of a transaction. The smart contracts help in automating the transactions and the most important advantage of smart contracts is they execute themselves without any human interference, once the predetermined conditions are met [26]. Smart contracts are considered to be an agreement between the entities in the network, which are made into program code and verified by the peers in the network. These smart contracts have the ability to speed up the processes and are more flexible as the transaction is not controlled by a third party [27]. Smart contracts along with the decentralized applications make the blockchain technology more easily accessible and more transparent technology. Decentralized Applications (DApp). The decentralized applications are P2P applications that run on the Blockchain network for example Ethereum [28]. DApps are decentralized applications which are used as an alternative for replacing the trusted third
30
K. Navendan et al.
parties in a business process [24]. DApps are user interfaces that are connected with smart contracts and BC networks [29]. The application layer in the BC is where the DApps are incorporated in order to obtain the data from the user or used as a page for the user to view the stored data. The transactions are carried out through the DApps [14]. DApps are considered to be identical to web classic web apps and [30] also describes that DApps adopts the same technology as the front end of the web applications.
2.4 Research Gap Analysis There have been few studies that involved the use of crowd logistics and Blockchain technology. In recent years the research on this topic has increased and researchers and companies are showing more interest in Blockchain technology because of its properties and functions that it provides to the development of businesses. One of the research carried out by the authors [31] is to bring trust among the entities with the help of the concept of proof of delivery within the transporters, buyers, and sellers. The authors used one of the Blockchain functionality i.e. the smart contracts for the proof of delivery which according to the authors enhanced the trust and security in the business [32]. The Blockchain along with the smart contract functionality helped in the tracking and traceability as the information is stored as logs in the smart contracts. The authors also introduced an arbitrator, an off-chain asset for carrying out claim management [33]. Another study on the physical delivery system called Lelantos using Blockchain technology by the authors [34] was carried out. The Lelantos consisted of on-chain and off-chain transactions and the on-chain transaction was the one that involved the use of smart contracts. This study also used smart contracts for recording the transactions, which in turn provided the peers in the network with traceability of the product. A package delivery system on Blockchain technology was developed by the authors [25]. The peers in the network have a clear idea of the transactions within the network, because of the three-block structures of the Blockchain. The three blocks mentioned by the authors in their work to maintain data integrity are block header which contains the previous data hash, block data which contains the time stamp, the information about the endorser, and the public keys and blocks metadata which consist of the identity and the signatures of the creator. This way the integrity of the data was maintained within the Blockchain network. The communication between the peers in the network is executed through smart contracts [35]. The authors [36] researched on crowdfunding concept involving smart contracts and Blockchain technology. According to the authors [36], the smart contracts in this study were used to automatically execute the transactions and the payments were stored in the smart contracts rather than using a trusted third party. This way the security was ensured. The smart contract was also used as a major function in the integration of Blockchain and e-commerce business. As stated by the authors [37] the smart contracts are used for payment transactions when it comes to an e-commerce business model. Blockchain technology is also used in the energy sector for the peer to peer virtual power trading and the authors [38] use smart contracts to place bids and withdraw funds after when an auction ends. The smart contracts were also used for the process of verification and validations of the IDs and the time stamps and digital signatures
Enhancement of Crowd Logistics Model in an E-Commerce Scenario
31
respectively before sharing the information to the peers in the network [39]. A hybrid Blockchain crowdsourcing platform called the zKCrowd was formulated by the authors [40]. The model ensures trust between the entities by making them pay a deposit and is stored in the smart contracts. The smart contracts help overcome the free-riding and the false reporting issues. The authors concluded that Blockchain had a solid foundation in providing trust and integrity among the peers in the networks and wanted to make them more user-friendly for the general public. In order to achieve this, the authors [14] used the concept called DApp i.e. the decentralized application in their Blockchain-based prototype for tracking and maintaining vehicle history. The DApp gives a user-friendly application layer through which the data can be updated and stored in the BC network and the smart contracts are used for the collection, validation, and exchange of vehicle history. The authors [41] also use the concept of DApps, they came up with a three-layered concept called the application layer for posting and receiving tasks, the Blockchain layer which provides consensus mechanisms, and the authenticity and the integrity of the data can be checked in the BC layer and also the entities in the network are rewarded through the smart contract mechanism and the storage layer. The concept of DApp and smart contract was well utilized in the research carried out by the authors [24]. The DApp for the androids were used as peer-to-peer architecture and the smart contract was used in the tracking of the process through their PuRSCA app. They hold the deposit in the smart contract and once the confirmation is carried out in the DApp the deposits are transferred back. This research includes the concept of DApp system and the function of smart contracts as mentioned in the work carried out by the authors [24] for holding the deposits and transferring the deposits, along with the few enhancements related to smart contracts according to the crowd logistics business model and also for the reverse crowd logistics model. As observed in the literature, no research has completely addressed smart contract mechanisms,security, traceability and tracking for last mile delivery, claim management System and incorporation to different trading platforms capability. This paper proposes a solution to cover the aforementioned themes. Moreover, enabling the token models in Blockchain, the financial transactions inside the model can be insured from financial perspective. Also, by using the mechanisms of Blockchain for creating accounts by Private /Public keys, the demanders and service providers are going to be managed in terms of their transaction. Also considering the shared logistics crowd concepts, even the already high reputation logistics service providers can share their unused capacities in platform and benefit from their unused service capacities.
3 Proposed Model 3.1 Conceptual Model A high-level abstraction of the Blockchain-based trading platform model is presented in Fig. 1. The Blockchain-based trading platform model for crowd logistics utilizes the security, tracking, and traceability features of the technology to create a solution for the delivery of the physical asset to the buyers from the sellers with the help of the crowd as the couriers.
32
K. Navendan et al.
Fig. 1. Blockchain based crowd logistics trading platform model.
The model gives a holistic view of the transactions that occur between the entities such as the buyers, sellers, couriers, and the involvement of the trading platform. The buyers and the couriers constitute the crowd; hence the buyers can choose the role of courier and the couriers can choose the role of buyers. This flexibility is provided in the trading platform. The other entities mentioned in the model are IPFS (InterPlanetary File System) for storing and sharing the details of the products in the distributed trading platform, which comprises of the pictures, specifications, and miscellaneous details of the products and finally the smart contracts that self-executes as per the agreement that is coded in the contract. Smart contracts eliminate the need for a trusted third party and they act as an escrow for automatic settlement of payments by transferring the share that was agreed upon by each entity when the package is successfully delivered. Smart contracts even have the ability to hold the payment when a dispute occurs in the process of transfer of goods or packages from the seller to the buyer and transfer the settlement as per the agreement when the dispute is handled. The model helps in incentivizing all the entities involved to act honestly by ensuring collateral, that is deposited by each
Enhancement of Crowd Logistics Model in an E-Commerce Scenario
33
entity in the model. This increases trust, security, and traceability and helps in dispute resolutions. 3.2 The Detailed Solution The crowd constitutes the buyers and the sellers. The buyers are the people who require a product and are willing to use the trading platform to order their desired product by checking the products information that is saved in the IPFS and uploaded in the form of hash into the Blockchain network by the partner companies who are considered to be the sellers as per the conceptual model. The model provides flexibility for the crowd to choose the role as per their requirements, i.e. when a person from the crowd wants a product he/she can choose to be a buyer and when the same person wants to deliver a product to another person from the crowd; can choose to be the courier. This flexibility will be given as an option in the decentralized application (DApp) platform. The partner companies are a group of sellers who have partnered with the trading platform and they share their products to the trading platform via an IPFS in the form of a hash; that carries all the information of the product. Apart from updating the product information regularly, the sellers are also responsible for preparing the order as per the buyer’s requirements and courier them on time as mentioned in the product information. The partner companies are completely responsible to deal with the reverse logistics of their products in case the buyer is not satisfied with the product. The courier is a part of the crowd and they have the flexibility to act as a buyer when they have a requirement. The courier collects the products from the partnered companies and delivers them to the buyers in the stipulated time mentioned by the sellers. The couriers are responsible for the safety and security of the physical asset during their time of possession and till the time of delivery. As discussed already the IPFS is used for storing and accessing files and data that is updated by the partnered companies or sellers. The hash generated in the IPFS is shared with the Blockchain networked which can be accessed with the help of the decentralized application. The DApp or the decentralized application which is the trading platform is the bridge that connects the sellers and the crowds (buyers and the couriers). The DApp user interface consists of the hash that contains all the details of the products and the sellers. All the processes mentioned above including the payment of deposits and return of the deposits are carried out with the smart contract mechanism that is designed and implemented within the Blockchains. The advantage of the Blockchain technology is its transparency, immutability, and the ability to log everything, therefore the smart contracts used in the model utilizes the property of the Blockchain to create the events and log the events, and help in tracking the orders made by the buyer, packaging made by the seller and the delivery of the physical asset by the courier. Thus, making the processes transparent.
4 Model Capabilities 4.1 Encryption Mechanism To ensure integrity, legitimacy, and authenticity; the buyers, sellers, and couriers create an account in the trading platform and obtain their private key and the public key for the
34
K. Navendan et al.
verification process. Identity provision is one of the major and key aspects of this DApp and is carried out using the QR reader concept through which the entities involved in the process read the other party’s identity by reading each other’s QR code image. 4.2 Model Transactions After creating their IDs and obtaining the keys, the sellers, buyers, and couriers can start using the DApp trading platform for their respective needs. Initially, the partnered companies or the sellers share and store their product information including the pictures, specifications, and their respective delivery information to the IPFS. The data concerning the products are stored in the IPFS to reduce the size that has to be stored in the Blockchain network. Only the hash created in the IPFS is updated in the Blockchain network and viewed through the user interface of the DApp. To deliver a product successfully from the seller to buyer three types of contracts are devised and they are initiated to create a chain of contracts. Apart from the three contracts, the first contract is created by the buyer which is discussed below. Every contract has its address and the addresses are mentioned in the other initiated contracts to make the traceability easy. The buyers after logging in with their credentials are allowed to browse the products and their specifications that are shared and stored by the sellers. Once the buyers are convinced with the products and their description, they can select the respective products; initiate a smart contract (Proof of Buyer – POB) in the Blockchain network, and pay the deposit, the buying price of the product, the delivery charge of the product and share the QR code to the seller for the identity verification. The seller receives the information of the buyer through the DApp and verifies the identity of the buyer, thus changing the contract state to “Order Received”. The seller then prepares the delivery package that has to be delivered to the buyer, and initiates a smart contract (Proof of Seller - POS), hashes a key, and pays the deposit. The seller after initiating the smart contract shares the id to the DApp for the verification process made by the courier. Along with the package that has to be delivered, the seller gives the physical copy of the verification key to the courier that has to be delivered to the buyer along with the package. The courier is assigned with the delivery job as per the machine learning techniques that can be used which are not discussed in the current paper since it is out of scope. The courier accepts the order by the seller, verifies the seller’s id and thus changing the contract state to “order accepted” and the couriers initiate a smart contract (Proof of Courier – POC) by paying the deposit and shares the QR code to the seller and buyer for identity verification. At the time of package collection, the seller verifies the courier id, thus changing the contract state “Package collected”. The courier collects the product and the physical verification key from the seller and delivers them to the buyer within the stipulated time. Finally, the buyer creates the final smart contract (Proof of Delivery – POD), where the buyer verifies the courier and confirms the product by verifying whether the hashed key and the physical key are handed over by the courier match. If the verification key matches then the buyer confirms the delivery, thus changing the contract state to “package received”. Figure 2 illustrates the chain of four contracts and the entities’ interaction involved. Finally, the payment transactions take place after a
Enhancement of Crowd Logistics Model in an E-Commerce Scenario
35
certain given period for the buyer to check the product, for example, a week for checking the product. Till then the payment and the deposits made by the seller, buyer, and the courier are held.
Fig. 2. Chain of four contracts showing the interaction among entities involved.
5 Conclusion This paper has presented the conceptual model of a Blockchain-based crowd logistics e-commerce business model using the functionalities of BC such as the smart contracts and the DApps. This can insure the challenges of crowd logistics system and fulfills trust issues between the peers and tracking and traceability concerns. So, while fulfilling the main challenges of crowd logistics models, it has increased the flexibility of the crowd logistics system with the help of Blockchain functionalities such as smart contracts and DApps. A series of smart contract structures have been introduced in the conceptual model for the tracking and traceability of the products through DApps and also for the financial transactions. The main focus of conceptual model has been on forward logistics and using the feature of trust creation is the BC technology through the DApp. However, there are even more complicated scenes to deal with like the involvement of a validator in the reverse logistic process to check the returned product and distinguish whether the product needs to be refurbished or demolished.
References 1. Akbaripour, H., Houshmand, M., Valilai, O.F.: Cloud-based global supply chain: a conceptual model and multilayer architecture. J. Manuf. Sci. Eng. Trans. ASME 137(4), 040913 (2015) 2. Lemoine, W., Dagnæs, L.: Globalisation strategies and business organisation of a network of logistics service providers. Int. J. Phys. Distrib. Logist. Manag. 33, 209–228 (2003) 3. Akyelken, N., Keller, H.: Framing the nexus of globalisation, logistics and manufacturing in Europe. Transp. Rev. 34, 674–690 (2014)
36
K. Navendan et al.
4. Sealey, R.: Logistics workers and global logistics: the heavy lifters of globalisation. Work Organ. Lab. Glob. 4, 25 (2010) 5. Mladenow, A., Bauer, C., Strauss, C.: ‘Crowd logistics’: the contribution of social crowds in logistics activities. Int. J. Web Inf. Syst. 12, 379–396 (2016) 6. Buldeo Rai, H., Verlinde, S., Merckx, J., Macharis, C.: Crowd logistics: an opportunity for more sustainable urban freight transport? Eur. Transp. Res. Rev. 9(3), 1–13 (2017). https:// doi.org/10.1007/s12544-017-0256-6 7. Carbone, V., Rouquet, A., Roussat, C.: The rise of crowd logistics: a new way to co-create logistics value. J. Bus. Logist. 38, 238–252 (2017) 8. Dayarian, I., Savelsbergh, M.: Crowdshipping and same-day delivery: employing in-store customers to deliver online orders. Prod. Oper. Manag. 29, 2153–2174 (2020) 9. Sampaio, A., Savelsbergh, M., Veelenturf, L., van Woensel, T.: Crowd-Based City Logistics. Sustainable Transportation and Smart Logistics: Decision-Making Models and Solutions (Elsevier Inc., 2018). https://doi.org/10.1016/B978-0-12-814242-4.00015-6 10. Aghamohammadzadeh, E., Fatahi Valilai, O.: A novel cloud manufacturing service composition platform enabled by Blockchain technology. Int. J. Prod. Res. 58, 5280–5298 (2020) 11. Ahmadi, E., et al.: Using Blockchain technology to extend the vendor managed inventory for sustainability. In: International Journal of Industrial and Manufacturing Engineering, ICSCLE 2021: International Conference on Supply Chain and Logistics Engineering, vol. 15, pp. 470– 474. World Academy of Science, Engineering and Technology (2021) 12. Ikeda, K. Hamid, M. N.: Applications of blockchain in the financial sector and a peer-to-peer global barter web. In: Advances in Computers, vol. 111. Elsevier Inc. (2018) 13. Sullivan, C., Burger, E.: E-residency and blockchain. Comput. Law Secur. Rev. 33, 470–481 (2017) 14. Chen, J., Ruan, Y., Guo, L., Lu, H.: BCVehis: a Blockchain-based service prototype of vehicle history tracking for used-car trades in China. IEEE Access 8, 214842–214851 (2020) 15. Christopher, M.: Logistics and supply chain management: creating value-adding networks (2011) 16. Aghamohammadzadeh, E., Malek, M., Valilai, O.F.: A novel model for optimisation of logistics and manufacturing operation service composition in Cloud manufacturing system focusing on cloud-entropy. Int. J. Prod. Res. 58, 1987–2015 (2020) 17. Sadeghi Aghili, S.A., Fatahi Valilai, O., Haji, A., Khalilzadeh, M.: Dynamic mutual manufacturing and transportation routing service selection for cloud manufacturing with multi-period service-demand matching. PeerJ Comput. Sci. 7, e461 (2021) 18. Luisa Dos Santos Vieira, C., Sérgio Coelho, A., Mendes Luna, M.M.: ICT implementation process model for logistics service providers. Ind. Manag. Data Syst. 113, 484–505 (2013) 19. Pokharel, S.: Perception on information and communication technology perspectives in logistics: A study of transportation and warehouses sectors in Singapore. J. Enterp. Inf. Manag. 18, 136–149 (2005) 20. Li, S., et al.: How do crowd logistics platforms create value? an exploratory case study from China. Int. J. Logist. Res. Appl. 22, 501–518 (2019) 21. Huckle, S., Bhattacharya, R., White, M., Beloff, N.: Internet of things, Blockchain and shared economy applications. Proc. Comput. Sci. 58, 461–466 (2016) 22. Tian, F.: A supply chain traceability system for food safety based on HACCP, Blockchain and internet of things. In: 14th International Conference Service System Service Management ICSSSM 2017 – Proceedings (2017). https://doi.org/10.1109/ICSSSM.2017.7996119 23. Gonzalez, L.: Blockchain, herding and trust in peer-to-peer lending. Manag. Financ. 46, 815–831 (2019) 24. Rafati Niya, S., Schüpfer, F., Bocek, T., Stiller, B.: A Peer-to-peer Purchase and Rental Smart Contract-based Application (PuRSCA). it - Inf. Technol. 60, 307–320 (2018)
Enhancement of Crowd Logistics Model in an E-Commerce Scenario
37
25. Ngamsuriyaroj, S., Likittheerameth, T., Kahutson, A., Pathummasut, T.: Package delivery system based on Blockchain infrastructure. In: Proceeding 2018 7th ICT International Student Project Conference ICT-ISPC 2018, pp. 1–6 (2018). https://doi.org/10.1109/ICT-ISPC.2018. 8523944 26. Alladi, T., Chamola, V., Sahu, N., Guizani, M.: Applications of Blockchain in unmanned aerial vehicles: a review. Veh. Commun. 23, 100249 (2020) 27. Tijan, E., Aksentijevi´c, S., Ivani´c, K., Jardas, M.: Blockchain technology implementation in logistics. Sustain. 11(4), 1185 (2019) 28. Kawa, A., Maryniak, A. (eds.): SMART Supply Network. E, Springer, Cham (2019). https:// doi.org/10.1007/978-3-319-91668-2 29. Ducrée, J., et al.: Open platform concept for Blockchain-enabled crowdsourcing of technology development and supply chains. Front. Blockchain 3, 1–13 (2020) 30. Thejaswini, S.: Peer to Peer Networks. pp. 600–606 (2020) 31. Hasan, H.R., Salah, K.: Blockchain-based proof of delivery of physical assets with single and multiple transporters. IEEE Access 6, 46781–46793 (2018) 32. Hasan, H.R., Salah, K.: Blockchain-based solution for proof of delivery of physical assets. In: Chen, S., Wang, H., Zhang, L.-J. (eds.) ICBC 2018. LNCS, vol. 10974, pp. 139–152. Springer, Cham (2018). https://doi.org/10.1007/978-3-319-94478-4_10 33. Hasan, H.R., Salah, K.: Proof of delivery of digital assets using Blockchain and smart contracts. IEEE Access 6, 65439–65448 (2018) 34. Altawy, R., Elsheikh, M., Youssef, A.M., Gong, G.: Lelantos: a Blockchain-based anonymous physical delivery system. In: Proceedings 2017 15th Annual Conference Privacy, Security Trust PST 2017, pp. 15–24 (2018). https://doi.org/10.1109/PST.2017.00013 35. Wang, S., Taha, A.F., Wang, J., Kvaternik, K., Hahn, A.: Energy crowdsourcing and peer-topeer energy trading in Blockchain-enabled smart grids. IEEE Trans. Syst. Man, Cybern. Syst. 49(8), 1612–1623 (2019) 36. Yadav, N., Sarasvathi, V.: Venturing crowdfunding using smart contracts in Blockchain. In: Proceedings 3rd International Conference Smart System Inventive Technology ICSSIT 2020, pp. 192–197 (2020). https://doi.org/10.1109/ICSSIT48917.2020.9214295 37. Ismanto, L., Ar, H.S., Fajar, A.N., Bachtiar, S.: Blockchain as E-commerce platform in Indonesia. J. Phys. Conf. Ser. 1179(1), 12114 (2019) 38. Seven, S., Yao, G., Soran, A., Onen, A., Muyeen, S.M.: Peer-to-peer energy trading in virtual power plant based on Blockchain smart contracts. IEEE Access 8, 175713–175726 (2020) 39. Khacef, K., Pujolle, G.: Secure peer-to-peer communication based on blockchain. In: Barolli, L., Takizawa, M., Xhafa, F., Enokido, T. (eds.) WAINA 2019. AISC, vol. 927, pp. 662–672. Springer, Cham (2019). https://doi.org/10.1007/978-3-030-15035-8_64 40. Zhu, S., Cai, Z., Hu, H., Li, Y., Li, W.: zkCrowd: A Hybrid Blockchain-based crowdsourcing platform. IEEE Trans. Ind. Informatics 16, 4196–4205 (2020) 41. Li, M., et al.: CrowdBC: a Blockchain-based decentralized framework for crowdsourcing. IEEE Trans. Parallel Distrib. Syst. 30, 1251–1266 (2019)
Synthesising COVID-19 Related Research from a Logistics and Supply Chain Perspective I¸sık Özge Yumurtacı Hüseyino˘glu1
, Ilja Bäumler2
, and Herbert Kotzab2(B)
1 Izmir University of Economics, Izmir, Turkey 2 University of Bremen, Bremen, Germany
[email protected]
Abstract. The COVID-19 pandemic research has emerged as a rapidly increasing field both from a practitioner’s and an academician’s perspective. In this paper, we present a synthesis of logistics and supply-chain related COVID-19 pandemic research and classify its results through a systematic literature review (SLR). To achieve this, the content of 87 papers was qualitatively assessed and classified according to their research themes and methodological approaches. The SLR findings pertain the changing mechanisms in logistics and supply chain management. The transformational change indicates insights for strategies, services, structures, and social issues. Accordingly, research avenues in logistics and supply chain management field are provided. Finally, propositions referring to research avenues and synthesis of the findings are suggested. Keywords: Supply chain management · Operations research · COVID-19
1 Introduction Since January 2020, the COVID-19 pandemic is dramatically impacting logistics and supply chain networks all over the world due to shortages in many areas of our lives as well as to a lack of medical protection (Liu et al. 2020). COVID-19 showed how fragile globally organised supply chains are as many companies have outsourced as well as offshored a lot of activities around the globe (Biedermann and Kotzab 2021). Especially in the beginning of the crisis, demand as well as supply situations have significantly changed, and many supply chains were not able to operate due to interrupted processes (see Hedtstück 2020). Due to the unknown effects of the virus, all over the world, national authorities executed shut-downs, meanwhile multiple times, in order to reduce the spread of the COVID-19 virus and many grocery stores experienced in the first days of the pandemic a ‘shopper tsunami’ which spilled over the stores within hours. We were able to observe long shopper queues with jam-fulled shopping carts and ‘late’ shoppers were then confronted with never experienced out-of-stock situations. Meanwhile, retailing observes clear winners and losers of the pandemic as some sectors lost immense due to lockdowns and shop-closing restrictions through the authorities. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 M. Freitag et al. (Eds.): LDIC 2022, LNLO, pp. 38–51, 2022. https://doi.org/10.1007/978-3-031-05359-7_4
Synthesising COVID-19 Related Research
39
After experiencing this immense situation, we observe an increasing number of papers and literature on COVID-19 pandemic in various disciplines and also in the management research area. A few literature reviews formulated research agendas based on the published COVID-19 papers such as the work by Chowdhury et al. (2021) that reviews COVID-19 studies in manifold supply chain disciplines and present four reoccurring research themes. These themes regard the impact, resilience strategies, the role of technologies and sustainability issues in the light of the COVID-19 pandemic. Alternatively, the work by Queiroz et al. (2020) focus on epidemic outbreaks and the consequences for logistics. They propose research questions dealing with COVID-19 effects from a modelling, organizational and technological perspective. In his editorial article for the Transportation Research Part E: Logistics and Transportation Review, Choi (2021) set up a research agenda and published risk analysis related studies and points to seven risk connected COVID-19 research themes. These are in the area of adaptation, general systems theory and theory in economics, new methods for risk analysis, social welfare, humanitarian logistics and technologies. However, efforts to compile, group, and review this body of work are still in their infancy. Consequently, we were interested to see how the field of logistics and supply chain management dealt with the pandemic situation in its research activities so far and consequently we formulated the following research questions (RQ): RQ1: What are the current issues of COVID-19 pandemic related logistics and supply chain research? RQ2: What are the future research avenues from a logistics and supply chain perspective? The remainder of the paper is as following: After the methodological chapter, where we present our methodological approach, and develop a research design according to SLR, we present the results of our applied SLR in a condensed and narrative manner. Therefore, we elaborated six category perspectives to quantitatively distinguish the research efforts in COVID-19 related logistics and supply chain management research. Afterwards we discuss the results and answer the research questions and offer future research avenues as well as research propositions. The paper closes with a critical reflection and some limitations.
2 Methodological Approach Systematic literature review (SLR) is a suitable tool in the social sciences to summarize existing investigations on a given topic (Stamm and Schwarb 1995). This is also true for the field of business and management research where this particular research approach is more and more used to gain new insights and interpretations from the research results of available studies (Eisend and Hampp 2020). Furthermore, Denyer and Tranfield (2009) argue that synthesizing previous research results and combining subsequent results unfolds new ways of looking at specific topics and opportunities for creating new knowledge. SLRs also allow the identification of existing contradictions, ambiguities, and research gaps (Rousseau et al. 2008; Rowe 2014). However, the method is not intended to consider and process all work within the entire topic area. Usually, the intention of a SLR is to pick up only a few sources that are relevant to the research project (Fettke 2006).
40
I. Ö. Y. Hüseyino˘glu et al.
In this context, Trapp (2012) distinguishes between quantitative-descriptive and qualitative-analytical SLR approaches. The research procedure of quantitativedescriptive analysis is clearly defined and proceeds according to the predefined criteria. Thus, based on applied rules and formalities, the sources to be included are selected according to keywords or lists describing the topics relevant to the analysis. The qualitative-analytic SLR is traditional and narrative in nature, but mostly fails to disclose the criteria for selecting sources and the procedures followed during the research process. The summary of existing and published knowledge without a prescribed methodology clouds the inter-subjective comprehensibility of the examined research as well as the following evaluation. In order to answer our research questions, we apply a quantitativedescriptive as well as qualitative-analytical SLR. After formulating our research questions which include the logistics and supply chain perspective, we searched for literature based on specific search criteria (see Table 1). Table 1. Criteria for article selection and inclusion for systematic review Article search criteria
Search
Database
Web of science core collection
Search field
Topic
Search string
Covid-19 AND (logistics OR “supply chain”)
Search date
21.12.2020
Years
2020–2021
Number of articles
1,327
Article selection criteria Type of document Article
1,223
Web of science categories Business, Business finance, Computer 124 science information system, economics, environmental studies, management, operations research management science, public administrations, transportation Following analysis of the abstracts and the document
102
Analysis and synthesis of the evidence 87
3 Results 3.1 Sample Characterisation Due to the actuality of the pandemic situation, it is no surprise that all sample papers were either granted early access or being published in 2020. Apparently, half of our sample articles build their argument on secondary data. These papers are conceptual,
Synthesising COVID-19 Related Research
41
theoretical, literature reviews or positional paper in nature. When it comes to primary research, we observe a dominance of papers from Europe, Asia, and North America and less frequently papers from Australia, Africa, and South America. Around 1/3 of the sample articles fall into a specific branch category, such as fast-moving consumer goods, car manufacturing or agriculture. Based on the SCIMAGO journal rating as well as ABS guide and journal interest areas, the findings of the selected papers were further categorized into the following six groups: Economics (n = 26), management (n = 8), marketing (n = 7), operations/operations research (n = 26), social sciences (n = 6) and sustainability (n = 14). We analysed the content of the sample articles based on their topical research interest as well as on their methodological approaches within the individual categories. Then, we compared the categories in order to identify communalities as well as differences. Next, we present exemplary the state of covid-19 related research within the area of operations/operations research and discuss the overall results of our comparative analysis. 3.2 Comparative Analysis Table 2 and Table 3 condense and evaluate the qualitative results of our analysis from a methodological as well as thematic orientation. Table 2. Concluding overview of the methodological approaches1 Perspective
n
Conceptual Pure
Economics
26
Management
8
Marketing
7
OPS/OR
26
Social sciences
6
Sustainability
14
Literature
Empirical Other
+ ++ + ++
+
-
+
-
+
-
+ +
-
+
+ + ++
-
-
+ ++
+
+
+
+ ++
-
+ ++
+
-
+ + + +
-
+ ++
-
+ + ++
-
-
+
+ ++ +
+ ++
Model-based
+
+ +
+ ++ + ++
+
+ ++
+ + + ++
+ ++
Simulation + +
+
+ +
Modelling Quantitative
+
+ ++ + ++
Qualitative
-
+ ++
+ + ++
+
+
1 Legend: Ordinal scaling from “ + + + ” to “-”where “ + + +” = predominant, meaning
the research field is dominated by this methodological or thematic orientation; “ + + ” = balanced, meaning in relation to other methods or research topics, a similar number of methods or research topics are used from the category perspective; “ + ” = rare, meaning in relation, the method or research topic was used or dealt with sporadically in the considered categories; “-” = non-existent, meaning within the research perspective under consideration, neither the methods nor the research topics from the respective columns were dealt with.
42
I. Ö. Y. Hüseyino˘glu et al.
Table 2 clearly shows the domination of a conceptual research orientation where either a pure descriptive-conceptual or literature-based conceptual COVID-19 analysis is pursued. Within the area of sustainability-focused research, issue analysis is also recognized as an applied method. When it comes to empirical work, marketing research exclusively applies questionnaires and interviews. From an operations and operations research perspective, conceptual, empirical, and model-based methods are equally used. However, model-based research is mostly applied in the Economics and OPS/OR. Table 3. Concluding overview on COVID-19 related -research themes (number in brackets indicate article mentions)2 Perspective
n
Economics (26)
Management (8)
Marketing (7)
OPS/OR (26)
Social sciences (6)
Sustain-ability (14)
Supply chain resilience and risk management
25
+ +
+++
+++
++
-
+
Pandemic management and perseverance
13
-
+
+
++
+
++
Agriculture and food supply chain
12
+++
-
-
++
-
-
Operations management related
11
-
+
+
++
++
++
Covid-19 impact and evaluation
9
+
-
-
++
++
+
Technology research and application
7
+
+
-
++
-
-
E-Commerce and online shopping behaviour
5
++
-
+
-
-
+
Global economy
5
++
-
-
-
-
-
Table 3 shows the thematic orientation of the 87 articles analysed from left to right in descending order based on the number of article mentions (see number in brackets). Hereby we observe that supply chain resilience and risk management (n = 25) are the major research themes, followed by papers with a focus on perseverance and management during COVID-19 times (n = 13). The third most frequent topic represents the agriculture and food supply chain (n = 12), which is mainly covered in the areas of Economics and OPS/OR. Hereby, we observe an even distribution of operations and operations research-related work in six out of eight research themes.
2 See footnote 1 for legend.
Synthesising COVID-19 Related Research
43
3.3 Covid-19 from an OPS/OR Point of View Synopsis of the Findings. Within this subarea we identified nearly 30 papers3 which include the in Table 4 presented themes as well as methodological approaches. The recognised research themes cover a variety of topics including strategies, mechanisms, and technologies to cope with the pandemic from an operations perspective and several specific research agendas. Furthermore, research deals with the COVID19’s impact on various supply chains, risk mitigation and disruption management as well as the effects of the COVID-19 pandemic for building up supply chain resilience. The papers can further be grouped into those that observe the current developments, those, which examine future outcomes, as well as those who compare current with prepandemic conditions. From a methodological point of view, research is dominated by conceptual and empirical work and some modelling work. Current Research Issues. Due to COVID-19 restrictions and consequences we observed that closed borders, sanitary and logistical controls as well as a lower volume of requests led to numerous disruptions in the movement of goods as well as transport bottlenecks. accordingly, a sharp decline in logistics demand, a lack of transportation capability, a breakdown in the logistics network, a shift in operation mode, and an increase in operational costs and numbers of loss-making businesses are observed (Chowdhury and Shumon 2020; Liu et al. 2020; Queiroz et al. 2020). The COVID-19 pandemic has increased food insecurity and food safety risks (Sharma et al. 2020c), elevated supply chain and logistics costs, dramatically altered customer behaviour, has raised concern about food waste and the value of homegrown foods (Singh et al. 2021). Depending on the complexity and size of the organization, supply threats, market risks, financial risks, logistical and technology risks, management and operating risks, legislation and regulatory risks, genetic and chemical risks all have a direct effect on agricultural supply chains following COVID-19 pandemic (Rejeb et al. 2020; Yadav et al. 2020). Therefore, risk mitigation can be accomplished by Industry 4.0, supply chain collaboration, and coordination. Blockchain technology can serve to agriculture ecosystem by ensuring food traceability, country of origin, certification and safety which can serve for seamless and reliable agricultural products flow from point of origin to point of consumption (Lin et al. 2020; Marzantowicz et al. 2020). Proper implementation of knowledge management can help with strategic preparations in a very efficient manner, minimizing gaps and shortages, and providing effective actions to reduce the biggest challenges that a pandemic can cause (Pinto 2020). Practical decision support systems are being developed to support demand management in the healthcare supply chain, reduce population tensions, interrupt the chain of 3 Published in: Annals of Operations Research (3), Brazilian Journal of Operations & Production
Management (1), IEEE Access (2), International Journal of Logistics-Research and Applications (3), International Journal of Operations and Production Management (4), International Journal of Physical Distribution & Logistics Management (1), International Journal of Production Research (2), Journal of Humanitarian Logistics and Supply Chain Management (1), LOGFORUM (3), Operations Management Research (2), Production and Operations Management (1), Transportation Research Part E- Logistics and Transportation Review (3).
44
I. Ö. Y. Hüseyino˘glu et al. Table 4. Themes and methodological approach in COVID-19 related operations research4
Research theme
Further divide
Source
Agriculture and food supply chain
Blockchain technology
Lin et al. 2020
Risk mitigation
Sharma et al. 2020 c
Internet of things
Yadav et al. 2020
Research agenda
Rejeb et al. 2020
Ambiguity in supply chains
Coping mechanisms
Gunessee and Subramanian, 2020
Corporate survivability
Capabilities and digital technologies
Niewiadomski 2020
Healthcare supply chain
Decision support system
Govindan et al. 2020
Impacts of epidemic outbreaks on supply chains
Research agenda
Queiroz et al. 2020
Food and beverage industry
Chowdhury and Shumon 2020
Global supply chain
Ivanov 2020a
Food chain
Singh et al. 2021
Impacts and trends
Liu et al. 2020
Last mile distribution
Robotic system
Sharma et al. 2020a
Manufacturing supply chain
Production recovery plan
Paul and Chowdhury 2020
Pandemic management
Strategy to stop pandemic spread
Baveja et al. 2020
Public distribution system
Blockchain technology
Kumar 2020
Service operations
Static versus mobile
Choi 2020
Supply chain management
Knowledge management
Pinto 2020
Supply chain resilience
Industry 4.0
Zhang et al. 2020
Managing disruptions
Handfield et al. 2020; Ivanov and Dolgui 2020
Risk mitigation
Marzantowicz et al. 2020
Research opportunities
Remko 2020
Viable supply chain model Ivanov 2020b Supply chain sustainability
Framework for survivability
Sharma et al. 2020b
Research guidance
Sarkis 2020 (continued)
4 Authors displayed in square brackets indicate a minor research fit. These articles do not focus
on COVID-19 research but offer insights for the COVID-19 pandemic and are therefore included for statistical purpose.
Synthesising COVID-19 Related Research
45
Table 4. (continued) Methodological approach
Further divide
Source
Conceptual
Purely conceptual/Discussion
Baveja et al. 2020; Gunessee and Subramanian 2020; Ivanov 2020b; Ivanov and Dolgui, 2020; Liu et al., 2020; Pinto, 2020; Sarkis, 2020; Sharma et al. 2020.a
Literature review
Queiroz et al. 2020; Rejeb et al. 2020; Zhang et al. 2020
Survey
Lin et al., 2020; Marzantowicz et al. 2020; Sharma et al. 2020b; Sharma et al. 2020 c; [Yadav et al. 2020]
Expert interviews/Case-study
Chowdhury and Shumon, 2020; Govindan et al. 2020; Handfield et al. 2020; Remko 2020
Empirical
Modelling
Expert discussion
Niewiadomski 2020
Discrete event simulation
Ivanov, 2020a
Simulation model
Singh et al. 2021
Analytical model
Choi 2020
News vendor model
Kumar 2020
Mathematical recovery model
Paul and Chowdhury 2020
spread of COVID-19 and more generally prevent disease outbreaks in the healthcare supply chain (Govindan et al. 2020). For pandemic management, the testing capacity, time interval from infection to isolation and isolation of contact’s network are challenges, which necessitate the administration of isolated zones. Therefore, food and critical drug supplies may be provided in the heavily restricted areas if a truck-drone simulated delivery mechanism is used as public delivery system (Baveja et al. 2020). In manufacturing supply chains, the demand for basic goods rises dramatically, but the availability of raw materials falls dramatically due to manufacturing capacity constraints, which necessitate production recovery plan to cope with dual disruptions (Paul and Chowdhury 2020). Consequently, it is important to ensure the secure delivery of food and basic resources to all residents to keep them afloat through difficult times. Hence, contactless robotic systems are proposed for delivery (Sharma et al. 2020a). In addition to this, implementing blockchain technology in a charitable supply chain will greatly minimize pilferage and ghost demand (Kumar 2020). Rather than the length of upstream disturbance or the pace of epidemic spreading, the timing of the closure and opening of facilities at various echelons can become a major factor influencing the epidemic outbreak effect on supply chain results. Moreover, the
46
I. Ö. Y. Hüseyino˘glu et al.
other critical variables in the supply chain include lead time, disease transmission intensity, and upstream and downstream disturbance durations. In this regard, the key criteria for handling buyer–supplier relationships and improving sustainable supply chain survivability before and after the COVID-19 pandemic is mainly on providing supply chain network viability (Ivanov 2020b; Ivanov and Dolgui 2020).
4 Research Avenues for Future OPS/OR Research and Conclusions 4.1 Rethinking Supply Chain Strategies and Structures With the COVID-19 pandemic, the low cost-sourcing strategy has started to be considered risky due to longer lead time, unforeseen events, and disruptions (Handfield et al. 2020). Firms have started to reconsider their supply chain strategies to change supply base, accelerate digitalization, back to localization, increase flexibility and responsiveness, and improve relationships with local suppliers. The innovations have a significant role to support current technologies, and to analyse existing technologies, for their future contribution to supply chain issues (Remko 2020). To encourage the agility of the supply chain during crisis, it is necessary to implement robust and transparent information technology (Kumar 2020). Therefore, supply chain strategies need to be aligned according to the changing dynamics and new normal conditions. In this regard, participation and value co-creation of supply chain members are more necessary than ever. Accordingly, scenario-planning approaches for the potential future pandemics can provide benefits across the supply chain. The existing supply chain strategies show their limits in the COVID-19 pandemic context (e.g. responsiveness, lean, agile and leagile and digital supply chain strategies), as there is only one new strategy named as „viable supply chain“ referring to the capability of a supply chain to sustain itself and function in an evolving world by redesigning systems and reorganizing success with long-term consequences (Ivanov 2020b). In the future, a taxonomy of COVID-19 pandemic supply chain strategies is necessary to compare strategies and structures in pre and post COVID-19 supply phases. Taking all these into account, OPS/OR research is essential for novel supply chain strategies and structures which enhance the flow of operations along the supply chains. In this regard, case-study research is applicable to reveal required supply chain strategies and enhance knowledge on different industries/countries and structures. 4.2 Supply Chain Risk Management The COVID-19 pandemic is considered as a natural disaster with many adverse effects on economy, human health, and business environment. From now on, supply chain research should focus on cost, response times and flexibility integrated in the context of COVID19 pandemic (Remko 2020) to mitigate risk. Accordingly, supply chains need to assess the resilience of their supply chains and formulate new strategies to mitigate risk also to protect human health, and potential job losses. Supply chain risk management needs to incorporate the changing dynamics and transformation processes during pandemic times as the two examples of the recent polymers and electronic chip crisis show (EuPC 2021; The Guardian 2021). In this regard, OPS/OR research can extend knowledge in the field
Synthesising COVID-19 Related Research
47
to mitigate risk, improve flexibility and control costs. The appropriate research strategy for this research avenue is modelling and simulation research to prepare scenarios and alternative plans to mitigate supply chain risks. 4.3 Circular Supply Chains The COVID-19 pandemic has increased the awareness on waste (e.g. Shurson 2020). In this regard, circular supply chain approach is applicable to convert waste, create value and improve sustainability of the supply chains. We expect more supply chain designs with circular structure that contribute to new business ecosystem and serve for sustainability purposes. Circular supply chain structures are likely to differ according to industry characteristics, therefore industry specific circular supply chain management can provide insights for the new dynamics and guide practitioners in the field. The circular supply chains are likely to be managed well with redesigned operations. Hence, research on OPS/OR will shed light on decreasing waste and transformation of business models. The qualitative (interviews, focus-group and case study) and quantitative (survey, modelling and simulation) research are promising research strategies to reveal the drivers of circular supply chains and extend the knowledge on the best practices. 4.4 Diversifying Logistics Services During the COVID-19 pandemic, the demand to seamless, integrated, and fast logistics flows have increased tremendously. Therefore, there has been a shift to air cargo (Li 2020) for urgent deliveries. Another recent issue is the shipping container crisis (UNCTAD 2021) due to the bottlenecks in logistics flow, which also increased the freight rates. Therefore, the need for intermodal transportation, warehousing space, integrated delivery network has gained significant role. Companies need diversification of logistics services, logistics solutions and control of logistics costs more than ever. In addition to this, the advances and acceleration of e-commerce transactions due to the COVID-19 pandemic, has increased attention to last mile logistics (e.g. Sharma et al. 2020a) and use of full automation in delivery services (e.g. Choi 2020). Hence, consumers need more innovative delivery solutions, which decrease human interaction, increases the speed of delivery, and ensure hygiene conditions. In brief, the companies and consumers are seeking for diversification of logistics services. The impact of COVID-19 pandemic on logistics services opens room for further investigation. The success of logistics services is dependent on the performance of OPS/OR which improves the logistics flow across the supply chains. Qualitative research approaches are adequate to design new services as these methods provide an in-depth understanding of the mechanisms, drivers, market conditions and changing dynamics. 4.5 Logistics for Social Welfare In response to a health problem with substantial economic and social consequences, public policy must take several aspects and variables into account (Coelho et al. 2020). The limited personal protective equipment and medical products have been an issue
48
I. Ö. Y. Hüseyino˘glu et al.
to control the spread of the pandemic. Moreover, the lack of fair distribution of vaccines to each country creates dramatic situations such as in India (India Today 2021). The integrated and coordinated logistics activities among countries are promising for vaccine and health products delivery to decrease the losses due to the COVID-19 pandemic. The humanitarian aid research can contribute to the planning of further steps to control the pandemic. Apart from the needs for essentials and healthcare products, less privileged/underdeveloped countries were in need for financial assets to cope with the emergent problems. Gonzalez et al. (2020) argue that the advances in technology can provide solutions to improve the flow of cash when people are isolated and have no access to cash. An integrated management of logistics can help to increase social welfare of people. The humanitarian aid research will provide more comprehensive insights when it is designed from an OPS/OR perspective. In this regard, logistics will be able to improve social welfare and quality of life. Consequently, mixed-method research designs are relevant to explore the humanitarian needs, validate the findings and disseminate the knowledge to meet the needs of the stakeholders.
5 Conclusion The purpose of this work was to shed some light into COVID-19 pandemic related logistics and supply chain research by the means of a systematic literature review. In a first step, we identified six different research angles, which deliver insights for the impact of COVID-19 pandemic. These angles refer to economics, management, marketing, OPS/OR, social sciences and sustainability (see Table 2). The major contributions of our study and consequently the answers to our two research questions are the identification of current issues in logistics and supply chain management (see Tables 3 and 4) as well as the development of potential research avenues in the field of COVID-19 pandemic research (see Sect. 4). Moreover, our work provides a holistic evaluation of how current logistics and supply chain research suggests overcoming the challenges of COVID-19 pandemic by developing strategies for economic growth and welfare. Another contribution is the documentation of the methodological orientation of this particular research field, where the majority of the papers is rather conceptual than empirical (see Table 3). The few empirical papers provide evidence for transformational changes. Due to the ongoing pandemic developments, we expect a significantly increasing interest and publication of research in the future.
References Baveja, A., Kapoor, A., Melamed, B.: Stopping Covid-19: a pandemic-management service value chain approach. Ann. Oper. Res. 289(2), 173–184 (2020). https://doi.org/10.1007/s10479-02003635-3 Biedermann, L., Kotzab, H.: Supply chain risk management und supply chain Resilienz. WiSt Wirtsch. Stud. 50, 4–12 (2021). https://doi.org/10.15358/0340-1650-2021-4-4 Choi, T.-M.: Risk analysis in logistics systems: a research agenda during and after the COVID-19 pandemic. Transp. Res. Part E Logist. Transp. Rev. 145, 102190 (2021). https://doi.org/10. 1016/j.tre.2020.102190
Synthesising COVID-19 Related Research
49
Choi, T.-M.: Innovative “bring-service-near-your-home” operations under corona-Virus (COVID19/SARS-CoV-2) outbreak: can logistics become the messiah? Transp. Res. Part E Logist. Transp. Rev. 140, 101961 (2020). https://doi.org/10.1016/j.tre.2020.101961 Chowdhury, P., Paul, S.K., Kaisar, S., Moktadir, M.A.: COVID-19 pandemic related supply chain studies: A systematic review. Transp. Res. Part E Logist. Transp. Rev. 148, 102271 (2021). https://doi.org/10.1016/j.tre.2021.102271 Chowdhury, P., Shumon, R.: Minimizing the gap between expectation and ability: strategies for SMEs to implement social sustainability practices. Sustainability 12, 6408 (2020). https://doi. org/10.3390/su12166408 Coelho, F.S., Corrêa, V., Lisboa, R.L., Resch, S.: The public administration’s ‘engine room’ in the fight against covid-19. Rev. Adm. Publica 54, 839–859 (2020) Denyer, D., Tranfield, D.: Producing a systematic review. In: Buchanan, D., Bryman, A. (Eds.), The SAGE Handbook of Organizational Research Methods. Sage Publications Ltd. (2009) Eisend, M., Hampp, R: Metaanalyse (2020) EuPC: Raw material shortages in Europe impacting production of plastic products [WWW Document]. Packag. Eur (2021). https://packagingeurope.com/api/content/cd41a638-7feb-11eba830-1244d5f7c7c6/ Accessed 5 Oct 2021 Fettke, P.: Wirtschaftsinformatik 48(4), 257–266 (2006). https://doi.org/10.1007/s11576-0060057-3 Gonzalez, L., et al.: Moedas complementares digitais e políticas públicas durante a crise da COVID-19. Rev. Adm. Pública 54, 1146–1160 (2020). https://doi.org/10.1590/0034-761220 200234 Govindan, K., Mina, H., Alavi, B., 2020. A decision support system for demand management in healthcare supply chains considering the epidemic outbreaks: A case study of coronavirus disease 2019 (COVID-19). Transp. Res. Part E Logist. Transp. Rev. 138, 101967. https://doi. org/10.1016/j.tre.2020.101967 Gunessee, S., Subramanian, N.: Ambiguity and its coping mechanisms in supply chains lessons from the Covid-19 pandemic and natural disasters. Int. J. Oper. Prod. Manag. 40, 1201–1223 (2020). https://doi.org/10.1108/IJOPM-07-2019-0530 Handfield, R.B., Graham, G., Burns, L.: Corona virus, tariffs, trade wars and supply chain evolutionary design. Int. J. Oper. Prod. Manag. 40, 1649–1660 (2020). https://doi.org/10.1108/ IJOPM-03-2020-0171 Hedtstück, M.: Corona und Lieferketten: Wie eine Kriegssituation [WWW Document] (2020). https://www.finance-magazin.de/cfo/strategie/corona-und-lieferketten-wie-eine-kriegs situation-2054131/. Accessed 5 Oct 2021 India Today: The vaccine crisis [WWW Document]. India Today (2021). https://www.indiatoday. in/magazine/cover-story/story/20210426-the-vaccine-crisis-1791822-2021-04-17. Accessed 5 Oct 2021 Ivanov, D.: Predicting the impacts of epidemic outbreaks on global supply chains: a simulationbased analysis on the coronavirus outbreak (COVID-19/SARS-CoV-2) case. Transp. Res. Part E Logist. Transp. Rev. 136, 101922. (2020). https://doi.org/10.1016/j.tre.2020.101922 Ivanov, D.: Viable supply chain model: integrating agility, resilience and sustainability perspectives—lessons from and thinking beyond the COVID-19 pandemic. Ann. Oper. Res. (2020b). https://doi.org/10.1007/s10479-020-03640-6 Ivanov, D., Dolgui, A.: Viability of intertwined supply networks: extending the supply chain resilience angles towards survivability. a position paper motivated by COVID-19 outbreak. Int. J. Prod. Res. 58, 2904–2915 (2020). https://doi.org/10.1080/00207543.2020.1750727 Kumar, A.: Improvement of public distribution system efficiency applying blockchain technology during pandemic outbreak (COVID-19). J. Humanit. Logist. Supply Chain Manag. 11, 1–28 (2020). https://doi.org/10.1108/JHLSCM-06-2020-0050
50
I. Ö. Y. Hüseyino˘glu et al.
Li, T.: A SWOT analysis of China’s air cargo sector in the context of COVID-19 pandemic. J. Air Transp. Manag. 88, 101875 (2020). https://doi.org/10.1016/j.jairtraman.2020.101875 Lin, W., et al.: Blockchain technology in current agricultural systems: from techniques to applications. IEEE Access 8, 143920–143937 (2020). https://doi.org/10.1109/ACCESS.2020.301 4522 Liu, W., Liang, Y., Bao, X., Qin, J., Lim, M.K.: China’s logistics development trends in the post COVID-19 era. Int. J. Logist. Res. Appl. (2020). https://doi.org/10.1080/13675567.2020.183 7760 Marzantowicz, Ł., Nowicka, K., Jedli´nski, M.: Smart “Plan B” - in face with disruption of supply chains in 2020. LogForum vol. 16. (2020). https://doi.org/10.17270/J.LOG.2020.486 Niewiadomski, P., 2020. Corporate maturity desiderata in the face of the COVID-19 pandemic the digital plane of logistics microfoundations. LogForum Vol. 16. https://doi.org/10.17270/J. LOG.2020.495 Paul, S.K., Chowdhury, P.: A production recovery plan in manufacturing supply chains for a high-demand item during COVID-19. Int. J. Phys. Distrib. Logist. Manag. 51, 104–125 (2020). https://doi.org/10.1108/IJPDLM-04-2020-0127 Pinto, C.A.S.: Knowledge management as a support for supply chain logistics planning in pandemic cases. Braz. J. Oper. Prod. Manag. 17, 1–11 (2020). https://doi.org/10.14488/BJOPM. 2020.033 Queiroz, M.M., Ivanov, D., Dolgui, A., Fosso Wamba, S.: Impacts of epidemic outbreaks on supply chains: mapping a research agenda amid the COVID-19 pandemic through a structured literature review. Ann. Oper. Res. (2020). https://doi.org/10.1007/s10479-020-03685-7 Rejeb, A., Rejeb, K., Keogh, J.G.: COVID-19 and the food chain? Impacts and future research trends. LogForum vol. 16. (2020). https://doi.org/10.17270/J.LOG.2020.502 van Remko, H.: Research opportunities for a more resilient post-COVID-19 supply chain – closing the gap between research findings and industry practice. Int. J. Oper. Prod. Manag. 40, 341–355 (2020). https://doi.org/10.1108/IJOPM-03-2020-0165 Rousseau, D.M., Manning, J., Denyer, D.: Evidence in Management and Organizational Science: Assembling the Field’s Full Weight of Scientific Knowledge through Syntheses (SSRN Scholarly Paper No. ID 1309606). Social Science Research Network, Rochester, NY (2008). https:// doi.org/10.2139/ssrn.1309606 Rowe, F.: What literature review is not: diversity, boundaries and recommendations. Eur. J. Inf. Syst. 23, 241–255 (2014). https://doi.org/10.1057/ejis.2014.7 Sarkis, J.: Supply chain sustainability: learning from the COVID-19 pandemic. Int. J. Oper. Prod. Manag. 41, 63–73 (2020). https://doi.org/10.1108/IJOPM-08-2020-0568 Sharma, A., Zanotti, P., Musunur, L.P.: Drive through robotics: robotic automation for last mile distribution of food and essentials during pandemics. IEEE Access 8, 127190–127219 (2020). https://doi.org/10.1109/ACCESS.2020.3007064 Sharma, M., Luthra, S., Joshi, S., Kumar, A.: Developing a framework for enhancing survivability of sustainable supply chains during and post-COVID-19 pandemic. Int. J. Logist. Res. Appl. (2020). https://doi.org/10.1080/13675567.2020.1810213 Sharma, R., Shishodia, A., Kamble, S., Gunasekaran, A., Belhadi, A.: Agriculture supply chain risks and COVID-19: mitigation strategies and implications for the practitioners. Int. J. Logist. Res. Appl. (2020). https://doi.org/10.1080/13675567.2020.1830049 Shurson, G.C.: “What a waste”—can we improve sustainability of food animal production systems by recycling food waste streams into animal feed in an era of health, climate, and economic crises? Sustainability 12, 7071 (2020). https://doi.org/10.3390/su12177071 Singh, S., Kumar, R., Panchal, R., Tiwari, M.K.: Impact of COVID-19 on logistics systems and disruptions in food supply chain. Int. J. Prod. Res. 59, 1993–2008 (2021). https://doi.org/10. 1080/00207543.2020.1792000
Synthesising COVID-19 Related Research
51
Stamm, H., Schwarb, T.M.: Metaanalyse. Eine Einführung. Ger. J. Hum. Resour. Manag. 9, 5–27 (1995). https://doi.org/10.1177/239700229500900101 The Guardian. Global shortage in computer chips “reaches crisis point” [WWW Document]. The Guardian (2021). http://www.theguardian.com/business/2021/mar/21/global-shortage-in-com puter-chips-reaches-crisis-point. Accessed 5 Oct 2021 Trapp, R.: Konvergenz des Rechnungswesens: Eine Inhaltsanalyse der Diskussion um eine Annäherung des internen und externen Rechnungswesens in deutschsprachigen Fachzeitschriften. Gabler Verlag (2012). https://doi.org/10.1007/978-3-8349-3884-8 UNCTAD: Shipping during COVID-19: Why container freight rates have surged | UNCTAD [WWW Document] (2021). https://unctad.org/news/shipping-during-covid-19-why-containerfreight-rates-have-surged. Accessed 5 Oct 2021 Yadav, S., Luthra, S., Garg, D.: Internet of things (IoT) based coordination system in Agri-food supply chain: development of an efficient framework using DEMATEL-ISM. Oper. Manag. Res. (2020). https://doi.org/10.1007/s12063-020-00164-x Zhang, F., Wu, X., Tang, C.S., Feng, T., Dai, Y.: Evolution of operations management research: from managing flows to building capabilities. Prod. Oper. Manag. 29, 2219–2229 (2020). https:// doi.org/10.1111/poms.13231
Challenges and Approaches of Non-pharmaceutical Interventions for Airport Operations During Pandemic Situations Ann-Kathrin Rohde1(B) , Birte Pupkes1 , Rafael Mortensen Ernits1 , Dennis Keiser1 , Michael Lütjen1 , and Michael Freitag1,2 1 BIBA – Bremer Institut für Produktion und Logistik GmbH at the University of Bremen,
Hochschulring 20, 28359 Bremen, Germany [email protected] 2 Faculty of Production Engineering, University of Bremen, Badgasteiner Straße 1, 28359 Bremen, Germany
Abstract. During the COVID-19 pandemic, an uncontrolled spread of the virus worldwide was observed. To reduce the world wide spread of highly contagious viruses in the future, it is essential to target situations with a high risk for spreading contagious diseases. The risk of rapidly spreading of diseases is characterized by the fact that prevention measures cannot be applied as required, e.g., due to faulty execution. This situation has been observed during the COVID-19 pandemic in passenger transportation, especially at airports. Air transportation enabled the rapid exchange of people to and from different locations, which may have contributed to the spread of the virus, especially at the beginning of the pandemic before hygiene measures and vaccines were available. It would be particularly interesting for airports to use innovative guidance and tracking strategies and assistance systems to allow people to travel further and react fast in circumstances similar to the one observed during the COVID-19 pandemic. This paper examines the challenges in preventing the uncontrolled spread of disease in air transportation of people and highlights the current state of the art. Furthermore, a new approach using technical systems is demonstrated and discussed in other scenarios, e.g., rail transport. Keywords: COVID-19 · Airports · Technical assistance systems · Passenger transportation · Prevention measures
1 Motivation Pandemic situations, like COVID-19, are one of the significant threats to our modern society. Besides the alarming number of deaths and associated health impacts, the adopted actions, e.g., lockdowns, hygiene measures, social distancing and restricted traveling, led to dramatic changes in our life [1]. Many businesses were directly impacted or adapted to the new circumstances, many of them associated with people’s mobility [2]. Therefore, one of the most affected industries is the transport industry, especially the flight business © The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 M. Freitag et al. (Eds.): LDIC 2022, LNLO, pp. 52–64, 2022. https://doi.org/10.1007/978-3-031-05359-7_5
Challenges and Approaches of Non-pharmaceutical Interventions
53
[3]. Based on 2019 data, for Germany, the number of flights in 2020 changed from 1,93 million to 0,79 million, a reduction of nearly 60% [4]. But also ground transportation services of rail and coach were affected in 2020. In long-distance rail transport, with 82 million passengers, 46% fewer people travelled. In regular long-distance bus services, passengers fell by as much as 71% to 6.1 million [5]. The German situation can be fairly transferred to western European countries. For example, the demand for EU27 passenger air transport decreased by approximately 100 million passengers in March-June from 2019 to 2020 [6]. The reasons for these changes are restrictions of national governments and individual prevention measures of the inhabitants. It is highly relevant to minimize the risk of disease spreading while providing mobility to people. Measures for containing the spread of the COVID-19 could be seen in different phases of the pandemic. Besides hygiene, one key aspect is maintaining a certain distance among people, as pathogens can be transmitted through air via droplets released during breathing, talking, sneezing, or coughing [7]. Even though it’s argued which distance is appropriate [8, 9], it can be said that the chance of spreading decreases with increasing distance. With a focus on the spread of COVID-19, there are two-way associations between transport and the spread of the virus [10]. Many scientific publications show a strong positive correlation between transportation and travelling and the number of infection cases for different countries as, e.g., China, Italy and Japan. Sun et al. found out that most country borders were likely closed too late during the pandemic [11]. Haug et al. [12] analyzed the effectiveness of worldwide COVID-19 government interventions by 63 categories and ranked them. Public transport and airport restrictions were in the mid-field by rank 19 and 21 as well as airport health checks were at rank 59. These interventions show us that transport services and transport hubs are sensitive areas that have to be monitored and controlled in pandemic situations prudently. It is important to identify conditions and locations at hubs with an increased risk of spreading infectious diseases and to reduce the likelihood of such spread within the identified situations. In general, it could be helpful to restrict and perhaps close the transport services at pandemic peaks, eventhough the arrival of possible virus variants may change the decision´s dynamics [12]. Still, usually, the resilience should be improved by purposeful measures in the meaning of non-pharmaceutical interventions. This paper is dedicated to analyzing challenges and given approaches of nonpharmaceutical interventions, measures apart from vaccinations and medicines, for airport operations during pandemic situations. Hereafter, transport hubs, in general, are considered as well as an own approach is presented. A literature review accompanies the analysis of air travel during the COVID-19 pandemic and the derivation of the associated requirements. The existing approaches and solutions of airport operations during pandemic situations are subsequently described. For this purpose, the paper is structured as follows: The second chapter deals with the problem analysis. Based on these findings, the third chapter transfers the problems to technical, organizational, and business requirements for a technical assistance system as a non-pharmaceutical intervention, while the fourth chapter analyses the current approaches in the literature, which satisfy the requirements. The fifth chapter presents the author’s approach for a novel technical assistance system, while the sixth chapter
54
A.-K. Rohde et al.
shows the transferability to other transport hubs. The last and seventh chapter is about the conclusion.
2 Problem Analysis A primary factor for the worldwide spreading of the COVID-19 virus in 2020 has possibly been air transportation [13, 14]. But even before the COVID-19 pandemic, it has been shown that air transportation has the potential for rapidly spreading virus due to the the instrinsic high mobility, exchange and contact among people [15, 16]. A high-level problem analysis has been conducted to understand the risk of illness spreading due to air transportation. Different risks for infectious disease being transmitted during air travel could be found in the literature, and they were grouped into categories based on a cause-effect diagram. Process. As many people from different cohorts meet at airports, the number of unknown contacts is relatively high. Therefore it is challenging to trace back sources of infections and possible infected persons [17]. Furthermore, there is no current standardization for managing passenger and travel data globally to support the containment of diseases [18, 19]. Passengers. Sick passengers without symptoms or transmitting in pre-symptomatic phases of illness make it possible for diseases to spread at airports undetected [17, 20– 22]. Passengers coughing and sneezing without covering mouth and nose leads to the spread of diseases by emitting droplets to the air, even while talking aerosole droplets occur [7]. Environment. The airport environment is a confined indoor space in which viruses and bacteria, especially airborne diseases, can spread quickly due to insufficient ventilation [23, 24]. Equipment. As the infrastructure is shared by healthy and ill passengers, infections can occur, e.g., using the same toilet or touch screen at self check-in. [25, 26]. In a nutshell, the special conditions at airports that foster the rapid transmission of disease, with particular attention to travellers, can be classified as follows: A. High rate of unknown contacts, e.g., during boarding or in corridors, due to high volume of people B. Contact to other persons in a confined space C. Contact to other persons who come from very different cohorts D. Strong urge to travel as planned, despite feeling ill E. Hygienic reasons, e.g., sneezing/coving without covering, insufficient hand disinfection and/or surface disinfection
Challenges and Approaches of Non-pharmaceutical Interventions
55
It could be stated that most of the prevention and control procedures found in the literature consider the infectiouous disease transmission during flight [15, 27] and the majority of the non-pharmaceutical approaches suggest the tracking of passengers. In order to achieve a proper tracking of passengers, it would be worthy to consider possible contacts inside the airports. Hereafter, a technical assistance system could provide the necessary support. The next chapter seeks to define the necessary requirements to be fullfilled by such an technical assistance system.
3 Requirements Based on the previosly presented problem analysis, requirements for a technical assistance system are derived. The technical assistance system tackles on the reducing the risk of infection under pandemic situations and are clusterd into three main categories: technical, organizational, and economic requirements. Technical and User Requirements. Due to the strong repercussions in the event of a failure or possible malfunction of the system, high reliability and durability of all subsystems must be considered as the number one requirement. Similar durability requirements can be applied as with critical infrastructure. These include high reliability and a failsafe performance, redundancy of subsystems, and high IT security [28]. For the development of critical infrastructure, Hellström (2007) provides a regulatory framework that must be taken into account [29]. The second technical requirement can be defined as the rapid adaptability of the system to a changing pandemic situation, for example, changes concerning distance regulations. In addition to rapid adaptability, the scalability of the overall system must also be ensured because scalability is a fundamental requirement in data processing especially in the case of airports with high-capacity utilization. Finally, the human-machine interaction (HMI) must also be taken into consideration to achieve the highest possible acceptance of the technical assistance system among workers. This is subsequently an important role for all technical systems with human machine interaction [30]. In this case, ISO 9241 part 110 [31] provides a suitable framework for the development of a fitting HMI by providing specific guidelines for the design. Organizational Requirements. The organizational requirements for the technical assistance system stem primarily from the basic airport operating model as well as the standard passenger behaviour. First of all, it must be taken into account that journeys are made by individuals as well as groups. The recognition of group members that belong together enables targeted guidance and the maintenance of groups so that future unintentional separation can be prevented. In order to enable targeted route planning, access to passenger information (e.g., destination gate) and route planning based on it represents the second organizational requirement. Despite optimal route planning and compliance with instructions, deviations from the route may occur. This is caused, for example, by spontaneous purchases by passengers. This results in the necessity of a dynamic adjustment of the route planning based on the behaviour of the individual passengers. Finally,
56
A.-K. Rohde et al.
continuous contact tracking should be provided as a further safety measure. The authors Shahroz et. al (2021) prove the great benefit of contact tracking, as it enables rapid and targeted containment measures to be initiated in the event of infection [32]. Business Requirements. As a third and final area of requirements, the paper identifies economically motivated requirements. This is because, despite a pandemic situation, the rapid introduction and long-term use of the system are greatly facilitated by the highest possible business value for the airports. First, the system must enable high utilization of the airport’s capacity, since the economic operation of such infrastructure is only made possible by a high utilization, in this case, high passenger numbers [33]. As a further important economic requirement, full automation can be determined. Thus, avoiding possible additional expenses due to specialized employees. In order to enable sensible and economic operation even in a non-pandemic situation, adaptation to further application scenarios is advised. Here, the prospect of personalized advertising through visualizations is worth mentioning. Finally, the legal framework conditions for personal data should be regarded to ensure a high level of acceptance and low decision-making obstacles at the management level. Table 1 summarises all requirements based on the classification made. Table 1. Overview of identified requirements for a novel technical assistance system on airports. Technical requirements
Organizational requirements
Business requirements
• High reliability and robustness of all components • Human-machine-interaction optimized for complex systems and individualized needs • Fast adjustments based on pandemic situation (e.g., increase of distance) • Scalability based on utilization of airports
• Recognition and guidance of individuals and also groups • Planning of route based on passenger information • Continuous tracking of passenger for contact screening and dynamic re-routing • Consideration of spontaneous changes of passenger direction and re-routing
• Enabler for a high utilization of airports • Fully automated system with low operating costs • Consideration of legal requirements (personal data) • Possibility to adapt new use-cases of assistance system (e.g., personalized advertisement)
4 Current Approaches During the pandemic, various approaches to combat COVID-19 were quickly developed and implemented. These included policy measures such as travel bans and the closure of public facilities [34]. In addition, the development of a vaccine was quickly initiated and vaccination campaigns were launched. A large number of publications can already be found on this higher-level intervention and its impact [35]. In addition to
Challenges and Approaches of Non-pharmaceutical Interventions
57
these measures, airport operators and airlines also introduced measures directly after the outbreak of the pandemic to maintain or enable air travel as soon as possible [3]. The world health organization (WHO) provided a guideline here at the beginning of the pandemic situation in 2020 [36]. Initially, these were limited to administrative and process-related measures, as these can be implemented very quickly. These include, on the one hand, low seat occupancy on the part of the airlines. Secondly, as in many other facilities, warning signs that point out on general rules (e.g., keeping distance to each other or hygienic behaviour) have been put up at airports and better passenger management has been introduced. The IATA summarizes in a guideline the approaches for ground handling, including recommendations from official organizations, industry recommendations as well as health authorities, e.g. WHO, ECDC, CDC, ICAO, EASA and FAA, all approaches include social distance measures, cleaning and hygiene concepts as well as vaccinaton for avoiding virus transmission [37, 38]. Additionally, innovative technological concepts are being tested and deployed. Firstly, the introduction of touchless technologies, which are specifically designed to prevent the spread of COVID-19 via surfaces [39, 40]. Secondly, the active contact tracking of passengers, which serves the purpose of tracing in case of a chain infection. Various exemplary technological implementation cases can be found in scientific literature, such as specially developed smartphone apps [41]. In some countries, secondary data sources such as mobile payment data are utilized [42]. Thirdly, various simulation models predict the spread [41]. In combination with these models, machine learning methods are applied, and clusters of identified persons are recognized based on big data approaches. This can then be used as a basis for deciding on localized lockdowns. Finally, access controls are used. These access controls are designed to detect sick passengers to separate them. Table 1 shows the technologies identified from the literature, provides implementation examples and shows the limitations of the approaches. It becomes apparent that the aforementioned options all either have low reliability (touchless technologies and approaches and access control) or are to be understood as reactive measures (tracking of contacts, forecast planning occurrence of infection, identification of clusters of infected people). Therefore, this assessment of approaches concludes that currently available strategies are proactive and, at the same time, not reliable. Technologies that prevent the initial contagion and spread of COVID-19 at airports cannot yet be implemented, or only to a limited extent. However, the identified approaches represent only an excerpt and are not based on a full systematic analysis. In response to the identified requirements in chapter three and the lack of a technological solution that fulfils these, the need for a new type of holistic system and procedure has emerged. In the next chapter, the authors, thus, propose a new proactive technical assistance system specifically for airports (Table 2).
58
A.-K. Rohde et al.
Table 2. Table with methods and technical solutions that are currently used to decrease the uncontrollable spreading of infectious diseases. Approach
Implementation example
Reference
Low reliability
[39, 40]
Tracking of contacts
Tracking Reactive Application for approach mobile phones, use of secondary data (e.g., mobile payment)
[40, 41, 43–45]
Forecast planning occurrence of infection
Machine learning Reactive models approach
[41, 44]
Identification Support of clusters of decision making infected people based on Big Data
Reactive approach
[40, 42, 44]
Access control temperature measurement at the entrance
Low reliability
[46]
Requirement fulfilment
Defined requirements for an technical assistance system at airports not fulfilled
Touchless Touchless technologies Check-In (e.g., and approaches Heathrow Airport in London)
Limitation
5 Proposed Approach As mentioned above, the non-pharmaceutical interventions for airport operations during pandemic situations have to combine administrative, process-related and technical measures in order to achieve highly efficient safety for passengers and employees. Considering the worked-out requirements, the following approach is proposed to track and navigate people through the airport while obeying the safety distance 1.5m to other persons/groups. Such a technical assistance system includes two main aims: first, “detection of compliance” and second, “anticipatory person navigation.“ Both objectives are interlinked and have mutual influence. Finally, the assistance system has to display the results to each individual. Figure 1 exemplifies the current boarding procedure. The red circles (b) represent the lack of necessary distance among the passengers. Figure 2 shows the optimized route (b) planning communicated by a dynamic ground projection as a dynamic output system, which can be used to identify the direction of movement to be followed.
Challenges and Approaches of Non-pharmaceutical Interventions
59
Fig. 1. Current boarding procedure.
Fig. 2. Boarding procedure with the proposed approach.
The assistance system integrates each persons’ destination and plans its’ optimal route to it under consideration of the infection prevention regulations and with the reaction on each person’s current behaviour about the infection prevention regulations and advises better behaviour under consideration of the person’s surroundings. The person’s surroundings may include other people, infrastructure (e.g., building elements), or obstacles (e.g., waste bin). To achieve this, each person in the system or within a predetermined area of the system must be reliably identified and continuously located. It is proposed to use an AI tracking system. Furthermore, the individual destination (e.g., departure gate) has to be detected. This can be executed via transport ticket recognition or by observation and forecast of peoples’ movement patterns. The integration of intermediate destinations (e.g., duty-free shop or toilet) has to be provided and can be executed automatically by the system or manually by the traveling person. Furthermore, the identified person, the identified destination, and the identified intermediate destinations must be assigned to one another and fed into the route planner. The route planner contains the information of all people currently in the system (e.g., current position, destination, planned route, current mode: resting, moving, etc.) as well as superordinate information (e.g., information on the infrastructure including ventilation systems and status of technical systems, current aerosol load, route, etc.) and rules for the execution of route planning. For simplification, the route planner is divided into two modules: In the first module, a global route defines waypoints within the terminal for the tracked person and thus sets individual waypoints, especially at turns. The local planning as
60
A.-K. Rohde et al.
the second module then proposes a route to the next waypoint of the people starting from their current position, whereby distance conditions to other people and residence restrictions are included in the local planning. Thus, a breach or a potential breach of the regulations are automatically detected, and solutions for pre-compliance are calculated and attended for route planning. The AI tracking system monitors the implementation of the route planning, and an optimal route planning solution based on the status is repeatedly searched for and transmitted. Finally, the system has to display the planned route to each person. The proposed system uses sensor and infrastructure data in conjunction with an AI identification system and an AI tracking system. Furthermore, it uses a route planer consisting of global and local calculations to create an optimal route planning for each person at a traffic junction and, if necessary, to derive infrastructural measures (e.g., closing a passage). The results are then shown via a dynamic output system.
6 Further Use Cases Considering the parameters determined regarding special conditions at airports that foster the rapid transmission of disease, the following other use cases have been identified to apply the approach. First railway stations are similar to those that are described in the airport scenario. As part of the pre-processes to the main transportation, which is managed by rail, the people to be transported enter the building at one or different entrances and follow an individual route to reach the appropriate platform. There are a large number of people from different cohorts that are in a confined space e.g., on the aisles, waiting at a platform for the onward journey or within a compartment of the train. In the railway station scenario, the route to the platform may also include intermediate destinations, e.g., a restaurant. The second potential use case for the proposed system is supermarkets. At Supermarkets, there are also various people from different cohorts in a limited space, and undercutting of minimum distances can occur. Multitude storage shelves characterize supermarkets and aisles between the rows of shelves through which customers and staff move to reach the goods on the shelves. The direction of using the aisles is not prescribed and does not necessarily result from a destination or intermediated destinations. Thus, particular attention should be paid to heavily frequented aisles and the waiting area at the cash desks as well as the entrance and exit doors. A persons’ route though the Supermarket is defined by the storage location of the goods to be bought as well as the purchasing list as well as personal knowledge of the storage locations. The last identified use case is business fairs. There are also various people from different cohorts in a limited space at business fairs, and undercutting of minimum distances can occur. Particular attention should be paid to heavily frequented corridors and the entrance and exit doors of the fair. Fairs are characterized by a multitude of presentation areas that are primarily in halls and thus with limited space. The number, order and specification of intermediate destinations, here specific booths, has a very high variability among the visitors and defines the individual route of a person. App-based systems could be used to predefine the individual route. The mentioned use cases meet a high number of the classified conditions of the airport scenario as the rate of unknown contacts, the confined space, contact with people from different cohorts, urge to act as planned and hygienic reasons.
Challenges and Approaches of Non-pharmaceutical Interventions
61
7 Limitations, Conclusion and Outlook The topic presented in this paper summarizes the challenges to avoid the virus spreading and current approaches observed during the COVID-19 pandemic, especially the methods used at airports. It can be stated that current measures are still lacking in consistency and may not contain and control future pandemics right from the beginning. The reasons were summarized after the presented analysis and are mostly related to tracking and coordinating people’s movement. Hereafter, the current approaches cannot deliver a robust and fast answer to the great variability inherent to airports associated with the combination and concentration of many people, together with a lack of distancing. After analyzing the current approaches, requirements for containing and controlling the spreading of the virus are derived, and current technical approaches are classified. A new approach is presented as a concept for tracking and coordinating people’s movement, and it aims to contribute to containing and controlling virus spreading. Even though the proposed approach foresees the implementation at an airport; the concept could be used in different scenarios without significant changes, as described, on railways stations, supermarkets, and even business fairs. The proposed approach is mainly based on a visual and optical solution. This leads to a fully automated system with low operating costs, whereby the investment costs may be high. The system can be enlarged through further implementation of technical equipment, as cameras or spots and thus, the technical systems that follows the approach is characterized by a high scalability. The recognition and guidance of persons and their personal destinations as well as intermediate destinations leads to a complex system, that meets individual needs. The constant tracking of every person leads to a continuous contact screening and the dynamic re-routing helps to avoid crowds. Thus, the mentioned approach is a high valuable tool to minimize the risk of disease spreading while providing mobility to people. Nevertheless, it is still necessary to be validated as a socio-technical system. A study to identify the added value of such technical assistance systems in combination with administrative and process-related measures should take place in real-world laboratory. In this sense, considering the challenges associated with diversity, it is essential to understand its effectiveness among the most different groups of people, e.g., unruly, blind, children and disabled. The evaluation of different scenarios, e.g., stress situations or sudden changes, must also be considered to allow a broad understanding of the system’s limitations. One challenge is the acceptance of the system by travellers. Therefore, acceptance studies should be carried out for this purpose or integrated in the studies regarding the socio-technical system. Besides, there are technical challenges with implementing such a system, primarily associated with real-time capabilities, as latency or lack of processing power could compromise tracking. Thus, a second paper in the development frameworks examines possible technologies for the proposed system [47]. Additionally challenges concern questions in the area of data protection and related to this in the area of data security. Detailed concepts concerning data security and data privacy have to be elaborated and integrated before marked launch. Furthermore, the concept’s economic viability must be verified to balance the effort for implementation and investigate further uses, e.g., marketing purposes for guiding customers throughout stores or even for
62
A.-K. Rohde et al.
coordinating processes as boarding more effectively. Hereafter, the concept could serve for multi-purposes and eventually become more attractive.
References 1. Yang, Q.-C., Chen, X., Chang, C.-P., Chen, D., Hao, Y.: What is the relationship between government response and COVID-19 pandemics? Global evidence of 118 countries. Struct. Chang. Econ. Dyn. 59, 98–107 (2021) 2. Wang, S., et al.: Observing the silent world under COVID-19 with a comprehensive impact analysis based on human mobility. Nature. Sci. Rep. 11, 14691 (2021) 3. Arora, M., Tuchen, S., Nazemi, M., Blessing, L.: Airport pandemic response: an assessment of impacts and strategies after one year with COVID-19. Transp. Res. Interdiscip. Perspect. 11, 100449 (2021) 4. Statistisches Bundesamt. Luftverkehr auf Hauptverkehrsflughäfen 5. Statistisches Bundesamt. Fahrgastzahl im Linienfernverkehr mit Bahnen und Bussen im Jahr 2020 halbiert, Pressemitteilung (2021) 6. Botlíková, M., Botlík, J., Stuchlíková, J.: Impacts of the global pandemic on the aviation industry of the European Union. In: The 20th International Scientific Conference Globalization and its Socio-Economic Consequences (2021) 7. Schijven, J., Vermeulen, L.C., Swart, A., Meijer, A., Duizer, E., de Roda Husman, A.M.: Quantitative microbial risk assessment for airborne transmission of SARS-CoV-2 via breathing, speaking, singing, coughing, and sneezing. Environ. Health Perspect. (2021) 8. Setti, L., et al.: Airborne transmission route of COVID-19: why 2 meters/6 feet of interpersonal distance could not be enough. Int. J. Environ. Res. Publ. Health (2020) 9. Sun, C., Zhai, Z.: The efficacy of social distance and ventilation effectiveness in preventing COVID-19 transmission. Sustain. Cities Soc. 62, 102390 (2020) 10. Zhang, J., Hayashi, Y., Frank, L.-D.: COVID-19 and transport: findings from a world-wide expert survey. Transp. Policy 103, 68–85 (2021) 11. Sun, X., Wandelt, S., Zhang, A.: On the degree of synchronization between air transport connectivity and COVID-19 cases at worldwide level. Transp. Policy 105, 115–123 (2021) 12. Monajjemi, M., Sayiner, H.-S., Kandemirli, F., Mollaamin, F.: An overview on lambda, epsilon, kappa, iota and zeta variants of covid-19 and its probability to merge with delta & delta plus, why it is a concern. Biointerf. Res. Appl. Chem. 5(12), 6895–6914 (2022) 13. Lau, H., et al.: The association between international and domestic air traffic and the coronavirus (COVID-19) outbreak. J. Microbiol. Immunol. Infect. 53(3), 467–472 (2020) 14. Gössling, S., Scott, D., Hall, C.-M.: Pandemics, tourism and global change: a rapid assessment of COVID-19. J. Sustain. Tour. 29, 1–20 (2020) 15. Mangili, A., Gendreau, M.-A.: Transmission of infectious diseases during commercial air travel. The Lancet 365(9463), 989–996 (2005) 16. Bobashev, G., Morris, R.-J., Goedecke, D.-M.: Sampling for global epidemic models and the topology of an international airport network. PLoS ONE 3(9), e3154 (2008) 17. Wilder-Smith, A., Chiew, C.-J., Lee, V.-J.: Can we contain the COVID-19 outbreak with the same measures as for SARS? Lancet Infect. Dis 20, e102–e107 (2020) 18. Allam, Z., Jones, D.-S.: On the coronavirus (COVID-19) outbreak and the smart city network: universal data sharing standards coupled with artificial intelligence (AI) to benefit urban health monitoring and management. Healthcare 8(46), 1–9 (2020) 19. Schenkel, K., et al.: Risk assessment guidelines for infectious diseases transmitted on aircraft. Technical Report commissioned by European Centre for Disease Prevention and Control (2009)
Challenges and Approaches of Non-pharmaceutical Interventions
63
20. Quilty, B.-J., Clifford, S., Flasche, S., Flasche, S., Eggo, R.-M.: Effectiveness of airport screening at detecting travellers infected with novel coronavirus (2019-nCoV). Euro Surveill. 25(5), 1–5 (2020) 21. Nishiura, H., Linton, N.-M., Akhmetzhanov, A.-R.: Serial interval of novel coronavirus (COVID-19) infections. Int. J. Infect. Dis. 93, 284–286 (2020) 22. Angelo, K.-M., Kozarsky, P.-E., Ryan, E.-T., Chen, L.-H., Sotir, M.-J.: What proportion of international travellers acquire a travel-related illness? A review of the literature. J. Travel Med. 24, 1–7 (2017) 23. Qian, H., Miao, T., Liu, L., Zheng, X., Luo, D., Li, Y.: Indoor transmission of SARS-CoV-2. Int. J. Indoor Environ. Health 29(1), 639–645 (2021) 24. Drury, J., Rogers, M.-B., Marteau, T.-M., Yardley, L., Reicher, S., Stott, C.: Re-opening live events and large venues after Covid-19 ‘lockdown’: behavioural risks and their mitigations. Saf. Sci. 139, 1–8 (2021) 25. World Health Organization (WHO). Management of ill travellers at POINT OF ENTRY (international airports, seaports, and ground crossings) in the context of COVID-19. Interim Guidance (2020) 26. Di Battista, A., Nicolaides, C., Georgiou, O.: Modelling disease transmission from touchscreen user interfaces. Roy. Soc. Open Sci. 26(3), 1513 (2020) 27. Mangili, A., Vindenes, T., Gendreau, M.: Infectious risks of air travel. Microbiol. Spect. (2015) 28. Alcaraz, C., Zeadally, S.: Critical infrastructure protection: requirements and challenges for the 21st century. Int. J. Crit. Infrastruct. Prot. 8, 53–66 (2015) 29. Hellström, T.: Critical infrastructure and systemic vulnerability: towards a planning framework. Saf. Sci. 45(3), 415–430 (2007) 30. Quandt, M., Stern, H., Grudpan, S., Beinke, T., Freitag, M., Malaka, R.: Evaluation of humancomputer-interaction design in production and logistics by using experimental investigations. In: Freitag, M., Haasis, H.-D., Kotzab, H., Pannek, J. (eds.) LDIC 2020. LNL, pp. 554–566. Springer, Cham (2020). https://doi.org/10.1007/978-3-030-44783-0_51 31. Deutsches Institut für Normung e.V.: DIN EN ISO 9421-110: Ergonomie der MenschMaschine-Interaktion - Teil 110: Grundsätze der Dialogestaltung (2008) 32. Shahroz, M., et al.: COVID-19 digital contact tracing applications and techniques: a review post initial deployments. Transp. Eng. 5, 1–5 (2021) 33. Rotondo, F.: An explorative analysis to identify airport business models. Res. Transp. Bus. Manag. 33, 100417 (2019) 34. Xue, L., Jing, S., Sun, W., Lui, M., Peng, Z., Zhu, H.: Evaluating the impact of the travel ban within mainland China on the epidemic of the COVID-19. Int. J. Infect. Dis. 107, 278–283 (2021) 35. Rozhnova, G., et al.: Model-based evaluation of school- and non-school-related measures to control the COVID-19 pandemic. Nat. Commun. 12, 1614 (2021) 36. World Health Organization (WHO). Operational considerations for managing COVID-19 cases or outbreak in aviation. Interim Guidance (2020) 37. International Air Transport Association (IATA). Guidance for ground handling during COVID-19 (2021) 38. International Air Transport Association (IATA). Safely Restarting Aviation - ACI and IATA Joint Approach (2020) 39. Amankwah-Amoah, J.: COVID-19 pandemic and innovation activities in the global airline industry: a review. Environ. Int. 156, 106719 (2021) 40. Thomas, M.-J., Lal, V., Baby, A.-K., Rabeeh, V.P.M., James, A., Raj, A.-K.: Can technological advancements help to alleviate COVID-19 pandemic? a review. J. Biomed. Inf. 117, 103787 (2021)
64
A.-K. Rohde et al.
41. Whitelaw, S., Mamas, M.-A., Topol, E., van Spall, H.-G.-C.: Applications of digital technology in COVID-19 pandemic planning and response. Lancet Digital Health 2(8), e435–e440 (2020) 42. Shaw, R., Kim, Y.-K., Hua, J.: Governance, technology and citizen behavior in pandemic: lessons from COVID-19 in East Asia. Progr. Disast. Sci. 6, 100090 (2020) 43. Fischer, S.-C., Kohler, K., Wenger, A.: Digitale Technologien im Corona-Krisenmanagement. Report ETH Zurich (2020) 44. Wang, Q., Su, M., Zhang, M., Li, R.: Integrating digital technologies and public health to fight Covid-19 pandemic: key technologies, applications, challenges and outlook of digital healthcare. Int. J. Environ. Res. Publ. Health 18(11), 6053 (2020) 45. Schwarzbach, P., Engelbrecht, J., Michler, A., Schultz, M., Michler, O.: Evaluation of technology-supported distance measuring to ensure safe aircraft boarding during COVID-19 pandemic. Sustainability 12(20), 8724 (2020) 46. Choi, J.-H.: Changes in airport operating procedures and implications for airport strategies post-COVID-19. J. Air Transp. Manag. 94, 102065 (2020) 47. Börold, A., et al.: Technology rewiew for guiding persons in airports and other hubs. In: Proceedings of 8th International Conference on Dynamics in Logistics (Accepted) (2022)
Identifying Common Elements Within Supply Chain Resilience and Sustainability - An Exploratory Study Based on Bibliographic Analysis Piotr Warmbier(B) and Aseem Kinra Global Supply Chain Management, University of Bremen, Hochschulring 20, 28359 Bremen, Germany {warmbier,kinra}@uni-bremen.de
Abstract. Over the last couple of years, there has been an increase in reciprocal discussion within the fields of supply chain resilience (SCRES) and sustainability (SSCM). Although some thematic overlap has been noted, SCRES and SSCM are generally still considered to be two separate domains. However current global SC events and legislative initiatives demonstrate why it is important to view SCRES and SSCM in combination to solve the problem of long-term supply chain risks and disruptions. The purpose of this paper is to explore whether SCRES and SSCM can be integrated, which overlapping elements need to be considered and which main congruent capabilities exist. The exploratory study applies a citation network and main path analysis based on a dataset of two previously conducted systematic literature reviews. The literature review revealed great potential for combining the two research domains based on four identified connecting elements - supply chain capabilities, practices, risks and performance. We also identified great potential for the application of two main congruent capabilities - transparency and big data analytics. The four elements identified provide us with a useful basis for creating a theoretical framework for integrating SSCM and SCRES. We also highlight the importance of the congruent capabilities that are required to overcome paradoxical tensions between resilience and sustainability. Keywords: Uncertainty · Supply chain resilience · Sustainable supply chain management · Supply chain capabilities · Bibliographic analysis
1
Introduction
Numerous SC disruption events in recent times, such as the semiconductor shortage in 2021, show the dependency and vulnerability of global supply chains (SC). They also highlight the strategic importance of supply chain management (SCM) and risk management (RM) in achieving a competitive and sustainable advantage [9,37]. However, one shortcoming of traditional RM is its inability to adequately c The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 M. Freitag et al. (Eds.): LDIC 2022, LNLO, pp. 65–81, 2022. https://doi.org/10.1007/978-3-031-05359-7_6
66
P. Warmbier and A. Kinra
identify and handle low-probability, high-consequence (LPHC) events [47,68]. Supply chain resilience (SCRES) is seen as a concept that can bridge this gap and supplement existing RM programs by enabling a SC to overcome unforeseen disruptions and create a sustainable, competitive advantage [68,69]. At the same time, new regulations, such as the initiative for a SC law currently being discussed in the European parliament, and our changing society are also driving the topic of supply chain sustainability (SSCM) onto the agendas of enterprises and academics alike [23,59,79]. This is generating both new opportunities and uncertainties that must be considered by manufacturing companies if they are to maintain their competitive advantage. In the field of SSCM, sustainability risks are regarded as a critical factor. Increasingly, they are seen as valuable elements that must be encompassed within RM to help organisations prevent losses [32]. Furthermore, unexpected risk events are not only relevant within traditional areas of operational and economic risk but also in social and ecological fields. One example that demonstrates the aforementioned arguments is the Covid 19 pandemic and its consequences for the SC [25,72]. Looking at the business articles published just after the Covid-19 pandemic outbreak, it is evident that an emphatic recommendation for companies occurred to bolster both their resilience and sustainability [25,48,72]. In these challenging times, companies cannot only focus on either SCRES or SSCM individually. Instead, they must merge the two streams and transform their SC accordingly to overcome long-term SC disruptions and retain their sustainable, competitive advantage [48]. However, we must ask ourselves how SSCM and SCRES can be merged and how companies can achieve a resilient, sustainable SC. Earlier research recommended transforming global SC using leagile, ecosilient strategies, LARG (lean, agile, resilient and green) strategies and, most recently, viability to tackle the problem of LPHC disruptions [4,12,38,62]. These developments demonstrate a partial integration of SCRES and SSCM. However, complete integration that will tackle the problem of long-term SC disruption is still at a very early stage. SC capabilities (SCC) are a decisive factor in evolving a SC in a certain direction and so can indeed play a vital role in the synthesis of the two research domains [69]. Yet, until now, research has focused on analysing SCCs in SCRES and SSCM separately. Given the perceived short comings of this previous research, the purpose of this paper is to explore if and how SCRES and SSCM can be integrated. We have formulated our research questions (RQ) on this basis and grouped them into two main question clusters: RQ1 - What are the common areas of overlap between SSCM and SCRES? RQ2 - What are the main congruent capabilities discussed in the associated literature?
2
Literature Review
Global SC networks are seen as complex systems due to the numerous parties involved in the value-added chain [10,46]. This complexity in the SC increases uncertainty and therefore the risk of disruptions [9]. The main elements of uncertainty that can be covered by traditional RM are related to supply, resource
Supply Chain Resilience and Sustainability
67
availability, production costs, operations, exchange rates, transportation, raw material prices and demand [71,89]. However extreme events, such as pandemic events, natural disasters or terrorist attacks, must also be taken into account as uncertainty factors [49]. The LPHC nature of these extreme SCR events challenge traditional RM and require RM programs to be supplemented with resilient strategies [68]. Tukamuhabwa et al. [88] defined SCRES as “the adaptive capability of a supply chain to prepare for and/or respond to disruptions, to make a timely and cost-effective recovery, and therefore progress to a post-disruption state of operations - ideally, a better state than prior to disruption”. An SC breakdown on the operational side can be mirrored by a negative impact on the economic side of a SC. In addition, longer periods operating in crisis mode can have a negative effect on the social and ecological aspects of SCM. Crisis management increases workloads and has a greater ecological impact [36,64]. This can be easily demonstrated in the post-Covid phase. In 2021, a significant number of backlogs at trading companies was caused by longer ocean lead times, which led to a higher number of air shipments to resolve critical bottlenecks [64]. This illustrates the first interrelation of SCRES and SSCM, in which resilience is a prerequisite for ensuring long-term sustainability and competitive advantage. At the same time, SSCM also became a primary research field within SCM. The focus shifted from cost-oriented and service-oriented SCs, in which operational excellence was paramount, to a holistic approach that addresses economic, social and environmental issues within the SC [13,22,78]. With regard to the three pillars of SSCM, SC performance (SCP) measures can be divided into three clusters within the advanced field of SCM - operational, ecological and social performance measures [58]. All three performance clusters can be interrelated and can be negatively affected by risk events. The definition put forward by Closs et al. [17] addresses SC risks (SCR) in the context of SSCM: “sustainability of supply chain is defined as the ability of an organization to mitigate, detect, respond, and to recover from growing global threats related to supply chain and to enhance the long-term value”. So sustainable SCR events can also have LPHC characteristics that resilient SCs could help to overcome. This also highlights the second correlation, in which the SCRES concept can be enriched by considering additional sustainable SCR clusters. SCCs play a vital role in both research domains as they are considered to be crucial factors for developing a sustainable resilient SC. Ponomarov and Holcomb [69] stated that “capability is considered as the major role of strategic management in adapting, integrating, and reconfiguring resources, organisational skills and functional competencies to respond to the challenges of the external environment”. Therefore, SCCs can be viewed as potential counterparts to SCRs and used to manage SC disruptions [68]. However, SCRES and SSCM research has previously only focused on analysing SCCs individually on either a researchbased view (RBV) or dynamic capability view (DCV) [8,33,57,68]. But current literature already shows potentially congruent SCCs in both SSCM and SCRES, such as agility, transparency, innovation, knowledge management, learning, SC integration, SC reengineering and collaboration [3,7,16,19,42,44,75]. So congruent SCC can be defined as SCC that appear in both research domains and
68
P. Warmbier and A. Kinra
that can be exploited to develop the respective field of SSCM or SCRES. Thus, capabilities that have a similar definition and similar impact on SC development can be seen as congruent SCCs. It would be advantageous to analyse congruent SCCs found in the associated literature for the purpose of providing companies with guidance for developing their SC in the direction of SCRES and SSCM simultaneously. In recent years, researchers have suggested several strategic directions in which a SC can be cultivated to tackle the problem of RM and performance development. These studies partially integrated both research domains. Azevedo et al. [4] delivered an assessment model displaying partial integration of SSCM and SCRES. They devised an integrated composite index, called the ecosilient index. It is used to assess the ecological aspects and resilience of companies and their SC. Moving on to decision-making models, Cabral et al. [12] developed a lean, agile, resilient and green (LARG) framework. Their framework and integrated analytic network process model focusses on facilitating the selection of the most appropriate supply chain management practices (SCMPs) and SCP KPIs to be implemented with regards to LARG. Ivanov [38] suggested another strategic orientation involving viability in the SC (VSC). His VSC model takes the three aspects of agility, resilience and sustainability into account. In this way, he creates a SC that is able to redesign structures and adapt its performance in changing environments, such as the Covid pandemic. However, other researchers consider agility to be one of the main SCCs already present in SCRES [16,44,75]. By considering the correlations between SCR and SCC, the work of Multaharju et al. [65] illustrates how organisations can manage social and ecological risks caused by service providers. The work conducted by Ivanov [37] was in a similar vein. He analysed both the conflicts and benefits of different SCMPs used in SCRES and SSCM strategies. His findings reveal a conflict of interest between some SCMPs. This highlights the research gap that we identified regarding the analysis of congruent SCCs. According to Ivanov [37], research related to an integrated framework of SC sustainability and resilience is still in its infancy. With regard to the research gap, a common foundation for creating an integrated, comprehensive view is still missing.
3
Research Methodology
Our study is an exploratory study focusing on answering both RQs by applying bibliographic analysis. Bibliographic analysis, a methodology for statistical evaluation of published scientific articles, provides robust techniques for classifying thematic areas and cluster-related research fields and for identifying emerging themes [86]. We employed CNA (citation network analysis) and MPA (main path analysis) to deliver the results we required to address our research questions by verifying the emerging trend of synthesis, analysing common elements and exploring the main congruent SCCs [93]. We used CNA to address both RQs and MPA was applied to verify our findings. We should note that we used CNA to focus on the particular text segments referred to and MPA to focus on the whole corresponding papers. We adapted the merged CNA and MPA structure
Supply Chain Resilience and Sustainability
69
from Colicchia and Strozzi [18] who developed a combined method of systematic literature review (SLR), CNA and MPA. The SLR data for both samples was collected from the Scopus and WoS databases on the 15th of March 2020. Because the citation data logic regarding cited references and the address information in Scopus is different from that of WoS, we had to address the problem of database selection [54]. We decided to use the WoS database as it is more commonly used for citation analysis [54]. We used the final SLR samples consisting of 114 SCRES and 171 SSCM articles to collect citation data from WoS on the 25th of January 2021. The screening process and results of each step are presented in Table 1. We first organised the collected citation data samples using the HistCite software package to enable us to analyse the citation network. Subsequently, we conducted the CNA using VOSviewer and the MPA in Pajek. CitaƟon network analysis (CNA) Main path analysis (MPA)
SystemaƟc literature review (SLR) 1. LocaƟng studies
1. Data collecƟon
3. Main path analysis
2. Study selecƟon and evaluaƟon
2. CitaƟon network analysis
3.1 QuanƟfying the traversal weight of citaƟons 3.2 Create global main path
2.1 DistribuƟon citaƟon network 2.2 ArƟcles with linkages in both citaƟon networks
4. ReporƟng results
3. Final SLR sample data
Fig. 1. Overview of methodology Table 1. Steps and output of CNA and MPA Steps
SCRES output
SSCM output
Data input - final SLR sample data
114 papers
171 papers
Citation data collection - citation network samples (−12) 102 papers (−10) 161 papers (collection data from WoS; Excluding articles with missing data in WoS; Data collection date: 25/01/21) Combination of citation network data from both samples (duplicates removed)
(−2) 261 papers
Citation network analysis (connecting articles)
(−2) 17 papers
Main path analysis
10 papers
7 papers
Starting from a static perspective, the selected papers can be studied by mapping the citation network of the samples separately or mapping both networks together. By visualising both samples in combination, we were able to identify and analyse related articles. Main path analysis is a dynamic approach and, as such, must be studied using a two-step algorithm [40,52]. This study followed the recommendation of Liu et al. [55] for calculating the traversal weight and applied the search path link count method. For the purpose of extracting the main path, the standard global main path is a useful method for determining the most significant path overall [56]. Therefore, we selected this method.
70
4
P. Warmbier and A. Kinra
Results
Out of the 261 papers included in the CNA, 215 publications show a connection within their corresponding research domain. The connected publications from both citation networks are visualised in Fig. 2. We identified 17 papers that cite publications from the respective other research domain. We should emphasise that thematic interaction was evident mainly after 2017. We analysed these 17 papers (Table 2) based on the referenced text segments. 4.1
The Four Connecting Elements in SSCM and SCRES Literature
The 17 correlating articles (Table 2) revealed insights into overlapping elements that were applied in both SSCM and SCRES. Shokouhyar et al. [82] and Jeble et al. [39] referenced Brandon-Jones et al. [10] to discuss complexity and uncertainty in the SC related to SSCM. According to Brandon-Jones et al. [10], a SC network involves numerous parties and, therefore, is a complex system. This supply base complexity increases the uncertainty of the SC, which in turn will result in disruptions sooner or later. All three SSCM publications, Shokouhyar et al. [82], Jeble et al. [39] and Mani et al. [61], demonstrated the need to take social and ecological SCR into consideration within the SCM field in addition to traditional operational and financial SCR. Continuing with possible countermeasures against SCR and uncertainty, Kilubi and Rogers [45] referenced the work of Huq et al. [35] on SSCM, stating that companies that focus on strategic SCC are better at managing SC disruptions. Shibin et al. [81], Garcia-Torres et al. [26] and Dubey et al. [20] also referenced the work of Brandon-Jones et al. [10] on complexity related to SCC. According to the results of Brandon-Jones et al. [10] information sharing and SC integration have a direct positive effect on transparency within the SC. Transparency also enhances resilience according to Brandon-Jones et al. [10] and sustainability in the SC according to Shibin et al. [81], Garcia-Torres et al. [26] and Dubey et al. [20]. A connection that is very often mentioned within SCRES and SSCM datasets is big data and predictive analytics (BDPA) capability. This ability can help to increase visibility to improve SCR management and increase SCP [20,31,60,92]. SCP plays a vital role in SSCM and SCRES. Golgeci and Ponomarov [27] as well as Defee and Fugate [19], emphasise the positive influence of a dynamic capability view and resource-based view on the SCP outcome. SCCs are not only important for handling SCR, they also have a positive effect on the SCP outcome [20,34,39,83]. The last connecting elements that we can present are SCMPs. SCMPs are discussed in the literature as a bundle of tools, processes and actions within an organisation that support the development of SCCs [84]. GarciaTorres et al. [26], summarised their findings as follows: certain SCCs, such as transparency, must be examined together with SCMPs, which can support the creation of specific and targeted SCCs. In this case, they stressed that investment in new SCMPs and technologies, such as RFID, blockchain and BDPA, can enhance visibility.
Supply Chain Resilience and Sustainability SCRES network
SSCM network
SCRES main path
71
SSCM main path
Fig. 2. Citation network and main path analysis of SCRES and SSCM research. Note: Left side visualises the citation network and the right side the main path
4.2
Capability of Transparency and Big Data Analytics
As already highlighted in the previous subsection, both transparency and BDPA have been heavily discussed within the interrelated literature in recent years. Ten out of seventeen related articles focused on these SCCs as ways to enhance resilience and sustainability in the SC (Table 2). The capability of transparency is considered to be a main driver of effective RM and resilience in the SC [10]. Having a complete view of all value creation steps in the SC can enhance proactive risk detection, reduce losses and increase business performance [16,82]. Companies that can ensure full and real-time transparency concerning their demand, inventory and delivery statuses across the SC can minimise uncertainty [39]. This enables one or more members of the SC to respond to changes swiftly[10]. The European Union and the United Nations have defined transparency in the SC as one of the key SCCs needed to improve sustainability and responsible management of global SCs [26,82]. Shibin et al. [81] identified that effective informationsharing and connectivity have a positive impact on improving SC visibility, which in turn improves resilience and SCP. One term that is often mentioned, especially in connection with SSCM, is traceability. ISO 9000 defines traceability as the “ability to trace the history, application or location of a product in relation to the origin of its materials and parts; the processing history; and the distribution and location after delivery” [26]. So, we can see that the duty to ensure the reliability of sustainability claims, such as occupational health and safety, human rights, ecological responsibility and anti-corruption policies, plays a vital role in SSCM [26]. Garcia-Torres et al. [26] illustrated that the potential to create transparency has increased in recent years through application of new enabling technologies. These technologies facilitate end-to-end disclosure and provide detailed information about a company’s operations, sources and impact. Traceability systems also simplify real-time record keeping and information-tracking processes throughout the entire SC. A study by Jeble et al. [39] identified BDPA capability as having a beneficial effect on environmental, social and economic sustainability performance. They added that well-defined business processes, databases and state-of-the-art information systems (including BDPA) can help firms to create
72
P. Warmbier and A. Kinra
transparency and reduce complexity [39]. The findings of Singh and Singh [83] highlighted that BDPA and IT capabilities can enable organisations to identify SCR in internal and external environments, assess the impact of SCR, develop risk management regulations and to monitor and review the effectiveness of risk mitigation strategies within the SC. In addition, the results of Mani et al. [61] show that companies can predict and mitigate various social problems, such as workforce safety, tracking of resource consumption and bribery, by using BDPA. Table 2. Connected articles within SSCM and SCRES citations networks Connected articles
Connecting element
Capabilities in ref. text segm.
Durach and Wiengarten [21]/R Busse et al. [11]/S
Referenced articles
SCC; SCMP
Governance and culture
Kilubi and Rogers [45]/R
Huq et al. [35]/S
SCC; SCR; SCP
–
Salam and Khan [73]/R
Huq et al. [35]/S
SCC; SCMP
Supplier relationship management
Scholten et al. [76]/R
Oelze et al. [67]/S
SCC; SCMP; SCP
Knowledge, Learning and Talent management
Hong et al. [34]/S
Lee and Rha [53]/R
SCC; SCR; SCP
Flexibility
Singh and Singh [83]/R
Singh and El-Kassar [84], SCC; SCR; SCP Ajamieh et al. [1], Benitez-Amado et al. [6]/S
BDPA; Information technology and management
Singh and El-Kassar [84]/S
Rajesh [70]/R
SCC; SCMP; SCP
BDPA
Mani et al. [61]/S
Sheffi and Rice Jr. [80]/R
SCC; SCR; SCP
BDPA
Mandal [60]/R
Hazen et al. [31]/S
SCC; SCR; SCP
BDPA
Shokouhyar et al. [82]/S
Brandon-Jones et al. [10]/R
SCC; SCR
BDPA; Transparency
Jeble et al. [39]/S
Brandon-Jones et al. [10]/R
SCC; SCR; SCP
BDPA; Transparency
Shibin et al. [81]/S
Brandon-Jones et al. [10]/R
SCC; SCMP; SCR; SCP Transparency; Information sharing; SC integration
Gupta et al. [30]/S
Rajesh [70]/R
SCC; SCP
BDPA
Gruchmann et al. [29]/S
Scholten and Schilder [74], Kwak et al. [51]/R
SCC; SCMP
Innovation; Collaboration; Information sharing
Golgeci and Ponomarov [27]/R Defee and Fugate [19]/S
SCC; SCP
Innovation
Garcia-Torres et al. [26]/S
Brandon-Jones et al. [10]/R
SCC; SCR; SCMP; SCP Transparency
Dubey et al. [20]/S
Brandon-Jones et al. [10]/R; Zhu et al. [92]/S and R
SCC; SCR; SCP
Transparency; Information sharing; BDPA
Note: R = SCRES literature; S = SSCM literature
4.3
Main Path of SCRES and SSCM Based on Capability Research
In summary, the four connecting elements that we identified in the interrelated literature were verified by the extracted main paths. This can be seen as a verification of results (Table 3). We should emphasise that social and ecological SCRs have only been added to SSCM within the last few years. However, LPHC risk events seem not to have been considered by SSCM researchers until now. The SCRES main path demonstrates that recent researchers see building resilience as a sustainable competitive advantage for SC systems as it prepares them to respond to unexpected events. Furthermore, a number of additional congruent SCCs, such as collaboration, SC reengineering and knowledge management, exist in both fields. However, several divergent SCC are also present.
Supply Chain Resilience and Sustainability
5
73
Discussion
The exploratory study identified four connecting elements within SSCM and SCRES literature - SCP, SCR, SCC and SCMP. Different SCMPs are needed to develop specific SCCs. In turn, a set of SCCs is required to create a sustainable and resilient SC system that can withstand disruptive events in all three SCR clusters - operational/financial, social and ecological risks. All this can have a positive effect on the three clusters of SCP and can ensure a long-term competitive advantage. We also identified two main congruent SCCs in the correlated SCRES and SSCM literature - transparency and BDPA. These two SCCs provide great potential for responding to LPHC events and successfully managing SCs through a crisis. Recent publications have discussed the trend towards a synthesis of SSCM and SCRES, which can be seen as confirmed by the findings of this study [24,28,37,38,77,91]. However, modern SC strategies, such as LARG, ecosilient and leagility strategies, do not cover both research domains equally. The 4 elements identified in this study are also confirmed by the findings of Zavala-Alcivar et al. [91], although the SCRES framework they developed only links to SSCM in a single way. Their research did not focus on investigating parallels between the two fields. The 4 elements identified in our study demonstrate that we examined both SSCM and SCRES comprehensively and equally. Cross-domain literature started to appear within the citation network from 2017 onwards. However, we should consider the drivers and barriers that resulted in 17 papers being cited in the alternative research field. On the one hand, social and ecological SCRs within SSCM have been investigated more and more often over recent years. This is the reason why the SSCM papers referenced SCRES research to deal with uncertainties within SCM. On the other hand, SCRES research was seen as a vital concept for lowering the impact and duration of disruption after a LPHC event and achieving a sustainable competitive advantage [66]. However, LPHC is still not yet firmly established within SSCM research as an SCR characteristic, which seems to be one reason for the lower number of correlated papers. Nevertheless, we can expect this number to increase, especially after the post-Covid phase. In addition, BDPA has appeared as an SCC in SCM literature since the middle of the last decade and is now considered to be an important SCC for improving transparency and supporting decision-making processes. It is seen as an ability that has the potential to prevent risks [50]. The interrelated papers focussed heavily on the topic of this SCC because it enables supply chains to improve SCRES and SSCM. A second topic that must be discussed are the challenges of integrating SSCM and SCRES. Both fields share complementary characteristics but merging the two can also lead to conflicts. Ivanov [37] pointed out the discrepancies between the two fields. He showed that various SCMPs, which can be coupled with SCCs, can have a negative effect on the other research domain. If we take the example of protective redundancy as a SCRES SCC with backup suppliers and facilities as the SCMP, it actually has a negative effect on resource efficiency and associated financial consequences [37]. Conflicts of interests and the disparate SCP goals of SSCM (outcome oriented) and resilience (process oriented) can also cause friction [5]. Furthermore, the
74
P. Warmbier and A. Kinra
simultaneous demand for dynamic capabilities that are able to be continuously adapted and stable routines with static operational capabilities underlines the complexity of problem [85]. This phenomenon has previously been described as a challenge for integration processes by the theory of paradox. In order to overcome opposing paradoxical conflicts, Smith and Lewis [85] suggest applying an iterative response process of splitting and integrating elements. This results in a composite product that will be confronted with new contradictory demands over time [85]. The recognition of these conflicts can be seen as a new insight into the process of integration. This is why we believe a comprehensive analysis of congruency and divergence (e.g. for SCCs) would be beneficial. It would enable us to form a clear picture of how a SC can be developed in direction of sustainability and resilience. Table 3. Analysis of main path articles within SSCM and SCRES MP articles SCRES
Element
SCC
MP articles SSCM
Element
SCC
Chopra and Sodhi [14]
SCC; SCR; SCP
Inventory and demand management; redundancy and capacity; agility/responsiveness; flexibility
Defee and Fugate [19]
SCC; SCP
Collaboration; knowledge management; innovation; SC integration; transparency; information sharing; efficiency; standardisation; IT management; effectivity
Sheffi and Rice Jr. [80]
SCC; SCR; SCP; SCMP
Redundancy; flexibility
Beske [7]
SCC; SCR; SCP
Knowledge management and learning; innovation; supplier relationship management; SC reengineering; risk management; continuity; reflexive control; co-evolving
Ponomarov and Holcomb [69]
SCC; SCR; SCP
Efficiency; risk management; redundancy; effectivity; IT management; standardisation; flexibility; agility; cost and benefit sharing; information sharing; knowledge management; social capital; logistic SCC
Beske et al. [8]
SCC; SCR; SCP
Effectivity; knowledge management and learning; innovation; supplier relationship management; SC reengineering; risk management; continuity; reflexive control; co-evolving
Juettner and Maklan [43]
SCC; SCR; SCP
Flexibility; velocity; transparency; collaboration
Touboulic and Walker [87]
SCC; SCR; SCP
Collaboration; information sharing; governance and culture; trust; cost and benefit sharing
Johnson et al. [41]
SCC; SCR; SCP; SCMP
Transparency; velocity; social capital
Wu et al. [90]
SCC; SCR; SCP; SCMP
Learning; innovation; information sharing; SC reengineering; sensing and seizing; customer orientation; acquisition and exit capabilities
Scholten and Schilder [74]
SCC; SCR; SCP; SCMP
Flexibility; collaboration; transparency; velocity
Akhavan and Beckmann [2]
SCC; SCR; SCP; SCMP
Collaboration; SC integration; supplier relationship management; governance and culture
Kamalahmadi and Parast [44]
SCC; SCR; SCP
Agility, collaboration; Mathivathanan transparency; redundancy and et al. [63] capacity; risk management; SC reengineering and design; efficiency; information sharing; velocity; innovation; trust; leadership
SCC; SCP; SCMP
Collaboration; knowledge management; innovation; SC integration; transparency; supplier relationship management; SC reengineering; efficiency; effectivity; IT management; information sharing; agility; governance and culture; flexibility; technology management; customer orientation; sensing and seizing; acquisition and exit
Chowdhury and Quaddus [15]
SCC; SCR; SCP; SCMP
Agility; flexibility; collaboration; transparency; redundancy and capacity; risk management; knowledge/learning and talent management; recovery
Liu et al. [57]
SCC; SCP; SCMP
Knowledge management; supplier relationship management; supply chain integration; product design; information technology management; flexibility
Hong et al. [34]
SCC; SCR; SCP
Knowledge management; innovation; SC reengineering; effectivity; social capital; customer orientation; trust
Supply Chain Resilience and Sustainability
6
75
Conclusion
Our analysis disclosed 4 potential elements - SCP, SCR, SCC and SCMP - all of which need to be examined if we are to integrate the two research streams. In addition, two main congruent capabilities - transparency and BDPA - deliver insights into how a sustainable resilient SC can be developed. However, the paradox theory indicates that congruent SCCs and the contrasting tensions that arise during integration must be comprehensively analysed in future research. The results offer a clear contribution to both theory and practice. The four connecting elements provide a productive basis for analysing the integration of SSCM and SCRES regarding complementary and contradictory behaviour between the opposing poles. Furthermore, we suggest that congruent SCCs research will be hugely important for overcoming conflicts. For business leaders a potential area was demonstrated in which their SC system can be transformed in order to create a long term competitive advantage. As a potential action field the development of transparency is needed with application of new technology and BDPA skills. However, future research should invest into a comprehensive analysis of resilient-sustainable SCCs and develop an assessment and decision-making model based hereon to support business leaders. This study was comprehensive but also had some limitations. As pointed out, we identified two main congruent capabilities. However, the literature analysis produced further congruent and potentially divergent SCCs that were not identified in the 17 correlating publications. This was because our methodology focussed only on interrelated papers and text segments found by the CNA. Furthermore, the research focus of the selected papers was limited to a restricted number of SCCs within SSCM or SCRES. We also used only one database when collecting citation data due to the different citation data logic used in the WoS and Scopus databases. For this reason, we were not able to process some papers that were included in the initial SLR dataset as their data did not exist in the WoS database. These limitations could be avoided by analysing the whole SLR data set using content analysis, which we highly recommend for future research. In summary, this study suggests several directions for further research. SC sustainability and resilience share common elements in which further investigation of contradictory and complementary behaviour between both poles as well as resolution strategies would be needed based on the paradox theory. In addition, a comprehensive analysis of congruent resilient-sustainable SCCs would be helpful to partially overcome tensions within sustainability and resilience. The authors of this study hope the findings provide a comprehensive basis for future studies in the area of supply chain sustainability and resilience.
References 1. Ajamieh, A., Benitez, J., Braojos, J., Gelhard, C.: It infrastructure and competitive aggressiveness in explaining and predicting performance. J. Bus. Res. 69(10), 4667– 4674 (2016)
76
P. Warmbier and A. Kinra
2. Akhavan, R.M., Beckmann, M.: A configuration of sustainable sourcing and supply management strategies. J. Purch. Supply Manag. 23(2), 137–151 (2017) 3. Ali, A., Mahfouz, A., Arisha, A.: Analysing supply chain resilience: integrating the constructs in a concept mapping framework via a systematic literature review. Supply Chain Manage. Int. J. 22(1), 16–39 (2017) 4. Azevedo, S.G., Govindan, K., Carvalho, H., Cruz-Machado, V.: Ecosilient index to assess the greenness and resilience of the upstream automotive supply chain. J. Clean. Prod. 56, 131–146 (2013) 5. Bellamy, L., Choudhary, A., Papaioannou, G., Zirar, A.: Dangling between sustainability and resilience supply chain practices: employing paradox theory to explore tensions. Loughborough University. Conference contribution. https://hdl.handle. net/2134/37993 (2019) 6. Benitez-Amado, J., Javier Llorens-Montes, F., Fernandez-Perez, V.: It impact on talent management and operational environmental sustainability. Inf. Technol. Manage. 16(3), 207–220 (2015) 7. Beske, P.: Dynamic capabilities and sustainable supply chain management. Int. J. Phys. Distrib. Logist. Manag. 42(4), 372–387 (2012) 8. Beske, P., Land, A., Seuring, S.: Sustainable supply chain management practices and dynamic capabilities in the food industry: a critical analysis of the literature. Int. J. Prod. Econ. 152, 131–143 (2014) 9. Bode, C., Wagner, S.M.: Structural drivers of upstream supply chain complexity and the frequency of supply chain disruptions. J. Oper. Manag. 36(1), 215–228 (2015) 10. Brandon-Jones, E., Squire, B., Autry, C.W., Petersen, K.J.: A contingent resourcebased perspective of supply chain resilience and robustness. J. Supply Chain Manag. 50(3), 55–73 (2014) 11. Busse, C., Schleper, M.C., Niu, M., Wagner, S.M.: Supplier development for sustainability: contextual barriers in global supply chains. Int. J. Phys. Distrib. Logist. Manag. 46(5), 442–468 (2016) 12. Cabral, I., Grilo, A., Cruz-Machado, V.: A decision-making model for lean, agile, resilient and green supply chain management. Int. J. Prod. Res. 50(17), 4830–4845 (2012) 13. Carter, C.R., Easton, L.P.: Sustainable supply chain management: evolution and future directions. Int. J. Phys. Distrib. Logist. Manag. 41(1), 46–62 (2011) 14. Chopra, S., Sodhi, M.S.: Managing risk to avoid supply-chain breakdown. MIT Sloan Manage. Rev. 46(1), 53 (2004) 15. Chowdhury, M.M.H., Quaddus, M.: Supply chain readiness, response and recovery for resilience. Supply Chain Manage. Int. J. 21(6), 709–731 (2016) 16. Christopher, M., Peck, H.: Building the resilient supply chain. Int. J. Logist. Manage. 15(2), 1–14 (2004) 17. Closs, D.J., Speier, C., Meacham, N.: Sustainability to support end-to-end value chains: the role of supply chain management. J. Acad. Mark. Sci. 39(1), 101–116 (2011) 18. Colicchia, C., Strozzi, F.: Supply chain risk management: a new methodology for a systematic literature review. Surg. Endosc. Other Interv. Tech. 17(4), 403–418 (2012) 19. Defee, C.C., Fugate, B.S.: Changing perspective of capabilities in the dynamic supply chain era. Int. J. Logist. Manag. 21(2), 180–206 (2010) 20. Dubey, R.: Examining the role of big data and predictive analytics on collaborative performance in context to sustainable consumption and production behaviour. J. Clean. Prod. 196, 1508–1521 (2018)
Supply Chain Resilience and Sustainability
77
21. Durach, C.F., Wiengarten, F.: Exploring the impact of geographical traits on the occurrence of supply chain failures. Supply Chain Manage. Int. J. 22(2), 160–171 (2017) 22. Emmett, S., Sood, V.: Green Supply Chains: An Action Manifesto. Wiley, Chichester (2010) 23. European Parliament: Meps: Hold companies accountable for harm caused to people and planet (2021), https://www.europarl.europa.eu/news/en/pressroom/20210122IPR96215/meps-hold-companies-accountable-for-harm-caused-topeople-and-planet 24. Fahimnia, B., Jabbarzadeh, A.: Marrying supply chain sustainability and resilience: a match made in heaven. Transp. Res. Part E Logist. Transp. Rev. 91, 306–324 (2016) 25. Forbes: Supply chain sustainability sparks resilience, recovery and growth (2020), https://www.forbes.com/sites/businessreporter/2020/05/26/supplychain-sustainability-sparks-resilience-recovery-and-growth/?sh=197360a26e6e 26. Garcia-Torres, S., Albareda, L., Rey-Garcia, M., Seuring, S.: Traceability for sustainability - literature review and conceptual framework. Supply Chain Manage. Int. J. 24(1), 85–106 (2019) 27. Golgeci, I., Ponomarov, S.Y.: Does firm innovativeness enable effective responses to supply chain disruptions? an empirical study. Supply Chain Manage. Int. J. 18(6), 604–617 (2013) 28. Golicic, S.L., Flint, D.J., Signori, P.: Building business sustainability through resilience in the wine industry. Int. J. Wine Bus. Res. 29(1), 74–97 (2017) 29. Gruchmann, T., Seuring, S., Petljak, K.: Assessing the role of dynamic capabilities in local food distribution: a theory-elaboration study. Supply Chain Manage. Int. J. 24(6), 767–783 (2019) 30. Gupta, S., Chen, H., Hazen, B.T., Kaur, S., Santibanez Gonzalez, Ernesto D. R.: Circular economy and big data analytics: A stakeholder perspective. Technological Forecasting and Social Change 144, 466–474 (2019) 31. Hazen, B.T., Skipper, J.B., Ezell, J.D., Boone, C.A.: Big data and predictive analytics for supply chain sustainability: a theory-driven research agenda. Comput. Ind. Eng. 101, 592–598 (2016) 32. Hofmann, H., Busse, C., Bode, C., Henke, M.: Sustainability-related supply chain risks: conceptualization and management. Bus. Strateg. Environ. 23(3), 160–172 (2014) 33. Hohenstein, N.O., Feisel, E., Hartmann, E., Giunipero, L.: Research on the phenomenon of supply chain resilience a systematic review and paths for further investigation. Int. J. Phys. Distrib. Logist. Manage. 45(1–2), 90–117 (2015) 34. Hong, J., Zhang, Y., Ding, M.: Sustainable supply chain management practices, supply chain dynamic capabilities, and enterprise performance. J. Clean. Prod. 172, 3508–3519 (2018) 35. Huq, F.A., Chowdhury, I.N., Klassen, R.D.: Social management capabilities of multinational buying firms and their emerging market suppliers: an exploratory study of the clothing industry. J. Oper. Manag. 46, 19–37 (2016) 36. ILO: Managing work-related psychosocial risks during the covid-19 pandemic (2020), https://www.ilo.org/wcmsp5/groups/public/---ed protect/---protrav/--safework/documents/instructionalmaterial/wcms 748638.pdf 37. Ivanov, D.: Revealing interfaces of supply chain resilience and sustainability: a simulation study. Int. J. Prod. Res. 56(10), 3507–3523 (2017)
78
P. Warmbier and A. Kinra
38. Ivanov, D.: Viable supply chain model: integrating agility, resilience and sustainability perspectives-lessons from and thinking beyond the covid-19 pandemic. Annals of operations research pp. 1–21 (2020) 39. Jeble, S., Dubey, R., Childe, S.J., Papadopoulos, T., Roubaud, D., Prakash, A.: Impact of big data and predictive analytics capability on supply chain sustainability. Int. J. Logist. Manag. 29(2), 513–538 (2018) 40. Jiang, X., Zhuge, H.: Forward search path count as an alternative indirect citation impact indicator. Journal of Informetrics 13(4), 100977 (2019) 41. Johnson, N., Elliott, D., Drake, P.: Exploring the role of social capital in facilitating supply chain resilience. Supply Chain Management-An International Journal 18(3), 324–336 (2013) 42. Ju, K.J., Park, B., Kim, T.: Causal relationship between supply chain dynamic capabilities, technological innovation, and operational performance. Management and Production Engineering Review 7(4), 6–15 (2016) 43. Juettner, U., Maklan, S.: Supply chain resilience in the global financial crisis: an empirical study. Supply Chain Management-An International Journal 16(4), 246– 259 (2011) 44. Kamalahmadi, M., Parast, M.M.: A review of the literature on the principles of enterprise and supply chain resilience: Major findings and directions for future research. International Journal of Production Economics 171, 116–133 (2016) 45. Kilubi, I., Rogers, H.: Bridging the gap between supply chain risk management and strategic technology partnering capabilities: insights from social capital theory. Supply Chain Management-An International Journal 23(4), 278–292 (2018) 46. Kinra, A., Hald, K.S., Mukkamala, R.R., Vatrapu, R.: An unstructured big data approach for country logistics performance assessment in global supply chains. International Journal of Operations & Production Management 40(4), 439–458 (2020) 47. Kinra, A., Ivanov, D., Das, A., Dolgui, A.: Ripple effect quantification by supplier risk exposure assessment. Int. J. Prod. Res. 58(18), 5559–5578 (2020) 48. Klein, C.: Companies must focus on resiliency, profitability and sustainability (2021), https://www.weforum.org/agenda/2021/01/companies-must-focus-onresiliency-profitability-and-sustainability/ 49. Klibi, W., Martel, A., Guitouni, A.: The design of robust value-creating supply chain networks: A critical review. European Journal of Operational Research 203(2), 283–293 (2010) 50. Kumari, S., Patil, Y., Jeble, S.: Role of big data and predictive analytics. International Journal of Automation and Logistics 2(4), 307 (2016) 51. Kwak, D.W., Seo, Y.J., Mason, R.: Investigating the relationship between supply chain innovation, risk management capabilities and competitive advantage in global supply chains. International Journal of Operations & Production Management 38(1), 2–21 (2018) 52. Lathabai, H.H., George, S., Prabhakaran, T., Changat, M.: An integrated approach to path analysis for weighted citation networks. Scientometrics 117(3), 1871–1904 (2018). https://doi.org/10.1007/s11192-018-2917-1 53. Lee, S.M., Rha, J.S.: Ambidextrous supply chain as a dynamic capability: building a resilient supply chain. Manag. Decis. 54(1), 2–23 (2016) 54. Leydesdorff, L., Thor, A., Bornmann, L.: Further steps in integrating the platforms of wos and scopus: Historiography with histciteTM and main-path analysis. El Profesional de la Informaci´ on 26(4), 662 (2017) 55. Liu, J.S., Lu, L.Y.Y., Ho, M.H.-C.: A few notes on main path analysis. Scientometrics 119(1), 379–391 (2019). https://doi.org/10.1007/s11192-019-03034-x
Supply Chain Resilience and Sustainability
79
56. Liu, J.S., Lu, L.Y.: An integrated approach for main path analysis: Development of the hirsch index as an example. Journal of the American Society for Information Science and Technology 63(3), 528–542 (2012) 57. Liu, Y., Zhu, Q., Seuring, S.: Linking capabilities to green operations strategies: The moderating role of corporate environmental proactivity. International Journal of Production Economics 187, 182–195 (2017) 58. Lu, Y., Zhao, C., Xu, L., Shen, L.: Dual institutional pressures, sustainable supply chain practice and performance outcome. Sustainability 10(9), 3247 (2018) 59. Luo, B.N., Yu, K.: Fits and misfits of supply chain flexibility to environmental uncertainty two types of asymmetric effects on performance. Int. J. Logist. Manag. 27(3), 862–885 (2016) 60. Mandal, S.: The influence of big data analytics management capabilities on supply chain preparedness, alertness and agility an empirical investigation. Information Technology & People 32(2), 297–318 (2019) 61. Mani, V., Delgado, C., Hazen, B.T., Patel, P.: Mitigating supply chain risk via sustainability using big data analytics: evidence from the manufacturing supply chain. Sustainability 9(4), 608 (2017) 62. Martin, C., Towill, D.R.: Supply chain migration from lean and functional to agile and customised. Supply Chain Management: An International Journal 5(4), 206– 213 (2000) 63. Mathivathanan, D., Govindan, K., Haq, A.N.: Exploring the impact of dynamic capabilities on sustainable supply chain firm’s performance using grey-analytical hierarchy process. J. Clean. Prod. 147, 637–653 (2017) 64. Matt, L.: Tight ocean market, extended lead times push shippers to airfreight (2021), https://www.supplychaindive.com/news/airfreight-volume-cargoplane-demand-rate-atlas-belly-capacity/596361/ 65. Multaharju, S., Lintukangas, K., Hallikas, J., Kahkonen, A.K.: Sustainabilityrelated risk management in buying logistics services an exploratory cross-case analysis. Int. J. Logist. Manag. 28(4), 1351–1367 (2017) 66. Munoz, A., Dunbar, M.: On the quantification of operational supply chain resilience. Int. J. Prod. Res. 53(22), 6736–6751 (2015) 67. Oelze, N., Hoejmose, S.U., Habisch, A., Millington, A.: Sustainable development in supply chain management: The role of organizational learning for policy implementation. Business Strategy and the Environment 25(4), 241–260 (2016) 68. Pettit, T.J., Fiksel, J., Croxton, K.L.: Ensuring supply chain resilience: Development of a conceptual framework. Journal of Business Logistics 31(1), 1–+ (2010) 69. Ponomarov, S.Y., Holcomb, M.C.: Understanding the concept of supply chain resilience. Int. J. Logist. Manag. 20(1), 124–143 (2009) 70. Rajesh, R.: Technological capabilities and supply chain resilience of firms: A relational analysis using total interpretive structural modeling (tism). Technological Forecasting and Social Change 118, 161–169 (2017) 71. Reich, J., Kinra, A., Kotzab, H., Brusset, X.: Strategic global supply chain network design - how decision analysis combining milp and ahp on a pareto front can improve decision-making. International Journal of Production Research 59(5), 1557–1572 (2021) 72. Saenz, H., Stephan, J., Terino, J., Bysong, T., Gnamm, J.: How to trace a path to resilient, sustainable supply chains (2020), https://www.bain.com/insights/ resilient-sustainable-supply-chains/ 73. Salam, M.A., Khan, S.A.: Achieving supply chain excellence through supplier management a case study of fast moving consumer goods. Benchmarking-An International Journal 25(9), 4084–4102 (2018)
80
P. Warmbier and A. Kinra
74. Scholten, K., Schilder, S.: The role of collaboration in supply chain resilience. Supply Chain Management-An International Journal 20(4), 471–484 (2015) 75. Scholten, K., Scott, P.S., Fynes, B.: Mitigation processes - antecedents for building supply chain resilience. Supply Chain Management-An International Journal 19(2), 211–228 (2014) 76. Scholten, K., Scott, P.S., Fynes, B.: Building routines for non-routine events: supply chain resilience learning mechanisms and their antecedents. Supply Chain Management-An International Journal 24(3), 430–442 (2019) 77. Sen, D.K., Datta, S., Mahapatra, S.S.: On evaluation of supply chain’s ecosilient (g-resilient) performance index: a fuzzy embedded decision support framework. Benchmarking 25(7), 2370–2389 (2018) 78. Seuring, S., M¨ uller, M.: From a literature review to a conceptual framework for sustainable supply chain management. J. Clean. Prod. 16(15), 1699–1710 (2008) 79. Shafiq, A., Johnson, P.F., Klassen, R.D., Awaysheh, A.: Exploring the implications of supply risk on sustainability performance. International Journal of Operations & Production Management 37(10), 1386–1407 (2017) 80. Sheffi, Y., Rice Jr., J.B.: A supply chain view of the resilient enterprise. Mit Sloan Management Review 47(1), 41–48+94 (2005) 81. Shibin, K.T., Gunasekaran, A., Dubey, R.: Explaining sustainable supply chain performance using a total interpretive structural modeling approach. Sustain. Prod. Consum. 12, 104–118 (2017) 82. Shokouhyar, S., Seddigh, M.R., Panahifar, F.: Impact of big data analytics capabilities on supply chain sustainability. World Journal of Science, Technology and Sustainable Development 17(1), 33–57 (2019) 83. Singh, N.P., Singh, S.: Building supply chain risk resilience role of big data analytics in supply chain disruption mitigation. Benchmarking-An International Journal 26(7), 2318–2342 (2019) 84. Singh, S.K., El-Kassar, A.N.: Role of big data analytics in developing sustainable capabilities. J. Clean. Prod. 213, 1264–1273 (2019) 85. Smith, W.K., Lewis, M.W.: Toward a theory of paradox: A dynamic equilibrium model of organizing. Academy of Management Review (Vol. 36, No.2), 381–403 (2011) 86. Thom´e, A.M.T., Scavarda, L.F., Scavarda, A.J.: Conducting systematic literature review in operations management. Production Planning & Control 27(5), 408–420 (2016) 87. Touboulic, A., Walker, H.: Theories in sustainable supply chain management: a structured literature review. International Journal of Physical Distribution & Logistics Management 45(1/2), 16–42 (2015) 88. Tukamuhabwa, B.R., Stevenson, M., Busby, J., Zorzini, M.: Supply chain resilience: definition, review and theoretical foundations for further study. Int. J. Prod. Res. 53(18), 5592–5623 (2015) 89. Vilko, J., Ritala, P., Edelmann, J.: On uncertainty in supply chain risk management. The International Journal of Logistics Management 25(1), 3–19 (2014) 90. Wu, K.J., Liao, C.J., Chen, C.C., Lin, Y., Tsai, C.F.M.: Exploring eco-innovation in dynamic organizational capability under incomplete information in the Taiwanese lighting industry. Int. J. Prod. Econ. 181, 419–440 (2016) 91. Zavala-Alc´ıvar, A., Verdecho, M.J., Alfaro-Saiz, J.J.: A conceptual framework to manage resilience and increase sustainability in the supply chain. Sustainability 12(16), 6300 (2020)
Supply Chain Resilience and Sustainability
81
92. Zhu, S., Song, J., Hazen, B.T., Lee, K., Cegielski, C.: How supply chain analytics enables operational supply chain transparency: An organizational information processing theory perspective. International Journal of Physical Distribution & Logistics Management 48(1), 47–68 (2018) ˇ 93. Zupic, I., Cater, T.: Bibliometric methods in management and organization. Organizational Research Methods 18(3), 429–472 (2015)
The Impact of Blockchain on Supply Chain Resilience Anna Kolmykova(B) University of Economics and Management, Linzerstr. 7, 28359 Bremen, Germany [email protected]
Abstract. This paper introduces the results of the study of the impact of blockchain technology on supply chain resilience. The study has been conducted using empirical research methodology. The qualitative research method of expert interviews has been applied to collect the data, and the text analysis method - for the data evaluation. Following capabilities of resilient supply chains have been studied: supply chain engineering, collaboration, agility, risk management culture and knowledge management. The experts in blockchain technology from the fields of logistics, production and consumer industry have been interviewed. The main finding of this study is the positive impact of blockchain technology on agility and collaboration in the supply chain. The implication for the supply chain management theory and practice is the understanding of blockchain technology contribution to enhance the resilience of supply chain and of the related challenges. Keywords: Supply chain resilience · Blockchain · Empirical research
1 Introduction The COVID-19 pandemic has disturbed the ability of most companies to produce and supply their products and disrupted some supply chains for electronic and consumer goods. Supply chain managers have started to reassess their strategies in order to be better prepared to perform during unexpected events such as the COVID-19 pandemic. This emergency situation has encouraged the exploration of new supply chain strategies and reinforced academic und practical discussion on building supply chain resilience against disruptions of any kind. First definitions and theoretical investigations of resilience applied to the supply chain was formulated in the works of well-known scientists in supply chain management like Christopher and Peck (2004), and Sheffi and Rice (2005) long before the coronavirus disaster. The scientific interest in supply chain resilience (SCRES) grew slowly over the last decade from 2–8 direct related journal articles in 2010 to 6–23 papers in 2014 and to 38 in 2016, as reviewed by Elleuch et al. (2016) and Biedermann et al. (2018). These were mostly motivated by the earthquake and tsunami in Japan in 2011 and by the global financial crisis. Since then, the number of articles has increased dramatically. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 M. Freitag et al. (Eds.): LDIC 2022, LNLO, pp. 82–91, 2022. https://doi.org/10.1007/978-3-031-05359-7_7
The Impact of Blockchain on Supply Chain Resilience
83
Most of them were the issues on the conceptualization that identified the properties of resilience. Based on that, further studies (Tukamuhabwa et al. 2015; Singh et al. 2019) were dedicated to theoretical framework development for the future SCRES research. However, all authors emphasize a lack on empirical research of resilience. From 91 direct and indirect related articles that Tukamuhabwa et al. (2015) analyzed in their study, the number of empirical papers – case studies or surveys – is limited to just 19. Considering the blockchain technology (BCT) and its impact on supply chain, there is a number of mostly theoretical investigations of BCT contribution to enhance flexibility and agility in supply chain operations (Gohil and Thakker 2021) and to reduce costs and time through improving information sharing and collaboration capabilities (Nandi et al. 2020). Wang et al. (2018) reviewed emerging practices and identified the BCT contribution to create transparency in global supply chains, to extend visibility and traceability, safety and security. Still, the empirical research in this field is limited by the amount of real use cases. However, regarding the relationship between SCRES and BCT, the research gap can be identified. There are only a few studies on this regard. Min (2019) emphasis the capability of blockchain as a peer-to-peer network using distributed ledgers to mitigate risks associated with intermediaries’ interventions and to enhance security. Dubey et al. (2020) found out that BCT can improve collaboration and thus an ability to get an operational relief in humanitarian supply chain. Frederico (2021) linked the following resilience improvements related to disruptive data, information and knowledge technologies, which include the BCT among others: efficiency, responsiveness, flexibility, reliability, transparency, visibility and traceability. Therefore, this empirical study was designed and conducted to answer the following research question: How does the BTC impact on the resilience of supply chain?
2 Conceptual Framework Supply chain resilience (SCRES) exposes companies’ capability to rebound from supply chain disruptions, and return to their original state after disturbances (Christopher and Peck 2004; Sheffi and Rice 2005). “Resilient supply chains incorporate event readiness, are capable of providing an efficient response, and often are capable of recovering to their original state or to even better state of operational performance after the disruptive event” (Ponomarov and Holcomb 2009). Allied with the system view on a supply chain, it is postulated that inherent system capabilities are decisive for building resilience. Following the resource-based view, supply chains should build capacities to respond to an unexpected event and to possess the power to return to its original state or shift to another even better one. SCRES is described as a capacity that enables proactive and reactive behavior against unanticipated disruptions and as an adaptive capability to recover from them (Ponomarov and Holcomb 2009). Christopher und Peck (2004) defined four capabilities of resilient supply chain: Supply Chain (Re)-Engineering, Supply Chain Collaboration, Agility (including Velocity and Visibility) and Creating a Supply Chain Risk Management Culture. Scholten et al. (2014) considered the fifth capability of Knowledge management additionally. In
84
A. Kolmykova
their literature review Tukamuhabwa et al. (2015) identified the range of most critical capabilities for improving resilience such as increasing flexibility, creating redundancy, forming collaborative supply chain relationships and improving supply chain agility. In a later literature review, Singh et al. (2019) identified and systematized seventeen indicators which could help to make a supply chain more resilient: agility, flexibility, robustness, redundancy, visibility, information sharing, collaboration, sustainability, awareness, risk management culture, velocity, market position, risk control, revenue sharing, adaptability, network design and security. On the other hand, BCT provides real-time data exchange and end-to-end visibility that could enhance transparency, traceability, efficiency (Francisco and Swanson 2018), and information security (Kshetri 2018) in supply chains. At the same time the BCT implementation in a supply chain facilitates supply chain collaboration by supporting communication, information sharing processes and joint knowledge creation (Rejeb et al. 2021). Hence, it is suggested, that the implementation of the BCT has a positive impact on SCRES-capabilities and thus on SCRES.
3 Methodology 3.1 Data Collection Method and Sample Description Being a social-technical complex system phenomena, the SCRES can be studied using qualitive research methods (Tukamuhabwa et al. 2015). A qualitive research is aimed to develop empirically substantial hypotheses for new theories (Glaser and Strauss 2010) such as the theory of SCRES and enables to explore new aspects of relationships between phenomena from the perspective of those experiencing it. That is why the qualitive research method have been chosen to explore the experts’ experience of adopting BCT and analyze it in terms of possible relationships to SCRES. The data collection method is the qualitative, semi-structured, open-ended interview. The process of data collection includes interview design, audio recording of interviews and transcription of the spoken data. The interview questions were generated around the SCRES-capabilities determined in Sect. 2: How does the BCT application contribute to Cooperation between supply chain partners? What is a role of the BCT in the supply chain Knowledge management? Does the BCT make a supply chain more agile and fast (Agility and Velocity), transparent and visible? How could the BCT affect supply chain design (Engineering)? What kind of influence could the BCT have on Supply Chain Risk Management awareness (Culture)? To obtain consistent and quality data for the study the sample size of 10 participants has been determined. Half of the group consists of high-level managers of large logistics and production companies with strong experience in implementing of BC technology and SCM. The second part contains third party developers of the BCT with multiple project experience in food and pharma industry. Finally, a total number of 9 interviews were conducted in the period from May till June 2021.
The Impact of Blockchain on Supply Chain Resilience
85
3.2 Data Analysis Method For data analysis the qualitative content analysis approach has been implemented. According to Flick, von Kardoff and Steinke, the aim of qualitative research is to describe living environments based on the subjective view of the people involved in order to get a better understanding of the interpretation patterns, structural features and processes (Flick et al. 2015). It is therefore primarily based on an understanding-interpretative reconstruction in the respective context, whereby the research design and thus the data evaluation method is based on non-numerical, mostly linguistically coded data material (Döring and Bortz 2016). Content analysis is therefore a scientific method to qualitative analysis of text data from primary and secondary sources (Mayring 2015). The qualitative content analysis enables a systematic extraction of relevant information from the interviews and is at the same time open to unexpected findings. The method is especially suitable to analyze the transcripts of multiple open-ended interviews. The exploration nature of the method is mostly suitable for specific questions for which no valid data is available for the theory building. Furthermore, the qualitative methodology is suitable for depicting opinions and knowledge in their entire complexity and enables very detailed statements to be obtained on the basis of small samples (Brosius et al. 2012). This research brings the BCT into the context of SCRES for the reasoning about their possible relationship based on practical experiences, opinions and knowledge of the selected respondents (see sample description in Sect. 3.1). Keeping the rule-based procedure is the main point of the methodology, however, it should be noted that the content analysis does not represent a standard instrument but rather addresses the specific question (Mayring 2015). The process of deductive procedure applicated in this research includes theoretical based formulation of main and sub categories, formulation of coding rules for the categories, coding, assignment of the category to a passage of text, subsumption of frequencies of coded categories, and interpretation of results (Mayring 2015). The system of categories is the central tool for the content analysis (Mayring and Fenzl 2019). The flexible category system consists of five, deductively developed, categories. The categories and sub-categories have been created in regard with dimensions of SCRES-capabilities: Collaboration (Information sharing, Transparency,), Agility (Visibility, Velocity), Supply Chain Reengineering (Efficiency, Flexibility), Knowledge management and SCRM Culture. The categories can be clearly defined with the help of anchor examples or the coding rules (Mayring 2015) (see Table 1). According to this, a statement from interviews is assigned to a certain category due to its context. The coding process is carried out with the help of the Qualitative Data Analysis (QDA) software MAXQDA.
4 Findings The following table (Table 1) summarizes the results with category rules, coding, and few selected anchor examples. The names of interviewees are codified.
86
A. Kolmykova Table 1. Categories and codes.
Category
Codes
A. Collaboration
“So that is the greatest strength of blockchain technology, … it makes a lot of sense if you operate it together with the other members of the supply chain …, so to speak, one also grows together …” (MB)
9/9
Trust
“And the blockchain technology enables you to couple complete trust and transparency with each other. That means, I determine what information I deliver and to whom I deliver it.” (AH)
3/9
Digital processes
“Also, as I told you the immutability of data, and how no one can change the data once saved. Once the transaction took place, no one can change it and this also improves the collaboration.” (DK)
4/9
“I think I’ve already said, that people can communicate with each other more easily. The more often they communicate with each other, the more valuable the information becomes and therefore the overall costs decrease.” (AH)
1/9
No intermediaries
Data immutability
Simplified communication Information quality
A2. Transparency
Total
Supply chain collaboration is ability to work effectively with other supply chain entities for mutual benefit (Christopher and Peck 2004; Ponomarov and Holcomb 2009) Increasing collaboration
A1. Information sharing
Rules and examples
Selective transparency Traceability
B. Agility
1/9
1/9
1/9
“Here one could say, however, that the 7/9 system only verifies him when he has 3/9 completely made his upload with all the information and the system simply says: he has done it. So, I know that what he said is true without telling me what exactly he did. I also find that very charming. So, this is selective transparency.” (RH) Supply chain agility is the ability to respond quickly to unpredictable changes in demand and/or supply (Christopher and Peck 2004)
Increase is possible
“This as an expression of the agility to 3/9 want to react to changes and to be able to react in a shorter time, because then ultimately I have a more dynamic system, which gives me enough standard so that I don’t somehow lose my level of quality.” (AH)
(continued)
The Impact of Blockchain on Supply Chain Resilience
87
Table 1. (continued) Category
Codes
Rules and examples
Total
No increase
B1. Visibility
Increased through collaboration
1/9
Information about supply chain
“Visibility yes, … a technology that is 4/9 used to make the supply chains more 1/9 visible and transparent. Yes, even our customer had no idea at the beginning 1/9 who the intermediate stages are, often they have an idea where the raw material might come from, e.g. Tanzania, but they don’t know how it gets from there: which steps, who is doing what. Blockchain is supportive tool here.” (CH)
BC is supportive tool Trust
B2. Velocity
Faster access on data Faster reaction on changes
Automation
C. Supply chain reengineering
No great use by unforeseen events Digitalization of documentation Less participants/intermediaries Connection of all participants (Smart Contract) C2. Flexibility
“The quality data are immediately available to all participants who are authorized to see this data, that is, if they think of a problem that arises everywhere, you can probably recognize the problem earlier, but you can order the project directly, so I have not to look now, where the problem arose.” (SG)
2/9
“…blockchain will help you with the 2/9 speed you need, because everything is automated. This is how speed is achieved.” (DK)
“…Blockchain cannot redesign the physical supply chain…” (AH)
1/9 1/9
“Peer-to-peer connection, will make 4/9 the blockchain more flexible and 1/9 efficient, because the less people are involved, the more efficient the supply 1/9 chain will be.” (DK)
More information for decision making “Agility and flexibility – yes, … you achieve maybe the better ability for decision making because of the more information you received.” (FT) Less participants No increase in the physical supply chain
D. Knowledge management
1/9
The supply chain network design, that means constructing the supply chain network for resilience by balancing redundancy and efficiency (Tukamuhabwa et al. 2015) No great benefit
C1. Efficiency
3/9
2/9 1/9 1/9
Developing knowledge and understanding of supply chain structure, and the ability to learn from changes as well as educate other entities (Christopher and Peck 2004; Ponomarov and Holcomb 2009)
(continued)
88
A. Kolmykova Table 1. (continued)
Category
Codes
Rules and examples
Total
Data collection
“So, I would now, if I refer to our case, I would say the blockchain solution helps gathering knowledge and also helps to analyze this knowledge. But that is now not “unique” with the blockchain.” (CH)
9/9
Data analysis Understanding of supply chain Learning E. SCRM Culture
5/9 1/9 1/9
Ensuring that all organizational members embrace supply chain risk management (Christopher and Peck 2004; Sheffi and Rice 2005) Anticipation of risks
5/9
Traceability
“So, for example you might have risk of the timing, you need to deliver things on time. This is a huge risk, Automation eliminates some risks because for some companies this is BC only as a support, not as a solution very crucial. How can blockchain change this? Blockchain speeds the time of the transactions and the automation allows things to happen on time and allows you to decide when exactly do you want your smart contract to take place.” (DK)
5/9 1/9 1/9
The findings of the content analysis result from an inductive reasoning process via coding. The total frequency of codes indicates the significance of the finding. The interviewees consider the integration of the BCT having no benefit for supply chain reengineering especially in a case of non-foreseeing event. They argue that smart contracts could not map unpredictable scenarios. However, supply chains become leaner and thus more efficient. The notable findings are the codes Trust and Less Intermediaries that contribute to significant increase of efficiency and collaboration in supply chains. The BCT enables tracking process, qualitative data and selective transparency that support a visibility in supply chains - one of the most important capabilities for resilience. The BCT increases velocity and enables a rapid reaction on and treatment of a failure in the supply chain. Concerning the agility, three from nine interviewees express the view that agility could not be increased by BCT. All of the interviewed experts emphasize the role of the BCT in knowledge management, mostly in data collection and analysis.
5 Discussion All participants of this study have the same opinion about the collaboration (category A), especially about a “direct collaboration without intermediaries”, as the most important contribution of the BCT for SCRES. Half of the participants argues that information sharing (category A1) is related to digital processes and enables trust. “Transparency” is associated with the “trust” and both are results of information sharing that is stimulated by the BCT. However, the experts emphasize the so called “selective transparency” (category A2). Collaboration is found to be the greatest achievement of the BCT implementation. The positive effect is evident for all interviewed experts. The Agility (category
The Impact of Blockchain on Supply Chain Resilience
89
B) is strongly associated with faster reaction and faster data access by participants who found a positive impact of the BCT on supply chain agility. The ability to identify the problem and its origin earlier is the contribution of the BCT in the category velocity (category B1). For the capability Supply Chain Reengineering (category C) no direct impact could be found. The “Less participants and intermediaries” (category C1) means to make a supply chain lean and thus both efficient and flexible. The flexibility is associated with improved information and accordantly improved decision making and thus it contributes to efficiency. Digitalization of documentation is seen as most significant contribution for both efficiency and flexibility. Automation with Smart Contracts provides the supply chain with a clear structure. Such automation along a supply chain (category C) contributes to both efficiency and flexibility. However, according to one interviewee, in a case of an unforeseen event a smart contract is useless, if a smart contract has not provided any solution for this possible case before. The lack of learning aspects becomes obvious in Knowledge Management (category D), when the BCT contributes mainly to data collection and data analysis. The learning process could be activated through enhanced communication. No direct impact on the SCRM Culture (category E) could be found. Enhancing the capability Velocity and automation of supply chain with the BCT improve the ability to anticipate risks on time. However, Soni et al. (2014) emphasizes the SCRM-culture as one of the most important drivers of SCRES among other capabilities, such as agility, visibility and collaboration. Summarized: The findings show positive effects of the BCT on SCRES-capabilities Collaboration and partly Agility, including Information sharing, Transparency, Visibility and Velocity. The positive impact on Supply Chain Reengineering through the improving efficiency and flexibility could be identified as well. The BCT contributes to knowledge development in a supply chain in terms of data collection and analysis. Less effect is found in the category SCRM-culture. There is no negative impact on SCRES detected at all. Suggesting, that the implementation of the BCT has a positive impact on SCREScapabilities and thus on SCRES, the research question can be answered as following: The BTC positively affects the resilience of supply chain, supporting collaboration, efficiency, and flexible supply chain design, at the same time making a supply chain more agile, fast, transparent and visible.
6 Conclusion The empirical study could provide evidence for positive impact of the BCT on SCRES. It focused attention on the role of the BCT in a supply chain for the strength of resilience - the most important task of supply chain management during and after the COVID-19 pandemic. Apparently, the integration of the BCT does not provide an ultimate solution for many aspects of resilience and does not have an impact on all capabilities of SCRES. The explorative approach that has been chosen enabled deeper insights in individual experience with the BCT in supply chain. The qualitative text analysis has been implemented to generate the answer for the research question. The research is obviously limited by the number of the interviewees. The scope of the conceptual framework and the number of samples disposed at Table 1 is limited by the space of this paper.
90
A. Kolmykova
The study contributes to theory building for SCRES in terms of its capabilities and their relationship with a disruptive technology such as BCT. The implication for the supply chain management practice is, first of all, the understanding of BCT contribution to enhance the resilience of supply chain and of the related challenges. Regarding the future research, there is a need of further analysis of the BCT impact on the cost efficiency as a SCRES-capability, especially in terms of transaction costs. The future research will be dedicated to this topic. The same concerns the necessity for technological deep insights on the BCT and its detailed implementation to enhance SCRES.
References Biedermann, L., Kotzab, H., Pettit, T.J.: Theory landscape and research perspectives in current supply chain resilience research. In: Freitag, M., Kotzab, H., Pannek, J. (eds.) LDIC 2018. LNL, pp. 26–33. Springer, Cham (2018). https://doi.org/10.1007/978-3-319-74225-0_4 Brosius, H.-B., Haas, A., Koschel, F.: Methoden der empirischen Kommunikationsforschung. Eine Einführung, 6th edn. Springer, Wiesbaden (2012). https://doi.org/10.1007/978-3-531-90762-8 Christopher, M., Peck, H.: Building the resilient supply chain. IJLM 15, 1–13 (2004). https://doi. org/10.1108/09574090410700275 Döring, N., Bortz, J.: Forschungsmethoden und Evaluation in den Sozial-und Humanwissenschaften, 5th edn. Springer, Heidelberg (2016). https://doi.org/10.1007/978-3-642-410 89-5 Dubey, R., Gunasekaran, A., Bryde, D.J., Dwivedi, Y.K., Papadopoulos, T.: Blockchain technology for enhancing swift-trust, collaboration and resilience within a humanitarian supply chain setting. Int. J. Prod. Res. 58(11), 3381–3398 (2020). https://doi.org/10.1080/00207543.2020. 1722860 Elleuch, H., El Mouloudi, D., El Mhamedi, A., Chabchoub, H.: Resilience and vulnerability in supply chain: literature review. IFAC-PapersOnLine 49, 1448–1453 (2016). https://doi.org/10. 1016/j.ifacol.2016.07.775 Flick, U., von Kardoff, E., Steinke, I.: QualitativeForschung. Ein Handbuch, 11th edn. Rowohlt Taschenbuch Verlag, Reinbeck bei Hamburg (2015) Francisco, K., Swanson, D.: The supply chain has no clothes: technology adoption of blockchain for supply chain transparency. Logistics 2(1), 2 (2018). https://doi.org/10.3390/logistics201 0002 Frederico, G.F.: Towards a Supply Chain 4.0 on the post-COVID-19 pandemic: a conceptual and strategic discussion for more resilient supply chains. Manag. J., 0972–9968 (2021). https://doi. org/10.1108/RAMJ-08-2020-0047 Glaser, B.G., Strauss, A.L.: Grounded Theory: Strategien qualitativer Forschung. Verlag Hans Huber (2010) Gohil, D., Thakker, S.V.: Blockchain-integrated technologies for solving supply chain challenges. Mod. Supply Chain Res. Appl. 3(2), 78–97 (2021). https://doi.org/10.1108/MSCRA-10-20200028 Kshetri, N.: Blockchain’s roles in meeting key supply chain management objectives. Int. J. Inf. Manag. 39, 80–89 (2018). https://doi.org/10.1016/j.ijinfomgt.2017.12.005 Mayring, Ph.: Qualitative Inhaltsanalyse. Grundlagen und Techniken, 12th edn. Beltz Verlag, Weinheim (2015) Mayring, P., Fenzl, T.: Qualitative Inhaltsanalyse. In: Baur, N., Blasius, J. (eds.) Handbuch Methoden der empirischen Sozialforschung, pp. 543–556. Springer, Wiesbaden (2014). https://doi. org/10.1007/978-3-531-18939-0_38
The Impact of Blockchain on Supply Chain Resilience
91
Min, H.: Blockchain technology for enhancing supply chain resilience. Bus. Horiz. 62(1), 35–45 (2019). https://doi.org/10.1016/j.bushor.2018.08.012 Nandi, M.L., Nandi, S., Moya, H., Kaynak, H.: Blockchain technology-enabled supply chain systems and supply chain performance: a resource-based view. Supply Chain Manag. 25(6), 841–862 (2020). https://doi.org/10.1108/SCM-12-2019-0444 Ponomarov, S., Holcomb, M.: Understanding the concept of supply chain resilience. IJLM 20, 124–143 (2009). https://doi.org/10.1108/09574090910954873 Rejeb, A., Keogh, J.G., Simske, S.J., Stafford, T., Treiblmaier, H.: Potentials of blockchain technologies for supply chain collaboration: a conceptual framework. Int. J. Logist. Manag. 32(3), 973–994 (2021). https://doi.org/10.1108/IJLM-02-2020-0098 Scholten, K., Sharkey S., Fynes, B.: Mitigation processes – antecedents for building supply chain resilience capabilities. Supply Chain Manag. 19 (2014). https://doi.org/10.1108/SCM-06-20130191 Sheffi, Y., Rice, J.B., Jr.: A supply chain view of the resilient enterprise. MIT Sloan Manag. Rev. 47(1), 41–48 (2005) Singh, C.S., Soni, G., Badhotiya, G.K.: Performance indicators for supply chain resilience: review and conceptual framework. J. Ind. Eng. Int. 15(1), 105–117 (2019). https://doi.org/10.1007/s40 092-019-00322-2 Soni, U., Jain, V., Kumar, S.: Measuring supply chain resilience using a deterministic modeling approach. Comput. Ind. Eng. (2014). https://doi.org/10.1016/j.cie.2014.04.019 Tukamuhabwa, B.R., Stevenson, M., Busby, J., Zorzini, M.: Supply chain resilience: definition, review and theoretical foundations for further study. IJPR 53, 5592–5623 (2015). https://doi. org/10.1080/00207543.2015.1037934 Wang, Y., Han, J.H., Beynon-Davies, P.: Understanding blockchain technology for future supply chains: a systematic literature review and research agenda. Supply Chain Manag. Int. J. 24 (2018). https://doi.org/10.1108/SCM-03-2018-0148
Towards Supply Chain Resilience in Mining Industry: A Literature Analysis Raúl Castillo-Villagra1,2,3(B)
and Klaus-Dieter Thoben1,2
1 Universität Bremen, Bibliothekstraße 1, Bremen, Germany
{car,tho}@biba.uni-bremen.de
2 BIBA-Bremer Institut für Produktion und Logistik GmbH, Hochschulring 20, 28359 Bremen,
Germany 3 Faculty of Engineering and Architecture, Arturo Prat University, Arturo Prat Avenue 2120,
Iquique, Chile
Abstract. The production of raw materials has become a backbone of the manufacturing industries and will play a relevant role in the transition process to global sustainability by 2050. Due to its nature, the mining industry (MI) is highly prone to disruptions, by triggering a shortage of mineral raw materials or commodities in the downstream segment of many supply chains (SC). Lately, this latent vulnerability of the early phases of today’s SC has drawn attention to better understand Supply Chain resilience (SCRes) in mining. However, there is no common understanding on the concept of SCRes in the mining industry context in literature. This paper aims to contribute to a conceptualization of SCRes in the mining industry by defining and operationalizing main resilience principles and elements from the mineral SC topics and by taking into account the industrial nature of the MI. The baseline of this paper is a literature review based on 27 relevant articles systematically selected from peer-reviewed papers from Scopus and Web of Sciences. Keywords: Supply chain resilience · Mining industry · Literature review
1 Introduction The mining industry (MI) or mining has a key role in future developments in industrial production. On the one hand, the increasing demand for high-tech products and systems [1] has increased the consumption of mineral raw materials such as rare earth. On the other hand, the intended transition to global climate neutrality by 2050 [2], based on an increased renewable energy generation, demands more mineral raw materials such as copper, lithium, iron, and aluminum. To successfully deal with these challenges, a goal of MI is to become a reliable partner for many value chains by securing a continuous supply of raw materials. Due to its inherent nature, the mining industry is prone to different kinds of uncertainties, e.g., natural disasters, geological failure, commodity market variability, transport infrastructure failures, among others [3]. These uncertainties could deviate mining © The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 M. Freitag et al. (Eds.): LDIC 2022, LNLO, pp. 92–103, 2022. https://doi.org/10.1007/978-3-031-05359-7_8
Towards Supply Chain Resilience in Mining Industry
93
operations performance and cause serious disruptions in the global supply chain (SC). Therefore, strong robustness and reliability of the mining sector as the backbone of many supply chains is needed. Supply chain resilience (SCRes) in general deals with these requirements, reducing the vulnerability of supply chains against disruptions. SCRes has been highlighted in SC management research domain [4, 5]. In recent years research on SCRes has focused on definitions, frameworks, and an in-depth understanding of terms and concepts [5] based on manufacturing. However, assuming the mining industry as a black box with an unlimited supply of raw materials [6] is not realistic when considering the mining nature uncertainties such as geological failure, geological quality or weather conditions, and the natural depletion of mineral resources [3, 7]. Therefore, there is still a lack of understanding of the mining industry’s impact on resilience of global supply chains. In addition it is necessary to better understand how existing concepts of SCRes fit within the context of mining. This paper aims to present a conceptualization of SCRes in the MI context. In order to achieve this, Sect. 2 presents a brief description of the mining industry. Next, Sect. 3 introduces the scope and methodology of this work. After that, in Sect. 4, the results of a literature analysis are presented, discussing a common understanding of resilience in mining. Finally, the last section gives a conclusion and an outlook for future research.
2 Characterizing the Mining Industry This section provides an overview of the main industrial characteristics of mining. We refer to the works by Zúñiga (2015) [3], Cameron and Standley (2017) [8] and Society for Mining, Metallurgy and Exploration (2011) [9], for more details of this industry. 2.1 Mining Industry in the Mineral Supply Chain The mining industry is a primary activity of the economy [10, 11]. MI differs considerably from other industries, e.g., manufacturing, due to its specific operational processes, mining life cycle, depletion of natural resources, and regulations [8, 12]. According to Sauer and Seuring, 2019 [13], mining is considered in the upstream segment of mineral SC, compromising all stages in order to produce and sell mining outputs or mineral commodities as concentrated and refined minerals. These mining outputs are inputs on the downstream segment of mineral SC, which considers all stages of manufacturing pre- and end-products as well as retail, use, recycling, and disposition of end-products. Figure 1 shows the generic mineral SC model and its segmentation. MI carries out several processes during its life cycle to produce a mineral commodity. These processes, which go from the discovery of the mining district, through mineral exploitation, to the ore’s depletion and definitive industry closure [3], are not depicted in the generic mineral SC model by Sauer and Seuring, 2019 [13]. Thereby, in this paper and to better understand MI, we extend the processes implied in the upstream segment of mineral SC, considering the mining processes of the EM model [14] as a baseline; namely, discovery, establish, exploit, and beneficiate, sell and rehabilitate (see Fig. 2).
94
R. Castillo-Villagra and K.-D. Thoben Downstream focal firm
Upstream focal firm
Extraction
Global minerals market
Beneficiation /Refinement
Reuse / Recycling Manufacturing
Upstream SC
Retail / Use
Downstream SC
Fig. 1. Generic mineral SC model [13]. Mining industry
Discover
Establish
Exploit
Beneficiate
Sell
Rehabilitate
Upstream SC
Fig. 2. Mining processes of the EM model [14] within the upstream segment of the mineral SC (adapted from Sauer y Seuring, 2019 [13]).
2.2 Mineral Deposit Main Source of Mining Industry The MI competitiveness is given by the geological characteristics of the mineral deposit -the mine- where the primary raw material -the ore- is originated [3, 15]. Inherent by its nature, the mine establishes its unique mineral resources’ quality, quantity, and location. The main characteristic of this industry that differentiates it from others is the depletion of the mineral resources, which define the mining life cycle. Therefore, continuous projects development -brownfield or greenfield- is necessary for the industry to survive, extending its life cycle [12]. Moreover, the geological parameters also define the sites where this industry is carried out [16, 17]. These sites are mostly inhospitable, located in remote and inaccessible locations, presenting high altitudes -3000 to 4600 m.a.s.l.- active earthquake zones, and tend to have high snow loads and other extreme environmental conditions [9]. Consequently, these sites’ characteristics make that this industry is more vulnerable to natural disturbances comparing others.
3 Scope and Methodology The scope of this literature review includes an overview of the current literature on SCRes, considering the MI perspective. In this paper, the MI refers to the industrial activities considered in “sector 21” of NAIC [18], i.e., mining, quarrying, and oil and gas extraction, which is justified due to the common characteristics of this industrial sector [8], The term mining industry (or mining) as defined in this paper is analogous as referred to in the literature as primary production, mining value chain, or mining supply chain.
Towards Supply Chain Resilience in Mining Industry
95
A systematic review (SR) method [19] was followed to conceptualize the term SCRes in mining. To identify, select and retrieve the research articles within literature related to this topic, we adopted a five-step process approach by Tavares et al. (2012) [20]: (i) database selection, (ii) identification of keywords for search, (iii) criteria for inclusion/exclusion of studies, (iv) manual review of selected abstracts, and (v) full-text review of selected papers. Two research questions (RQ) are established to lead this SR (RQ1): “What is understood by resilience in the upstream SC segment in mining context?”, and (RQ2): “What elements of SCRes are described in the literature inside the mining context?”. Two databases, Scopus and WoS, were selected. Based on the general guidelines suggested by Chen et al. (2014) [21], several keywords combinations (shown in Table 1) were established for searching based on the research purpose. The inclusion criteria considered peer-review articles, conference proceedings, and reviews, all of these in English, which consider the term resilience in the upstream SC segment processes or activities of mining. Our first search (May 2021) retrieved 290 articles, 106 from WoS and 184 from Scopus. After removing duplicates 198 papers were identified. Afterwards, abstracts were assessed based on the inclusion criteria. The result was 49 articles to be read in full text. Finally, 27 papers were selected for answering the research questions proposed above. Table 1. Research protocol. Criteria Keywords combination
Details descriptions Supply Chain Resilience considering mining context Main term combination: Main terms must consider the keyword “resilience” in combination with “supply chain”, “primary production”, and “value chain” Terms inside mining industrial context: Main terms combinated with keywords of the mining industrial context “mining”, “quarry”, “oil”, “Gas”, “commodities” or “extractive industries” Search String: ((“Resilien*”) AND (“primary production” OR “value chain” OR “Supply Chain”)) AND (“Mining” OR “Quarry” OR “Oil” OR “Gas” OR “commodit*” OR “Extractive industr*”)
Publication type:
Articles, reviews and proceedings. (peer review)
Inclusion criteria:
Papers that do not consider or describe upstream SC segment activities or processes (black-box perspective, unlimited resources supply)
Exclusion criteria:
Papers that not consider or describe upstream SC segment activities or processes (black box perspective, unlimited resources supply) e.g. (ii) Economics effects of multiple disaster in mining areas (Economic perspective)
96
R. Castillo-Villagra and K.-D. Thoben
4 Conceptualizing Supply Chain Resilience in Mining In this section, we present results of the literature analysis answering both research questions. Firstly, answering RQ1 in Subsect. 4.1, three perspectives, based on the current literature of SCRes in the mineral raw material SC context, were analyzed and discussed in order to propose a common definition of SCRes in mining. Secondly, in Subsect. 4.2 answer to RQ2 is given, in which the operationalization of SCRes elements in the mining is addressed. In order to do this, we consider definitions and comprehending of the SCRes elements from the more recent and relevant works by Biederman (2018) [4] and Tukamuhabwa et al. (2015) [22] and fit them to the mining, considering the industry nature and the mine operation context. 4.1 Defining SCRes in Mining Mineral SC’s Sustainability Perspective. Resiliency has become a sustainability branch in the mineral SC context since it has to protect itself against uncertainties and unforeseen disruptions. Resiliency carries out the economic perspective from a sustainability perspective, aiming to meet the stakeholder requirement against current and future vulnerabilities at a minimal cost [23]. To achieve this goal, it is relevant that mineral SC improves its operations [13] at all its levels: strategic, tactical, and operational. SC’s operational levels are needed to improve the mineral industry’s long-term resilience and sustainability facing the dynamics of ever changing challenges SCs are concerned with. Thus, high levels of capital investment are required during the mining life cycle [16, 17, 24] in order to adapt to current and future challenges. Most of these investments are focus on infrastructure development and on maintaining operational continuity, e.g., improving irresponsible operations reducing environmental impacts. Resilience is a backbone to achieve sustainability in the mineral SC. Although there is not a clear concept detached from sustainability, some statements are highlighted which support the definition SCRes in the mining context: • Resilience as a proactive capacity aims to protect SC against ongoing and future disruptions. • Resilience capability in the mining aims for operational improvement and operational continuity. • Resilience as a long-term perspective needs to be adaptable over time. Critical Raw Material Perspective. The critical raw materials (CRM) initiative [25] ranks minerals criticality considering the importance to the EU economy and the high risk associated with their supply. In the CRM context, SCRes is a concept that deals with assessing mineral criticality regarding how resilient is its current SC or how resilient it will be in the future [26, 27]. A common conceptualization of SCRes is identified in criticality mineral studies, which is provided by Sprecher et al. (2015) [26]. This conceptualization is based on the system’s resistance to supply disruption, its fast recovery from the disruption, and flexibility enough to adopt alternative supply strategies or find mineral substitutes. Based
Towards Supply Chain Resilience in Mining Industry
97
on this common understanding, mechanisms of resilience have been evaluated in critical mineral studies, e.g., material substitution in neodymium [28], recycling and stocking in tantalum [29], dynamic factors such as societal and geo-political in rare earth elements [30], and primary production in cobalt [31]. Although these studies focus on establishing SC resilience as an indicator for mineral criticality analysis, none propose a fixed value; hence, the SC resilience understanding is a muscle to build up rather than a value that achieving [30]. The conceptualization of SCRes discussed above focuses on the mineral SC; in other words, from the primary production or mining industry -upstream segment- to the final consumer -downstream segment-. In this sense, some perspectives do not fit in the mining context. For instance, mining is just considered one of the resilience mechanisms for mineral SC, which aims to increase the production from new projects or restore operations closure [31] against disruptions in the mineral SC. For this reason, it is paramount to bear in mind the considerations and constraints from CRM perspective to conceptualize SCRes in the mining context: • Resilience as a long-term perspective offers insight into the SC behavior against current and future disruptions. • The substitution of mineral material, as a resilience mechanism, is not realistic in the MI since mining is developed with a known ore body condition. • Resilience is a capability that has to build up in mining. Mining Life Cycle Perspective. The mining life cycle (or mine life cycle) describes the industry stages: mineral exploration, mine development, mine operation, and closure [3, 9, 14]. While the two former stages regard the processes of the mining development project, the latter two stages consider the whole processes for transforming ore into a commodity. During the whole stages of the mining life cycle can happen disruptions affecting the mineral SC. However, the literature mainly has focused on disruption in the operation mine [16, 17, 24] and the closure stage [7] since could cause a shortage in the supply of commodities in the SC downstream, as shown in Fig. 3. Mining industry Mineral ExploraƟon Mining life cycle stages
Mine Development Mine OperaƟon
Discover Establish Exploit
Beneficiate
Sell
Rehabilitate
Closure
Upstream SC Disruptions
Fig. 3. Disruptions in mine operation stage and closure stage.
The mine development stage establishes mining operational conditions both for mineral commodity production and final rehabilitation. These operational conditions are
98
R. Castillo-Villagra and K.-D. Thoben
established several years before the beginning of both the mine operation stage -at least five years- and the closure stage -at least fourteen years- [3]. This temporality and the high probability of disruption in the mining life cycle make the topic of resilience to be considered in the mine development stage. Considering resilience in the mine development stage offers effective resource utilization in the mine operation stage [32] and designing more robust operations to reduce uncertainties over time [33]. This is because resilience considers various disruptions scenarios and not just the common uncertainty parameter - economic and geological[7, 34]. In the operation mine stage, the resilience concept focuses on disruption management. This focus considers daily disruptions or emerging threats -operation mine- rather than a holistic viewpoint - the mineral supply chain- [23, 24]. This management viewpoint better understands drivers and barriers to cope with disruptions set by the mining nature. Thus, the mining industry achieves its short-term goals and establishes a long-term competitive positioning. Some principles from the mining life cycle perspective reinforce a common understanding of SC resilience in the mining industry as: • Resilience is focused on operational disruption management -daily disruptions- to achieve short-term goals that support its long-term competitive positioning. • Operational conditions have to be considered during operation and closure stages in mining. Defining Supply Chain Resilience in Mining. Despite the increasing attention paid to mineral raw material supply chain resilience studies, there is still no common understanding. The literature analyzed in this work still provides a wide variety of definitions and concepts. Most of these researches do not specify the role of the mining industry the upstream segment - in the resilience of the mineral supply chain, leading to confusion or misinterpretation to practitioners and academics. Bearing this in mind and the considerations discussed previously; we define SCRes in mining, based on the definition and concepts of SCRes in the material supply chain by Sprecher et al. (2015) [26], as: “The capability of the upstream mineral SC segment to continuously adapt during the life cycle of the mine, building-up capabilities to withstand disruptions and recover from them, aiming to ensure the continuity of its operations by delivering sufficient mining outputs to the downstream mineral SC segment.” 4.2 Operationalizing SCRes Elements in Mining The definition of SCRes in mining previously proposed depends on proactive, reactive, and adaptive capabilities. The proactive capability can withstand disruptions, absorb the impact, and maintain operational performance at acceptable levels. If there is an operational performance collapse, the reactive capability initiates. The reactive capability has to restore the operation and reach the desired level of performance, which can commonly take a high level of time and cost. Once the operating performance levels have
Towards Supply Chain Resilience in Mining Industry
99
been restored, the adaptive capability has to prepare the mining against future disruptions based on the previous events. To build up these capabilities in mining, some SC resilience elements must define and fit, considering the industry’s nature and operating context. Figure 4 depicts these specific elements involved in the building of capabilities of SCRes in the mining, which is discussed below.
Resilience -Mining-
Robustness
AdaptaƟon
Recovery
-ProacƟve-
-AdapƟve-
-ReacƟve-
Redundancy
F l ex i b i l i t y
Adaptability
Agility
S ta n d b y mode
Fig. 4. SCRes in mining context (adapted from Ivanov, 2021 [35]).
Building Proactive Capability. Robustness is the main precursor to achieving resiliency since it can withstand the shock of disruptions without being highly impaired in its operation [7, 23, 27, 31, 34, 36–38]. Two SCRes elements can be classified to build a proactive capability: “Redundancy” and “Flexibility.” Redundancy. It highlights its capability to develop capacities surplus - safety stock, multiple sourcing, and unused spare capacity-to increase SC resilience [4, 22]. In mining, context is not completely fit due twofold. Firstly, developing a surplus of unprocessed ore results in an extensive surface in order to stock. Secondly, handling material is the highest operational cost in mining [39], hence growing cost due to double movement. For this reason, redundancy in mining focuses on building capacity surplus in operational tiers through flexible operations to resist and respond against disruptions. Flexibility. It is the firm’s ability to take measures in a disruption event enabling rapid reaction and adaptation [4, 22]. In the mining context, flexibility is an element considered in the mine development stage. In this sense, flexibility eases operational design in alternative disruptions during the mining life cycle, aiming to withstand and respond to disruptions for operational continuity. Building Reactive Capability. The recovery principle deals with reactive capability, aiming to return to normal operation after disruptions. In a mining context, this scenario is not desired since it entails high costs and times [13, 17, 24, 31, 40–42]. In order to build up reactive capability, two elements are described. “Agility” is discussed in the current literature on SCRes elements, and the “standby mode” is discussed and conceptualized as a SCRes element in a mining context. Agility. It is a principle that pursues resource allocation promptness and process tuning as flexible reactions to unpredictable disruptions in the SC through transparency dissemination of information [22, 42]. This concept fits within the mining context; however, some consideration is important to keep in mind as the velocity to react -temporality in the mining life cycle- [3, 16].
100
R. Castillo-Villagra and K.-D. Thoben
Standby Mode. The literature analyzed presents the primary production re-starting as a resilience mechanism [26, 31]. This mechanism refers to the finishing of temporary closure. This kind of closure is a capacity of the mining industry, which suspends its activities for a limited period [43, 44]. In this period, the ‘care and maintenance phase’ is carried out to prepare the site for returning to active production because there is still a mineral resource to be exploited. This situation may be due to financial pressures, environmental incidents, social incidents, regulatory changes, and structural failure [44]. In the current literature of SCRes, there is no elements with this characteristic. Therefore, we define the standby mode element as the mining capability to start a latency state to respond against unexpected disruptions, maintaining its basic operating conditions to re-start its operations when needed. Building Adaptive Capability. Adaptation in mining is vital to continuous growth, building up a more protected industry in the long term against future disruptions [13, 17, 24, 32, 33, 38]. Adaptability. It is an element that involves many principles, e.g., flexibility, velocity, visibility, and agility, as an SC’s structural property focusing on recovery from disruptions [4, 22]. In the mining context, adaptability aims long-term perspective that depends on knowledge and learning gained during its life cycle [16, 24]. To do this, adaptability in mining depends on the mineral SC collaboration -visibility and sharing informationto reconfigure its operation to prepare and withstand disruptions.
5 Conclusion This paper aimed to conceptualize SCRes in the mining industry context. Twenty-seven systematically reviewed papers from SCRes in the industrial mining context were the basis for this study. The literature analysis conducted resulted in a common definition of SCRes in mining and a better understanding of the SCRes’ elements operationalization considering the nature of this industry. The SCRes in mining definition proposed builds on three specific perspectives analyzed in the literature, namely sustainability in the mineral SC, critical raw material, and mining life cycle. From these perspectives, the principles and limitations of SCRes understanding were discussed and highlighted, focusing on the upstream segment of the mineral SC. Although most of the principles of this definition match with current SCres literature, the focus of ensuring the continuity of its operations concerning daily threats differs from the resilience research published so far. Three capabilities sustain the definition of SCRes in mining proposed. Our literature analysis conceptualized five SCRes elements to build up these capabilities. Four of them fairly matched SCRes current literature, for which they were fitted and redefined them considering the industry nature and mining reality. One of them, standby mode, was not found in the current SCRes literature; thereby, we proposed this element as a specific characteristic to build up resilience capabilities in the mining context. To our knowledge, this work is the first attempt to comprehend the SCRes concept at the starting point of many global supply chains, being a preliminary work on a more
Towards Supply Chain Resilience in Mining Industry
101
extended research agenda. Future works can include the conceptualization proposed in this research to develop frameworks and strategies in the SCRes in the mining context from empirical and theoretical perspectives. Acknowledgments. The authors thank the International Graduate School for Dynamics in Logistics (IGS) of the University of Bremen. Mr. Castillo-Villagra´s work was supported by Agencia Nacional de Investigación y Desarrollo –ANID, Chile under grant no. 72200439.
References 1. Gislev, M., et al.: Report on Critical Raw Materials and the Circular Economy. Publications Office of the European Union, Luxembourg (2018) 2. International Council on Mining & Metals: Adapting to a changing climate: building resilience in the mining and metals industry. International Council on Mining & Metals (2019) 3. Zuniga, R.: Modeling of supply chain processes of the mineral raw materials industry from the perspective of EM, SCOR and DCOR models. Dissertation (2015). https://media.suub. uni-bremen.de/handle/elib/918 4. Biedermann, L.: Supply Chain Resilienz. Springer, Wiesbaden (2018). https://doi.org/10. 1007/978-3-658-23516-1 5. Pettit, T.J., Croxton, K.L., Fiksel, J.: The evolution of resilience in supply chain management: a retrospective on ensuring supply chain resilience. J. Bus. Logist. (2019). https://doi.org/10. 1111/jbl.12202 6. Aboah, J., Wilson, M.M., Rich, K.M., Lyne, M.C.: Operationalising resilience in tropical agricultural value chains. Supply Chain Manag. (2019). https://doi.org/10.1108/SCM-052018-0204 7. Dominy, S., O’Connor, L., Parbhakar-Fox, A., Glass, H., Purevgerel, S.: Geometallurgy— a route to more resilient mine operations. Minerals (2018). https://doi.org/10.3390/min812 0560 8. Cameron, P.D., Stanley, M.C.: Oil, Gas, and Mining. A Sourcebook for Understanding the Extractive Industries. World Bank Group, Washington DC (2017) 9. Society for Mining, M.A.E.: SME Mining Engineering Handbook, 3rd edn., vol. 49. SME, Englewood (2011) 10. Kenessey, Z.: The primary, secondary, tertiary and quaternary sectors of the economy. Rev. Income Wealth (1987). https://doi.org/10.1111/j.1475-4991.1987.tb00680.x 11. Radetzki, M.: A Handbook of Primary Commodities in the Global Economy. Cambridge University Press, Leiden (2008) 12. Zuñiga, R., Wuest, T., Thoben, K.-D.: Comparing mining and manufacturing supply chain processes: challenges and requirements. Prod. Plann. Control (2015). https://doi.org/10.1080/ 09537287.2013.855335 13. Sauer, P.C., Seuring, S.: Extending the reach of multi-tier sustainable supply chain management – insights from mineral supply chains. Int. J. Prod. Econ. (2019). https://doi.org/10. 1016/j.ijpe.2018.05.030 14. The Open Group Exploration, Mining, Metals, and Minerals (EMMM) Forum: The exploration and mining business reference model: concepts and definitions (2013) 15. Sachs, P.F.T.: Cadeia de valor mineral e tecnologia da informação: alinhamento estratégico como gerador de eficácia em empresas de mineração. Universidad de São Paulo, Tese de Doutorado (2009)
102
R. Castillo-Villagra and K.-D. Thoben
16. Hodgkinson, J.H., Hobday, A.J., Pinkard, E.A.: Climate adaptation in Australia’s resourceextraction industries: ready or not? Reg. Environ. Change 14(4), 1663–1678 (2014). https:// doi.org/10.1007/s10113-014-0618-8 17. Lim-Camacho, L., et al.: Complex resource supply chains display higher resilience to simulated climate shocks. Glob. Environ. Chang. (2017). https://doi.org/10.1016/j.gloenvcha. 2017.08.011 18. Executive Office of the President-Office of Management and Budget: NAIC - North American Industry Classification System, United States (2017). https://www.census.gov/naics/refere nce_files_tools/2017_NAICS_Manual.pdf. Accessed 13 July 2021 19. Grant, M.J., Booth, A.: A typology of reviews: an analysis of 14 review types and associated methodologies. Health Info. Libr. J. (2009). https://doi.org/10.1111/j.1471-1842.2009.008 48.x 20. Tavares Thomé, A.M., Scavarda, L.F., Fernandez, N.S., Scavarda, A.J.: Sales and operations planning: a research synthesis. Int. J. Prod. Econ. (2012). https://doi.org/10.1016/j.ijpe.2011. 11.027 21. Chen, L., Olhager, J., Tang, O.: Manufacturing facility location and sustainability: a literature review and research agenda. Int. J. Prod. Econ. (2014). https://doi.org/10.1016/j.ijpe.2013. 05.013 22. Tukamuhabwa, B.R., Stevenson, M., Busby, J., Zorzini, M.: Supply chain resilience: definition, review and theoretical foundations for further study. Int. J. Prod. Res. (2015). https:// doi.org/10.1080/00207543.2015.1037934 23. Alfaqiri, A., et al.: A systemic approach for disruption risk assessment in oil and gas supply chains. Int. J. Crit. Infrastruct. (2019). https://doi.org/10.1504/IJCIS.2019.100428 24. Lim-Camacho, L., Jeanneret, T., Hodgkinson, J.H.: Towards resilient, responsive and rewarding mining: an adaptive value chains approach. Resour. Policy (2019). https://doi.org/10.1016/ j.resourpol.2019.101465 25. European Commission: Study on the EU’s list of Critical Raw Materials - Final Report (2020) 26. Sprecher, B., Daigo, I., Murakami, S., Kleijn, R., Vos, M., Kramer, G.J.: Framework for resilience in material supply chains, with a case study from the 2010 Rare Earth Crisis. Environ. Sci. Technol. (2015). https://doi.org/10.1021/acs.est.5b00206 27. Dewulf, J., Blengini, G.A., Pennington, D., Nuss, P., Nassar, N.T.: Criticality on the international scene: Quo vadis? Resour. Policy (2016). https://doi.org/10.1016/j.resourpol.2016. 09.008 28. Sprecher, B., et al.: Novel indicators for the quantification of resilience in critical material supply chains, with a 2010 rare earth crisis case study. Environ. Sci. Technol. (2017). https:// doi.org/10.1021/acs.est.6b05751 29. Mancheri, N.A., et al.: Resilience in the tantalum supply chain. Resour. Conserv. Recycl. (2018). https://doi.org/10.1016/j.resconrec.2017.10.018 30. Mancheri, N.A., Sprecher, B., Bailey, G., Ge, J., Tukker, A.: Effect of Chinese policies on rare earth supply chain resilience. Resour. Conserv. Recycl. (2019). https://doi.org/10.1016/ j.resconrec.2018.11.017 31. Van den Brink, S., Kleijn, R., Sprecher, B., Tukker, A.: Identifying supply risks by mapping the cobalt supply chain. Resour. Conserv. Recycl. (2020). https://doi.org/10.1016/j.resconrec. 2020.104743 32. Dell, J.J., Pasteris, P.: Adaptation in the oil and gas industry to projected impacts of climate change. In: The SPE International Conference on Health, Safety and Environment in Oil and Gas Exploration and Production (2010). https://doi.org/10.2118/126307-MS 33. Katopodis, T., Sfetsos, A.: A review of climate change impacts to oil sector critical services and suggested recommendations for industry uptake. Infrastructures (2019). https://doi.org/ 10.3390/infrastructures4040074
Towards Supply Chain Resilience in Mining Industry
103
34. Coward, S., Dowd, P.: Geometallurgical models for the quantification of uncertainty in mining project value chains (2015) 35. Ivanov, D.: Introduction to Supply Chain Resilience. Springer, Cham (2021). https://doi.org/ 10.1007/978-3-030-70490-2 36. Al-Haidous, S., Al-Ansari, T.: Sustainable liquefied natural gas supply chain management: a review of quantitative models. Sustainability (Switzerland) (2020). https://doi.org/10.3390/ SU12010243 37. Scholz, R.W., Wellmer, F.-W.: Losses and use efficiencies along the phosphorus cycle. Part 1: Dilemmata and losses in the mines and other nodes of the supply chain. Resour. Conserv. Recycl. (2015). https://doi.org/10.1016/j.resconrec.2015.09.020 38. Uzcategui, M., Mathison, J., Soto, A.J.: Design of resilient production facilities through innovation and risk management. In: The SPE Latin American and Caribbean Health, Safety, Environment and Sustainability Conference (2015). https://doi.org/10.2118/174104-ms 39. Icarte, G., Berrios, P., Castillo, R., Herzog, O.: A multiagent system for truck dispatching in open-pit mines. In: Freitag, M., Haasis, H.-D., Kotzab, H., Pannek, J. (eds.) LDIC 2020. LNL, pp. 363–373. Springer, Cham (2020). https://doi.org/10.1007/978-3-030-44783-0_35 40. Hossain, N.U.I., Jaradat, R., Marufuzzaman, M., Buchanan, R.K., Rinaudo, C.: Assessing and enhancing oil and gas supply chain resilience: a Bayesian network based approach. In: IIE Annual Conference. Proceedings, pp. 1115–1120 (2019) 41. Jager, B., Hjelle, H.M.: Handling multi-party complexities in container flows in the upstream oil and gas supply chain: potential lessons for an application to intercontinental container supply chains. In: The 3rd International Conference on Transportation Information and Safety, pp. 661–668 (2015) 42. Shqairat, A., Sundarakani, B.: An empirical study of oil and gas value chain agility in the UAE. Benchmarking (2018). https://doi.org/10.1108/BIJ-05-2017-0090 43. Australian Goverment: Mine closure and completion. Leading practice sustainable development program for the mining industry. Department of Industry Tourism and Resources, Canberra (2006) 44. International Council on Mining & Metals: Integrated Mine Closure. Good Practice Guide, 2nd edn. ICMM (2019)
Data Quality in Social Media Analytics for Operations and Supply Chain Performance Management Fabian Siekmann1(B)
, Aseem Kinra1
, and Herbert Kotzab1,2
1 Universität Bremen, Bremen, Germany
{fab.siekmann,kinra,kotzab}@uni-bremen.de 2 Universiti Utara Malaysia, Sintok, Malaysia
Abstract. Social media analytics (SMA) is claimed to be an opportunity for practical inquiry to create new knowledge and possibilities but is only slowly finding its way into practice due to uncertain information quality. Good Operations and Supply Chain Management (OSCM) decisions are just as good as the data they are based upon. A more detailed consideration of data quality is needed, especially when natural language data is processed for decision-making. Motivated by recent calls in the domain, the purpose of this study is to investigate how big data quality is considered in SMA for operations and supply chain performance. The study employs a directed qualitative content analysis of 56 research contributions based on the re-analysis of a previous systematic literature review. The results reveal that within performance-oriented SMA literature, intrinsic and contextual data quality are not comprehensively addressed by OSCM-research to date. More particularly it is shown, that contextual data quality assessment remains a challenge for the analysis of textual social media data. The study contributes by reporting how data quality is considered for SMA in operations and supply chain performance management (OSCPM) literature from an intrinsic and contextual perspective. Based on the results of this analysis, data relevancy and data believability are identified as levers to reduce information uncertainty in SMA-aided decisionmaking, paving the way for future research on contextual social media data quality in OSCM. Keywords: Operations and Supply Chain Management · Social media analytics · Data quality
1 Introduction The growth of quantity and types of data generated all around the world has led to the development of advanced analytics tools (Arunachalam et al. 2018; Gandomi and Haider 2015). An increasing number of companies employ big-data-related technologies in their operations (Choi et al. 2018). Data-driven companies are supposedly more productive and more profitable compared to less data-driven peers (McAfee and Brynjolfsson 2012). Accordingly, industrial and academic communities in operations and © The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 M. Freitag et al. (Eds.): LDIC 2022, LNLO, pp. 104–116, 2022. https://doi.org/10.1007/978-3-031-05359-7_9
Data Quality in Social Media Analytics for Operations
105
supply chain management (OSCM) are paying increasing attention to social media data research, which is associated with the opportunity to improve daily decision making by analyzing the available data (Sheng et al. 2017; Waller and Fawcett 2013). In the context of operations and supply chain performance management (OSCPM), social media data have been used to supplement sourcing (Lin et al. 2017), product development (Irani et al. 2017), delivery (Albuquerque et al. 2016), reverse logistics (Bhattacharjya et al. 2016) and demand forecasting (Swain and Cao 2019) through the means of social media analytics (SMA). The degree to which social media data can be used is largely determined by their quality and poor data quality is a growing problem for decision-makers (Hazen et al. 2014). Furthermore, the meaning of social media data depends strongly on the context within which it is used (Wenzel and Van Quaquebeke 2018). The foundations for the control of data quality have now been laid out in the field (Ershadi et al. 2021; Hazen et al. 2014; Parssian et al. 2004). Data quality can be categorized into intrinsic- and contextual quality (Hazen et al. 2014). Despite data quality being highly relevant for big data analytics (BDA), it has received little attention in OSCM-research even though SMA requires, yet lacks the ability to understand nuances in human communication and language (Patton et al. 2020). The purpose of this study is to investigate how social media quality aspects are thematized in SMA for OSCPM. To the best of our knowledge, no past research has performed a structured literature review illumining social media data quality in OSC research from a performance management perspective. Thus, a directed quantitative and qualitative content analysis is carried out to shed light on the topic by answering the following research questions: To what extent is the data quality of different social media data platforms considered for SMA in OSCPM-literature? What are the data quality dimensions impeding SMA-enabled OSCPM-research? The remainder of this paper is structured as follows. In Sect. 2, OSCPM is linked to social media data as a new information opportunity. In Sect. 3 data quality and its effect on performance management is carved out against the backdrop of social media data. In Sect. 4, the performed research method is introduced. Section 5 presents the findings of the content analysis, directed by big data quality categories and performance dimensions. In Sect. 6, the findings are discussed followed by a summary of this study in Sect. 7.
2 OSC Performance and Unstructured Social Media Data OSCM is defined as ‘(…) the management of processes by which goods and services are designed, procured, produced, and delivered.’ (Seyedghorban et al. 2021). Performance is a widely used term across the management fields, which refers to the extent to which customers’ requirements are met and how economically the firm’s resources are utilized to achieve a given level of requirements (Neely 2007; Neely et al. 1995). Kamble and Gunasekaran (2019) differentiate performance measures along the dimensions of cost, quality, time, flexibility, and innovativeness to quantify efficiency and effectiveness of
106
F. Siekmann et al.
action. From an OSCM-perspective, improving their efficiency lends itself as social media data technologies allow for greater utilization of real-time data to reduce costs and risks (Sheng et al. 2017). Big data technologies benefit organizations by providing reliable and timely information for decision-makers (Matthias et al. 2017; McAfee and Brynjolfsson 2012). As a consequence, BDA is seen as a differentiating competitive factor between companies (McAfee and Brynjolfsson 2012; Wamba et al. 2016b) that transforms the way operations and supply chains are managed and designed (Waller and Fawcett 2013). Virtually every social interaction among the consumers on social media can be observed and analyzed (Baskerville et al. 2020; Wamba et al. 2016a) and organizations have begun to gain significant value from social media (Wamba et al. 2016a). One promising source of rich information lies in posts or comments written by customers in natural language (Kinra et al. 2020a). Consequently, organizations utilize this rich qualitative information (Kinra et al. 2020b) to measure performance and decide how to optimally allocate resources. Examples of the wide range of applications are provided by Lin et al. (2017), who develop a decision model for supplier performance evaluation. Gu et al. (2016) use social media data to detect traffic incidents in real-time. Abrahams et al. (2012) illustrate how SMA can be applied for product defect discovery from discussion forums. Lin et al. (2017) develop a decision model for supplier performance evaluation and Zavala and Ramirez-Marquez (2019) utilize tweets to reduce the response time for product recalls. However, the decisions made are only as good as the data which the SMA tools are fed.
3 Data Quality Issues for OSCPM with Social Media Data Owning a huge amount of data complicates the adaptation of analytics algorithms (Taleb et al. 2021) and quality problems are increasing as larger amounts of acquired data are analyzed (Parssian et al. 2004) ultimately impairing decision quality (Dubey et al. 2019; Hazen et al. 2014). On the other hand, if social media data quality is ensured, it is considered a key differentiator, leading to advantageous decisions (Taleb et al. 2021). OSCM-literature suggests considering data quality along two categories: intrinsic and contextual (Cai and Zhu 2015; Hazen et al. 2014; Taleb et al. 2021). The dimensions of intrinsic data quality refer to the quality of the data itself, ensuring that it is timely and of good quality for interpretability (Wilkin et al. 2020). Within this study, the following are considered in more detail as they match both purpose and literature sample: Accuracy answers the question of freedom from data error and whether data ambiguity was excluded. Timeliness describes the delay between data generation within reasonable time intervals. Consistency refers to the usable data format and the extent to which the data matches other data sources. Completeness indicates that no data relevant to the analysis is missing (Cai and Zhu 2015; Hazen et al. 2014). Prior research reflects issues of intrinsic data quality. Gu et al. (2016) analyze tweets for real-time traffic detection and highlight that tweets often contain typos, grammatical errors and abbreviations and Kinra et al. (2020b) show how to automatically retrieve information from unstructured country reports with mixed results due to inconsistent input data. Contextual data quality dimensions are described as relevancy, value-added,
Data Quality in Social Media Analytics for Operations
107
quantity, believability, accessibility, and reputation of the analyzed social media data. Within this study, the following are considered in more detail: Relevancy describes the fitness of the data to expound investigated aspect. Accessibility is understood as the provision and ease of use of data interfaces. Believability refers to the reputation (Beheshti-Kashi 2020) of the data source and the credibility of the data values (BeheshtiKashi 2020; Cai and Zhu 2015; Hazen et al. 2014). Prior research has picked up the issue of contextual data quality. Jiang et al. (2017) capture reviews based on contextual criteria and Mudambi and Schuff (2010) show that the helpfulness of product reviews depends on the product type. If natural language processing systems are not designed to understand context, it is unlikely the system will be able to accurately interpret the meaning of the data (Boyd and Crawford 2012; Patton et al. 2020). Consequently, if SMA does not consider contextual data quality, it fails to provide targeted information (Jiang et al. 2017). Amplifying the challenges mentioned above, the speed and volume of the data generated make it difficult to judge data quality with reasonable efforts despite technological advancements (Cai and Zhu 2015). Given the increasing utilization of SMA in the field of OSCPM and the importance of data quality for extracting information, this study investigates the extent to which data quality is considered and addressed in SMA-aided OSCPM-research. The underlying problem for decision-making is illustrated in Fig. 1 with reference to the garbage-in-garbage-out paradigm commonly applied in the context of data processing. Following this, poor intrinsic or contextual data quality leads to a limited decision-making quality, despite the assumed flawlessness of the analysis models applied. To understand this issue, this study examines the underlying data quality of the current OSCPM-research.
Fig. 1. Implications of limited data quality for decision-making quality.
4 Methodology To develop a comprehensive data quality perspective on SMA-enabled OSCM-research, this study applies systematic approaches (Booth et al. 2012) to a structured literature review (Paul and Criado 2020). A directed latent content analysis was selected as the suitable technique to explore data quality explicitly as well as implicitly (Berelson 1952) within the examined literature and to disclose its interrelations to performance dimensions derived from OSCM-literature (Hsieh and Shannon 2005; Krippendorff 2004).
108
F. Siekmann et al.
A peer-reviewed, systematically collected literature sample (Denyer and Tranfield 2009) of 1253 articles identified by Huang et al. (2019) in which 168 papers have been purposefully collected from to investigate SMA-aided performance improvement in OSCM serves as the literature basis for the review. In contrast to their work, the focus of this paper is on the underlying data quality of the studies, as well as data formats and data sources. Relevance sampling (Bengtsson 2016; Krippendorff 2004) was applied with a broad definition of OM (Mentzer et al. 2008) setting the focus on internal business processes in this study to reduce the number of contributions to a manageable size (Krippendorff 2004) to explore the extent to which data quality has been considered in research in-depth. Based on the focus on predominantly internal processes, 120 sampled papers focusing on externally linked processes were excluded, which for example have a focus on CRM or demand management. Thus, this study addresses the OSCM-activities sourcing, product development and production, delivery, product returns and reverse logistics and general activities, (Huang et al. 2019) covering the subareas of logistics and production planning (Mentzer et al. 2008). The data collection process is shown in Fig. 2.
Fig. 2. Method flow chart.
The resulting 48 papers were reanalyzed through immediate coding (Krippendorff 2004), along the dimensions OSCM-activity (Huang et al. 2019), operations performance dimensions (Kamble and Gunasekaran 2019) and data quality (Hazen et al. 2014). For example, quality and timeliness were coded as follows: ‘(…) in industrial processes for measuring and controlling quality during manufacturing processes’ and ‘(…) reducing decision-makers response time’ (Zavala and Ramirez-Marquez 2019). The presence of the category was noted with a checkmark in Table 1. A frequency comparison was combined with a qualitative analysis of content (Bengtsson 2016; Hsieh and Shannon 2005) allowing to investigate how intrinsic and contextual data quality (Hazen et al. 2014) are covered in the current literature across data sources, data formats and performance dimensions.
Data Quality in Social Media Analytics for Operations
109
To examine the intrinsic data quality, Hazen et al.’s (2014) categorization into accuracy, timeliness, consistency, and completeness was applied. Similarly, to examine the contextual data quality, a categorization into relevancy, accessibility, believability, and reputation and was applied. The categories value-added and quantity were not included since they are implicit in the selected sample. Finally, the contributions were analyzed in terms of performance categories addressed to allow for a linkage between data quality and performance. The latter is further understood along the established performance dimensions of cost, quality, time, flexibility, and innovativeness (Kamble and Gunasekaran 2019). The results of the associated content analysis offer an overview and are presented in the following sections based on Table 1.
5 Findings 5.1 Quantitative Analysis Table 1 illustrates the results of the directed content analysis. The publication years of the examined samples show that starting with isolated publications in the years 2010 to 2012, the number of publications increases beginning with the year 2013. The number of publications then rises to a peak of 12 in 2017 and 2018, before falling off due to the time constraints of the sample. Observing the social media data utilized, it is evident that text-based data from reviews and twitter is dominant in current research. In line with this, Gandomi and Haider (2015) state, that the analysis of image-intensive media is still in its infancy. Taking a closer look at the textual data it strikes that, with regard to the social media platforms used, different data sources are often combined to obtain information along with all OSCMactivities. At the same time, especially Twitter and product reviews are frequently used as single data sources. Twitter data is often exploited to increase efficiency by improving time and flexibility (Albuquerque et al. 2016; Chen et al. 2016; Cottrill et al. 2017; Gkiotsalitis and Stathopoulos 2016; Zavala and Ramirez-Marquez 2019), while product reviews serve as a data source primarily with a focus on effectiveness through quality and innovativeness information (Minnema et al. 2016; Ramanathan et al. 2017; Sahoo et al. 2018; Tan et al. 2015; Wood et al. 2016; Yuan et al. 2018). Considering the performance dimensions, costs is addressed in 52% of the contributions. The frequent occurrence of the dimension relates to the linkages to Time, Quality, Flexibility and Innovativeness (Gal-Tzur et al. 2014; O’leary 2011; Sodero and Rabinovich 2017), 33% of the articles analyzed refer to the performance dimension of time. Flexibility is addressed at 27% and Innovativeness is addressed at 31% within the investigated literature. The quality dimension is most frequently addressed among the non-cost measures at 50%. Along with the intrinsic categories, accuracy is the most frequently addressed dimension at 50%. It is argued to be less controllable in social media data but a large amount of data compensates for this disadvantage (Roden et al. 2017). The timeliness of the social media data is thematized in 40% of the contributions. It is stressed, that timeliness is often in tradeoff with accuracy and the balance is determined by the purpose of the analysis (Yuan et al. 2018). Consistency is addressed in 33% of the contributions, which is reasonable given the unstructured nature of textual data. The completeness of
110
F. Siekmann et al.
the data analyzed within the studies was thematized in 21% of the papers. Ashton et al. (2015) stress that information completeness is difficult to achieve with social media data. Along with the contextual categories, relevancy is the most frequent data quality dimension addressed, with indicators appearing in 69% of the papers and researchers face the challenge of extracting relevant information in the abundance of data due to low information density within the analyzed documents (Ashton et al. 2015). Accessibility is thematized in 25% of the examined contributions, mainly related to the ease of use of data interfaces, as the public accessibility lies in the nature of social media (O’leary 2011). Believability is a data quality dimension frequently taken up in the literature at 48%, which refers to both the reputation of the data source and the credibility of the values generated from the data. Based on the frequency analysis above, it is evident that accuracy, relevancy and believability are the most frequently quality addressed categories. In the following, more in-depth content analysis will be used to examine them in more detail. 5.2 Qualitative Analysis: Accuracy, Relevance and Believability Various studies deal with the issue of data accuracy. Sodero and Rabinovich (2017) ensure that their data sample is not biased by restricting it to only one product category. Going more into detail, Swain and Cao (2019) recommend the removal of spelling errors and punctuations from the text data. The required data accuracy depends on the target of the analysis because products are evaluated according to different features and from multiple perspectives limited (Ashton et al. 2015; Hou et al. 2017). Zavala and Ramirez-Marquez (2019) limit their study to twitter data to avoid data ambiguity and Lee et al. (2017) indicate possible accuracy conflicts between operations and the marketing function of the organization. However, it is also stressed, that technological-processing capability increases the firm’s ability to evaluate information quality (Cheng and Krumwiede 2018). Researchers also face the challenge of extracting relevant information in the abundance of data (Zavala and Ramirez-Marquez 2019). Illustrating the issue, Gu et al. (2016) find that a small part of the analyzed data already provides a comprehensive analytical picture. However, the relevance of data may also depend on whether incremental improvement or innovation is the goal of the analysis (Hartmann et al. 2016), making general aggregation of data inappropriate (Papadopoulos et al. 2017). Ashton et al. (2015) stress the issue of low information density within the analyzed documents. Chae (2015) points out, that user visibility may also influence information sharing behavior, ultimately affecting results depending on the purpose of analysis. The believability of social media data is also mentioned frequently. Lee et al. (2017) point out that data manipulation through artificially generated reviews reduces the believability of this data for the operations function and Zavala and Ramirez-Marquez (2019) opt for Twitter instead of Facebook data for reputational reasons. At the same time, Chae (2015) points to problematic fake tweets on Twitter, which is circumvented by Bhattacharjya et al. (2016) including providing verification signals on the platforms. Abrahams et al. (2015) link credibility to the data itself by examining the volume of the documents analyzed. Taking a look at the
Data Quality in Social Media Analytics for Operations Table 1. Data extraction sheet.
111
112
F. Siekmann et al.
stakeholders, Sigala (2014) notes different roles that customers take on the social internet depending on the platform they use and O’leary (2011) states, that certain platforms are seen as more trustworthy than others.
6 Discussion This study revealed that unstructured textual data provides OSCM with an interface to the consumer of the products and services, especially useful for insights on product and service quality. This finding is in line with Kamble and Gunasekaran (2019), who found the quality to have the highest representation in social media data-aided performance measurement. The results further show that data quality, despite being detected along the dimensions broken down for the field by Hazen et al. (2014), plays a subordinate role according to the current state of OSCM-research (Parssian et al. 2004). It is striking that although the work of Hazen shows a way to control data quality, it is limited to intrinsic dimensions and thus does not do justice to the richness of context in social media data, which in turn is a much-mentioned topic in the literature. Overall, intrinsic data quality is more frequently addressed in the literature as compared to contextual data quality. However, Choi et al. (2018) note that in OM literature, heuristics are developed in this context that match the problem. The results further show that contextual data quality tends to be neglected in the literature, despite domain knowledge being crucial (Waller and Fawcett 2013) for SMA. In line with Parssian et al. (2004), the quality of information derived remains unclear. In particular, the three most frequently mentioned categories accuracy, relevancy, and believability show that despite evolving models and methods (Abrahams et al. 2012; Cheng and Krumwiede 2018; Yuan et al. 2018), SMA-based decision quality remains unclear and therefore limited. While pretreatment and prioritization of data promise technical remedy (Swain and Cao 2019; Wilkin et al. 2020), the contextual dimensions of relevance and believability pose an overlying problem. The assessment of the relevance and believability of the respective social media data requires expert knowledge (Waller and Fawcett 2013) at the interface of SMA and OSCM. In line, Hardy et al. (2020) and Banerjee et al. (2020) argue to consider the human factor’s influence on the generation and interpretation of social media data. Therefore, it appears appropriate to develop orientation aids to show how and for what purpose individual social media data sources can be used and to examine prevailing patterns of interpretation in order to be able to take these into account in decision-making. Consequently, this study stresses that to mitigate SMA limitations, OSCM-research needs to shed light on contextual data quality for social media-complemented decision making.
7 Conclusion The purpose of this study was to examine how data quality is reflected in SMA literature for OSCPM. The first research question was ‘To what extent is the data quality of different social media data platforms considered for SMA in OSCPM-literature?’. Based on the findings of this study with quality dimensions thematized in 16% to 63% of the examined contributions, it can be stated that data quality is not addressed holistically
Data Quality in Social Media Analytics for Operations
113
in the literature on SMA in OSCM. The categorization of data quality into intrinsic and contextual enabled a differentiated view to answer the second research question ‘What are the data quality dimensions impeding SMA-enabled OSCPM-research?’. The intrinsic dimensions of SMA-accessible social media data are irregularly addressed, with data accuracy most frequently thematized. With regard to contextual data quality, relevance, and believability of social media data that are mainly thematized in research, which in turn together interfere with the insights gained from SMA. The findings of the study point to data quality aspects that require special attention when analyzing social media data for OSCPM. Thus, this study contributes to the understanding of SMA-aided OSCPM-research by examining it from a data quality perspective. The results can further suggest notions for scholars at the interface of OSCM, SMA, and Data Science and can provide initial orientation for the use of social media data in OSCM business practice. However, this study has certain limitations. Despite the applied methodology suitable for highlighting striking aspects of data quality, the results are limited by the underlying data sample. Also, the assessment of data quality within the contributions is strongly dependent on the purpose and method of analysis. Altogether, the findings of this study offer several avenues for the scientific community. On the one hand, there is a need for a more in-depth illustration of the intrinsic and contextual quality dimensions of social media data in order to further advance the work initiated here. On the other hand, there is a need for quantification of data quality, for example along the lines of Hazen (2014)’s model, to provide both researchers and practitioners with means to measure and objectively assess the underlying data quality of the information obtained from SMA. In doing so, future research in this direction will contribute to the understanding of SMA’s boundaries in OSCM.
References Abrahams, A.S., Fan, W., Wang, G.A., Zhang, Z., Jiao, J.: An integrated text analytic framework for product defect discovery. Prod. Oper. Manag. 24(6), 975–990 (2015) Abrahams, A.S., Jiao, J., Wang, G.A., Fan, W.: Vehicle defect discovery from social media. Decis. Support Syst. 54(1), 87–97 (2012) Albuquerque, F.C., et al.: A methodology for traffic-related Twitter messages interpretation. Comput. Ind. 78, 57–69 (2016) Arunachalam, D., Kumar, N., Kawalek, J.P.: Understanding big data analytics capabilities in supply chain management: unravelling the issues, challenges and implications for practice. Transp. Res. Part E Logist. Transp. Rev. 114, 416–436 (2018) Ashton, T., Evangelopoulos, N., Prybutok, V.R.: Quantitative quality control from qualitative data: control charts with latent semantic analysis. Qual. Quant. 49(3), 1081–1099 (2015). https://doi. org/10.1007/s11135-014-0036-5 Banerjee, A., Ries, J.M., Wiertz, C.: The impact of social media signals on supplier selection: insights from two experiments. Int. J. Oper. Prod. Manag. 40(5), 531–552 (2020) Baskerville, R.L., Myers, M.D., Yoo, Y: Digital first: the ontological reversal and new challenges for information systems research. MIS Q. Manag. Inf. Syst. 44(2), 509–523 (2020) Beheshti-Kashi, S.: Development of a social media process model for fashion and apparel supply chain decisions. Cyber-Phys. Syst. 6(2), 76–95 (2020)
114
F. Siekmann et al.
Bengtsson, M.: How to plan and perform a qualitative study using content analysis. NursingPlus Open 2, 8–14 (2016) Berelson, B.: Content Analysis in Communication Research. Free Press, New York (1952) Bhattacharjya, J., Ellison, A., Tripathi, S.: An exploration of logistics related customer service provision on Twitter: the case of e-retailers. Int. J. Phys. Distrib. Logist. Manag. 46(6/7) (2016) Booth, A., Papaioannou, D., Sutton, A.: Systematic Approaches to a Successful Literature Review, 2nd edn. In: Steele, M. (ed.). Sage Publications Limited, Los Angleses, London, New Dehli, Singapore, Washington DC (2012). https://doi.org/10.5596/c13-009 Boyd, D., Crawford, K.: Critical questions for big data: provocations for a cultural, technological, and scholarly phenomenon. Inf. Commun. Soc. 15(5), 662–679 (2012) Cai, L., Zhu, Y.: The challenges of data quality and data quality assessment in the big data era. Data Sci. J. 14, 1–10 (2015) Chae, B.: Insights from hashtag #supplychain and Twitter analytics: considering Twitter and Twitter data for supply chain practice and research. Int. J. Prod. Econ. 165, 247–259 (2015) Chen, X., Elmes, G., Ye, X., Chang, J.: Implementing a real-time Twitter-based system for resource dispatch in disaster management. GeoJournal 81(6), 863–873 (2016). https://doi.org/10.1007/ s10708-016-9745-8 Cheng, C.C.J., Krumwiede, D.: Enhancing the performance of supplier involvement in new product development: the enabling roles of social media and firm capabilities. Supply Chain Manag. 23(3), 171–187 (2018) Choi, T.M., Wallace, S.W., Wang, Y.: Big data analytics in operations management. Prod. Oper. Manag. 27(10), 1868–1883 (2018) Cottrill, C., Gault, P., Yeboah, G., Nelson, J.D., Anable, J., Budd, T.: Tweeting transit: an examination of social media strategies for transport information management during a large event. Transp. Res. Part C Emerg. Technol. 77, 421–432 (2017) Denyer, D., Tranfield, D.: Producing a systematic review. In: The SAGE Handbook of Organizational Research Methods, pp. 671–689 (2009) Dubey, R., Gunasekaran, A., Childe, S.J., Blome, C., Papadopoulos, T.: Big data and predictive analytics and manufacturing performance: integrating institutional theory, resource-based view and big data culture. Br. J. Manag. 30(2), 341–361 (2019) Ershadi, M.J., Niaki, S.T.A., Azizi, A., Esfahani, A.A., Abadi, R.E.: Monitoring data quality using hoteling T2 multivariate control chart. Commun. Stat. -Simul. Comput. (2021). https://doi.org/ 10.1080/03610918.2021.1887232 Gal-Tzur, A., Grant-Muller, S.M., Kuflik, T., Minkov, E., Nocera, S., Shoor, I.: The potential of social media in delivering transport policy goals. Transp. Policy 32, 115–123 (2014) Gandomi, A., Haider, M.: Beyond the hype: big data concepts, methods, and analytics. Int. J. Inf. Manage. 35(2), 137–144 (2015) Gkiotsalitis, K., Stathopoulos, A.: Joint leisure travel optimization with user-generated data via perceived utility maximization. Transp. Res. Part C Emerg. Technol. 68, 532–548 (2016) Gu, Y., Qian, Z., Chen, F.: From Twitter to detector: real-time traffic incident detection using social media data. Transp. Res. Part C Emerg. Technol. 67, 321–342 (2016) Hardy, C., Bhakoo, V., Maguire, S.: A new methodology for supply chain management: discourse analysis and its potential for theoretical advancement. J. Supply Chain Manag. 56(2), 19–35 (2020) Hartmann, P.M., Zaki, M., Feldmann, N., Neely, A.: Capturing value from big data – a taxonomy of data-driven business models used by start-up firms. Int. J. Oper. Prod. Manag. 36(10), 1382– 1406 (2016) Hazen, B.T., Boone, C.A., Ezell, J.D., Jones-Farmer, L.A.: Data quality for data science, predictive analytics, and big data in supply chain management: an introduction to the problem and suggestions for research and applications. Int. J. Prod. Econ. 154, 72–80 (2014)
Data Quality in Social Media Analytics for Operations
115
Hou, F., Li, B., Chong, A.Y.L., Yannopoulou, N., Liu, M.J.: Understanding and predicting what influence online product sales? A neural network approach. Prod. Plann. Control 28(11–12), 964–975 (2017) Hsieh, H.F., Shannon, S.E.: Three approaches to qualitative content analysis. Qual. Health Res. 15(9), 1277–1288 (2005) Huang, S., Potter, A., Eyers, D.: Social media in operations and supply chain management: stateof-the-art and research directions. Int. J. Prod. Res. 58(6), 1893–1925 (2019) Irani, Z., Sharif, A.M., Papadopoulos, T., Love, P.E.D.: Social media and Web 2.0 for knowledge sharing in product design. Prod. Plann. Control 28(13), 1047–1065 (2017) Jiang, C., Liu, Y., Ding, Y., Liang, K., Duan, R.: Capturing helpful reviews from social media for product quality improvement: a multi-class classification approach. Int. J. Prod. Res. 55(12), 3528–3541 (2017) Kamble, S.S., Gunasekaran, A.: Big data-driven supply chain performance measurement system: a review and framework for implementation. Int. J. Prod. Res. 58(1), 65–86 (2019) Kinra, A., Beheshti-Kashi, S., Buch, R., Nielsen, T.A.S., Pereira, F.: Examining the potential of textual big data analytics for public policy decision-making: a case study with driverless cars in Denmark. Transp. Policy 98(May), 68–78 (2020) Kinra, A., Hald, K.S., Mukkamala, R.R., Vatrapu, R.: An unstructured big data approach for country logistics performance assessment in global supply chains. Int. J. Oper. Prod. Manag. 40(4), 439–458 (2020) Krippendorff, K.: Content Analysis: An Introduction to its Methodology, 2nd edn. Sage Publications Limited, Thousand Oaks, CA (2004). https://doi.org/10.2307/2288384 Lee, S.-Y., Qiu, L., Whinston, A.: Sentiment manipulation in online platforms: an analysis of movie Tweets. Prod. Oper. Manag. 27(3), 393–416 (2017) Lin, K.P., Hung, K.C., Lin, Y.T., Hsieh, Y.H.: Green suppliers performance evaluation in belt and road using fuzzy weighted average with social media information. Sustainability (Switzerland) 10(1), 1–11 (2017) Matthias, O., Fouweather, I., Gregory, I.: Making sense of big data – can it transform operations management? Int. J. Oper. Prod. Manag. 37(1), 37–55 (2017) McAfee, A., Brynjolfsson, E.: Big data: the management revolution. Harv. Bus. Rev. 90(10), 60–66 (2012) Mentzer, J.T., Stank, T.P., Esper, T.L.: Supply chain management and its relationship to logistics, marketing, production, and operations management. J. Bus. Logist. 29(1), 31–46 (2008) Minnema, A., Bijmolt, T.H.A., Gensler, S., Wiesel, T.: To keep or not to keep: effects of online customer reviews on product returns. J. Retail. 92(3), 253–267 (2016) Mudambi, S.M., Schuff, D.: What makes a helpful online review? A study of customer reviews on Amazon.com. MIS Q. Manag. Inf. Syst. 34(1), 185–200 (2010) Neely, A. (ed.): Business Performance Measurement, 2nd edn. Cambridge University Press, New York (2007) Neely, A., Gregory, M., Platts, K.: Performance measurement system design: a literature review and research agenda. Int. J. Oper. Prod. Manag. 8(1), 1–14 (1995) O’leary, D.E.: The use of social media in the supply chain: survey and extensions. Intell. Syst. Account. Finance Manag. 18(2–3), 121–144 (2011) Papadopoulos, T., Gunasekaran, A., Dubey, R., Altay, N., Childe, S.J., Fosso-Wamba, S.: The role of big data in explaining disaster resilience in supply chains for sustainability. J. Clean. Prod. 142, 1108–1118 (2017) Parssian, A., Sarkar, S., Jacob, V.S.: Assessing data quality for information products: impact of selection, projection, and Cartesian product. Manage. Sci. 50(7), 967–982 (2004) Patton, D.U., Frey, W.R., McGregor, K.A., Lee, F.T., McKeown, K., Moss, E.: Contextual analysis of social media: the promise and challenge of eliciting context in social media posts with natural
116
F. Siekmann et al.
language processing. In: AIES 2020 - Proceedings of the AAAI/ACM Conference on AI, Ethics, and Society, February 2020 Paul, J., Criado, A.R.: The art of writing literature review: what do we know and what do we need to know? Int. Bus. Rev. 29(4), 101717 (2020) Ramanathan, U., Subramanian, N., Parrott, G.: Role of social media in retail network operations and marketing to enhance customer satisfaction. Int. J. Oper. Prod. Manag. 37(1), 105–123 (2017) Roden, S., Nucciarelli, A., Li, F., Graham, G.: Big data and the transformation of operations models: a framework and a new research agenda. Prod. Plann. Control 28(11–12), 929–944 (2017) Sahoo, N., Dellarocas, C., Srinivasan, S.: The impact of online product reviews on product returns. Inf. Syst. Res. 29(3), 723–738 (2018) Seyedghorban, Z., Tahernejad, H., Meriton, R., Graham, G.: Supply chain digitalization: past, present and future. Prod. Plan. Control. 31(2–3), 96–114 (2020). https://doi.org/10.1080/095 37287.2019.1631461. Taylor & Francis Sheng, J., Amankwah-Amoah, J., Wang, X.: A multidisciplinary perspective of big data in management research. Int. J. Prod. Econ. 19(1), 97–112 (2017) Sigala, M.: Customer involvement in sustainable supply chain management: a research framework and implications in tourism. Cornell Hosp. Q. 55(1), 76–88 (2014) Sodero, A.C., Rabinovich, E.: Demand and revenue management of deteriorating inventory on the internet: an empirical study of flash sales markets. J. Bus. Logist. 38(3), 170–183 (2017) Swain, A.K., Cao, R.Q.: Using sentiment analysis to improve supply chain intelligence. Inf. Syst. Front. 21(2), 469–484 (2019). https://doi.org/10.1007/s10796-017-9762-2 Taleb, I., Serhani, M.A., Bouhaddioui, C., Dssouli, R.: Big data quality framework: a holistic approach to continuous quality management. J. Big Data 8(1), 1–41 (2021). https://doi.org/10. 1186/s40537-021-00468-0 Tan, K.H., Zhan, Y.Z., Ji, G., Ye, F., Chang, C.: Harvesting big data to enhance supply chain innovation capabilities: an analytic infrastructure based on deduction graph. Int. J. Prod. Econ. 165, 223–233 (2015) Waller, M.A., Fawcett, S.E.: Data science, predictive analytics, and big data: a revolution that will transform supply chain design and management. J. Bus. Logist. 34(2), 77–84 (2013) Wamba, S.F., Akter, S., Kang, H., Bhattacharya, M., Upal, M.: The primer of social media analytics. J. Organ. End User Comput. 28(2), 1–12 (2016) Wamba, S.F., Gunasekaran, A., Akter, S., Ren, S.J.F., Dubey, R., Childe, S.J.: Big data analytics and firm performance: effects of dynamic capabilities. J. Bus. Res. 70, 356–365 (2016) Wenzel, R., Van Quaquebeke, N.: The double-edged sword of Big Data in organizational and management research: a review of opportunities and risks. Organ. Res. Meth. 21(3), 1–44 (2018) Wilkin, C., Ferreira, A., Rotaru, K., Gaerlan, L.R.: Big data prioritization in SCM decision-making: its role and performance implications. Int. J. Account. Inf. Syst. 38, 100470 (2020) Wood, L.C., Reiners, T., Srivastava, H.S.: Think exogenous to excel: alternative supply chain data to improve transparency and decisions. Int. J. Logist. Res. Appl. 20(5), 426–443 (2016) Yuan, H., Xu, W., Li, Q., Lau, R.: Topic sentiment mining for sales performance prediction in e-commerce. Ann. Oper. Res. 270(1–2), 553–576 (2018). https://doi.org/10.1007/s10479-0172421-7 Zavala, A., Ramirez-Marquez, J.E.: Visual analytics for identifying product disruptions and effects via social media. Int. J. Prod. Econ. 208, 544–559 (2019)
Managing Interfaces Between Smart Factories and Digital Supply Chains Bennet Zander1,2(B) , Kerstin Lange1 , and Hans-Dietrich Haasis2 1 Department of Maritime and Logistics Studies, Jade University of Applied Sciences, 26931
Elsfleth, Germany {bennet.zander,kerstin.lange}@jade-hs.de 2 University of Bremen, 28359 Bremen, Germany [email protected]
Abstract. The digital transformation of industry and the accompanying virtualization of supply chains are creating new physical, contractual and organizational boundaries between the physical and virtual world. These interfaces need to be analyzed in order to solve potential challenges at an early stage. The aim of this paper is to provide an overview of future interfaces between smart factories, systems and parties in the supply chain. Furthermore, it is fundamentally investigated how they are managed and communicated. The approach is based on a comprehensive literature review, in which the design preconditions of smart factories and adaptive environments are examined. Subsequently, the review results are analyzed using the three interface management steps definition, control and communication to identify their future controlling. The paper demonstrates how five technology network levels can make a significant contribution to successful operations by handling interfaces and defining responsibilities between transformed processes in the supply chain. Keywords: Smart factory · Digital supply chain · Interface management
1 Introduction The manufacturing industry is still undergoing a structural change as the result of digital transformation. Driven by the fourth industrial revolution, companies are investing in the development of smart factories with the aim of integrating digitalization into fabrication structures [36]. By creating an autonomous production and logistics environment in which machines, systems and vehicles communicate independently with each other, they hope to remove humans from manufacturing processes in order to operate more efficiently [38]. Studies show that in this course, a cumulative value creation potential of up to e1.25 trillion may be possible for Europe by 2025 [9]. The German economy, in particular, is affected by growth due to its high industrial share and could reach a rise of up to e425 billion in European value creation potential [36]. Globally, digitalization, automation and artificial intelligence (AI) are expected to add $13 trillion to the global GDP by 2030, as these innovative technologies create significant first-time business opportunities and reinvest productivity gains into economies [29]. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 M. Freitag et al. (Eds.): LDIC 2022, LNLO, pp. 117–129, 2022. https://doi.org/10.1007/978-3-031-05359-7_10
118
B. Zander et al.
However, restructuring and digitizing the industry takes time and is not possible without further precautions. One of the essential measures is to adapt interface management. Along the supply chain, interfaces inevitably and unavoidably occur in every association. They are where the most interesting things happen – and where most things go wrong [15]. Interfaces can be either defined as the physical, contractual or organizational boundary between two states of a medium or as the part of a system that is used for communication. They result from the principle of division of labor, in which activities within or between organizations are broken down and assigned to relatively autonomous organizational departments. The purpose is to use the specialization of the different subsystems for certain task areas [44]. To deal with the results of digital transformation, interface management must be fundamentally revised, as completely new adaptive manufacturing structures based on sensor technology are emerging, which are able to make independent decisions. Common interfaces in physical manufacturing have recently been complemented by systematic interfaces of digital twins. Every process or manufacturing step has to be made available digitally and in real-time so that the supply chain can be accessed virtually and adjusted at an early stage to avoid errors. Before smart factories are fully integrated into the market, it is important to analyze the changes in interfaces between them and current supply chains. Unless new interfaces are acknowledged and analyzed to understand how to handle them and how they are integrated into the entire system, serious interface problems may occur during the realization of smart factories. Thus, this paper aims to provide a first identification of new interfaces between smart factories and the parties in the supply chain respectively the digital value network. We also want to find out how the associated interface management transforms in future and what is decisive from now on. The outline of this paper is as follows. Section 2 describes the methodology which is applied in the further process. Section 3 presents related work of interface challenges and their management, the structure of smart factories, as well as new interfaces in adaptive environments by reviewing relevant literature. Section 4 relates the results identified in the literature review to provide an overview of interface management between smart factories and supply chains. Lastly, Sect. 5 discusses the results, illustrates the recommendations for future research and concludes the paper.
2 Methodology This paper is being carried out by the following four steps illustrated in Fig. 1. 1. Literature Review
Interface Challenges and Management
2. Literature Review
3. Compare & Pair
Design Principles of a Smart Factory &
Interface Challenges and Management
Interfaces in Adaptive Environments
Design Principles & Adaptive Environm.
Fig. 1. Research process.
4. Discussion
Managing Interfaces Between Smart Factories
119
First, a comprehensive literature review on interface challenges and management is conducted in order to describe the general procedure for identifying and controlling interfaces. Step number one is complemented by another literature review on the design principles of smart factories as well as interfaces in adaptive environments to find out what precautions are necessary for successful future interface management of digital supply chains. Third, the new interfaces of smart factories found in the literature are compared and paired with interface management methods to investigate how interfaces in digital supply chains can be closed. Subsequently, the results are discussed. Both literature reviews were performed according to [2] and were conducted using Scopus, Elsevier and EBSCO-host in July 2021 focusing on English and German literature from 2014 onwards. The following variety of keywords was used: • ( OR
OR ) AND • ( OR AND • ( OR OR )
OR
Number of publications
The joint consideration of categories one and two shows that interface management has been given little attention in the development of smart factories so far, as the focus has been on production technologies. Categories one and three show that, in contrast to smart factories, there has already been some research regarding the integration of digital twins. However, only in a few cases the publications relate to the manufacturing industry, but rather to the construction and transport sector. Only smart factories and digital twins have been deeply researched together, since digital representations are an essential basis for the construction of such a factory. The final combination of the three categories shows that there are hardly any cross-thematic works depicting interface management, smart factories and adaptive environments in one (see Fig. 2). The scientific relevance of this work results from this research gap, which is to be closed by emphasizing the future interface management of digital and adaptive environments.
3.600
95 45 Category 1+2
8 Category 1+3
Category 2+3
Category 1+2+3
Fig. 2. Number of relevant publications.
3 Related Work In the following sections, the results of the literature review are presented and the background for this paper is described to reach a common understanding. The sections are
120
B. Zander et al.
organized as follows. First, we discuss known interface challenges and management techniques. Second, we outline the structure and design principles of a smart factory. Third, we emphasize interfaces in adaptive production and logistics environments. 3.1 Interface Challenges and Management Challenges occurring across interfaces can have a massive impact on the entire organization. Small time delays in the process flow, not clearly coordinated mutual requirements and expectations in the cooperation can result in a deteriorated order processing, an increased lead time of orders, as well as increased inventories [5]. In digital transformation, the selection of a suitable data format for simulation and production and the determination and configuration of interfaces are both major challenges. Error-free communication is only possible if the system interfaces of the agents match [28]. The change from traditional manufacturing techniques to intelligent factories requires an accompanying adaptation of logistics throughout the supply chain [14]. In terms of smart factories and digital supply chains, there has merely been fundamental research addressing the future management of interfaces so far. [21] made a distinction between computational data, machine or hardware, network, software and user interfaces, which connect systems with mechanical, electrical or physical information. [20] developed an architectural framework on how digital transformation can increase customer value in the logistics of manufacturing companies. Further, [35] present an architecture of a 3D visualization system for shop floor management by defining the interfaces of a real-time visualization system based on sensors in radio frequency identification (RFID) chips connected to the Internet of Things (IoT). [4, 9] focus on interface information content that needs to be addressed at each level of system decomposition. All papers recognize the importance of new forms of collaboration, augmented reality, robotics or semi-autonomous processes requiring coordinated interfaces and consistently integrated data to be fully exploited. Nevertheless, there seems to be no work specifically related to their management. According to [31], “Interface management includes the activities of • defining, • controlling and • communicating the information needed to enable unrelated objects (including systems, services, equipment, software, and data) to co-function. Most new systems or services require external interfaces with other systems or services. All of these interfaces must be defined and controlled […].” It further attempts to avoid problems caused by interfaces and to ensure that processes run smoothly [8]. The approach of the three presented management techniques is also followed in Sect. 4 of this paper in order to clearly illustrate new digital interfaces along the supply chain of a smart factory.
Managing Interfaces Between Smart Factories
121
However, if things don’t run as planned, the following reasons are frequently cited as the nature of problems in interface management: “lack of communication and coordination between project participants”, “incomplete project plan”, “poor definition of interfaces”, “mismanagement of responsibilities”, or “misunderstanding of integration and fusion between project participants as system components” [10, 12]. In the end, poor interface management practice can lead to interfacial issues like design errors, mismatched parts, systems performance failures and construction conflicts [21]. Many of these complications are related to communication, coordination, and visualization issues that new digital connections between the real and digital worlds can solve. 3.2 Design Principles of a Smart Factory A smart factory can be defined as a flexible and adaptive production facility of Industry 4.0, which is based on self-organized units and in which machines independently coordinate manufacturing processes and driverless transport vehicles autonomously complete logistics orders [11]. It combines the use of AI, mechanical engineering and information technologies. The objects involved, such as machines, tools, load carriers and vehicles, form cyber-physical systems (CPS) and exchange information in real-time via the internet [19, 36, 48]. When a smart factory is implemented, the IoT is used to integrate underlying device resources, giving the manufacturing system sensing, connectivity and data integration capabilities [7]. This enables a dynamic production structure in which processes are no longer statically predefined, but are adapted depending on the order by restructuring objects and processes [32, 46]. The result is a flexible manufacturing system that uses a continuous stream of data from networked operating and production systems to learn and adapt to new requirements [43]. Smart factories possess over four characteristics that make them “smart”. First, devices and their sensors can self-organize, learn and maintain their own information to analyze their behavior and capabilities [41]. Second, they have interoperability and real-time control of the IoT. Interconnection between devices can improve coordination and make configuration protocols more flexible [40]. Third, they are integrated through the use of CPS and AI, giving workers control over manufacturing systems [26]. Fourth, different techniques facilitate human-machine integration, based on which computer, signal processing, animation, prediction, or simulation information of manufacturing can be retrieved to virtualize, analyze and affect manufacturing processes [41]. To complete the four characteristics, [17] present six design principles to help developers build new factories or make existing factories smart (see Table 1). In smart factories, significant amounts of data are generated by sensors, actuators, machine protocols or manufacturing processes, which need to be controlled as part of future interface management. However, due to their variable metrics, these cannot be used directly and need to be brought into a unified data language beforehand. In addition, the factory must evaluate and combine its internal data with external data of customers, products or the supply chain in service platforms [49].
122
B. Zander et al. Table 1. Design principles of a smart factory.
Design principle
Definition
Modularity
A high degree of modularity refers to the simple design and ability to couple and configure system components from different manufacturers so that modules can be added to the production line
Interoperability
This refers to the ability to seamlessly exchange technical information within system components and between parties in the supply chain
Decentralization
System elements make decisions independently without being subordinated to a controller. Sensors enable CPS to interact with their environment, allowing processes to be readjusted for each task
Virtualization
Virtualization is used to create a virtual factory environment that has the ability to monitor and simulate physical processes. This model can test design implementations and allows the digital design to be modified before being transferred to the physical system
Service orientation
In the future, companies are forced to also become service providers, as their products have reached competitive parity and are sold to customers at almost no profit. Products and services are now integrated
Real-time capability The system should have the adaptive capability to respond in a timely manner to changes in the environment and meet requirements with available resources by reconfiguring or cooperating with others
3.3 Interfaces in Adaptive Environments In Industry 4.0 digitization reaches a new level through innovative technologies, intelligent software tools and cloud-based communication. One of the most important trends is the digital twin, which is a virtual representation of an engineered object that is updated from real-time data and thus spans its lifecycle by using simulation and machine learning [16, 47]. It is inevitable for realizing the design principles of a smart factory. Smart factories and their digital value networks function as a representation of physical reality. Consequently, to bring traditional production processes and supply chains into the digital age, their CPS have to create a loop of physical and digital transfer [27]. This loop starts with the generated data and manages its evolution from the physical to the digital world. The information is collected and aggregated to create a digital fingerprint of the physical environment. In the inner loop, data is analyzed by new technology-enabled capabilities, such as AI or machine learning, resulting in meaningful insights. The gained insights then inform opportunities for machine-driven action in the physical world. With the help of automation technologies, physical actions can be triggered by digital instructions to achieve the desired outcome (see Fig. 3) [33, 42]. Suppliers, manufacturers, customers and service providers use, generate and share information with each other that lead to a multitude of challenges and opportunities. Current supply chains are in constant evolution which is driven by changes in the markets and emerging needs of the fourth industrial revolution era [6]. In the adaptive environments of Industry 4.0, interfaces are critical elements supporting the complex nature of future systems, which are becoming more distributed and
Managing Interfaces Between Smart Factories
Physical Value Creation
123
Digital Value Network
Fig. 3. Physical-digital circuit [42].
interconnected with other systems along the supply chain. Consequently, the connection between the real and digital world adds a new interface that was previously known but not yet considered under characteristics of interface management. In particular, the interface is composed of the three technologies CPS, IoT and digital twins, which in turn consist of smaller physical and organizational interface agents, such as sensors, actuators, control processing units and communication devices. These are the drivers as well as prerequisites for enabling the design principles of a smart factory. Their future interaction is fundamentally examined in the next chapter according to the specification of the three interface management steps definition, control and communication.
4 Managing Interfaces Between Smart Factories and Digital Supply Chains In the previous chapter, CPS, IoT and digital twins as the connection between the physical and digital world were identified as new computational interfaces. Since there have already been studies on already known interfaces (data, machine or hardware, network, software and user interfaces), these are not the focus of discussion. At this point, the aim is to emphasize which interface management provisions are necessary to ensure the information transfer between two states along the supply chain. 4.1 Definition [5] define the unique identification of a physical object and its current status as a basic prerequisite for linking them with its digital image. Only when an action can be clearly identified by sensors it is possible to locate and track it. Various documentation systems and standards have been established for marking an object with a number or character sequence [25]. In this context, [14] mention intelligent sensor technology, which is necessary to derive the data for the transformation of machines and systems to CPS. For the control of autonomy as well as for the planning, tracking and optimization of processes along the supply chain, not only the position of the object is of interest, but also other information such as temperature, pressure, degree of capacity utilization or energy consumption. In the area of in-plant logistics, the plant and machine network, consisting of conveyor and sorting technologies, driverless transport systems and industrial robots, have to be automated in a way that they can be seamlessly linked to the other levels of the technology networks (see Fig. 4) [14, 25]. The technology networks levels reflect the supply chain that need to be visualized for the next higher (management) level to influence the physical value creation by controlling
124
B. Zander et al. Cloud
Sensor
Digital Integration
Fig. 4. Technology network levels.
digital twins. The layers are necessary to create a digital representation of the supply chain that can automatically influence every action [30]. The integration of sensor information from the plant and machine network of upstream suppliers and downstream customers into the company’s own production and enterprise network is known as digital integration along the value chain. However, growing exchange of information between systems can also lead to an increased dependence of production and logistics on suitable sensor technology [14]. All levels of the technology network are surrounded by the cloud, which stores large volumes of data from all production and logistics processes and their digital twins in a decentralized manner and distributes it along the supply chain [22, 24]. This capacity is even more important when considering that sensors and actuators produce more data than humans and that they must always be connected to all machinery and equipment in order to enable autonomy [1, 45]. 4.2 Control The main interface challenge across the SC are media discontinuities and different data formats. Currently, both manufacturers and software providers try to create platforms as hubs for data integration and interface management. Examples are Axoom by Trumpf, Siemens Mindshere, the Logistics Interface by Jungheinrich, Predix by General Electric or the SAP Cloud Platform IoT [20]. In today’s practice, business processes are controlled by semi-autonomous enterprise resource planning systems (ERP) and visualized in digital form. These design softwares can simulate the behavior of a process, but not the lifecycle [39]. Consequently, this data should now be further transformed with product lifecycle (PLM) and customer relationship management (CRM) systems into a common system to close existing interfaces and manage increasing product complexity [3]. In the future, this can for example be achieved by configuration lifecycle management (CLM) systems. [34] describe CLM as a software for managing all of a company’s configuration models (e.g. sales, engineering, manufacturing) as well as associated data across all
Managing Interfaces Between Smart Factories
125
lifecycle phases of a product in an IoT platform. These systems provide an essential prerequisite for digital value networks, as a link is created between the supply chain and the digital lifecycle of a configurable product. Product requirements are already defined in marketing and product management and the configuration model is then developed in research and development and subsequently expanded to include the requirements of procurement, manufacturing, sales and after-sales management [34]. 4.3 Communication The key feature of digital supply chains is to enable data continuity and shared use of virtual models through standard interfaces, thus supporting simultaneous work of different disciplines. Furthermore, standard interfaces offer an easy transfer of data without complex conversion, which avoids data loss and increases quality [18]. Logistics is the focus of the changes proposed by Industry 4.0, as it represents the cross-sectional connection between machines, systems and processes and links external parties in the supply chain. In particular, the aim is to both vertically and horizontally link the five technology network levels (see Fig. 4) with the individual manufacturing steps along the supply chain through the subsystems of logistics to create an adaptive digital environment. Emerging interface problems resulting from the complexity of many different tasks and from the cooperation of different departments can thus be avoided [23]. Hence, through virtual organisations and standardized interfaces, professionally lean empowered suppliers get synchronized with manufacturers [37]. Figure 5 shows in simplified form where interfaces of a smart factory are located along the supply chain and how they are held together by higher-level networks. Depending on the level, each block represents a single sensor, machine, department or company. The higher the technology network level, the more processes it supervises. Technology Network Levels
Supply Chain Network Enterprise Network
Raw-material Supplier
Supplier
Manufacturer / Smart Factory
Distributor
Retailer
Customer
Production Network Plant- and Machine Network Sensor
Production Plan Supply Chain Process
Fig. 5. Interfaces between the smart factory and supply chain along the technology network levels.
At the lowest network level, it is the function to link individual sensors and actuators of the production program (e.g. machines, tools, load carriers and vehicles) through the IoT with each other (e.g. by using OPC Unified Architecture, Device-to-Device (D2D) communication or edge computing) to form plant and machine networks. There is complete autonomy among them. Edge computing or D2D communication can support
126
B. Zander et al.
control systems that ensure centralized data exchange between devices. Information from individual production steps can be bundled and processed vertically and horizontally to the next level. However, it is important to weight which information is made available, since even small differences can have serious consequences. The bundled information from the sensor level are then digitally integrated into the other levels with the help of the CLM system and linked to all production steps of the various parties along the supply chain. With this solution, it may be possible to realize intelligent factories according to their design requirements, which are able to react to any information from the supply chain by their own and close interfaces.
5 Discussion and Conclusion Admittedly, research on interfaces and their management seems in some aspects outdated, as the last definitive books were written years ago. Since then, it looks clear how to deal with challenges in this regard. However, the results of this paper show that digital transformation and introduction of smart factories are leading to innovations that have not yet been considered. Interfaces no longer exist only between contracts, systems or companies, but now also in all sensors that are connected through CPS to the IoT and generate a connection between the real and digital world as digital twins. Thus, interface management is changing from people, departments or companies to data management of sensors and machines. It is a prerequisite for enabling the design principles of a smart factory. This means that interfaces will soon be less noticed by the general public, but predominantly only by IT experts. Nevertheless, it is likely that supply chains will not be fully integrated in the near future, as end-to-end connectivity and visibility are still hampered by existing barriers in the areas of incompatible data formats, lack of interfaces between technological network levels, integration of ICT systems and IT security concerns. Therefore, further standardization attempts such as Plattform Industrie 4.0 [13] are desirable. The aim of this paper was to identify and fundamentally define interfaces between smart factories and the supply chain that are emerging as a result of digital transformation. As a result, a first conceptual recommendation for its management was made. The literature review indicated that smart factories and future supply chains are both based on digital twins, allowing different production scenarios to be simulated in advance in order to subsequently manage processes in real-time. The results of our study show that interfaces can be defined using technology network levels, controlled by CLM systems and communicated with the help of edge computing or D2D communication. Future research could focus on what respective technologies would need to be applied in each level to enable end-to-end visualization of processes and current states.
References 1. Alam, K.M., El Saddik, A.: C2PS: a digital twin architecture reference model for the cloudbased cyber-physical systems. IEEE Access 5, 2050–2062 (2017) 2. Baker, M.: Writing a literature review. Mark. Rev. 1(2), 219–247 (2000)
Managing Interfaces Between Smart Factories
127
3. Batchelor, J., Andersen, H.R.: Bridging the product configuration gap between PLM and ERP - an automotive case study. In: 19th International Product Development Management Conference, Manchester (2012) 4. Bianculli, D., Giannakopoulou, D., Pasareanu, C.S.: Interface decomposition for service compositions. In: 2011 33rd International Conference on Software Engineering (ICSE), Honolulu, pp. 501–510 (2011) 5. Bousonville, T.: Logistik 4.0: Die digitale Transformation der Wertschöpfungskette. Springer, Wiesbaden (2017). https://doi.org/10.1007/978-3-658-13013-8 6. Büyüközkan, G., Göçer, F.: Digital supply chain: literature review and a proposed framework for future research. Comput. Ind. 97, 157–177 (2018) 7. Chen, B., Wan, J., Shu, L., Li, P., Mukherjee, M., Yin, B.: Smart factory of Industry 4.0: key technologies, application case, and challenges. IEEE Access 6, 6505–6519 (2018) 8. Chen, Q.: An object model framework for interface management in building information models, Virginia (2007) 9. Daagi, M., Ouniy, A., Kessentini, M., Gammoudi, M.M., Bouktif, S.: Web service interface decomposition using formal concept analysis. In: 2017 IEEE International Conference on Web Services (ICWS), Honolulu, pp. 172–179 (2017) 10. Eray, E., Golzarpoor, B., Rayside, D., Haas, C.: An overview on integrating interface management and building information management systems. In: 6th CSCE-CRC International Construction Specialty Conference 2017 - Held as Part of the Canadian Society for Civil Engineering Annual Conference and General Meeting 2017, vol. 1, pp. 196–205 (2017) 11. Federal Ministry for Economic Affairs and Energy: FAQ. https://www.bmwi.de/Redaktion/ DE/FAQ/Industrie-40/faq-industrie-4-0-03.html. Accessed 30 July 2021 12. Federal Ministry for Economic Affairs and Energy: Industrie 4.0 und Digitale Wirtschaft. https://www.bmwi.de/Redaktion/DE/Publikationen/Industrie/industrie-4-0-unddigitale-wirtschaft.pdf%3F__blob%3DpublicationFile%26v%3D3. Accessed 16 Apr 2021 13. Federal Ministry for Economic Affairs and Energy: What is the Plattform Industrie 4.0? https:// www.plattform-i40.de/PI40/Navigation/EN/Home/home.html. Accessed 10 Aug 2021 14. Fürstenberg, K., Kirsch, C.: Intelligente Sensorik als Grundbaustein für cyber-physische Systeme in der Logistik. In: Vogel-Heuser, B., Bauernhansl, T., ten Hompel, M. (eds.) Handbuch Industrie 4.0 Bd.3, 2nd edn., pp. 271–297. Springer, Heidelberg (2017). https://doi.org/10. 1007/978-3-662-53251-5_83 15. Hause, M.: Systems interface management with MBSE: from theory to modeling to reality. In: 28th Annual INCOSE International Symposium, Washington (2018) 16. Herlyn, W., Zadek, H.: Mastering the supply chain by a concept of a digital control-twin. In: Data Science and Innovation in Supply Chain Management. Proceedings of the Hamburg International Conference of Logistics (HICL), Hamburg, vol. 30, pp. 661–698 (2020) 17. Hermann, M., Pentek, T., Otto, B.: Design principles for Industrie 4.0 scenarios: a literature review (Working Paper No. 01/2015). Technische Universität Dortmund (eds.) (2015) 18. Institut für Integrierte Produktion Hannover: Digitale Fabrik als Fabrik der Zukunft. https:// www.iph-hannover.de/de/information/fabrikplanung/digitale-fabrik/. Accessed 1 July 2021 19. Jacobi, H.-F.: Computer Integrated Manufacturing (CIM). In: Westkämper, E., Spath, D., Constantinescu, C., Lentes, J. (eds.) Digitale Produktion, pp. 51–92. Springer, Heidelberg (2013). https://doi.org/10.1007/978-3-642-20259-9_10 20. Junge, A.L.: Conceptualizing and capturing digital transformation’s customer value – a logistics and supply chain management perspective. Universitätsverlag der TU Berlin, Berlin (2020) 21. Keerthanaa, K., Shanmugapriya, S.: Role of interface management in construction industry. Int. Res. J. Eng. Technol. (IRJET) 4(2), 1217–1220 (2017)
128
B. Zander et al.
22. Klostermeier, R., Haag, S., Benlian, A.: Digitale Zwillinge – Eine explorative Fallstudie zur Untersuchung von Geschäftsmodellen. HMD Praxis der Wirtschaftsinformatik 55, 297–311 (2018) 23. Koch, S.: Logistik: Eine Einführung in Ökonomie und Nachhaltigkeit. Springer, Heidelberg (2012). https://doi.org/10.1007/978-3-642-15289-4 24. Krüger, J., et al.: Daten, Information und Wissen in Industrie 4.0. In: Reinhart, G. (eds.) Handbuch Industrie 4.0: Geschäftsmodelle, Prozesse, Technik, pp. 89–110. Carl Hanser, Munich (2017) 25. Lieberoth-Leden, C., Röschinger, M., Lechner, J., Günthner, W.A.: Logistik 4.0. In: Reinhart, G. (eds.) Handbuch Industrie 4.0: Geschäftsmodelle, Prozesse, Technik, pp. 451–512. Carl Hanser, Munich (2017) 26. Lu, Y.: Artificial intelligence: a survey on evolution, models, applications and future trends. J. Manag. Anal. 6(1), 1–29 (2019) 27. Lu, Y., Liu, C., Wang, K.I.-K., Huang, H., Xu, X.: Digital twin-driven smart manufacturing: connotation, reference model, applications and research issues. Robot. Comput. Integr. Manuf. 61, 101837 (2020) 28. Mareis, C.: Design als Wissenskultur. Transcript, Bielefeld (2014) 29. McKinsey: Twenty-five years of digitization: ten insights into how to play it right. https://www.mckinsey.com/~/media/mckinsey/business%20functions/mckinsey%20digital/ our%20insights/twenty-five%20years%20of%20digitization%20ten%20insights%20into% 20how%20to%20play%20it%20right/mgi-briefing-note-twenty-five-years-of-digitizationmay-2019.as. Accessed 15 Mar 2021 30. Mishra, S.: The OSI reference model: an overview. Int. J. Linguist. Comput. 1(3), 9–12 (2017) 31. Mitre: Systems Engineering Guide. https://www.mitre.org/publications/systems-eng ineering-guide/se-lifecycle-building-blocks/systems-integration/interface-management. Accessed 2 Aug 2021 32. Mosler, A.: Integrierte Unternehmensplanung: Anforderungen, Lösungen und Echtzeitsimulation im Rahmen von Industrie 4.0. Springer, Hamburg (2017). https://doi.org/10.1007/9783-658-08752-4 33. Mussomeli, A., Laaper, S., Gish, D.: The rise of the digital supply network. https://www2. deloitte.com/us/en/insights/focus/industry-4-0/digital-transformation-in-supply-chain.html. Accessed 15 Mar 2021 34. Myrodia, A., Randrup, T., Hvam, L.: Configuration lifecycle management-an assessment of the benefits based on maturity. In: 20th International Configuration Workshop, Graz (2018) 35. Qian, X., Tu, J., Lou, P.: A general architecture of a 3D visualization system for shop floor management. J. Intell. Manuf. 30(4), 1531–1545 (2017). https://doi.org/10.1007/s10845-0171335-1 36. Roland Berger Strategy Consultants: Die digitale Transformation der Industrie. https://bdi. eu/media/presse/publikationen/information-und-telekommunikation/Digitale_Transform ation.pdf. Accessed 25 Sept 2020 37. Sanders, A., Elangeswaran, C., Wulfsberg, J.P.: Industry 4.0 implies lean manufacturing: research activities in industry 4.0 function as enablers for lean manufacturing. J. Ind. Eng. Manag. 9(3), 811–833 (2016) 38. Scheer, A.-W.: Industrie 4.0: Von der Vision zur Implementierung. In: Obermaier, R. (ed.) Industrie 4.0 als unternehmerische Gestaltungsaufgabe, pp. 35–52. Springer, Wiesbaden (2016). https://doi.org/10.1007/978-3-658-08165-2_2 39. Schneider, B.: Unternehmenssoftware als Forschungsfeld ökonomischer Bildung. Springer, Wiesbaden (2017). https://doi.org/10.1007/978-3-658-19083-5 40. Shariatzadeh, N., Lundholm, T., Lindberg, L., Sivard, G.: Integration of digital factory with smart factory based on Internet of Things. Procedia CIRP 50, 512–517 (2016)
Managing Interfaces Between Smart Factories
129
41. Shi, Z., Xie, Y., Xue, W., Chen, Y., Fu, L., Xu, X.: Smart factory in Industry 4.0. Syst. Res. Behav. Sci. 37(4), 607–618 (2020) 42. Sinha, A., Bernardes, E., Calderon, R., Wuest, T.: Digital Supply Networks: Transform Your Supply Chain and Gain Competitive Advantage with Disruptive Technology and Reimagined Processes. McGraw-Hill Education, New York (2020) 43. Sjödin, D., Parida, V., Leksell, M., Petrovic, A.: Smart factory implementation and process innovation: a preliminary maturity model for leveraging digitalization in manufacturing. Res. Technol. Manag. 61(5), 22–31 (2018) 44. Specht, G.: Schnittstellenmanagement. In: Tietz, B., Köhler, R., Zentes, J. (eds.) Handwörterbuch des Marketing, pp. 2265–2275. Schäffer Poeschel, Stuttgart (2017) 45. Velasquez, N., Pesado, P.: Cloud computing, big data and the Industry 4.0 reference architectures. J. Comput. Sci. Technol. 18(03), 258–266 (2018) 46. Weyer, S., Schmitt, M., Ohmer, M., Gorecky, D.: Towards Industry 4.0 - Standardization as the crucial challenge for highly modular, multi-vendor production systems. IFAC-PapersOnLine 48(3), 579–584 (2015) 47. Wright, L., Davidson, S.: How to tell the difference between a model and a digital twin. Adv. Model. Simul. Eng. Sci. 7(1), 1–13 (2020). https://doi.org/10.1186/s40323-020-00147-4 48. Yao, X., Zhou, J., Lin, Y., Li, Y., Yu, H., Liu, Y.: Smart manufacturing based on cyber-physical systems and beyond. J. Intell. Manuf. 30(8), 2805–2817 (2017). https://doi.org/10.1007/s10 845-017-1384-5 49. Zander, B., Lange, K., Haasis, H. D.: Designing the data supply chain of a smart construction factory. In: Adapting to the Future. Proceedings of the Hamburg International Conference of Logistics (HICL), vol. 31, pp. 41–62 (2021)
Maritime Logistics and Port Operations
Container Flow Generation for Maritime Container Terminals Marvin Kastner(B)
, Ole Grasse , and Carlos Jahn
Institute of Maritime Logistics, Hamburg University of Technology, Hamburg, Germany [email protected]
Abstract. In maritime logistics, mathematical optimization and simulation are widely-used methods for solving planning problems and evaluating solutions. When putting these solutions to test, extensive and reliable data are urgently needed but constantly scarce. Since comprehensive real-life data are often not available or are classified as sensitive business data, synthetic data generation is a beneficial way to rectify this deficiency. Even institutions which already own comprehensive container flow data are dependent on synthetic data, due to the need to adapt and test their business models to uncertain future developments. A synthetic data generator that creates incoming and outgoing containers from the perspective of a maritime container terminal has already been proposed. However, since its publication more than 15 years have passed and the industry has changed. This justifies to rethink, rework, and improve the existing solution. This paper presents a synthetic container flow generator which allows the user to create synthetic but yet realistic data of container flows for maritime container terminals. After the introduction and motivation, this paper provides an overview about the state of the art of synthetic data generators. Then, the conceptual model of the generator is presented. Furthermore, an exemplary visual validation of the generated output data is shown. The paper closes with a discussion and outlook on planned future developments of the software. Keywords: Synthetic data · Data generation · Container terminal · Maritime logistics · Container
1 Introduction Over 80% of world merchandise trade is transported by sea, making the maritime supply chain a vital link between continents for the economy (UNCTAD 2020). In the last decades, containerized trade has outperformed other types in terms of growth, coinciding with an increasing average transport capacity of container vessels (UNCTAD 2020). This development affects the operations at container terminals as less but larger port calls put stress on terminal operations as well as the hinterland. The described situation does not affect all container terminals alike. Economic cycles, changing tariff policies, and events such as the current global pandemic change the transport demands between the ports in the world. In 2020, the number of weekly port calls, container services, and direct connections between ports decreased, and 35 ports lost their position in the network © The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 M. Freitag et al. (Eds.): LDIC 2022, LNLO, pp. 133–143, 2022. https://doi.org/10.1007/978-3-031-05359-7_11
134
M. Kastner et al.
of global liner shipping services (UNCTAD 2020). While the volume of transported containers seems to further grow in the future, not all ports (and all container terminals) will grow alike. For each port, factors such as the local economy and the hinterland connection to relevant economic regions affect the transport demands for import and/or export, while the position of the port in the continuously changing global transportation network might offer the opportunity to operate as a transshipment hub (e.g., Wang et al. 2019). When a new container terminal is planned or when an existing container terminal undergoes major adjustments (e.g., a conversion between equipment types or an aerial expansion), many uncertainties regarding the future transport demands and technological developments exist (Twrdy and Beskovnik 2008). The most common way to put a planned container terminal to test before the actual construction is by means of simulating the logistics processes taking place at the container terminal beforehand (Kastner et al. 2020). This allows to answer the question of whether the different functional areas of the container terminal are properly dimensioned so that no bottleneck throttles the performance of the future container terminal (see e.g. Kastner et al. 2021; Schwientek et al. 2020). A simulation study requires a simulation model of the planned container terminal and container flow data which correspond to the expected future transport demands. During such a planning stage, by definition no real-world operational data of that container terminal can exist so a substitute is sought for. A straight-forward approach is to synthetically generate data that meets the expectations of the planners regarding future operations. Thus, synthetic container flow data is crucial for the planning phase of container terminals. Furthermore, synthetic container flow data is also vital for academic research beyond the previously described planning problem. At times, real-world data might exist but it cannot be used for research because it is confidential business data. Even if the terminal operator shares some of the data with chosen scientists, restrictions might exist to publish the data or derived metrics. Under such circumstances, synthetic container flow data is useful as a substitute to run experiments, e.g. with simulation models or mathematical models, and to create new, publishable scientific insights.
2 State of the Art The topic of synthetic data generation has received some attention over the years in the scientific community. In literature, several multi-purpose synthetic data generators can be found. Mannino and Abouzied (2019), for example, propose a data generation tool called Synner. The tool enables users to generate synthetic multi-purpose datasets by processing declaratively set data properties like statistical distributions, domains, and relationships for each field. The tool has an integrated visualization feature to show the inherent randomness of statistical data (Mannino and Abouzied 2019). It is based on the free and universal data generator tool Mockaroo, a tool which allows to create test data in the browser. Mockaroo was originally created to assist software testing in clinical applications, but is not limited to this domain (Mockaroo LLC 2021). Another software example is the commercial software package Benerator, which includes multiple data generators for various types of datasets (rapiddweller 2021). However, the software aims to generate elementary test datasets without the option to create semantically connected scenarios.
Container Flow Generation for Maritime Container Terminals
135
Expósito-Izquierdo et al. (2012) develop an instance generator for the container stacking problem. This problem is commonly found in the yard of container terminals. The created instances vary in their degree of difficulty. The mathematical model that is presented alongside the generator uses the generated instances for optimization runs. The model aims to minimize the number of repositioning moves. Briskorn et al. (2019) develop a test data generator which can be used for simulating container handling processes of yard cranes. Their generic model creates test instances of crane scheduling problems as an input for simulation models or to test algorithms. Meisel and Bierwirth (2011) present an approach for the evaluation of quay crane scheduling models and algorithms. For this purpose, they create an instance benchmark generator, which aims to create test scenarios for demonstrating the potentials and for comparison of models that solve the quay crane scheduling problem. Related to container flow data generation, Bovenkerk (2005) presents a framework called SMILE+ where an economic input/output model translates economic scenarios on a regional level into forecasts of container flows. The approach considers freight generation, freight distribution, modal split, and traffic conversion and assignment. However, since this approach is based on economic relations and NUTS cells (French: nomenclature des unités territoriales statistiques, which translates to uniformed nomenclature of territorial statistics) it has a granularity which is much lower than for container flows based on individual container movements. Major work in the context of container flow generation was accomplished by Hartmann (2004). The developed tool generates scenarios for the entire sea port container terminals. The generator depicts container flows for the four transport modes deep sea vessels, feeders, trains, and trucks. The generator accepts various general inputs to shape the generation, e.g. the time horizon, number of containers, relations of container flows, and dwell time can be set. The tool works in four stages: in the first stage, the arriving deep sea vessels, feeders, and trains are scheduled according to the set number of containers. Trucks can only carry one container and are generated per container. In the second stage, the arrival dates and times for each deep sea vessel, feeder, train, and truck are generated according to previously set time distributions for each mode of transport. In the third stage, the container pickups are created. First, a transport mode which picks up the container is drawn. Then, a pickup day and time is defined by considering the given maximum dwell time and its distribution. For the determined day, a suitable means of transportation is picked. For containers picked up by trucks, the corresponding vehicle is created on the spot. In the fourth stage, individual properties are assigned to each generated container, e.g. weight, type, size or if it contains dangerous goods. The generated output can serve as an input for comprehensive simulation models as well as for the testing of algorithms. However, Hartmann decided to only share the concepts of the scenario generator which creates the need for each possible user to re-implement these concepts on their own.
3 Conceptual Model In this publication, the work of Hartmann (2004) is picked up and further elaborated. This is necessary because since 2004 the maritime supply chain has undergone significant
136
M. Kastner et al.
change. Between 2004 and 2020, the number of transported twenty-foot equivalent units (TEU) worldwide (excluding intraregional flows) increased by 75% from 80 million TEU to 140 million TEU (UNCTAD 2020), while the container vessel sizes increased tremendously (UNCTAD 2019). In the beginning of 2004, the largest container vessel had a capacity of 8,063 TEU, while today vessels with a capacity beyond 15,000 TEU are frequently seen (UNCTAD 2020), the largest ones having a capacity of nearly 24,000 TEU (Chambers 2021). The largest category of vessels is called ultra-large container vessel (ULCV) and once such a vessel calls a port, this creates major peaks at the destined container terminal. Compared with 2004, this leads to a less steady container flow from and to the hinterland – a detail that should be reflected in the generated synthetic data. As the generator of Hartmann instantiates all arriving vehicles independent from each other, the departure of an ULCV cannot create a peak of trucks that deliver export containers as it is seen in reality. The second issue to elaborate on is the distribution of vessel arrival times. As simulation often targets at specific questions (e.g., “could the container terminal handle two ULCVs in a short time window?” or “would the yard capacity be sufficient for a given set of services and train schedules?”), simply providing a distribution to generate vessels might be unsatisfactory. Instead, the arrival times of the vessels are determined by the experimenter beforehand, e.g. by reading in sailing lists and manipulating the arrival times according to their research question. These points, as well as the fact that no code was shared by Hartmann, motivated to rethink the concept of that generator, implement it, and share the code (see Kastner and Grasse 2021). In the following, first the reader is presented with the required input data and generated output data of the newly developed generator. Then, the interaction between the user and the data generator is described. In the last step of this chapter, the underlying data generation process is explained. At the outset, the necessary input data and the generated output data is presented. The corresponding input-output diagram is depicted in Fig. 1. On the left, the required input data is grouped into three sections: first, container services are added, e.g. based on the information published by the shipping companies. In case a sailing list is read in, the frequency of port calls is neglected. Second, container property distributions for its length, weight, or type are determined. For export and transshipment containers, in addition the next destination is added. In case the synthetic data is later used to examine container stacking procedures, the destination supports to identify container groups (i.e., containers with the same destination might be kept in the same bay in the same yard block). Third, certain container flow properties are set. This includes the time range of the synthetically generated data, the vehicle-type-dependent modal split (i.e., given the vehicle type the container is delivered with, how frequently a container is picked up by a vehicle of a specific vehicle type), and the dwell time properties of import, export, and transshipment containers. The output of the data generation tool is stored in several tables: for each of the vehicle types, an independent table exists. In that table, each row represents one vehicle including its static and journey-specific properties. In the last table, all containers including their static properties are described. In addition, for each container two vehicle ids are stored, the id of the vehicle that delivers the container and the id of the vehicle that picks up the container. Thus, the journey-specific properties of the container need to be retrieved from these two vehicles.
Container Flow Generation for Maritime Container Terminals
137
Input
Landside
Seaside
Vehicle type Deep sea vessel Feeder vessel Barge
Output Moving according to schedule Every k days Number moved containers Probability of delay Probability of cancellaƟon
Train Truck
Arriving on demand for vessel (both inbound and outbound)
Container properƟes (picked according to distribuƟon): Lengths: 20’, 40’, 45’, and other Weights: 2 to 30 tons in steps of 2 Type: standard, empty, reefer, and dangerous goods Next desƟnaƟon (depending on schedule) Container flow properƟes: First and last day of the flow data Vehicle type dependent modal split Minimum and maximum dwell Ɵme for containers
List of containers, each having these properƟes: Weight Length Type Delivered by vehicle id Picked up by vehicle id Next desƟnaƟon List of vehicles, each having these properƟes (excerpt): Vehicle id Announced arrival Ɵme Realized arrival Ɵme (or arrival cancelled) Moved capacity (in TEU) Total capacity (in TEU)
Fig. 1. Input output diagram
For the user, two options exist to set the input data, trigger the generation process, and export the container flow data for further usage. The first option is an application programming interface (API) written in Python. The exported container flow data is represented as a set of tables. Each table can be saved either as a comma-separated value file or as an Excel file. The API is clearly separated from the internal code which is not intended to be called directly. The second option is a graphical user interface (GUI) for which one exemplary dialog window is shown in Fig. 2. At the time of writing, the implementation of the GUI is still ongoing. In both the API and the GUI, the user mainly provides distributions, minimum and maximum values, and similar. For a quick start, default values have been provided that are meant to be overwritten with the use-case specific assumptions by the user. The container weight distribution has been taken from MacGregor (2016). The maximum dwell times are taken from HHLA (2021a) assuming that all members of the supply chain avoid additional storage charges. The generator further introduces the concept of a minimum dwell time that resembles the cut-off of the shipping companies as well as a safety buffer between delivery and pickup of a container. The origin and destination mode of transport of a container are determined by the transshipment model of the Institute of Shipping Economics and Logistics (2015) with an update on the modal split in hinterland traffic (Hafen Hamburg Marketing e.V. 2021). The truck arrival distribution as well as the distribution of container lengths have been determined by expert interviews for the port of Hamburg. Following Hartmann (2004), each vehicle accepts 20% more containers on its outbound journey compared to the inbound journey as long as this would not exceed the total capacity of the vehicle. For both the Python API and the GUI, the general workflow is the same and is depicted in Fig. 3. During the entire process, intermediate information and other meta
138
M. Kastner et al.
Fig. 2. Mockup of user interface
data is stored in a relational database. Thus, as the first step a database must be selected. To keep the installation procedure of this software simple, the presented implementation uses an SQLite database while throughout the project the object-relational mapper peewee (Leifer 2021) has been used. This abstraction layer allows an easy shift to e.g. PostgreSQL or MySQL. In the second step, the input data is added, modified and/or deleted either via the API or via the GUI. By triggering the generation process (in Fig. 2 that corresponds to clicking the button “Save and generate container flow”), a seven-step procedure is started. During this process, extensive logging, visualizations and statistical reports inform the user about the generated data as well as the progress. The seven-step procedure depicted in Fig. 3 starts with creating deep sea vessels, feeders, barges, and trains based on the schedules. At this stage, vehicle properties (both static and journey-dependent) are set. If applicable, they are drawn from random distributions. For each of these vehicles, in the next step the containers which this vehicle delivers to the terminal are generated. At this point, also the vehicle type of the vehicle that eventually picks up the container is determined (but not yet the actual vehicle). In the third step, for each container the vehicle that picks up the container is selected. It must satisfy the intended vehicle type chosen in the previous step as well as the container dwell time restrictions. For the random vehicle selection, each vehicle in the permissible time range is assigned the probability linear to the number of empty TEU slots on that vehicle. Compared to Hartmann, this approach is simplified as no distribution for the intended container dwell time is assumed. While the new approach preserves the vehicle type at the costs of distorted container dwell time distributions, Hartmann restarts the assignment process for a given container if at the selected day (based on the container
Container Flow Generation for Maritime Container Terminals
139
dwell time distribution) no vehicle is available, leading to longer execution times. This restart, however, does not guarantee that the intended container dwell time distribution is actually approximated in the empirical distribution of the output data because some of the drawn values are repeatedly discarded. If no vehicle of the specified vehicle type with sufficient capacity is left, for the container a vehicle of a different vehicle type is selected and a flag is set at the container. As the data is randomly generated, this re-assignment is acceptable to happen once in a while. But if it frequently happens, it means that some of the input parameters are not yet properly tuned and require re-adjustment. In the fourth step, trucks are generated that pick up the containers delivered by the vehicles generated in the first step. This way, after the arrival of a ULCV its discharged containers are quickly distributed to the hinterland. Lastly, the containers delivered by trucks are placed on the vehicles listed in the vehicle-type-dependent modal split. Typically, these are the export containers that go from a truck to a vessel. In case of an arrival of a ULCV, this ensures that the free container slots are filled before the vessel leaves the terminal. In the next step, the trucks are generated that deliver these containers. Their arrival time at the terminal is drawn from the truck arrival distribution and always obey the maximum container dwell times. Lastly, for each container placed on a vessel, barge, or train, the next destination for each of the containers is chosen. The last step is optional and if no destination is selected, this entry remains empty. Choose SQLite Database for persistence (also keeps meta data)
Add, modify and delete input data
Generate flow data
Export flow data (e.g., without schedules and distribuƟons)
Create all vehicles within Ɵme range according to schedule
Create containers which are delivered by these vehicles
Assign next vehicle for each of these containers
Generate trucks to pick up containers
Generate trucks to deliver containers
Allocate containers that are delivered by trucks on available vehicles
For vehicles moving according to schedule that transport container groups: Choose next desƟnaƟon for each container
Fig. 3. Flowchart of execution
4 Visual Validation In this chapter, some of the generated data properties are highlighted by exemplary data based on the default values presented in the previous section if not indicated otherwise. In the first case of validation, the number of containers a vessel discharges is compared to the number it loads. This serves as evidence for a balanced assignment of containers to outbound journeys. Second, the temporal dependency between truck arrivals and vessel arrivals is examined. The visualized data demonstrates that the generator is capable of generating peaks at the truck gate when deep sea vessels load many containers.
140
M. Kastner et al.
With the presented generation procedure, intended prominent container flows (e.g., transshipment or import/export) are well preserved. By creating the containers for the vehicles that arrive at the container terminal and determining the vehicle type for pickup, the modal split is well approximated. This approach, however, might lead to unbalanced container flows, i.e. a certain vehicle type might deliver more containers than it picks up (or vice versa). While such imbalances also exist in real life, it should be subject to validation. One such imbalance in synthetic data is presented in Fig. 4. Each of the subplots follow the same scheme: each data point represents a vessel and if a vessel lies on the light gray line (ratio of 1:1), the vessel discharges the same number of containers as it loads. If it lies below the gray line, more containers are discharged from the vessel than loaded. No vessel can exceed the ratio of 1:1.2 due to the user-defined constraint. Subplot (b) shows the default mode of transport distribution as presented in Sect. 3. For subplot (a), 10% points are subtracted from the feeder traffic and 10 are added to the deep sea vessels before creating the synthetic data. For subplot (c), the reverse shift in traffic is executed. Publicly accessible information for Container Terminal Altenwerder in July 2021 are used (e.g., HHLA 2021b). When comparing the three subplots of Fig. 4, it turns apparent that the vessel arrivals and the origin and destination mode of transport need to match so that the inbound and outbound container flows are balanced. If containers that arrive from the hinterland are set to be picked up by deep sea vessels more often – as in subplot (a) – then the deep sea vessels load more containers on their outbound journey than on their inbound journey which is capped at the ratio of 1:1.2. The reverse is true for feeders in subplot (c). This is a warning that the transport capacities of a vehicle are completely consumed.
Fig. 4. Ratio of containers delivered and picked up by vessel type
In subplot (b) both feeders and deep sea vessels mostly show a ratio of 1:1 of containers that are discharged and loaded. This indicates that the origin and destination mode of transport distribution, the assumed vehicle capacities, and the vehicle arrival patterns do match – for the vessels, the inbound and outbound journey are in balance. Lastly, one obstacle with synthetic container flow data is the beginning and the end of the considered time range. One vehicle must be the first to arrive at the container terminal which is completely empty. One vessel must arrive first and it cannot pick up containers that have been delivered by a previous vessel as there was none. This warm up period
Container Flow Generation for Maritime Container Terminals
141
can be detected in Fig. 5 in the end of June and beginning of July due to the peak of truck arrivals, i.e. in the beginning the vessels are filled with containers delivered by trucks. During the month, the dips to zero truck arrivals per hour are frequent. They indicate Sundays – the day of the week when transportation by truck is highly restricted by law in Germany. Similarly, during the night the truck gate is open but as many industries do not operate 24/7 and truck drivers have restricted driving time for each day, only a few trucks arrive with containers during this time period. In the end of the time range, a reverse effect is seen: as transshipment containers must be assigned to one of the last leaving vessels, these vessels have less free TEU slots for containers delivered by truck. This behavior is not explicitly coded but is an effect of the previously described generation procedure. It further requires the user of the synthetic data to be careful. Most likely, removing these phases from further analysis is expedient.
Fig. 5. Relationship between vessel arrivals and truck deliveries
5 Discussion and Outlook In the previous two sections, a procedure to generate synthetic container flow data has been presented both on a conceptual level and with first practical insights in the shape of visualizations. Unlike Mockaroo or Benerator, this is a special-purpose synthetic data generator which only allows restricted user input as the domain of the container terminal is already extensively modelled in code. Part of this domain model are the relationships between the objects, such as “container is delivered by X”, “container is picked up by Y”, or “vessel moves according to sailing schedule Z” which make it necessary to properly model these relationships. Unlike the problem instance generators described by Expósito-Izquierdo et al. (2012) or Meisel and Bierwirth (2011), the presented synthetic data generator generates container flow data on a container terminal level. The output data contains the vehicles and corresponding containers. A vehicle either delivers or picks up a container. This is conceptually different from synthetic data which only covers one of the functional areas of the terminal as in the previous two cases. Furthermore, in the first timestep of the container flow data, the first container enters an empty terminal and in the last time step
142
M. Kastner et al.
the last container leaves the container terminal, leaving it empty again. As the warm-up and cool-down phase (each of them lasting for several days) might not suit to the research question at hand, they might need to be removed later from further analysis. Thus, this synthetic data is applicable to research questions regarding larger-scale interactions of different areas of the container terminal over a longer time period (e.g., one or several weeks). In the output data of the generator, the containers are positioned on vehicles without an explicit representation of the arrangement of these containers. Related problems such as stowage planning are not considered. Unlike Hartmann’s generator, the order of the containers in the database or the exported files (comma-separated value or Excel files) does not imply that they must be discharged or loaded in that sequence. In the future, further extensions of the synthetic container flow data generator are planned. For import containers, the information regarding the truck arrival time is frequently updated as many container terminals nowadays use truck appointment systems (see e.g. Lange et al. 2018). This arrival information is vital for an optimized housekeeping and should thus be part of the synthetic data. Similarly, the arrival time information for vessels, barges, and trains change over time (a port call might even be cancelled) and more realistic synthetic data would profit from covering those cases. Furthermore, a related but separated tool could be used to generate schedules for vessels, barges, and trains alike to reach a specific terminal capacity and not harming restrictions such as quay length, berth layout, number of rail tracks, and similar. If the reader needs any kind of functionality which is currently not supported, they are invited to fork the project which is open source and available at (Kastner and Grasse 2021).
References UNCTAD: Review of Maritime Transport 2020. Shipping in times of the Covid-19 pandemic, Geneva (2020) Wang, P., Mileski, J.P., Zeng, Q.: Alignments between strategic content and process structure: the case of container terminal service process automation. Marit. Econ. Logist. 21, 543–558 (2019). https://doi.org/10.1057/s41278-017-0070-z Twrdy, E., Beskovnik, B.: Planning and decision-making to increase productivity on a maritime container terminal. PROMET 20, 335–341 (2008) Kastner, M., Lange, A.-K., Jahn, C.: Expansion planning at container terminals. In: Freitag, M., Haasis, H.-D., Kotzab, H., Pannek, J. (eds.) LDIC 2020. LNL, pp. 114–123. Springer, Cham (2020). https://doi.org/10.1007/978-3-030-44783-0_11 Kastner, M., Nellen, N., Schwientek, A., Jahn, C.: Integrated simulation-based optimization of operational decisions at container terminals. Algorithms 14, 42 (2021). https://doi.org/10.3390/ a14020042 Schwientek, A.K., Lange, A.-K., Jahn, C.: Effects of terminal size, yard block assignment, and dispatching methods on container terminal performance. In: Bae, K.-H., et al. (eds.) Proceedings of the Winter Simulation Conference 2020, pp. 1408–1419 (2020) Mannino, M., Abouzied, A.: Is this real? Generating synthetic data that looks real. In: Guimbretière, F., Bernstein, M., Reinecke, K. (eds.) Proceedings of the 32nd Annual ACM Symposium on User Interface Software and Technology, pp. 549–561. ACM, New York (2019). https://doi. org/10.1145/3332165.3347866 Mockaroo LLC: Mockaroo (2021). https://www.mockaroo.com/. Accessed 30 Aug 2021
Container Flow Generation for Maritime Container Terminals
143
rapiddweller: Benerator. The smart way. rapiddweller GmbH, Hamburg (2021). https://www.ben erator.de/. Accessed 30 Aug 2021 Expósito-Izquierdo, C., Melián-Batista, B., Moreno-Vega, M.: Pre-marshalling problem: heuristic solution method and instances generator. Expert Syst. Appl. 39, 8337–8349 (2012). https://doi. org/10.1016/j.eswa.2012.01.187 Briskorn, D., Jaehn, F., Wiehl, A.: A generator for test instances of scheduling problems concerning cranes in transshipment terminals. OR Spectrum 41(1), 45–69 (2019). https://doi.org/10.1007/ s00291-018-0529-z Meisel, F., Bierwirth, C.: A unified approach for the evaluation of quay crane scheduling models and algorithms. Comput. Oper. Res. 38, 683–693 (2011). https://doi.org/10.1016/j.cor.2010. 08.001 Bovenkerk, M.: SMILE+: the new and improved Dutch national freight model system. In: Proceedings of ETC 2005, Strasbourg, France, 18–20 September 2005 - Transport Policy and Operations - Freight and Logistics - Freight Modelling I. Strasbourg, France (2005) Hartmann, S.: Generating scenarios for simulation and optimization of container terminal logistics. OR Spectrum 26, 171–192 (2004). https://doi.org/10.1007/s00291-003-0150-6 UNCTAD: Review of Maritime Transport 2019. Sustainable Shipping, Geneva (2019) Chambers, S.: Evergreen takes the biggest boxship mantle away from HMM by just 28 TEU (2021). https://splash247.com/evergreen-takes-the-biggest-boxship-mantle-away-from-hmmby-just-28-teu/ Kastner, M., Grasse, O.: ConFlowGen (2021). https://www.github.com/1kastner/conflowgen. Accessed 1 Dec 2021 MacGregor: Container securing systems. Product catalogue (2016). https://www.macgregor.com/ globalassets/picturepark/imported-assets/65120.pdf/. Accessed 30 Nov 2021 Institute of Shipping Economics and Logistics: Prognose des Umschlagpotenzials und des Modal Splits des Hamburger Hafens für die Jahre 2020, 2025 und 2030. Band 1: Umschlagpotenzialprognose (2015). https://www.hamburg-port-authority.de/fileadmin/user_upload/End bericht_Potenzialprognose_Mai2015_5.pdf/. Accessed 30 Nov 2021 Hafen Hamburg Marketing e.V.: Modal Split in hinterland traffic 2020 (2021). https://www.hafenhamburg.de/en/statistics/modal-split/. Accessed 30 Nov 2021 HHLA: Quay Tariff, September 2021a. https://hhla.de/fileadmin/download/HHLA_Kaitarif_2 021a_09_01_en.pdf. Accessed 30 Nov 2021a HHLA: Report Vessel Operations: Sailing list (2021b). https://coast.hhla.de/segelliste. Accessed 1 Dec 2021b Leifer, C.: Peewee (2021). https://github.com/coleifer/peewee. Accessed 30 Nov 2021 Lange, A.-K., Branding, F., Schwenzow, T., Zlotos, C., Schwientek, A.K., Jahn, C.: Dispatching strategies of drayage trucks at seaport container terminals with truck appointment system. In: Freitag, M., Kotzab, H., Pannek, J. (eds.) LDIC 2018. LNL, pp. 162–166. Springer, Cham (2018). https://doi.org/10.1007/978-3-319-74225-0_21
Simulation-Based Port Storage Dimensioning to Mitigate Operational Instability Yuri Triska(B)
and Enzo Morosini Frazzon
Industrial and System Engineering Department, Federal University of Santa Catarina, Florianópolis, Brazil [email protected], [email protected]
Abstract. Port storage is an important resource in port operations, complex to manage and increasingly scarce. Considering the great disturbance in port operations due to lack of storage, this paper aims to propose a simulation-based approach for dimensioning the storage of port terminals. Discrete event simulation (DES) was applied in a test case of a container terminal, whose scope contains truck gates, container yard and ship berths. First, operational stability regimes were identified, which are dependent on the availability of storage static capacity. This availability, instead of directly affecting mean operational indicators, is related to the probability of storage shortage over time. Later, an approach for dimensioning port storage by means of simulation was proposed, which is based on recurrence time for a maximum required storage in the year. This approach is oriented to statistics and generalizable for port terminals with different characteristics. The results of this work may assist port managers to plan terminals with different cargo types, providing greater assertiveness of investment, minimizing the occurrence of congestion and hence improving port performance. Keywords: Transportation infrastructure · Port expansion planning · Port capacity
1 Introduction Ports and terminals are logistic nodes whose function, among others, is to change cargo transport mode (Ligteringen and Velsink 2012). Their importance is related to the low cost of the waterborne mode compared to the others for transporting high volumes of cargo over long distances. The infrastructure of a single port has the potential to affect several supply chains around the world, and consequently the economic and social development of several countries. Research on port terminal yard operations has received attention from the scientific community, as shown in some literature reviews (Steenken et al. 2004; Stahlbock and Voß 2008; Carlo et al. 2014). The efficiency of operations in port terminal yards is considered a competitive advantage because it has an effect on the overall performance of the terminal (Chen et al. 2003). Steenken et al. (2004) point out storage as an increasingly scarce resource in container terminals. Alcalde et al. (2015) state that the storage yard is the © The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 M. Freitag et al. (Eds.): LDIC 2022, LNLO, pp. 144–155, 2022. https://doi.org/10.1007/978-3-031-05359-7_12
Simulation-Based Port Storage Dimensioning
145
most complex element of container terminals, considering the different resources (e.g. dock cranes, yard cranes) directly affected or demanded. When it comes to port planning, the throughput capacity is one of the main pieces of information that a manager has for planning and controlling operations (Olba et al. 2017), since it allows assessing if the available assets, as well as the way they are used, enable the satisfactory handling of expected demand. In this sense, Triska et al. (2020) proposed a method for assessing capacity and planning port expansion through simulation. The logic behind the proposed method is to verify the influence of the availability of resources on operational indicators and financial costs. The authors found that storage availability does not directly affect these values, but rather the system stability, so in the test case the method was adapted to plan the expansion of this specific resource. However, the article did not detail how this was done, nor did it explore this definition of system stability, which is a research opportunity to be explored in this article. That said, this paper aims to propose a simulation approach for dimensioning port terminal storage. The following research question is raised: • How can the required static capacity of a port terminal be evaluated to prevent operational instability caused by insufficient storage? First, the work contributes with the identification of operational stability regimes for port terminals related to the availability of static storage capacity. As it is discussed later in the paper, instead of directly contributing to the improvement of the terminal’s operational indicators, storage availability is more related to coping with the probability of storage shortage during the planning horizon, which can generate great disturbance on terminal operations. Secondly, it contributes with a proposition of a simulation-based approach for dimensioning port storage, that is oriented to statistics, less dependent on practical rules, generalizable to port terminals with different characteristics and brings the advantages of computational simulation. The rest of the article is structured as follows. Section 2 shows the review of the literature that assesses or dimensions port storage capacity. Section 3 displays the methodological approach, describing the developed simulation model and specifics about the simulation experiment. Section 4 is dedicated to the results, where the effect of storage shortage on the operational instability of a port terminal is analysed in a simulation model; an approach to dimensioning port storage considering this behaviour is presented; and such approach is discussed in the light of the literature. The fifth and last section is reserved for conclusions, resuming the main contributions of the work, limitations of the study and suggestions for future work.
2 Literature Review Steenken et al. (2004) reviewed the literature on operations research in container terminals. Several problems were analysed and revised, including berth allocation problem, quay crane scheduling and stowage plan. In relation to storage operations, it was observed that the focus of the literature is usually on the allocation of storage slots for containers that demand their use. This same characteristic was observed in the literature
146
Y. Triska and E. M. Frazzon
review updated by the authors (Stahlbock and Voß 2008) and also by Carlo et al. (2014), who reviewed the literature on container terminal yard operations. Other problems that are often addressed include the allocation of yard equipment and container reshuffling decisions (Carlo et al. 2014). In this literature review, we show studies that consider the storage capacity as a problem to be decided, according to the classification proposed by Carlo et al. (2014) for problems in storage yard operations. We classify the studies in three categories: practical rules, analytical approaches and simulation approaches. The practical rules are generally simple equations based on empirical values that are used to evaluate the capacity or to dimension the storage of a terminal. The formulae presented by UNCTAD (1985), PIANC (2014) and Chu and Huang (2005) are based on average operational indicators and safety factors. China (Ding 2010) and Brazil (Brazil 2020) used them for national planning practical rules to determine port storage capacity. PIANC (2014) and the Brazilian government (Brazil 2020) also used the comparison between static capacity and maximum expected lot size as a possible criterion for verifying that the demand for storage is met. Ligteringen and Velsink (2012) showed in their book the coefficient of 0.6 to 1.0 TEU/year/m2 for dimensioning the container terminal yard. Analytical approaches require mathematical modelling and are prevalent in the literature. Taleb-Ibrahimi et al. (1993) considered that the required storage size is greater than the maximum number of containers simultaneously in the yard, so that extra handling of cargo in storage is minimized. This difference depends on the pattern of container arrival at the terminal. Kim (1998) evaluated the optimal amount of storage of import containers in a port terminal considering deterministic and constant demand. The problem also optimizes the number of yard cranes for import containers. Kim (2002) developed a similar approach to Kim (1998), proposing several cost models for determining the required storage and number of transfer cranes in import container yards. Lee and Kim (2010) proposed an analytical model to determine optimal block dimensions considering costs related to the cycle time of yard cranes and the cost of ground space for the block, exploring four optimization models. Do Ngoc and Moon (2011) developed a tool for planning the expansion of container depots, through mathematical formulation based on mixed-integer programming. Alcalde et al. (2015) developed an integrated yard planning to determine optimal storage utilization considering the effects of yard congestion on terminal’s performance. The formulation involves stochastic variables, and the approach considered peak values over the years and recurrence time. All these analytical studies, except Taleb-Ibrahimi et al. (1993), considered the objective of the problem as the minimization of financial costs. The simulation approach is also treated with relevance in the literature. Triska et al. (2020) systematically reviewed the literature on port capacity assessment through simulation, in which some of the identified studies assessed storage capacity (Rusgiyarto et al. 2018; Keceli et al. 2013; Huang et al. 2008). Huang et al. (2008) suggest the container overflow level in the yard as a capacity assessment criterion, however they did not explore this characteristic in detail since berths were the bottleneck of the terminal. Other studies considered the yard occupation rate as a criterion for determining terminal capacity, with a threshold of 70% in Ro-Ro terminal (Keceli et al. 2013), threshold of
Simulation-Based Port Storage Dimensioning
147
80% in a container terminal (López-González et al. 2020) or without clearly specifying the threshold (Rusgiyarto et al. 2018). Triska et al. (2020) proposed approaches for planning terminal expansion and capacity assessment, including storage facilities, however they did not explore in detail the specifics for dimensioning storage. One can observe that the practical rules and simulation approaches used operational criteria, while the analytical approaches were oriented to costs for the dimensioning of storage. Nevertheless, with the exception of Alcalde et al. (2015) and Taleb-Ibrahimi et al. (1993), the consulted literature did not consider the probability of occurrence of storage shortage for the dimensioning of the facilities.
3 Methodological Approach According to Taha (2008), simulation is the second-best procedure for assessing the behaviour of a real system, after observing the system itself. The main advantage of using simulation is the fact that many real situations are too complex to be solved using mathematical models (Hillier and Lieberman 2006; Figueira and Almada-Lobo 2014). Recently, leaps in computational capacity have made simulation-based studies more viable (Figueira and Almada-Lobo 2014), contributing to increase the attractiveness of this method. Discrete-event simulation (DES) was used in this work, considering its applicability in port operations (Stahlbock and Voß 2008; Angelodious and Bell 2011; Dragovi´c et al. 2017) and the desired level of abstraction. 3.1 Description of the Developed Simulation Model The model proposed for the test case was inspired by a container terminal in the southern region of Brazil. The data for quay operations, that traditionally have the greatest impact on port performance (UNCTAD 1985), were all extracted from the terminal statistics, which makes the model appropriate for testing the proposed procedures. Some adaptations in the operational configuration were made, considering data availability. Although a precise representation of the reference terminal is not intended, the magnitude of the variables and the results of the test case are compatible with the operational reality of a container terminal. The scope of the evaluated model considers operations at the quay, yard and reception/dispatch for other modes of transport, as identified by Dragovi´c et al. (2017) as the main subsystems of a port system. This approach is also endorsed by Dekker et al. (2013), which described the main stages of a container terminal considering processes within these subsystems. It should be noted that, at the level of detail adopted in this work, the influence of reefer sockets, yard equipment and human resources on the storage operation is not evaluated, and the modelling of berth operations considers the berth and its respective quay cranes as a single productive resource. In the proposed model, the following entity types were generated: ships, trucks, containers and container lots. In addition, three resource pools were considered: berths, storage slots and truck gates. The number of berths is fixed for a given scenario, i.e. it does not depend on the relation between quay length and vessels’ size. The diagram in Fig. 1 shows, in a simplified way, the flow of ships, containers and trucks at the terminal.
148
Y. Triska and E. M. Frazzon
Fig. 1. Simplified flow of the operations in the test case.
3.2 Specifics Regarding the Simulation Experiments Two types of simulation experiments were performed: • Type 1 (restricted flow): considers each TEU slot in storage as a resource of the simulation model, and containers need to seize this resource to enter storage; • Type 2 (free flow): does not impose this restriction, as the seize block in the storage service is deactivated (the number of seized resources is set to zero). The lack of storage never interrupts the terminal operations (as considered by López-González et al. 2019). The first type of experiment was performed to illustrate the different stability regimes in Fig. 3. The second type of experiment is used to illustrate the variation in storage levels during a simulation period, as shown in Fig. 2, and to propose the approach to dimension port storage coping with operation instability, described in Sect. 4.2. The following parameters are used on both experiment types: simulation period of 1 year; sample size of 100 rounds; warm up period of 60 days (defined through preliminary testing); confidence interval of 95%; 2 berths; and 6 truck gates. The type 2 experiment considered a simulation demand equal to 730,645 containers/year, while the type 1 experiment considered a demand of 950,000 containers/year, in order to explore the instability of the system. As mentioned previously, these two types of experiments have different purposes and hence are not compared between each other. The type 1 experiment considered the values of static storage capacity of 33,000, 35,000 and 40,000 TEU, respectively for the scenarios of high instability, low instability and stable.
Simulation-Based Port Storage Dimensioning
149
For removing outliers, the approach of Tukey (1977) was used, which consists of discarding values outside the range of 1.5 times the interquartile range (IQR) above the upper quartile (Q3) and below the lower quartile (Q1). If the lower limit is negative, it is set to zero due to the non-negativity of the amount of storage slots. For illustrating the stability regimes in a port terminal and dimensioning port storage, the needed sample size was calculated assuming a normal distribution. The experiments were built in the AnyLogic software and run in a computer with processor i7 of 3.20 GHz, Windows 10, 128 GB of SSD and 16 GB of RAM.
4 Results 4.1 Effect of Storage Shortage on the Operational Instability of a Port Terminal The level of required storage in a port terminal can vary over the course of an operational year, as shown in Fig. 2. Minimum, average and maximum values of storage level are shown in each simulation round with free flow in storage (type 2 experiment). During a simulation year, the storage level fluctuates in the range of 14,569 to 38,592 TEU, with average values around 24,000 TEU. Still, it is possible to visually verify the greater variance of the maximum values compared to the minimum and average values. Analysing Fig. 2, it can be seen that the static capacity of 35,000 TEU is sufficient for most of the 100 simulation rounds in the experiment, however in the years when there is a need for greater use, this static capacity is insufficient in order to meet the demand during the whole year. 45.000 40.000
Containers in storage (TEU)
35.000 30.000 25.000 20.000 15.000 10.000 5.000 0 0
10
20
Minimum storage level
30
40 50 60 Simulation round Average storage level
70
80
90
100
Maximum storage level
Fig. 2. Example of variation in storage levels during a simulation period
That said, the operational instability of a simulation model is defined here as the occurrence of sudden variations in the output parameters (especially the outflow) in a set of simulation rounds with the same input parameters. In the case of the problem evaluated in this work, operational instability depends on the availability of static storage capacity (more specifically, TEU slots, since it is a container terminal).
150
Y. Triska and E. M. Frazzon
For a better presentation of this concept, three types of stability regime are defined: the stable regime, the low instability regime and the high instability regime. Figure 3 illustrates these regimes, considering results from 100 simulation rounds with restricted flow in storage (type 1 experiment). a) Stable regime
c) High instability regime
b) Low instability regime
Outflow (containers/year)
900.000 800.000 700.000 600.000 500.000 400.000 300.000 200.000 100.000 0
20
40
60
Simulation round
80
100 0
20
40
60
Simulation round
80
100 0
20
40
60
80
100
Simulation round
Fig. 3. Operational stability regimes and outflow values along 100 simulation rounds
In a stable regime (Fig. 3-a), there is little or no sudden variation in the output parameters of the simulation model. This occurs when there is no storage shortage in any simulation round, during the entire simulation period. None or few simulation rounds are identified as outliers. In a low instability regime (Fig. 3-b), there are already sudden variations in output parameters in some simulation rounds due to the lack of storage. This may occur for a short time during the simulation period. However, all or most anomalies of operational indicators are identified as outliers (also by the interquartile deviation criterion). Thus, the average values of these indicators are not contaminated by the discarded outliers. This may hide a significant chance that operations will be affected by the lack of available storage, considering that until the occurrence of 25% of outlier values in the set of simulations, the value of the lower quartile (Q1) may not be affected. In a regime of high instability (Fig. 3-c), anomalous values are no longer identified as outliers by the criterion of interquartile deviation. In addition to the problems already described for the low instability regime, there is also contamination of the average values of the indicators, which hinders the analysis of the model’s behaviour. In addition, the increase in the coefficient of variation of the sample significantly increases the needed sample size, and this considerably increases the time for carrying out the experiment, which may turn it unfeasible. The needed sample size calculated for the data in Fig. 3-c, with a significance level of 5%, relative error of 1% and excluding outliers according to Sect. 3.2, is 5,504 values. For comparison, the initial sample size used by Triska et al. (2020) is 10 simulation rounds. For port planning, there is no interest in studying the terminal in a regime of high instability or even low instability, since the probability of storage unavailability is relevant. In the event of a great drop in the outflow (as shown in Fig. 3-c), the throughput capacity is reduced in the same proportion. In addition, there are great financial costs associated with congestion in operations, due to the exponential increase in the average turnaround times of ships, containers and trucks (UNCTAD 1985).
Simulation-Based Port Storage Dimensioning
151
That said, the dimensioning of port storage is more related to decreasing the likelihood of terminal congestion due to storage shortage rather than improving the average values of operational indicators. This characteristic is in conformity with the main role of port storage according to literature, which is to minimize the interruption of quay operations (UNCTAD 1985; PIANC 2014; Brazil 2020). 4.2 Approach to Dimension Port Storage Coping with Operation Instability The approach aims to answer the research question: “how can the required static capacity of a port terminal be evaluated to prevent operational instability caused by insufficient storage?” The research question is adapted to the following problem formulation: • Suppose that the Maximum Required Storage in the Year (MRSY) variable, an output of a simulation run, has a normal distribution, with the mean and standard deviation estimated by a simulation experiment. What is the static capacity (SC) corresponding to the 95% probability that in the next tR years, keeping the current cargo demand pattern, there is no MRSY higher than the SC? With the mean and standard deviation, the MRSY value associated with a probability of occurrence can be obtained through the inverse function of the normal distribution. Note that the probability of an event occurring in consecutive years is a problem that can be treated by conditional probability and consequently using the multiplication rule. For example, if the probability that an event will not occur in one year is 95%, the probability that it will not occur in any of the next 3 years is 0.953 = 0.857. Applying this reasoning, one can construct the Eq. 1 to find a probability p1 to be used in the inverse function of the normal distribution, which is a function of a given confidence level p2 and a recurrence time tR . 1/ t p1 = p2 R
(1)
Figure 4 illustrates the storage dimensioning performed by this approach for different recurrence times (1 year, 25 years and 100 years) and 95% confidence level. The required storage capacity for a recurrence time of 25 years, corresponding to the terms of port terminals’ leasing contracts in Brazil (Brazil 2020), is 38.264 TEUs. After 30 simulation rounds, the results converged to stabilization. While increasing the tR from 1 year to 25 years requires an additional static capacity of 2,074 TEUs, increasing the tR from 25 years to 100 years requires only 699 more TEU slots. With the experiment data and adopting a criterion for storage dimensioning, it is possible to estimate the expected average utilization rate of the storage. Considering the 25-year recurrence time, normal distribution of the average storage level shown in Fig. 2 and 95% confidence level, the average storage utilization rate in a given year is in the range between 61.5% and 64.4%, with an expected value of 63.5%. The proposed approach can illustrate the undesirability of the low instability regime for port planning purposes. For example, considering that 10% of the simulation rounds were outliers and that in fact there is a 10% chance that storage will be insufficient in
152
Y. Triska and E. M. Frazzon 45.000
Static capacity (TEU)
40.000 35.000 30.000 25.000 20.000 15.000 10.000 5.000 0
20
40
60
80
100
Simulation rounds tR = 1 year
tR = 25 years
tR = 100 years
Fig. 4. Storage dimensioning considering normal distribution and recurrence time during 100 simulation rounds
the next year, for the 25-year recurrence time this probability rises to 93%, applying Eq. 1. For this probability to be 5% in this recurrence time, the probability of insufficient storage in a given year should be only 0.2%, that is, in a sample of 1,000 simulation rounds, only 2 rounds would be expected to be outliers, which already characterizes a stable regime. 4.3 Discussion of Results The proposed approach allows a dimensioning with greater precision and considering more specificities of the evaluated terminal when compared to others in literature. The consideration of the terminal’s recurrence time enables the formulation of a more objective and clear statistical problem. In comparison with the analytical approaches identified in the literature (e.g. Kim 1998, Kim 2002; Lee and Kim 2010; Do Ngoc and Moon 2011), the proposed approach is more generalizable for different situations (including different types of terminals, e.g. containers, Ro-Ro, dry bulk, liquid bulk), as the system’s operating premises are in the simulation model instead of the method for dimensioning the facilities. Mathematical models are full of premises in the developed formulae. Kim et al. (2002), for example, pointed out that several simplifications and assumptions were made, and that real operations can deviate from these assumptions. There are models developed only for import yards (Kim et al. 1998; Kim et al. 2002), while the proposed simulation-based approach can consider import and export containers jointly or separately, as needed. In relation to existing studies that apply simulation, the analysis of the effect of storage on the operation instability could show that the storage occupation rate alone may not be the best parameter for dimensioning port storage. Considering the proposed dimensioning approach with a recurrence time of 25 years, the expected utilization rate in the experiment, 63.5%, is lower than the 80% used by López-González et al. (2020). Since the adoption of the 80% value was not justified in the referred work, the storage may have been under-dimensioned when not considering the peak situation of storage utilization.
Simulation-Based Port Storage Dimensioning
153
5 Conclusion This work proposed a novel method for dimensioning port storage, which is based in simulation and aims to prevent terminal congestion due to insufficient storage. Port managers can use the results of this work to design new or existing terminal storage facilities, of any cargo type. The proposed approach constitutes a tool to support decision making in the context of port planning, and in this sense contributes to greater assertiveness of investment and minimizing the occurrence of congestion. The work limitations are inherent to the simulation approach, due to the availability of data and time for model development. Besides, although the recommended approach allows the dimensioning of port storage with certain precision, the simulation models are not a mirror of reality and may also present some deviation from the actual operation of the system. We recommend the following research directions: • Dimensioning the storage of real terminals using the approach recommended in this work; • Study of the stability regime of waterway channels; • Combination of storage shortage analysis with criteria that consider the service level of the operations, in view of the increase in cargo handling movements in a high occupied container yard; • Elaboration of practical rules for calculating port storage capacity based on a more robust simulation study, considering, for example, different terminal models and sensitivity analysis.
Acknowledgements. This study was supported by CAPES - Brazilian Federal Agency for Support and Evaluation of Graduate Education within the Ministry of Education of Brazil, under the finance code 001 and by the International Cooperation Program PIPC.
References Alcalde, E.M., Kim, K.H., Marchán, S.S.: Optimal space for storage yard considering yard inventory forecasts and terminal performance. Transp. Res. Part E Logistics Transp. Rev. 82, 101–128 (2015). https://doi.org/10.1016/j.tre.2015.08.003 Angeloudis, P., Bell, M.G.: A review of container terminal simulation models. Marit. Policy Manag. 38(5), 523–540 (2011). https://doi.org/10.1080/03088839.2011.597448 Brazil. Ministério dos Transportes, Portos e Aviação Civil. Secretaria Nacional de Portos. (Planos Mestres - Versão Completa). https://www.gov.br/infraestrutura/pt-br/assuntos/planejamento-egestao/planos-mestres-portos. Accessed Sept 2020 Carlo, H.J., Vis, I.F., Roodbergen, K.J.: Storage yard operations in container terminals: literature overview, trends, and research directions. Eur. J. Oper. Res. 235(2), 412–430 (2014). https:// doi.org/10.1016/j.ejor.2013.10.054 Chen, C., Hsu, W.J., Huang, S.Y.: Simulation and optimization of container yard operations: a survey. In: Proceedings of International Conference on Port and Maritime R and D and Technology, pp. 23–29 (2003)
154
Y. Triska and E. M. Frazzon
Chu, C.Y., Huang, W.C.: Determining container terminal capacity on the basis of an adopted yard handling system. Transp. Rev. 25(2), 181–199 (2005). https://doi.org/10.1080/014416404200 0244608 Dekker, R., van der Heide, S., van Asperen, E., Ypsilantis, P.: A chassis exchange terminal to reduce truck congestion at container terminals. Flex. Serv. Manuf. J. 25(4), 528–542 (2012). https://doi.org/10.1007/s10696-012-9146-3 Ding, Y.Z.: Throughput capacity of a container terminal considering the combination patterns of the types of arriving vessels. J. Shanghai Jiaotong Univ. (Sci.) 15(1), 124–128 (2010). https:// doi.org/10.1007/s12204-010-8174-y Do Ngoc, A.D., Moon, I.K.: The storage capacity expansion and space leasing for container depots. Flexible Serv. Manuf. J. 23(4), 364–384 (2011). https://doi.org/10.1007/s10696-011-9075-6 Dragovi´c, B., Tzannatos, E., Park, N.K.: Simulation modelling in ports and container terminals: literature overview and analysis by research field, application area and tool. Flex. Serv. Manuf. J. 29(1), 4–34 (2016). https://doi.org/10.1007/s10696-016-9239-5 Figueira, G., Almada-Lobo, B.: Hybrid simulation-optimization methods: a taxonomy and discussion. Simul. Model. Pract. Theory 46, 118–134 (2014). https://doi.org/10.1016/j.simpat.2014. 03.007 Hillier, F.S., Lieberman, G.: J. Introdução à Pesquisa Operacional, 8th edn. McGraw Hill Brasil, São Paulo (2006) Huang, S.Y., Hsu, W.J., Chen, C., Ye, R., Nautiyal, S.: Capacity analysis of container terminals using simulation techniques. Int. J. Comput. Appl. Technol. 32(4), 246–253 (2008). https://doi. org/10.1504/IJCAT.2008.021379 Keceli, Y., Aksoy, S., Aydogdu, Y.V.: A simulation model for decision support in Ro-Ro terminal operations. Int. J. Logistics Syst. Manage. 15(4), 338–358 (2013). https://doi.org/10.1504/ IJLSM.2013.054896 Kim, K.H., Kim, H.B.: The optimal determination of the space requirement and the number of transfer cranes for import containers. Comput. Ind. Eng. 35(3–4), 427–430 (1998). https://doi. org/10.1016/S0360-8352(98)00125-9 Kim, K.H., Kim, H.B.: The optimal sizing of the storage space and handling facilities for import containers. Transp. Res. Part B Methodol. 36(9), 821–835 (2002). https://doi.org/10.1016/ S0191-2615(01)00033-9 Lee, B.K., Kim, K.H.: Optimizing the block size in container yards. Transp. Res. Part E Logistics Transp. Rev. 46(1), 120–135 (2010). https://doi.org/10.1016/j.tre.2009.07.001 Ligteringen, H., Velsink, H.: Ports and Terminals. VSSD, Delft (2012) López-González, A., Medina-León, S.V., González-Angeles, A., Mendoza-Muñoz, I., GilSamaniego-Ramos, M.: Assessment of a container terminal expansion using simulation. DYNA 87(214), 129–138 (2020). https://doi.org/10.15446/dyna.v87n214.82822 Olba, X.B., Daamen, W., Vellinga, T., Hoogendoorn, S.P.: Network capacity estimation of vessel traffic: an approach for port planning. J. Waterw. Port Coast. Ocean Eng. 143(5), 04017019 (2017). https://doi.org/10.1061/(ASCE)WW.1943-5460.0000400 PIANC - Permanent International Association of Navigation Congresses. Masterplans for the Development of Existing Ports. PIANC Report nº 158 - Maritime Navigation Commission. Bruxelles, PIANC (2014). http://www.pianc.org/downloads/publications Rusgiyarto, F., Sjafruddin, A., Frazila, R.B., Suprayogi, S., Burhani, J.T.: Inland container depots effect for import container terminal performance at Koja container terminal, Jakarta based on optimization-simulation model. In: AIP Conference Proceedings, vol. 1977(1), p. 020038. AIP Publishing LLC (2018). https://doi.org/10.1063/1.5042894 Stahlbock, R., Voß, S.: Operations research at container terminals: a literature update. OR Spectrum 30(1), 1–52 (2008). https://doi.org/10.1007/s00291-007-0100-9
Simulation-Based Port Storage Dimensioning
155
Steenken, D., Voß, S., Stahlbock, R.: Container terminal operation and operations research-a classification and literature review. OR Spectrum 26(1), 3–49 (2004). https://doi.org/10.1007/ s00291-003-0157-z Taha, H.A.: Pesquisa Operacional, 8th edn. Pearson Prentice Hall, São Paulo (2008) Taleb-Ibrahimi, M., de Castilho, B., Daganzo, C.F.: Storage space vs handling work in container terminals. Transp. Res. Part B Methodol. 27(1), 13–32 (1993). https://doi.org/10.1016/01912615(93)90009-Y Triska, Y., Frazzon, E.M., Silva, V.M.D.: Proposition of a simulation-based method for port capacity assessment and expansion planning. Simul. Model. Pract. Theory 103, 102098 (2020). https://doi.org/10.1016/j.simpat.2020.102098 Tukey, J.W.: Exploratory Data Analysis. Addison-Wesley, Reading (1977) UNCTAD - United Nations Conference on Trade and Development. Port Development: A Handbook for Planners in Developing Countries, 2nd edn. New York, United Nations (1985)
Integration of Renewable Energies at Maritime Container Terminals Felix Schütze(B)
, Anne Kathrina Schwientek , Ole Grasse , and Carlos Jahn
Institute of Maritime Logistics, Hamburg University of Technology, Hamburg, Germany [email protected]
Abstract. Maritime container terminals play an important role in global supply chains. In addition to the rapid handling of containers, the reduction of CO2 emissions is also increasingly crucial for terminal operators. This can be achieved by integrating renewable sources such as photovoltaic or wind energy. While energy supply and demand must be in balance, the amount of energy produced through renewable sources cannot be controlled as it depends exclusively on variable weather conditions. One option for efficient use of renewable energy sources is to modify energy consumption by intelligently controlling processes. This study aims to answer the question to what extent energy intensive consumption processes at container terminals can be adapted to a volatile energy supply. A discrete event simulation study is conducted to analyze handling processes by quay cranes as main energy consumers depending on the availability of photovoltaic energy. Therefore, the operating times of quay cranes are partially limited to daylight hours. Only a low number of quay cranes is deliberately deployed when a time window occurs between the predicted end of loading of the vessel and a departure time of the vessel determined by the tidal range. The simulation results show that by flexibly using certain quay cranes only during daylight hours the percentage of energy produced by renewable resources can be increased by up to 50%. As a result, handling-related CO2 emissions can be reduced. The study offers an approach to a sustainable energy supply on terminals by reconciling energy use and environmentally friendly generation. Keywords: Discrete event simulation · Quay cranes · Energy consumption · Renewable energies · Container terminal
1 Introduction Climate change is considered one of the greatest challenges of the 21st century. In maritime supply chains, processes are more and more reorganized to reduce the overall carbon footprint to a feasible minimum. Moreover, container terminals as trimodal transport nodes are becoming increasingly important and offer enormous potential to save on CO2 emissions and noise [1]. However, an increase in electrification results in an increase of the terminal’s overall electricity demand. Changeable weather conditions dictate when renewable energies, such as solar or wind energy, can be generated. Under unfavorable conditions, the supplied power is provided by the main grid [1]. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 M. Freitag et al. (Eds.): LDIC 2022, LNLO, pp. 156–167, 2022. https://doi.org/10.1007/978-3-031-05359-7_13
Integration of Renewable Energies at Maritime Container Terminals
157
In parallel to emission reduction initiatives, the energy market submitted to further change due to the fluctuating energy provision from renewable resources and the opening of electricity markets to further suppliers. Therefore, a clear management of energy use and production is becoming increasingly important [2, 3]. Various management concepts, such as battery swapping systems and charging strategies for Automated Guided Vehicles (AGVs) [4] or the use of short-term storage technologies [2], have already been investigated in container terminal literature for effective energy use. A sustainable alternative to fully power terminal operations with renewable energy has not yet been considered in detail before and offers potential for further research. Additionally, the process of digitalization enables smart grids and microgrids. Such intelligent energy management systems provide a link between generation, storage, and demand of electricity (Fig. 1). This way, the power fluctuations common with energy provided by renewable energy sources can be balanced out. GeneraƟon Wind Power
PV
Fuel Cell
Etc.
ConsumpƟon
Energy Management System
Energy Storage
Forecast LogisƟc Info Weather Info ...
Cranes
Reefers
Buildings
others
Energy Billing
Fig. 1. Components of a smart microgrid in a maritime terminal [15].
In addition to electricity, data is also transferred in a smart grid. As a result, electricity generators and consumers can communicate with each other [2, 5]. Therefore, smart grids and microgrids play a central role in energy optimization processes in port operations. They represent an important criterion for future research in the organization and the handling of transport processes within terminals [2]. Based on this understanding, the aim of this study will be to integrate microgrid concepts into terminal processes to increase the share of renewable energies and reduce CO2 emissions. Moreover, energy intensive consumption processes will be adapted in the best possible way to the energy supply from renewable resources without hindering everyday terminal operations.
2 Problem Statement 2.1 Energy Demand of Quay Cranes as Main Energy Consumers at Maritime Container Terminals Maritime container terminals are considered complex systems that can be divided into three main areas [6, 7]. These include the waterside transshipment area, the yard area, and the landside transshipment area. For handling and transport within a container terminal and between the areas it is important to select a suitable combination of handling equipment best suited to the terminal. The choice of handling equipment depends on various factors, such as their energy consumption. The amount of consumption of the individual devices depends on the handled quantity (see Fig. 2).
158
F. Schütze et al.
To identify the main energy consumers, several container terminals such as Noatum Container Terminal Valencia [8] or Pier E at Port of Long Beach [9] were analyzed in terms of total electrical consumption and distribution. In all three terminals analyzed, QCs cause a high share of total energy consumption and can be considered as energy intensive. An exemplary distribution of total energy consumption of the container terminal Hamburg-Altenwerder can be found in Fig. 2 in form of an energy demand forecast. In this case, depending on traffic volume, QCs account for up to 20–30% of total consumption. Higher utilization leads to more QC moves. This also increases the total share of energy consumption since more energy is required for handling. Energy ConsumpƟon [kW/h]
40% 30% 20% 10% 0% Yard Cranes
Others
Quay Cranes
Reefer
LighƟng
Low UƟlizaƟon: 32,130 Containers handled Medium UƟlizaƟon: 38,539 Containers handled High UƟlizaƟon: 70,105 Containers handled
Portal Cranes Landside
BaƩery Exchange StaƟon
Fig. 2. Energy demand forecast of the container terminal Hamburg-Altenwerder [10].
Energy ConsumpƟon [kW]
A general distinction can be made between two types of container cranes. On the one hand, cranes are used as handling instruments within the block storage facilities, and on the other hand, cranes can be used for shoreside handling. This study focuses mainly on the second type of container cranes also known as QCs. Energy consumption behavior will consequently be examined in more detail in the following (Fig. 3).
A
800 600 400 200 0 -200 -400
E
B C
F G D
0
20
A: Hoisting container (peak load) B: Hoisting container C: Horizontal Transport with container D: Lowering of the container
40
60
80
100
120
Time (seconds) E: Hoisting without container F: Horizontal movement without container G: Lowering without container
Fig. 3. Simplified load profile of a QC when handling one container based on [10–13]
Integration of Renewable Energies at Maritime Container Terminals
159
Peak loads refer to short periods in which the demand for power is significantly higher than the average supply level. Cyclical peak loads result in high energy costs. This can be justified by increased energy supply prices for peak loads of the electricity supply companies, which are usually accompanied by high costs for terminal operators [14]. As already elaborated, QCs take up a significant share on the total caused energy consumption. In addition, QCs also contribute to high peak loads when hoisting containers simultaneously and are therefore mainly responsible for short-term cost-intensive peak loads [15]. Figure 2 shows the simplified load profile of a QC when handling one container. Energy consumption differs depending on the current movement. To transport one container from the vessel to the quay and then return to the starting position the crane needs six movements in total that are executed with the help of an attached spreader [12]. 2.2 Forecast of Energy Supply Through Renewable Resources
Daily Irradiance [W/m2]
In contrast to the prediction of wind speeds, the daily amount of solar radiation can be predicted well [16]. Accordingly, a forecast of the electricity yield of an individual plant can be derived. In the following, an exemplary irradiation analysis and a yield forecast were carried out for the container terminal Hamburg-Altenwerder. The data originates from the database “Photovoltaic Geographical Information System” (PGIS) of the European Commission. It stores radiation data for Europe, Southwest Asia and Africa. By collecting radiation data over several years, average values have been determined and used as a basis for calculating the yield of photovoltaic (PV) systems. 500 400 300 200 100 0
Hour [hh:mm] Direct
Diīuse
Global
Fig. 4. Daily average irradiance at the container terminal Hamburg-Altenwerder in June [17].
Figure 4 shows the analysis of the container terminal Hamburg-Altenwerder with regard to the daily average radiation intensity using PGIS in the month June. Diffuse radiation is the portion of solar radiation that reaches the earth’s surface or the object after reflection from other surfaces. Direct radiation describes the portion that directly hits the earth’s surface or object without being scattered, while global radiation is the sum of both radiation components. For electricity production, only the direct radiation is predominantly usable [18]. The radiation analysis of the two months January and June shows that there is a significantly higher radiation intensity over a longer period in June compared to January. This can be explained by shorter day times in the winter months
160
F. Schütze et al.
as well as flatter irradiation angles of the solar radiation [18]. This indicates that a high proportion of renewable energies can potentially be used for energy supply, especially during the summer months.
3 Simulation Procedure Angeloudis and Bell [6] define maritime container terminals as complex systems. Due to the number of equipment used, the general size of the system, and the interaction of all actors, it is generally difficult to make accurate predictions regarding different configurations and to test them. This is why simulation models are essential to create test scenarios and to analyze the dynamics within terminals. In this study, following Ha et al. [19] and Angeloudis and Bell [6], an equipment-focused simulation model with a low degree of detail is used. This type of simulation does not commit itself to the structure or processes of a specific terminal and can be transferred to various simulation models. It also allows statements to be made on the extent to which new or additionally deployed equipment can affect terminal processes and the overall terminal productivity. To create a comprehensive test environment and attempt to solve the problem of high energy consumption by QCs mentioned above as well as the adaptability to a renewable energy supply, a maritime container terminal is simulated over a period of one week (168 h). The terminal simulated is an open tidal port, which means that some vessels can only enter and leave during a certain period of time. One such example would be the port of Hamburg. The waterside handling is examined in detail with the arrival of feeder and seagoing vessels. For both, the operating times of certain QCs are limited to daylight hours. Furthermore, fewer QCs are used in a targeted manner if a time window occurs between the predicted end of loading and the departure time of a seagoing vessel determined by the tidal range. 3.1 Simulation Model Description The model is developed in a Plant Simulation software environment. It is structured as follows: a 1000 m quay wall separates the port basin from the AGV area. The quay wall is divided into two berths. One berth is used by feeder vessels and the other by seagoing vessels. A different number of QCs is available depending on the time of day. QCs that are only used during daytime operations thus represent a symbolic energy supply by PV systems and are supplied by green energy. This study follows the approach that two to four QCs can be used for feeder vessels and three to six for seagoing vessels. Two QCs each are responsible for handling one container area on a vessel (import and export): one QC is permanently deployed and the other is being used only during daytime. In practical operation, full daylight limited use is only of limited sense, as this would have an immense impact on the service quality of the terminal. For economic reasons, one quay crane is permanently in operation. The other one is only used to accelerate handling speed if renewable energy is available. The container areas on the vessels are additionally defined so that the handling of containers is also possible by one QC. Therefore, constant operations can be achieved with a smaller number of QCs.
Integration of Renewable Energies at Maritime Container Terminals
161
The number of containers to be handled for a container area varies depending on the traffic volume. In addition, the traffic volume for a feeder vessel is divided approximately 50% between the two container areas. For seagoing vessels, the distribution between the container areas is approximately 33%. It must be considered that the handling of export containers for one container area can only be started as soon as all import containers for one container area have been unloaded. A total of 50 AGVs are used for horizontal container transport between the block yards and the QC area. Five AGVs are assigned to one QC. This corresponds to an average time of 12 min per container move between yard area and quay wall per AGV [20]. Due to optimization aspects and to avoid long empty runs and waiting times, there is usually no fixed assignment of AGVs to QCs in practice [20]. However, this is not critical for the research question and how equipment is dispatched highly depends on the terminal’s general objectives [21, 22]. Thus, a fixed allocation is selected here and calculated with the average data available from the literature, in which only the disturbance behavior of the AGVs is taken into account. In addition to the simulated transport of containers, the energy consumption caused by QCs is monitored and calculated. For this purpose, the overall handling process at the quay wall is divided into small sub-processes in the model to map the energy consumption of the individual movements in as much detail as possible (Fig. 2). Accordingly, for one container move by a single QC, a total static energy consumption is calculated by adding up the energy consumption for each individual subprocess. The total energy consumption therefore increases linear with the traffic volume. 3.2 Variation of Input Parameters Various experiments are carried out in which parameters, such as arrival times of vessels, duration of daylight and number of containers handled, are varied. The overall aim is to find out how energy consumption processes of QCs can be optimally adapted to a renewable energy supply. 50 simulation runs are performed per experiment. Arrival Times of Vessels: Two options are tested for the vessel arrival times (A1, A2). For option A1, the vessels arrive randomly during the day to reflect a usual business day, and for option A2, the arrival times are adjusted to the different time of sunrise to optimally use the renewable energy sources. In total, the arrival of three seagoing and 12 feeder vessels is simulated over the period of one week. Duration of Daylight: Both the arrival times and the duration of daylight within the experiment are varied. For this purpose, the months of January, April, July, October and December are considered in more detail and the respective average daylight durations are used. For example, while the month of January has an average daylight duration of only 6.5 h, in July daylight is available for 16 h. Traffic Volume: In general, three different traffic volumes are used for the experiments. The exact data is shown in Table 1. A distinction is made between import and export containers. The container volumes vary depending on the vessel type and are selected within a defined range (Feeder: ± 100 cont., seagoing vessel: ± 500).
162
F. Schütze et al. Table 1. Number of containers handled per vessel depending on traffic volume.
Traffic volume
Number of containers handled per vessel Import feeder
Export feeder
Import seagoing vessel
Export seagoing vessel
Total volume
High
±300
±400
±1,500
±2,000
15,700
Medium
±200
±300
±1,200
±1,700
12,500
Low
±100
±200
±1,000
±1,500
9,730
4 Results and Discussion 4.1 Results for Throughput Times
21:21 19:40 18:00 16:19 14:38 12:57 11:16 9:36 January April July October December January April July October December January April July October December January April July October December January April July October December January April July October December
Throughput Ɵmes of export containers seagoing vessels [hh:mm]
The throughput time in this simulation model is defined by the average time for one container to travel from the vessel to the block yard (import container) or in the opposite direction (export container). The results of the study show that the average throughput time correlates with the traffic volume. Accordingly, the average throughput time increases with a higher number of containers handled. Furthermore, the results show that the average throughput time also correlates with the duration of daylight. With a long daylight period, a higher number of QCs is used for a longer period, which has the effect of reducing the mean throughput time. However, the handling of export containers on seagoing vessels is an exception. Due to a tide-related time window and the targeted use of a lower number of QCs, we see a reverse behavior (Fig. 5). Here, throughput times increase with increasing daylight hours.
A1
A2
Traĸc Volume - High
A1
A2
Traĸc Volume - Medium
A1
A2
Traĸc Volume - Low
Fig. 5. Average throughput time of export containers (seagoing vessels) at different traffic.
Integration of Renewable Energies at Maritime Container Terminals
163
Even with a high volume of traffic, all containers can be handled for each parameter set. This shows that a configuration of the container terminal presented in the paper with two berths for feeder and seagoing vessels is generally suitable for handling the required volumes in the considered period. In addition, when vessel arrivals are adjusted to daylight hours, average throughput times are also reduced. This is because more QCs are being used during a longer period. This shows that especially during the summer months an additional use of QCs is beneficial to decrease throughput times. 4.2 Energy Consumption Analysis
Percentage of renewable energy consumpƟon in total consumpƟon [kWh]
The results of the energy consumption analysis show that the share of energy consumption from renewable resources, especially in the summer months, can be increased by up to 50% of the total consumption (Fig. 6). Accordingly, 50% of the energy supply from the main grid could be replaced by renewable energies to supply the QCs with green energy. This indicates that the implemented concept for daylight restricted QCs utilization expanses the share of sustainably operated QCs. Consequently, an increase of the share of renewable energy, as seen in the presented terminal configuration with daylight limited usage of certain QCs, is an alternative for terminal operators to integrate renewable energies into terminal operations and still deliver sufficient service quality. Whether the share of renewables can be further increased depends on how durable the service quality of the terminal operators is and how the terminal defines its own objectives. The results also show that when vessel arrivals at a low, medium or high traffic volume are adapted to daylight hours as best as possible it can further increase the share of sustainably operated QCs. The simulation outcomes indicate that the traffic volume correlates with the total energy consumed. The lower the traffic volume is, the lower is the total energy consumption. Furthermore, it is depicted that the share of renewable operated QCs can account for up to 50% of total energy consumption regardless of the traffic volume but depending on the available daylight hours. To operate more sustainably from a terminals operator perspective, it is therefore not the traffic volume that is decisive, but much more the number of daylight hours available. 100% 80% 60% 40% 20% 0% A1
A2
January
A1 April
A2
A1
A2 July
A1
A2
October
A1
A2
December
Energy consumpƟon of quay cranes - used at dayƟme and at night [kWh] Energy consumpƟon of quay cranes - only used at dayƟme kWh]
Fig. 6. Share of renewable energy consumption in total consumption at a high traffic volume.
164
F. Schütze et al.
6,5 12 16 10,5 7 hrs hrs hrs hrs hrs
Following the analysis of energy consumption, the evaluation is also used to calculate potential CO2 emission savings. In this section, the energy consumption data obtained is used to evaluate the total CO2 emissions. For the calculations, different emission factors for the respective supply systems are multiplied by the energy consumption. It should be noted that PV systems also cause CO2 emissions [23]. This includes for example the production, installation etc. While the supply by the main grid (e.g. electricity mix Hamburg) has an emission factor of 643.7 g/kWhel [24], the average supply from PV systems has an emission factor of 50 g/kWhel [16, 23]. December October July April January 0
10
20 30 CO2-Emissions [t/kWh]
40
50
CO2-Emissions quay cranes - used at dayƟme and at night [t/kWh] CO2-Emissions quay cranes - only used at dayƟme [t/kWh]
Fig. 7. Total CO2 -emissions caused by all QCs.
Figure 7 visualizes the occurred total CO2 emissions at a high traffic volume according to arrival strategy A2. The total CO2 emissions decrease with increasing daylight hours. Furthermore, the simulation results indicate that the implemented terminal configuration with the associated QCs operation time can lead to a reduction of the total CO2 emissions, especially during the summer months. Thereby, the share of CO2 emissions, which is caused by the daylight operated QCs, is significantly lower compared to those permanently in use. To reduce the total CO2 emissions, it is therefore recommended to adjust the QCs operating times as much as possible, especially in the summer months, to the daylight hours.
5 Conclusion and Outlook 5.1 Conclusion The aim of this study was to investigate how flexible energy intensive consumption processes at maritime container terminals can be adapted to a non-constant energy supply from renewable energies in form of PV energy. It was first investigated which consumers account for a high share of the total caused energy consumption at maritime container terminals and to what extent they prove to be flexible in terms of time. It was found that QCs contribute significantly to the total energy consumption during cargo handling processes and can be used flexibly in terms of time provided that:
Integration of Renewable Energies at Maritime Container Terminals
• • • •
165
the tidal range extends the laytime of the vessel being unloaded/loaded a flexible crane assignment strategy is deployed a storage medium is used the size of the vessel allows simultaneous use of QCs.
Based on this, a discrete event simulation model was developed which reconstructed the container handling processes of seagoing and feeder vessels by QCs to the block storage facilities. Thereby, the model also enables an integration of renewable energies in the form of PV systems for sustainable energy supply. In this simulation, the operating times of certain QCs are also limited to daylight hours and fewer QCs are deliberately used if a time window occurs between the predicted end of vessel loading and the departure time determined by the tidal range. To achieve the greatest possible share of renewable energies within this time window, priority is given to those QCs that are powered by renewable energies. Various experiments have been conducted as part of a simulation study, varying the number of containers to be handled, the daylight hours, and the arrival time of the vessels. The results of the study show that a flexible crane assignment strategy offers an opportunity for terminal operators to reduce CO2 emissions while maintaining the same handling volume. In addition, by flexibly using certain QCs during daylight hours and utilizing the entire tidal restricted berthing time of the seagoing vessels, the percentage of energy from renewable resources can be increased by up to 50% during the summer months. As a result, handling-related CO2 emissions can be reduced. At the same time, the results show evidence that implementing storage media during the winter months would not prove to be advantageous. Furthermore, it is found that the throughput rates at the quayside can be increased by adapting vessel arrivals to daylight hours. However, this is accompanied by higher peak loads and must be brought into line with the terminal’s service quality. 5.2 Outlook The results of the study show a first approach to the integration of renewable energies on maritime container terminals by adjusting the QC operating times to daylight hours. Based on the results presented, further measures could be investigated that would lead to a further reduction in CO2 emissions and further increase the share of renewable energies. For this purpose storage media could be implemented, which store the excess energy generated by PV systems during the day to release it again in the evening. It would also be worthwhile to investigate whether the implementation of storage media would also allow an autonomous energy supply. Further research is needed on the financial feasibility of the concept. Such an economic analysis should especially examine those costs that arise from the downtimes of certain QCs at night and also include a parameter for measuring the service quality provided by terminal operators. Another starting point for extending the simulation model would be the implementation of operation rules to reduce peak loads that occurred in the model. For example, Hej [25] shows that peak loads can be reduced by limiting the number of QCs lifting simultaneously and by limiting the maximum energy demand of all operating QCs. Introducing such operational strategies could result in reducing costly peak loads and
166
F. Schütze et al.
would simultaneously lead to a financially more attractive solution from the perspective of terminal operators.
References 1. Wieschemann, A.: Forschung und Entwicklung batteriebetriebener Schwerlastfahrzeuge (AGV) und Erprobung in einem Feldversuch im Container-Terminal Altenwerder in Hamburg - B-AGV: gemeinsamer Abschlussbericht Düsseldorf (2011). https://doi.org/10.2314/GBV: 726781827 2. Iris, Ç., Lam, J.S.L.: A review of energy efficiency in ports: Operational strategies, technologies and energy management systems. Renew. Sustain. Energy Rev. 112, 170–182 (2019). https://doi.org/10.1016/j.rser.2019.04.069 3. Kumar, J., Parthasarathy, C., Västi, M., Laaksonen, H., Shafie-Khah, M., Kauhaniemi, K.: Sizing and allocation of battery energy storage systems in Åland Islands for large-scale integration of renewables and electric ferry charging stations. Energies 13, 317 (2020). Vaasa 4. Schmidt, J., Meyer-Barlag, C., Eisel, M., Kolbe, L.M., Appelrath, H.-J.: Using battery-electric AGVs in container terminals—assessing the potential and optimizing the economic viability. Res. Transp. Bus. Manag. 17, 99–111 (2015). https://doi.org/10.1016/j.rtbm.2015.09.002 5. Umweltbundesamt: Was ist ein “Smart Grid”. https://www.umweltbundesamt.de/service/ubafragen/was-ist-ein-smart-grid. Accessed 12 Feb 2021 6. Angeloudis, P., Bell, M.G.H.: A review of container terminal simulation models. Maritime Policy Manage. 38, 523–540 (2011). https://doi.org/10.1080/03088839.2011.597448 7. Brinkmann, B.: Seehäfen - Planung und Entwurf. Springer, Heidelberg (2005). https://doi. org/10.1007/b138397 8. Sapiña, R., et al.: Greentechnologies and ecoefficient alternatives for cranes and operations at port container Terminals. GreenCranes Consortium and TEN-T EA, Project Code: 2011EU-92151-S (2013) 9. Kermani, M., Parise, G., Martirano, L., Parise, L., Chavdarian, B.: Utilization of regenerative energy by ultracapacitor sizing for peak shaving in STS crane. In: IEEE International Conference on Environment and Electrical Engineering and 2019 IEEE Industrial and Commercial Power Systems Europe (EEEIC/I&CPS Europe), Genoa (2019) 10. Grundmeier, N.: Simulationsbasierte Energiebedarfsprognose in Seehafen ContainerTerminal. Carl von Ossietzky Universität Oldenburg, Oldenburg (2016) 11. Le, K.: Analysing electric yard cranes with simulation. Port Technology International, vol. 53, Oakland (2012) 12. Geerlings, H., Heij, R., van Duin, R.: Opportunities for peak shaving the energy demand of ship-to-shore quay cranes at container terminals. J. Shipping Trade 3(1), 1–20 (2018). https:// doi.org/10.1186/s41072-018-0029-y 13. Kermani, M., Parise, G., Chavdarian, B., Martirano, L.: Ultracapacitors for Port Crane applications: sizing and techno-economic analysis. Energies 13, 2091 (2020). https://doi.org/10. 3390/en13082091 14. van Duin, J.H.R., Geerlings, H., Verbraeck, A., Nafde, T.: Cooling down: a simulation approach to reduce energy peaks of reefers at terminals. J. Clean. Prod. 193, 72–86 (2018). https:// doi.org/10.1016/j.jclepro.2018.04.258 15. Tao, L., Guo, H., Moser, J., Mueller, H.: A Roadmap towards smart grid enabled harbour terminals. In: CIRED Workshop, Rome (2014) 16. Wirth, H.: Aktuelle Fakten zur Photovoltaik in Deutschland. Fraunhofer ISE (2021) 17. EU Science Hub: Monthly Radiation. https://ec.europa.eu/jrc/en/PVGIS/tools/monthly-rad iation. Accessed 17 Feb 2021
Integration of Renewable Energies at Maritime Container Terminals
167
18. Watter, H.: Regenerative Energiesysteme: Grundlagen, Systemtechnik und Analysen ausgeführter Beispiele nachhaltiger Energiesysteme. Springer Fachmedien Wiesbaden, Wiesbaden (2019). https://doi.org/10.1007/978-3-658-23488-1 19. Ha, B.-H., Park, E.-J., Lee, C.-H.: A simulation model with a low level of detail for container terminals and its applications. In: Proceedings of a Simulation Conference, 9–12 December 2007, J.W. Marriott Hotel, Washington, D.C., USA. IEEE, Piscataway, N.J. (2007) 20. Ranau, M.: Planning approach for quayside dimensioning of automated traffic areas and impact on equipment investment. In: Böse, J.W. (ed.) Handbook of Terminal Planning. ORSIS, pp. 301–318. Springer, Cham (2020). https://doi.org/10.1007/978-3-030-39990-0_14 21. Schwientek, A.K., Lange, A.-K., Jahn, C.: Effects of terminal size, Yard block assignment, and dispatching methods on container terminal performance. In: Bae, K.-H., et al. (eds.) Proceedings of the Winter Simulation Conference 2020, pp. 1408–1419, Online (2020) 22. Schwientek, A., Lange, A.-K., Jahn, C.: Simulation-based analysis of dispatching methods on seaport container terminals. In: Proceedings ASIM SST 2020, pp. 357–364. ARGESIM Publisher Vienna (2020). https://doi.org/10.11128/arep.59.a59050 23. Lauf, T., Memmler, M., Schneider, S.: Emissionsbilanz erneuerbarer Energieträger. Umweltbundesamt, Dessau-Roßlau (2019) 24. AEE. https://www.foederal-erneuerbar.de/landesinfo/bundesland/HH/kategorie/strom/aus wahl/733-spezifische_co2-emis/. Accessed 3 June 2021 25. Hej, R.: Opportunities for Peak Shaving Electricity Consumption at Container Terminals Applying New Rules of Operation to Achieve a More Balanced Electricity Consumption. Delft University of Technology, Rotterdam (2015)
An Appraisal of the Northern European LNG Bunker Ship Fleet Antje Roß1(B) and Kerstin Lange2 1 NOW GmbH, Fasanenstr. 5, 10623 Berlin, Germany
[email protected]
2 Department of Maritime and Logistics Studies, Jade University of Applied Sciences,
26931 Elsfleth, Germany [email protected]
Abstract. LNG as ship fuel is an emerging market. LNG infrastructure is a prerequisite for the adoption of LNG technology. One important element of LNG infrastructure is LNG bunker ships. By global comparison, Northern Europe is a region with dense traffic of LNG vessels. The bunker ship fleet in this region consists of several units in operation, with numerous others on order. The operational profile, assessed based on AIS data for the months June to September 2019, confirms that LNG as ship fuel is still a niche market. Ship-to-ship LNG deliveries are concentrated in several hotspots. The fleet of LNG vessels supplied by the Northern European fleet of LNG bunker ships is rather diverse in terms of ship types. The LNG bunker ships with a larger carrying capacity also engage in feedering LNG to smaller-scale LNG terminals. With a future denser net of LNG infrastructure and a larger LNG bunker ship fleet, LNG bunker ships are expected to specialize increasingly with regard to the services they supply. Keywords: LNG bunkering · LNG infrastructure · LNG vessel
1 Introduction 1.1 LNG as Ship Fuel and Associated Supply Infrastructure Liquefied Natural Gas (LNG) is a favourable alternative to conventional ship fuels, because it features lower emission levels when combusted. For this reason, more and more vessels sail on LNG (Sharples 2019). At the same time, the use of LNG is not trivial; it requires purpose-built installations on board (including specialised engines, tanks and fuel supply systems) as well as infrastructure ashore (European Marine Safety Agency 2018). With the designation of the North Sea and the Baltic Sea as emission control areas (ECAs), relatively stringent emission regulations were put in place in Northern Europe. This has stimulated the adoption of LNG technology by vessels and, hence, the development of LNG infrastructure. Furthermore, the latter is politically supported through the Directive 2014/94/EU on the Deployment of Alternative Fuels Infrastructure, which © The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 M. Freitag et al. (Eds.): LDIC 2022, LNLO, pp. 168–178, 2022. https://doi.org/10.1007/978-3-031-05359-7_14
An Appraisal of the Northern European LNG Bunker Ship Fleet
169
obliges EU member states to ensure that LNG bunkering infrastructure is established in major seaports by 2025. Next to LNG liquefaction plants and LNG terminals, small-scale LNG carriers like bunker ships, trucks or containers are vital elements of LNG infrastructure. LNG bunker ships are suitable for suppling a wide range of volumes. Moreover, bunkering can take place simultaneously with other operations. Therefore, ship-to-ship (STS) transfers can be regarded as a particularly promising mode of LNG bunkering, which makes LNG bunker ships a very relevant object of enquiry (European Marine Safety Agency 2018). 1.2 Research Gap The application of LNG technology in vessels has been studied in large variety of publications from environmental (e.g. Stenersen and Thonstad 2017; Sharafian et al. 2019), technical (e.g. Köhler et al. 2018), and commercial (e.g. Schinas and Buttler 2017) perspectives. The literature regarding LNG bunker ships is much smaller. Baresic et al. (2018) as well as Park and Park (2019) examine the mix of infrastructure necessary for bunkering LNG vessels in specific areas. Baresic et al. estimate the infrastructural requirements and associated costs in the European Union against the background of attaining greenhouse gas reduction targets. Park & Park developed a model to assess the demand for LNG infrastructure and apply this model to the port of Busan, South Korea. In both cases, LNG bunker ships are integrated into the respective models as fixed input parameters (e.g. in terms of carrying capacity, discharging capacity, operational costs). They are not the phenomenon under research. The European Maritime and Safety Agency (EMSA) (2018) and the classification society American Bureau of Shipping (ABS) (2017) published guidelines on LNG bunkering. These documents provide overviews on the application of LNG as ship fuel with special emphasis on LNG bunking. They include typologies of LNG bunker ships. Empirical research on the fleet of LNG bunker ships is very scarce. Corkhill (2018) and Sharples (2019) have compiled basic vessel information (e.g. year of entry into operation, owner/charterer, carrying capacity) for the existing and commissioned units. A more detailed appraisal of the state of the Northern European LNG Bunker Ship (NELBS) fleet and especially its operational profile so far has not been undertaken. The research question here reads as follows: How does the fleet of LNG bunker ships deployed to the Northern Europe operate and how is this expected to change in the future? The research results will foster an understanding of the state of LNG supply infrastructure for LNG vessels in general, which is of interest i.a. to government organisations co-funding LNG infrastructure as well as ship owners investing in LNG technology. 1.3 Methodological Approach The state of play of LNG bunker ship operation in Northern Europe will be assessed based on ship operation data provided by VesselTracker.com, a company that maintains a database with information on vessels, including vessel locations extracted from automatic identification system data (AIS).
170
A. Roß and K. Lange
From the dataset, it was possible to gather two types of events. First, ship-to-ship encounters between LNG bunker ships and LNG vessels, which reflect bunker operations. An encounter was identified based on proximity (vessels are within 30 m of one another) and speed (less than 5 km). Second, entries and exits of LNG bunker ships in areas marked as LNG storage and distribution facilities, which reflect loading and unloading operations. The investigated timeframe covers the months June to September 2019. Since the coverage of the AIS stations is rather comprehensive for the area under study, data availability is generally good. The dataset contained several inconsistencies (entries and exists of bunker ships to ports without an associated operation) suggesting missing events. The missing ship interactions could be harvested manually from the VesselTracker.com platform. Although all identified inconsistencies could be eliminated in this way, further “invisible” gaps cannot be ruled out completely. The relevant information that could be extracted from the dataset included the type of operation (bunking, loading/unloading), location, type of LNG vessel involved in the bunkering operation and, for some LNG bunker ships, the quantity of LNG transferred during loading and unloading operations. The future development of the fleet of LNG bunker ships was projected based on the status quo assessed via the ship operation data as well as anticipated developments of supply infrastructure and demand for LNG as ship fuel.
2 Supply Infrastructure for LNG Bunker Ships and Demand for LNG as Ship Fuel – State of Play and Future Developments LNG bunker ships satisfy the demand of LNG as ship fuel. They shuttle LNG between sources of supply, i.e. LNG liquefaction facilities or LNG terminals (incl. floating storage and regasification units), and LNG vessels in need of ship fuel. The state of supply sources as well as demand for LNG are key determinants of the state of the LNG fleet. A brief overview will be given on the status quo and the expected development of these two factors. Figure 1 is the output of desk research (mainly from operators’ websites, press releases, business news) on LNG supply infrastructure for LNG bunker ships. It covers the North Sea and Baltic Sea, i.e. the main part of Northern European waters. It is clear that supply infrastructure for LNG bunker ships is available. LNG liquefaction plants only exist in Norway and Russia, which are gas exporting countries. The distances between the individual facilities vary. In the future, the network of LNG storage and distribution facilities serving bunker vessels will become denser (Calderón et al. 2016). This growth is mainly made up by new LNG terminals, since these facilities (with few exceptions) primarily serve as points of entry for gas imports and many countries seek to diversify their sources of gas supply through LNG terminals. With 143 units in early 2019 (Sharples 2019), the global fleet of LNG vessels that rely on bunkering services (i.e. excluding LNG-propelled tankers transporting LNG as cargo) is still very small. A major hotspot is Northern Europe, due to the stringent emission regulations in the ECAs. However, only 1% of the overall regional fleet is LNG-propelled. LNG as ship fuel still constitutes only a niche market. Databases, e.g. by DNV GL, show that number of LNG vessels will increase in the immediate future
An Appraisal of the Northern European LNG Bunker Ship Fleet
171
Fig. 1. Map of LNG storage and distribution facilities and loading infrastructure for LNG bunker vessels in the North Sea and Baltic Sea (as of December 2019)
(Sharples 2019). Various studies (e.g. Danish Maritime Authority 2012; Calderón et al. 2016; Adolf et al. 2019) expect a tremendous long-term growth of LNG-fuelled vessels and, consequently, LNG as ship fuel. Although demand for LNG as ship fuel is and will be met via different bunkering modes, STS bunkering is expected to play a dominant role.
3 Operational Profile of the Northern European LNG Bunker Ship Fleet 3.1 Composition of the Fleet Desk research revealed that as of December 2019, nine units (originally) designated to Northern Europe had the capacity to supply LNG vessels with fuel. Seven more units were scheduled for delivery. It needs to be stressed that the size and composition of any fleet represents only a snapshot in time. In the investigated time period of June to September 2019, seven of the nine ships were actually deployed on bunkering operations in Northern Europe. These seven units constitute the NELBS fleet, whose operating profile is analysed in the following. LNG bunker ships come in various shapes. The main differentiating technical parameters, which also impact the operational profile, are carrying capacity, external vs. selfpropulsion as well as the classification as a seagoing or inland water vessel. The technical configuration of the NELBS fleet can be seen in Table 1.
172
A. Roß and K. Lange Table 1. Composition of the NELBS fleet (as of December 2019)
Ship name
Year of delivery
Carrying capacity
Propulsion
Classification IMO number, if applicable
Seagas
2013
180 m3
Self-propelled
Seagoing vessel 7382691
Engie Zeebrugge
2017
5,000 m3
Self-propelled
Seagoing vessel 9750024
Cardissa
2017
6,500 m3
Self-propelled
Seagoing vessel 9765079
Coralius
2017
5,800 m3
Self-propelled
Seagoing vessel 9769128
Kairos
2018
7,500 m3
Self-propelled
Seagoing vessel 9819882
FlexFueler1
2019
1,500 m3
External
Inland water vessel
LNG London
2019
3,000 m3
Self-propelled
Inland water vessel
With increasing demand for LNG as ship fuel, the NELBS fleet will grow, too. Seven units are scheduled for delivery by 2021 (see Table 2). Carrying capacities of these additional units ranges from very small to very large. An even further growth of the fleet is expected in the mid-term future, given a steady increase in demand for ship fuel as well as a greater dispersion of demand across the whole of Northern Europe. Table 2. Units scheduled for delivery/deployment in Northern Europe (as of December 2019) Operator
Expected year of delivery
Carrying capacity
Area of operation
Joint venture Avenir LNG 2019
7,500 m3
Northern Europe
Titan LNG
2020
760 m3
Antwerp and ARA region
Gasnor, subsidiary of Shell
2020
850 m3
Port of Bergen and beyond
Totale and Mitsui OSK Lines
2020
18,600 m3
Northern Europe
Elenger
2020
6,000 m3
Estonia
Gazprom Neft
2020
5,800 m3
Baltic Sea
Titan LNG
2021
8,000 m3
Antwerp and ARA region
An Appraisal of the Northern European LNG Bunker Ship Fleet
173
3.2 Type of Operations The maps in Figs. 2 and 3 depict STS bunkering operations of the NELBS fleet as well as loading and unloading operations at LNG storage and distribution facilities from June to September 2019. The design of LNG bunker ships enables them to supply ship fuel. However, they are essentially small-scale LNG tankers and are principally able to supply LNG terminals, too. This is why unloading operations are accounted for (see Sect. 3.2.4). 3.2.1 Loading Operations The larger five units of the NELBS fleet load LNG directly at storage and distribution facilities. Alternatively, bunker ships can be supplied by truck or STS by other (bunker) ships. The latter phenomenon occurred twice in the investigation period. The Cardissa supplied the LNG London and the LNG London supplied the Flexfueler1. Although a growth of the number of LNG terminals is anticipated, there will not be a storage and distribution facility in every port, as these are expensive investments. Hence, not all bunker ships will directly load LNG at such facilities, but will instead need to be supplied by alternative means. Since truck loading is less efficient (especially when sourcing the LNG over greater distances), bunker ships will increasingly be supplied by other ships. These may be larger bunker ships or other small-scale LNG carriers. In comparison to LNG terminals. LNG bunker ships feature lower CAPEX and OPEX (Barsic et al. 2018). This is why they can supply other bunker ships at relatively low cost. STS loading is also more flexible since ships can easily be deployed to areas where they are needed. 3.2.2 Bunkering Operations – Geographical Coverage 233 STS LNG bunkering operations of in total 31 LNG vessels were carried out by the NELBS fleet in the investigation period. All but one bunkering operation took place in Northern Europe (see Figs. 2 and 3). These operations are concentrated in several hotspots, which are either ports at which various LNG vessels call (Rotterdam, Gothenburg) or ports with liner traffic of individual LNG vessels/fleets (Zeebrugge, Visby/Nynäshamn, Stockholm). Next to operations in ports, 25 offshore bunkering manoeuvres could be identified. LNG vessels served offshore were either waiting for port entry or stopped for fuel supply on their way to other destinations. It is noteworthy that all bunkering operations were conducted within the North Sea and Baltic Sea as ECAs. In Norway, where LNG ship traffic is particularly dense, no STS bunkering operations took place. The units of the NELBS fleet satisfying the prevailing demand for STS LNG bunkering showed very different operating ranges. Some LNG bunker ships (Seagas, ENGIE Zeebrugge, LNG London) remained within a single port. Others travelled moderately to serve LNG vessels at two or three locations (Coralius, Flexfueller001). Two LNG bunker ships supplied fuel to LNG vessels also across wider distances. Of these two, one (Cardissa) undertook one exceptionally long trip to the Mediterranean. The second one (Kairos) supplied fuel at nine different locations within the Baltic Sea.
174
A. Roß and K. Lange
Fig. 2. Ship operations in Northern Europe – along the Southern Norwegian and Swedish coast and in the Baltic Sea
Fig. 3. Ship operations in Northern Europe – ARA region (June to September 2019)
For a bunker ship to be operated long-distance, it needs to be classified as seagoing. Furthermore, it needs to be self-propelled, because the cost of external propulsion would prohibitively drive up the costs of transportation. Apart from that, travelling over greater distances is economically only worthwhile if satisfactorily large volumes of LNG are supplied. Therefore, these LNG bunker ships need to feature a larger carrying capacity. The number of bunkering operations as well their distribution across Northern Europe reflect that LNG as ship fuel is currently still a niche market. The geographical coverage of the LNG supply offer will increase in the future. LNG will be supplied in a larger number of ports across the region. However, there will still be the need for supplying LNG over certain distances, since there will not be a terminal in every port. In a situation of a denser supply network across Europe, extremely long voyages to supply relatively small volumes are not likely to occur anymore.
An Appraisal of the Northern European LNG Bunker Ship Fleet
175
3.2.3 Bunkering Operations – Vessels Served LNG bunker ships served LNG vessels of different types in the investigation period (see Fig. 4). Ro-Ro cargo ship RoPax ferry Product tanker Oil tanker Container ship Car carrier Bulk carrier 0
2
4
6
8
10
12
No. of vessels Fig. 4. LNG vessels supplied by NELBS fleet bunker ships in Northern Europe (June to September 2019)
An important parameter for the operation of bunker ships is the parcel size of the ship fuel delivered. Information on volumes transferred by STS cannot be derived from the VesselTracker.com dataset. Although LNG vessels may take up volumes of less than a full load, the size of LNG vessels’ tanks serves as a proxy to identify the relation between volumes bunkered and carrying capacity of the LNG bunker ship. Figure 5 visualises this relation for the set of LNG vessels for which information on tank size could be researched. Apparently, the trend of carrying capacity of the LNG bunker ship increasing with tank size of the LNG vessel is not particularly strong, since large bunker ships also deliver LNG to vessels with small tanks. Generally, the range of tank sizes in this sample is narrow in contrast to the volumes of 400–20,000 m3 that are classified as suitable for STS bunkering by the European Maritime and Safety Agency (2018). Although today LNG bunker ships already supply a wide range of LNG vessel types, this portfolio of vessel types will still diversify in the future. In particular, the range of loads delivered to LNG vessels will broaden as tank sizes and, thus, the upper limit of volumes demanded will increase. The answer to this development is bunker ships of a larger carrying capacity. In a situation of increased demand for LNG as ship fuel, bunker ships will to a certain extent specialise with regard to the volumes they supply in accordance to their carrying capacity. In particular, larger bunkerships, which feature higher costs of transportation, will avoid providing small loads. 3.2.4 Feedering Operations Although LNG bunker ships are designed to perform STS fuel deliveries, three ships of the NELBS fleet also supply smaller terminals with LNG from liquefaction plants or large-scale terminals.
176
A. Roß and K. Lange
Carrying capacity of bunker ships [m³]
8000 7000 6000 5000 4000 3000 2000 1000 0 300
400
500
600
700
800
900
1000
1100
Tank size of LNG vessels [m³] Fig. 5. Correlation tank capacity of LNG vessels and carrying capacity of NELBS fleet units (June to September 2019)
Table 3. Volumes loaded and unloaded (June to September 2019) Cardissa
Coralius
Kairos
Range of volumes unloaded [m3 ]
2,215–3,988
683–2,732
1,605–4,011
Total volume unloaded [m3 ]
19,938
19,465
10,831
Total volume loaded [m3 ]
27,913
29,709
33,696
Total volume unloaded in relation to total volume loaded
71%
66%
32%
The dataset includes information on draught changes when entering and exiting LNG storage and distribution facilities. This information allows for an estimation of the quantity of LNG loaded and discharged, employing the concept of displacement. Volumes transferred were calculated based on bunker ships’ dimensions, an assumed block coefficient of 0.8 (Pak et al. 2020) and draught changes derived from the dataset. Aggregated information on volumes loaded and unloaded in the investigation period can be found in Table 3. In the investigation period, 26 supply trips to smaller terminals (two of them to Central Norway that are not included in the maps of Figs. 1 and 2) were identified. For the three aforementioned LNG bunker ships, this means that feeder services are currently an integral part of their business strategy. The Cardissa and the Coralius supply the bulk of the LNG they load to smaller terminals, not LNG vessels. With 32%, this share is still significant for the Kairos, too. Feedering can also be carried out by small-scale LNG tankers, instead of the technically more sophisticated LNG bunker ships that feature higher costs of transportation. Hence, the fact that bunker ships engage in feedering activities is an indication for too low demand for LNG STS supply.
An Appraisal of the Northern European LNG Bunker Ship Fleet
177
With a greater demand for LNG as ship fuel, it is anticipated that in the future LNG bunker ships will concentrate more on their primary purpose of supplying LNG vessels at the expense of feedering. In particular, this is expected to be the case for feedering over longer distances, because the increased costs of transportation play a greater role, the longer the voyage.
4 Conclusion Northern Europe is a frontrunner with regard to traffic of LNG-fueled vessels. Therefore, a relatively dense network of LNG infrastructure has developed. LNG bunker ships function as a backbone of this infrastructure, linking sources of supply (liquefaction plants, LNG terminals) and customers (LNG vessels). In contrast to conventional ship fuel, LNG is still a niche market, and the fleet of LNG bunker ships operating in Northern Europe is rather small. With a growing fleet of vessels running on LNG, a rapid growth of LNG bunker ships can be expected in the immediate future (see Table 2) and beyond. This implies that the geographical coverage of ship fuel provision will grow compared to the current situation, in which bunker ships supply fuel primarily in a limited number of hotspots. However, since LNG reloading points will not be available in every port, ships of larger capacity will supply LNG to end users as well as smaller units of LNG bunker ships. LNG bunker ships supply fuel to vessels of various types, but with similar tank sizes. As this portfolio diversifies and loads demanded get larger, the units of the NELBS fleet will specialise with regard to the volumes they supply in relation to their own carrying capacity. A surprising phenomenon is that some LNG bunker ships heavily engage in feedering services and supply smaller scale LNG terminals. This strategy secures profits in a situation of an emerging fuel market. However, in a thriving market, it is expected that feedering will play a lesser role.
References Baresic, D., Smith T., Raucci, K., Rehmatulla,C., Narula, N., Rojon, I.: LNG as a marine fuel in the EU. Market, bunkering infrastructure investments and risks in the context of GHG reduction. University Maritime Advisory Services, London (2018) American Bureau of Shipping: LNG bunkering. Technical and operational advisory (2017). https://ww2.eagle.org/content/dam/eagle/advisories-and-debriefs/ABS_LNG_Bunker ing_Advisory.pdf. Accessed 13 Nov 2021 Calderón, M., Illing, D, Veigaet, D.: Facilities for bunkering of liquefied natural gas in ports. Transp. Res. Procedia 14, 2431–2440 (2016). Elsevier, London Corkhill, M.: LNG bunker vessels. Fleet review (2018). https://www.rivieramm.com/opinion/opi nion/lng-bunker-vessels-fleet-review-23723. Accessed 28 Aug 2021 Danish Maritime Authority, North European LNG Infrastructure Project. A feasibility study for an LNG filling station infrastructure and test of recommendations (2012). https://www.anave.es/images/seguridad/danish_maritime_authority-north_european_ lng_infrastructure_project-mar_12.pdf. Accessed 28 Aug 2021
178
A. Roß and K. Lange
Köhler, J., Kirsch, D, Klukas, A., Timmerberg, S., Kaltschmitt, M.: Studie über die Marktreife von Erdgasmotoren in der Binnen-und Seeschifffahrt (2018). https://www.bmvi.de/Shared Docs/DE/Anlage/G/MKS/studie-marktreife-erdgasmotore-schifffahrt.pdf?__blob=publicati onFile. Accessed 13 Nov 2021 Pak, K.-R., Song, G.-S., Kim, H.-J., Son, H.-J., Park, H.-G.: Hull form design for resistance minimization of small-scale LNG bunkering vessels using numerical simulation. Int. J. Naval Archit. Ocean Eng. 12, 856–867 (2020). Society of Naval Architects of Korea, Seoul Park, N.K., Park, S.K.: A study on the estimation of facilities in LNG bunkering terminal by simulation. Busan Port case. J. Mar. Sci. Eng. 7, 354 (2019). MDPI, Basel European Marine Safety Agency: Guidance on LNG Bunkering to Port Authorities and Administration (2018). https://www.parismou.org/sites/default/files/EMSA%20Guidance%20on% 20LNG%20Bunkering.pdf. Accessed 28 Aug 2021 Schinas, O., Buttler, M.: Feasibility and commercial considerations of LNG-fueled ships. Ocean Eng. 122, 84–96 (2016). Elsevier, Amsterdam Sharafian, A., Blomerus, P., Mérida, W.: Natural gas as a ship fuel. Assessment of greenhouse gas and air pollutant reduction potential. Energy Policy 131, 332–346 (2019). Elsevier, Amsterdam Sharples, J.: LNG supply chains and the development of LNG as a shipping fuel in Northern Europe. In: OEIS Paper, NG 140. The Oxford Institute for Energy Studies. Oxford (2019) Stenersen, D., Thonstad, O.: GHG and NOx emissions from gas fuelled engines. Map-ping, Verification, reduction technologies. SINTEF Ocean AS, Trondheim (2017). https://midc.be/wp-con tent/uploads/2018/06/methane-slip-from-gas-engines-mainreport-1492296.pdf. Accessed 13 Nov 2021 European Union, European Parliament and European Council: Directive 2014/94/EU of the European Parliament and of the Council on the deployment of alternative fuels infra-structure (2014). https://eur-lex.europa.eu/legal-content/EN/TXT/HTML/?uri=CELEX:32014L0094& from=en. Accessed 28 Aug 2021
Investigating the Requirements of Automated Vehicles for Port-internal Logistics of Containers Hendrik Rose1(B) , Ann-Kathrin Lange2 , Johannes Hinckeldeyn1 , Carlos Jahn2 , and Jochen Kreutzfeldt1 1
2
Institute for Technical Logistics, Hamburg University of Technology, Hamburg, Germany [email protected] Institute of Maritime Logistics, Hamburg University of Technology, Hamburg, Germany
Abstract. With its potential of increasing economic efficiency, safety and process reliability automated driving offers substantial benefits for road freight logistics. Limited by technical and regulatory hurdles, first use cases of automated driving in logistics are in pilot testing stage on private areas or include traffic scenarios with fixed, foreseeable environmental conditions. These conditions partially exist in seaports. This article addresses the feasibility and specific requirements of automated trucks when implemented in seaport systems. A hybrid study consisting of a semi-structured interview survey and an extended project review is conducted to analyze the applicability of fully automated vehicles for the transportation of containers in port areas and its resulting requirements. Firstly, a review of currently existing automated road freight projects and their findings about container transportation is given. Secondly, port-related requirements for automated road freight transportation are derived as a result of the interview survey. The authors find main requirements in the technical feasibility, the operation of the vehicles as well as the organizational and process integration into the port system. Keywords: Autonomous vehicles transport · Port logistics
1
· Automated driving · Road freight
Automated Container Transportation in Port Logistics
Automated trucks have become the next great technological expectation for logistics. The work of truck driving has been a manual effort for decades and little has changed since the invention of the heavy-goods-vehicles in the nineteenth century. Riding, steering and navigating are still daily tasks of truckers, although new driver support systems have improved the working conditions. However, regular pauses and periods of resting of truck drivers as well as the higher cost for night shifts and the resulting lower number of tours at night add up to unused capacity of trucks. Furthermore, the current driver shortage threatens to further c The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 M. Freitag et al. (Eds.): LDIC 2022, LNLO, pp. 179–190, 2022. https://doi.org/10.1007/978-3-031-05359-7_15
180
H. Rose et al.
limit existing capacity. Hence, the replacement of manual work with automation would unleash great potential for improvement of productivity in transportation. A recent study estimates that automated trucks will halve delivery time and costs, while doubling the utilization of capacity [13]. The highest savings are expected removing the safety driver, a Swedish study finds [2]. Consequently, the development of automated technology is being pushed throughout the trucking industry by start-ups and original equipment manufacturers (OEMs). Besides the technical challenges of automated driving as such, it is essential to provide the necessary infrastructure and put appropriate rules and regulations in place [20]. For this reason, many companies have started to conduct research projects with automated trucks using special permission on public streets for test and validation purposes [1]. Using these findings, companies and researchers target to integrate driverless trucks into daily transportation processes. The logistics character of road transportation, however, is divers and shows many different facets. It ranges from long haul transports between remote hubs to short distances of stop and go for urban deliveries in city logistics. As road freight transportation includes a variety of different environmental conditions creating different levels of traffic complexity, the feasibility of automated vehicles (AVs) is highly application dependent. Each use case for AVs poses a set of different requirements on the vehicle technology and the system it is applied to. These requirements have to be analysed and considered, planning the integration of AVs which is subject to this research. This study focusses on container transportation in seaports. When it comes to transportation within international supply chains, seaports play a vital role. Besides containers being transported to and away from the sea port system, the majority of container transports between the logistics nodes within the port area is carried out by trucks. The logistics system of ports is different in comparison to other logistics systems. Ports and their surrounding facilities usually show a high density of logistics nodes on a relatively small area, different to hub-to-hub transportation and more comparable to cities. This implies short distances and many stops for trucks. At the same time, the size and weight of goods, which are usually transported in containers among these nodes, is rather big in comparison to city logistics. Therefore, the required trucks have to be able to carry heavy loads on short distances, which poses special requirements on trucks in comparison to city logistics. In addition, traffic situations seem less complex, especially compared to city logistics. Automating port-internal container transportation, hereafter called inter-terminal transport (ITT), therefore, poses specific requirements to the vehicle technology and the port system itself. This implies the following research questions: – RQ1: What is the state of the art technology for automated port-internal container transportation on the road? – RQ2: Which requirements can be derived when implementing automated container transportation using trucks? The above research questions are to be answered giving a project overview for AVs used in ITT. Further an interview study shows a set a requirements and
Automated Container Transport Systems
181
remarks to be addressed when implementing AVs for container transportation. The remaining paper is structured as follows: the next section will show that the current state of research has missed to investigate the use of automated trucks in container port logistics. Section three outlines the research methodology, which is used to answer the research questions. The results are presented in section four, before the paper concludes with some interpretations, limitations and final remarks in section five.
2
State of Research
The variety of logistics use cases for automated transportation opens a wide field of study for researchers focusing on different topics. These topics include general overviews of technical feasibility, cost analysis for automated road freight use cases and investigations about the implications of automated trucks on the road freight industry. Graf and Anner [11] present an overview of current vehicle providers offering automated driving solutions for transportation. They identify potential applications of automated technologies in trucking and last mile delivery, without linking technology aspects to domain specific logistics operations. Considering economic benefits, Engholm et al. [7] investigate the cost efficiency of driverless trucks categorized by weight, stating a potential cost decrease of up to 58 percent. Several other studies also conclude high cost benefits [2,17,18]. Cost calculations however, do not consider the specific node costs in port systems including different handling equipment from usual logistics hubs. Besides high economic benefits, automation technology will have high impact on processes, infrastructure and the way transport business is conducted. Therefore, many studies analyze implications of AVs on the trucking industry [25,28,29,31] identifying impacts on safety, accountability, compliance [25] and technological barriers [28,29]. Furthermore, policy recommendations of how to cope with issues like cyber security, safety and infrastructure challenges are given [15,18]. Studies describing the impacts of automated trucks take a general approach focussing on technical, social and regulatory topics without specific relation to port operations. Currently, the technical possibilities of applying automated technologies in public traffic are limited. Considering the level of traffic complexity, automated trucks seem to be most feasible on fixed routes with foreseeable traffic situations or on private fenced areas. A possible application of automated trucks in controlled environments and fixed conditions can be found in seaports. The consideration and improvement of innovative transport systems within ports are continuously increasing, both in practice and in science, in view of the progressive shortage of personnel in the transport sector and the increasing time and quality demands on the various actors in the port. Possibilities here range from automated continuous conveyors to underground rail systems to AVs in mixed traffic on the roads in the port [23]. Shin et al. [27] provide an overview of examples and developments in intermodal transport systems. One focus is the consideration of transports on the road
182
H. Rose et al.
with special attention on sustainability aspects. The systems analyzed here are only partially automated. Heilig and Voß [12] also focus on container transports between different handling companies in the port, which are only automated under certain circumstances. In doing so, they highlight current technologies and future trends, and analyze the methods used to improve these technologies. Nellen et al. [23] detail these considerations by limiting them to automated transport systems that have been studied or already implemented worldwide. However, the focus is only among other things on the studies of automated systems on the road. Altogether, the authors are not aware of a detailed overview of automated container transport on public roads in ports and the consequences for logistics companies in such systems. Given the potential economic benefits using AVs and the challenging need for optimal transport and handling processes in sea ports, the overall objective of this study is to assess the general feasibility of AVs for ITT and its resulting requirements answering the above mentioned research questions (see Sect. 1). Automated container transportation on the road is not being implemented yet. Therefore, RQ1 is posed to analyse the current stage of technical development for automated transportation as well as its technical limitations and challenges to the port system. Further, handling processes at the logistics nodes pose requirements to the system integration of AVs. Therefore, RQ2 addresses relevant requirements of logistics companies in ports for using AVs. In the following both questions are answered conducting a detailed project review of automated road freight systems and a semi-structured interview survey analyzing implications of AVs for container transportation.
3
Methodology
A two-phased research approach was chosen in order to provide an overview on the state of automated road freight vehicle systems and an outlook on mid-future technology development in general and in specific for ITT (see Fig. 1). The reason for the two phases was the interdisciplinary character of the research projects and the scattered sources of information, which can be found both in literature and practice. At first, an extensive project and literature review was conducted. Technical solutions of AVs, press releases and articles on AVs were analysed. Literature assessing the topic of automated freight vehicles for heavy loads on private or public areas of at least level 4 according to SAE (Society of Automotive Engineers) automation standards was marked as relevant to the topic. The analysed vehicle projects were categorized according to specific application, its relation to port operation and its driving limitation. The general considerations of automated road transport were complemented by an extensive literature analysis on port-internal transport systems for containers. This analysis is based on the results of Nellen et al. [23] and was supplemented by new findings. The literature databases Google Scholar, Scopus and IEEE were used for searching. In the second phase, 28 semi-structured interviews with different actors from the automotive sector as well as from port logistics companies were conducted
Automated Container Transport Systems
1. State of the Art
183
2. Port & System
Phase Truck
Method
Result
Tractor
Project & Literature Review
Semi-structured Interview Survey
Categorization of automated freight vehicle projects
Vehicle Projects
Analysing requirements from manufacturer and port system
Port Perspective
Technical State of the Art Relation to port system
Requirements from vehicles Requirements from port stakeholders
Fig. 1. Analysis concept
based on the findings of the literature review. Interviewees in the automotive sector stem OEMs of trucks, automated driving technology start-up companies, manufacturers for automated tractor-trailer systems and related research institutes. Here, the main topics were the specific company’s existing and planned technical solution for automated driving, including level of automation and applied technology, existing technical and road infrastructure requirements as well as future developments of automated technology for on road logistics. Furthermore, the topics of future operation and business models, the challenges of automated driving and the effects of high automation on the overall vehicle price were addressed. In the case of ITT, relevant interview partners were identified as logistics nodes in the port and haulage companies. Besides container terminals, logistics nodes also include empty container depots, container packing stations and customs and veterinary service providers. Furthermore, there are many companies in the port that act as shippers. Also relevant is input from port authorities and other public stakeholders. For the interviews, individual companies as well as interest groups of a port in the North Range were chosen for each type of company in order to get the broadest possible overview of the range of interests of port stakeholders. The findings of the literature review and the semi-structured interviews are analyzed in Sect. 4.
4
Results
The results of the study show a variety of challenges, which have to be addressed in order to successfully use automated trucks for container logistics in the port area. These challenges arise from different backgrounds, technologies, concepts of the operating system and integration, infrastructure requirements and business models. Altogether, ten automated driving projects or conceptual vehicles were identified as relevant to the topic. This was complemented by interviews of
184
H. Rose et al.
14 industry experts from the field of automated driving, intelligent traffic management systems and tractor-trailer manufacturers. A short summary about the most relevant topics of the AVs projects, infrastructure requirements, the operation model, service and pricing models, as well as the future development of automated trucking of SAE level 4 is given. The project review (see Table 1) shows that approaches for automated trucking are currently tested in freight applications for hub-to-hub and container transportation partially including public road infrastructure. Table 1. Automated trucking projects and vehicles Project
Company
Application Port
Domain
Source
POD AET 3-4
Einride
Hub2Hub
–
Highway/ rural
[6, 30, 32]
HH truck pilot
MAN
Container transport
Terminal
private terminal area
[21]
ANITA
MAN
Container transport
ITT
Limited public [21] road
Highway pilot
Plus
Hub2Hub
–
Highway
[24]
Hub2Hub
TuSimple & Scania
Hub2Hub
–
Highway
[10]
Autonomous freight network
TuSimple
Hub2Hub
–
Highway
[1]
Vera
Volvo
Container transport
Terminal ITT/
Port area
[14]
Shanghai port area
TuSimple
Container transport
Terminal/ ITT
Port area
[8]
Automated Konecranes Container terminal tractor transport
Terminal (fenced)
No human interference
[19]
Autonomous yard tractor
Terminal
Private area
[9]
GAUSSIN
Logistics yard
The technical solutions for automation consist of upgrading conventional trucks with automation technology (sensing and processing) and the development of tractor vehicles without a conventional cabin. First highly automated approaches without a safety driver are currently being tested in specific operational design domains such as highways and interstate routes in the US and China [1,8] . Full automated tractors are mostly operating exclusively in private area including logistics yards or terminals with only short driving distances on public roads near the facilities [14].
Automated Container Transport Systems
185
In terms of projects assessing container transportation, vehicle companies are conducting extensive research. MAN Truck and Bus has been carrying out extended research activities on integrating an automated heavy truck into terminal processes and driving partly on public roads in the Hamburg Truck Pilot and ANITA projects [21,22]. The technology start-ups TuSimple and Plus have successfully operated automated trucks in port operations transporting containers within the container terminals and partially on public roads [8,24]. Volvo is further investigating automated container transportation in their Vera projects conducting pilot tests between the sea port of Gothenburg and a Volvo site [14]. The existing vehicle projects show that container transportation using AVs is under ongoing investigation the industrial landscape. While yard and terminal transportation were successfully tested and partially implemented, ITT including public roads is limited due to the technical requirements for save driving in port traffic. To further investigate the requirements and challenges of automated trucking for ITT scenarios, a semi-structured interview survey with 14 experts from the automated trucking industry and 14 stakeholders from the port landscape was conducted. Industry experts for automated trucks were asked about the technical feasibility of automated trucks for ITT, infrastructure requirements, operating and business models and automation costs for transportation. Port stakeholders were interviewed about requirements for the integration of AVs into the existing processes and their implications. The main findings are summarized and categorized in Fig. 2.
Fig. 2. Requirements for automated container transportation
In terms of the technical feasibility for automated trucks in port transportation at this point of time, there are still serious hurdles to take. If public traffic with cyclists and pedestrians is included, sensors, data processing and decision making algorithms do not show the required maturity to handle the complexity of traffic situations. Although, seaports mostly have only vehicle traffic with
186
H. Rose et al.
rather fixed traffic scenarios, complex situations including pedestrians can occur at the logistics nodes. To support the AVs with additional traffic information and create redundant safety cases, additional sensor infrastructure including Vehicleto-Everything (V2X) communication and traffic monitoring is seen as essential. Enabling automated driving in port areas was assessed highly dependent on environmental conditions. In general, the majority of interviewees was confident that managing complex environmental conditions including traffic situations in ports will be possible within the next decade. Considering that automated trucks imply different application dependent infrastructure measures and will firstly be applied on a fixed route network, they will be part of a system including an operations center. The functions of an operations center include the management of orders and the fleet of trucks, in particular the teleoperation of single vehicles. Depending on the technical development of automated driving, different concepts for teleoperation ranging from passive maneuver decisions to complete remote operation in case of complete system failure are possible. Various concepts for vehicle control result in different requirements for data transmission. The location of the operations center in the port itself is not required. However, assuring functional logistics processes at the logistics nodes requires the integration of the fleet’s order dispatching and coordination into existing data transmission processes. Beside the integration into logistics processes the assumption of a shift of business models was indicated throughout the interviews. Requiring high technical expertise for testing, calibrating and deploying the vehicle fleet, vehicle manufacturers will increasingly offer services, such as transportation operation and vehicle maintenance. Alternatives range from operating a customer owned vehicle fleet to manufacturer owned and operated fleets. Although automation costs are assumed to decrease in the future, it was assumed by the interviewees that AVs will firstly be applied in large vehicle fleets. This results to the expectation to put a higher level of competitive pressure on small logistics providers not being capable of competing against the decreasing transportation prices. In addition to container terminals, packing stations and empty container depots, there are also small logistics providers in seaports who face major challenges by the increasing price competition due to the automation of large fleets. As a result, consolidation effects in the port landscape might become apparent. The structural conditions must take into account the limited availability of space within port areas. When AVs are used in an existing port, it is often not possible to create exclusive roads or handling areas at the logistics nodes for the AVs, due to the severely limited open space. In this case, the only solution is that the AVs share the public roads. If this is not possible for the entire infrastructure, individual routes could also be enhanced for automated transports. This can be done, for example, by means of additional sensors or markings on roads. If it is not yet possible to have mixed operation with manual vehicles due to insufficiently advanced technical development, potential alternatives are spatial or temporal separation, for example by banning manual vehicles from driving at night. Furthermore, it may not be possible to integrate all required logistics
Automated Container Transport Systems
187
nodes in such a system, because not all roads are capable for the spatial or temporal separation of manual and automated transports. When implementing technical and structural changes, special attention must be paid to the organizational requirements. The effects on logistics flows in the port are important, such as the competition of temporal separation for automated transport to heavy transports at night. Another important basic prerequisite for automated transports in the port is to ensure flawless data transfer between the various players. A high level of data transparency is essential for the operation of such a system. Examples of such data are the expected arrival times of the containers at the nodes or cargo-related information. Both are legally required or operationally needed at the nodes, where the container is transferred. Through this data transfer, congestion in front of and on the logistics nodes can be reduced or avoided altogether. This enables the precise scheduling of the various transports and thus a good utilization of the automated transport system. To ensure smooth handling of automated trucks at the different logistics nodes, process adaptions are required. High safety standards must be taken into account, when installing the necessary sensors and equipment to perform the tasks that are otherwise carried out by humans in automated trucks, since the final control instance is eliminated. An example of this is the notification by the truck driver that the container restraints have been released by the crane or reach stacker prior to acceptance. If this was not fully released, the container would partially lift the truck with it, causing considerable economic and possibly also human damage. The release and subsequent check must be carried out with a very high level of reliability by corresponding systems in the case of automated transports. Furthermore, it must be ensured that the node’s processes cannot be impeded in the event of system failures. This can be achieved, for example, by manual control in exceptional cases or by creating redundancies in routes or handling equipment. In summary, the stakeholders in the port were positive about automated transports in the port, as long as the described preconditions are met. In particular, the port companies see automated transport as a way to counteract the acute driver shortage and increase the efficiency of the processes in the ports by reducing the idle time of trucks at night and increasing transparency within the logistics chain. Many of the necessary adjustments to both internal and crosscompany processes and structures have already been started in individual pilot projects and will need to be further explored and tested in the coming years.
5
Conclusion
As automated driving is currently evolving and continuously overcoming the technical challenges to be solved, it is important to understand how AVs can be used for different application scenarios such as ITT in seaports. Technical and process requirements are crucial when it comes to integrating automated container transportation into existing port systems.
188
H. Rose et al.
The results of this study show that the use of AVs for ITT can be expected in the midterm future. The project review implicates companies conducting tests and pilot projects integrating AVs into on-road container logistics in ports. Still limited in scope and application area, it is likely that manufacturers start with some prototypes and proof-of-concept projects before scaling its application to complete systems capable of operating in the whole port area. There are a few advantages, which will simplify automated transports in ports. The area of a port is usually relatively small and space is limited, but characterized with a high density of logistics nodes. This will allow the definition of particular routes and equipping them with the necessary infrastructure for automated trucks, if sufficient space is available. Hence, automated trucking will not be possible without adequate infrastructure and some use of additional space. However, it will further be easier for a control center to manage the transports within such a limited area. The fixed route system will keep enabling sensor and communication infrastructure measures within predictable dimensions. At the same time, heavy containers can be well handled by automated trucks. In doing so, automated trucks look like a good fit for interterminal transports with resulting requirements of the technical feasibility, the fleet operation as well as different business models. However, some challenges have to be resolved before automated trucks ride smoothly between container terminals and other logistics nodes in the port. Up to now, the list of unresolved technological problems has been long and it also contains serious questions beyond autonomy, for example the use of new propulsion technology (hydrogen vs. batteries). Furthermore, also legal prerequisites have to be developed, such as the law of automated driving in Germany [5]. On the other side the feasibility of automated ITT is highly dependent of the integration requirements of the port system. Implications of AVs at logistics nodes, processes at the terminals and the overall organization are far-reaching and have to be well understood automating ITT. It is doubtful that all stakeholders will be able to continue their business in the same manner with just using automated trucks. However, these questions were not the objective of this study and have to be addressed in further research. This study shows that automated transportation is a disruptive force, which has got the potential to radically change the logistics landscape. The use of automated trucks in ports is promising and large potential for improvement of efficiency is expected, due to the organizational and spatial structure of these systems. Further research has to be conducted to fully understand this technology and its implications for communities, companies and society.
References 1. Ackerman, E.: Robot trucks overtake robot cars: this year, trucks will drive themselves on public roads with no one on board. IEEE Spectr. 58(1), 42–43 (2020) 2. Andersson, P., Ivehammar, P.: Benefits and costs of autonomous trucks and cars. J. Transp. Technol. 9, 121–145 (2019). https://doi.org/10.4236/jtts.2019.92008
Automated Container Transport Systems
189
3. Chottani, A., Hastings, G., Murnane, J., Neuhaus, F.: Distraction or disruption? Autonomous trucks gain ground in US logistics. McKinsey and Company (2018). https://www.mckinsey.com/industries/travel-logistics-and-infrastructure/ourinsights/distraction-or-disruption-autonomous-trucks-gain-ground-in-us-logistics, Accessed 07 Aug 2021 4. Dawkins, T., G¨ undo˘ gdu, C.: Autonomous Trucks: An Opportunity to Make Road Freight Safer. Cleaner and more Efficient, World Economic Forum (2021) ¨ 5. Deutscher Bundestag: Entwurf eines Gesetzes zur Anderung des Straßenverkehrsgesetzes und des Pflichtversicherungsgesetzes - Gesetz zum autonomen Fahren. Drucksache 19/27439 (2021) 6. Einride: Einride launches next-generation Pod. makes autonomous electric transport commercially available wordwide (2020). https://www.einride.tech/press/ einride-launches-next-generation-pod-makes-autonomous-electric-transport. Accessed 09 Sept 2021 7. Engholm, A., Pernest˚ al, A., Kristoffersson, I.: Cost analysis of driverless truck operations. Transp. Res. Rec. 2674(9), 511–524 (2020). https://doi.org/10.1177/ 0361198120930228 8. Gasgoo: TuSimple gets Shanghai’s nod to road test self-driving trucks (2021). https://autonews.gasgoo.com/icv/70018084.html. Accessed 09 Sept 2021 9. GAUSSIN: Automated Yard Tractor (2021). https://www.gaussin.com/atmautonomous. Accessed 09 Sept 2021 10. Gilroy, R.: TuSimple, Trato Group Develop Autonomous Trucks in Europe (2020). https://www.ttnews.com/articles/tusimple-traton-group-develop-autonomoustrucks-europe. Accessed 09 Sept 2021 11. Graf, L., Anner, F.: Autonomous vehicles as the ultimate efficiency driver in logistics. In: Disrupting Logistics, pp. 191–206 (2021). https://doi.org/10.1007/978-3030-61093-7 12. Heilig, L., Voß, S.: Inter-terminal transportation: an annotated bibliography and research agenda. Flex. Serv. Manuf. J. 29(1), 35–63 (2017). https://doi.org/10. 1007/s10696-016-9237-7 13. Hodes, C., McCabe, P., McGeeney, A., Smith, S., Sun, I., Awwad, M.: The future of autonomous trucks in the supply chain industry. In: Proceedings of the IIE Annual Conference, pp. 203–208. Institute of Industrial and Systems Engineers (IISE) (2020) 14. Hoffmann, J.: Autonomer Lkw u ¨ r den Hafen: Vera startet Einsatz (2019). https:// www.eurotransport.de/artikel/autonomer-lkw-fuer-den-hafen-volvo-vera-startetersten-einsatz-10809169.html. Accessed 09 Sept 2021 15. ITF: Managing the Transition to Driverless Road Freight Transport. International Transport Forum Policy Papers, No. 32. OECD Publishing (2017) 16. Kassai, E.T., Azmat, M., Kummer, S.: Scope of using autonomous trucks and lorries for parcel deliveries in urban settings. Logistics 4(3), 17 (2020). https://doi. org/10.3390/logistics4030017 17. Keese, N., Berhardt, W., Dressler, N., Baum, M., Rentzsch, W.: Automated Trucks. The next big disruptor in the industry? Roland Berger. M¨ unchen/ Chicago (2016) 18. Keese, S., Aulbur, S., van Marwyk, K., Rentzsch, W.: Shifting up a gear. Automation, electrification and digitalization in the trucking industry. Roland Berger (2018) 19. Konecranes: Automated Terminal Tractor. https://www.konecranes.com/ equipment/container-handling-equipment/automated-terminal-tractor. 09 Sept 2021
190
H. Rose et al.
20. Mallozzi, P., Pelliccione, P., Knauss, A., Berger, C., Mohammadiha, N.: Autonomous vehicles: state of the art, future trends, and challenges. In: Automotive Systems and Software Engineering, pp. 347–367. Springer, Cham (2019). https://doi.org/10.1007/978-3-030-12157-0 16 21. MAN Truck and Bus: Das Terminal 4.0 im Test (2020). https://www. mantruckandbus.com/de/innovation/das-terminal-40-im-test.html. Accessed 09 Sept 2021 22. MAN Truck and Bus: Hamburg TruckPilot feiert Erfolg: Prototypen-Lkw bew¨ ahren sich im autonomen Praxistest (2021). https://press.mantruckandbus. com/corporate/de/hamburg-truckpilot-feiert-erfolg-prototypen-lkw-bewaehrensich-im-autonomen-praxistest/. Accessed 09 Sept 2021 23. Nellen, N., Grafelmann, M., Ziegenbein, J., Lange, A.-K., Kreutzfeldt, J., Jahn, C.: Literature classification on container transport systems for inter-terminal transportation. In: Freitag, M., Haasis, H.-D., Kotzab, H., Pannek, J. (eds.) LDIC 2020. LNL, pp. 52–61. Springer, Cham (2020). https://doi.org/10.1007/978-3-03044783-0 5 24. Plus: Plus Completes Driverless Level 4 Semi Truck Highway Demonstration (2021). https://plus.ai/press-release/Plus-Completes-Driverless-Level-4Semi-Truck-Highway-Demonstration.html. Accessed 09 Sept 2021 25. Pudasaini, B., Shahandashti, S.M.: Impacts of automated trucking technologies on the critical issues of trucking industry. In: Zhang, G. (eds.) International Conference on Transportation and Development 2020, Seattle, Washington (Conference Cancelled), 26–29 May 2020, pp. 37–47. American Society of Civil Engineers, Reston, VA (2020) 26. SAE International: Taxonomy and Definitions for Terms Related to Driving Automation Systems for On-Road Motor Vehicles. International Standard 2018. SAE International (2018) 27. Shin, S., Roh, H.-S., Hur, S.-H.: Technical trends related to intermodal automated freight transport systems (AFTS)*. The Asian J. Shipping Logist. 34(2), 161–169 (2018). https://doi.org/10.1016/j.ajsl.2018.06.013 28. Short, J., Murray, D.: Identifying Autonomous Vehicle Technology Impacts on the Trucking Industry. American Transportation Research Institute (ATRI) (2016) 29. Slowik, P., Sharpe, B.: Automation in the long haul: challenges and opportunities of autonomous heavy-duty trucking in the United States. ICCT - The International Council On Clean Transportation (2018) 30. Vaish, E: Driverless electric truck starts deliveries on Swedish public road (2019). https://www.reuters.com/article/us-einride-autonomous-swedenidUSKCN1SL0NC. Accessed 09 Sept 2021 31. van Meldert, B., de Boeck, L.: Introducing autonomous vehicles in logistics: a review from a broad perspective. Working Papers of Department of Decision Sciences and Information Management, Leuven 543558, KU Leuven. Faculty of Economics and Business (FEB), Department of Decision Sciences and Information Management, Leuven (2016) 32. Verpraet, I.: SKF and Einride trial autonomous truck for cross-facility shipments (2021). https://www.automotivelogistics.media/inventory-management/ skf-and-einride-trial-autonomous-truck-for-cross-facility-shipments/41677.article. Accessed 09 Sept 2021
Unmanned Vessels and the Law Eva Ricarda Lange(B) Universität Bremen, Fachbereich Rechtswissenschaft, Universitätsallee, GW1, 28359 Bremen, Germany [email protected]
Abstract. Unmanned and autonomous vessels are a multi-disciplinary research topic. The related legal research analyses international maritime law and depicts several legal challenges to be met. The article provides an overview of the legal literature, focusing on three aspects concerning the commercial shipping industry. First, the legal status of unmanned and autonomous vessels in maritime law will be discussed. Almost unanimously, such vessels are considered as ‘ships’ within the scope of maritime law. Second, the minimum safe manning requirement governed by several conventions will be examined. The main issue is whether the statutory provisions prevent unmanned operations at all, or whether a more favourable interpretation can be applied. Third, the duty to maintain a proper lookout will be examined. Scholars are reluctant to interpret the respective rules broadly and call for amendments considering the technical peculiarities. Keywords: Unmanned vessels · Autonomous vessels · Minimum safe manning requirement · Lookout duty · IMO
1 Introduction Autonomous and unmanned vessels are no longer found only in science fiction movies. They have become reality. Multiple research projects have developed prototypes and are conducting trials. This technical development upends maritime law at all levels and is likely to cause tensions with long developed and acknowledged legal principles. This literature review shall give an overview of the current legal situation and the research on three aspects; whether unmanned and autonomous vessels are ships within the meaning of maritime law, the requirement of minimum safe manning, and the duty to maintain a proper lookout. To give an overview of the present situation, some of the ongoing research projects regarding unmanned maritime vessels (UMV) are presented, including unmanned remotely controlled vessels and unmanned autonomous vessels. 1.1 Current Legal Situation Up until now, neither new legal instruments regulating UMV nor amendments to existing international conventions have been adopted. However, the International Maritime Organization (IMO) has not been entirely idle and the Maritime Safety Committee (MSC) of © The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 M. Freitag et al. (Eds.): LDIC 2022, LNLO, pp. 191–203, 2022. https://doi.org/10.1007/978-3-031-05359-7_16
192
E. R. Lange
the IMO adopted interim guidelines for ‘Maritime Autonomous Surface Ships’ (MASS) trials at its 101st session in 2019. [1] Between 2018 and 2021 the MSC and its member states conducted a regulatory scoping exercise to provide an analysis of international maritime law provisions possibly affected by MASS and revealing the most effective way of approaching MASS. It was concluded by adopting a circular during the MSC’s 103rd session in May 2021. [2] Eventually, the IMO Strategic Plan for the period 2018 to 2023 includes the aim to integrate new and advancing technologies in the regulatory framework [3]. National legislators, in contrast, were less active and have not updated their legislation. [4] However, the classification societies DNV and Lloyd’s Register have adopted codes and guidelines dealing with UMV. [5] Lloyd’s Register even issued the first classification certificate to a UMV in June 2021 [6]. 1.2 Technology and Current Research Projects The following overview is not comprehensive but shall illustrate some ongoing research projects. The Norwegian company Yara runs trials with the container vessel “Yara Birkeland” since November 2021. [7] In Trondheim, Norway, a full-scale prototype of the autonomous and zero-emission passenger ferry “MilliAmpère 2” was launched in May 2021. [8] It is expected to cross the river Nidelva by 2022. [9] The Norwegian Maritime Authority has announced a test area for autonomous vessels in the Trondheim Fjord already in 2016. [10] The above-mentioned certificate was issued for the SEA-KIT 12m X-Class, which took part in a 22 days remotely supervised trial southwest of the UK 2020. [11] The autonomous Finnish car ferry “Falco” completed its maiden voyage in December 2018 and is now further tested [12].
2 UMV as ‘Ship’ and ‘Vessel’ Within the Meaning of International Maritime Law The design and equipment of vessels have changed fundamentally over the centuries; however, the presence of a crew on board was a constant in this development. Until recently, it was not conceivable that ships would sail the ocean unmanned. Hence, the first legal question raised when dealing with UMV is whether UMV are ‘ships’ within the meaning of maritime law. As the globally accepted legal framework provides certainty for commercial operations, the application of the framework has to be determined. The United Nations Convention on the Law of the Sea (UNCLOS) is often referred to as the “constitution of the seas”. It provides a legal framework for all marine and maritime activities and is ratified by 174 states. [13] UNCLOS prescribes rights of the member states relating to shipping, their area of jurisdiction and grants ships several rights crucial for commercial use. The most important rights are subsequently illustrated. If UNCLOS were inapplicable, UMV operations would take place in a legal vacuum and legal predictability would not exist.
Unmanned Vessels and the Law
193
2.1 Rights Granted to Ships and Vessels UNCLOS grants ships the freedom of navigation in straits used for international navigation and on the high sea in Art. 38 (2) and 87 (1) (a). In territorial seas and archipelagic waters, ships have the right of innocent passage. Every member state may issue a flag for vessels that enable port calls for (un)loading cargo because certificates issued by a flag state are mutually recognized according to Art. 217 (3) UNCLOS. This saves ships from the obligation to comply with the national law of the concerned port state. Hence, without being in the scope of application of UNCLOS, port calls and international trade are factually impossible for UMV. The next section examines whether UNCLOS applies to UMV. 2.2 Definition in International Maritime Law UNCLOS neither defines ‘ship’ nor ‘vessel’, but the Convention uses both terms interchangeably. [14] However, other international treaties governing shipping provide definitions considering their respective objectives and one can refer to them to define the scope of UNCLOS. [15] According to Reg. 3 (a) of the Convention on the International Regulations for Preventing Collisions at Sea (COLREGS), a vessel is “every description of watercraft […] used or capable of being used as a means of transportation on water.” Other treaties provide similar definitions [16] and none of them refer to a crew or master, so the manning is not decisive. All of them can therefore cover UMV [17]. Although the wording is quite clear, the issue is discussed as all conventions were enacted having manned vessels in mind. The Comité Maritime International (CMI) affirms the question because UMV are vessels or ships by virtue of size, features, and functions when engaged in international trade. [18] It is unjustified to discriminate UMV conducting the same task as traditional vessels. [19] Moreover, UNCLOS does not explicitly exclude UMV from its scope. Finally, the IMO dealt with UMV extensively. Hence, the IMO assumes that regulating UMV is within its field of competence [20]. To conclude, UNCLOS and international maritime law apply to UMV. As UMV are a new development not mentioned in any convention explicitly, it should be clarified in the statutory texts whether UMV are ships or vessels in legal terms [21]. 2.3 Definition in National Maritime Law Irrespective of international law, every member state of UNCLOS is entitled to issue a flag according to its legislation. Therefore, the national definition is decisive for a UMV to obtain a flag certificate. In Germany, the terms ‘vessel’ and ‘ship’ are not defined by legislation, but in 1951, the Federal Court of Justice held that “a ship in the legal sense is understood to be any buoyant vessel of not entirely insignificant size, provided with a cavity, the purpose of which entails that it is moved on the water”. [22] This definition is widely accepted by scholars [23] and applied by the national authorities. Other national definitions neither refer to a crew, except for the French. [24] UMV operators may then be optimistic that UMV can fly flags. However, Suri marks correctly that also national law should expressly include UMV to streamline the administration process and to save time during court procedures [25].
194
E. R. Lange
In conclusion, national and international maritime law most probably apply to UMV. Clarification is nevertheless needed and could be best achieved by expressly including a clear definition of and reference to UMV in the legal texts.
3 Requirement of Safe Manning Several international conventions require a vessel to be safely manned. Traditionally understood, this calls for human presence on board. As the essence of UMV are unmanned operations, this requirement seems to be most challenging. 3.1 Regulatory Framework Art. 94 IV (b) UNCLOS provides for the flag state to ensure “that each ship is in the charge of a master and officers who possess appropriate qualifications”. The Convention for the Safety of Life at Sea (SOLAS) requires in Ch. V/14 sufficient manning of each ship from the point of view of safety of life at sea. It does, however, not prescribe a concrete number and leaves the vessel-specific assessment to the flag state. The non-binding IMO Principles of Minimum Safe Manning [26] serve as guidelines to the flag states. The International Convention on Standards of Training, Certification and Watchkeeping for Seafarers (STCW Convention) prescribes minimum qualification standards for mariners. Ch. VIII/2–2.1 of the Annex requires that “the officers in charge of the navigational watch […] shall be physically present on the navigation bridge or in a directly associated location […]” [emphasis added]. The Hague-Visby-Rules and the Rotterdam Rules – contractual clauses governing the carrier’s liability – are incorporated into most of the maritime freight contracts and also refer to a minimum manning level. 3.2 Status of Opinion The scholarly discussion of the manning requirement examines each above-mentioned convention separately. As an introduction, a theoretical approach shall be presented. The Level of Autonomy, the Position of the Crew, and the Manning Level Ringbom outlines three factors to be distinguished when dealing with UMV and the manning requirement. Firstly, the level of automation is decisive; secondly, the location of the crew has to be considered. Thirdly, the number of crew members, even if closely related to the level of automation, gives rise to legal challenges, independent of the location of the crew. [27] Although the factors affect each other and are intertwined, they have to be analysed separately to identify concrete legal questions. Ringbom points out that these three factors form the axes of a three-dimensional coordinate system having its point of origin at a fully manned, manually operated vessel. Along the respective axes, all steps of automation, remote operation, and reduction of crew can be mapped separately for each vessel function. Eventually, Ringbom points out that each operational action has to be assessed separately as every operational task legally requires a different level of onboard manning or human involvement. [28] His approach allows for the
Unmanned Vessels and the Law
195
assessment of which factors are concerned by a certain technical development and where new technology causes legal tension. Such tensions can then be addressed more precisely. Ringbom gives two examples of technical navigation-assistance systems to highlight that new technology can easily be used as long as no fundamental legal principle is affected. [29] The autopilot and the dynamic positioning system help the crew keeping a given course or position. These systems were added to the human onboard presence and did neither affect the responsibility of the master nor the requirement of safe manning. However, UMV challenge the manning requirement and are therefore more demanding to scope [30]. Having this structure and challenge to the manning requirement in mind, UNCLOS, SOLAS, and the STCW Convention shall be examined. The Safe Manning Requirement in UNCLOS According to Art. 94 IV (b) UNCLOS, the manning must be appropriate regarding the size and equipment of the vessel. Lutter concludes that crew members can be replaced by technical means and, hence, the crew may be reduced. She questions, however, whether the master and officers can be replaced [31] and touches thereby upon the essence of the issue. Technical research is guided by the objective to relocate the crew or even replace personnel by technical means to realise an economic advantage. When the master and officers have to be present on board, the maximum profit potential of UMV cannot be realised. Unfortunately, Lutter leaves the question raised unanswered. Jessen argues based on the wording of Art. 94 (4) (b) UNCLOS that “in charge of” does not necessarily require the onboard presence of crew and master. However, he refers to the sense of UNCLOS and deducts the legislator’s assumption of human presence on board. He and other authors conclude by urging for clarification of the wording of UNCLOS [32]. The Safe Manning Requirement in SOLAS According to Art. 94 (5) UNCLOS, the flag state administration has to consider the generally accepted international regulations (such as SOLAS Reg. V/14 [33] and the legally non-binding IMO Resolution A.1047 (27) on principles of minimum safe manning) when determining the necessary minimum manning level of a vessel. The foremost goal of these provisions is the safety of life at sea. Would SOLAS be violated if a flag state accepts zero crew members? A purpose-based interpretation could be applied. The main cause of accidents at sea is mental and physical fatigue of the officers in charge. [34] Based ashore, a crew has a reduced risk of fatigue and can operate the vessel more safely. Automated processes are not at all at risk of fatigue or inattentiveness. UMV could hence achieve the aforementioned goal more effectively than traditional manned vessels and do therefore not necessarily conflict with the objectives of SOLAS. [35] The proof of this predicted performance of UMV - to the conviction of the flag state administration - is not as easy as it seems to be. In the 1990s, the IMO supervised trials on one-man bridge operations. Even though the IMO sub-committee on Safety of Navigation examined the vessels participating in the trials and could not identify any decrease in the safety level of the navigational watch, many states, foremost the United States, expressed their reservations against the proposed development. Policy concerns led to the termination of the trials. [30] The resistance against this development could give a foretaste of what has to be
196
E. R. Lange
expected when trials on UMV are performed. Chircop [36] and Noma [37] propose to develop criteria for determining the equivalence of a technical bridge operation with a manned bridge. Such criteria could help the operators to convince the flag state administration and obtain the certificate. Ntovas et al. add, that UMV do not put the crew at risk. [38] However, the life of third persons may be affected. Jessen [39] provides a systematic argument and points out that Annex 2, 1.1.3 of the IMO Resolution A.1047 (27) mentions the level of ship automation as a relevant factor for establishing the minimum safe manning. Lutter acknowledges the relevance of the mentioned factors but is reluctant in accepting zero crew members and refers to the wording. [31] Likewise Lutter, Cioto argues that SOLAS Ch. V/14 provides for the minimum safe manning, so some manning is the threshold. Zero crew members would therefore violate SOLAS. An interpretation of a provision overriding the wording may not be adopted, so it depends on the meaning of ‘minimum’ whether zero crew members can be included. Pritchett presents another point of friction: SOLAS Annex Ch. III, Part B/III-31, 1.1.1 prescribes that a survival craft must be located on each side of a cargo or passenger ship, and regulation 7.10.2 requires a sufficient number of crew members onboard to operate the survival crafts. Pritchett points out that a strict application of the provision would lead to a situation where crew members are needed just for the operation of survival crafts [40]. Jo et al. also detect this problem, though they assume that the provisions concerning lifesaving facilities will be inapplicable when the vessel is unmanned [41]. Chircop [42] deals with this issue as well, but refers to SOLAS Ch. 1/4 (2) that provides for the flag state administration the option to grant an exemption from SOLAS Ch. III to vessels of novel kind. UMV are of such novel kind and, thus, likely to receive such exemption. To conclude the analysis of SOLAS on the requirement of minimum safe manning, open legal questions exist. However, an interpretation of the provisions in favour of UMV is possible and the tacit amendment procedure prescribed in Art. VIII (8b) (vi) (2) SOLAS is a rather frictionless way of amending SOLAS. Amendments are deemed accepted unless more than a third of the contracting governments declare their objection to the amendment within two years. The MSC suggests reconstruction of regulations to accommodate unmanned operations. [43] Large-scale amendments of current regulations, however, can cause tensions when applied to conventional ships. A separate instrument to address challenges specific to MASS is therefore preferred. Alternatively, the MSC could amend the IMO Resolution A.1047 (27) by majority vote to incorporate a clause for unmanned operations. Such a clause would not be legally binding for the member states but could provide a legal basis during the amendment process for the Conventions. Even if unattended bridges and far-reaching automation of bridge functions are compatible with SOLAS and COLREGS, Ch. VIII of the STCW Convention presents a considerable barrier to unmanned vessels. The Safe Manning Requirement in the STCW Convention and Code The STCW Convention applies as per Art. 3 to all “seafarers serving on board seagoing ships”. No one is serving on board a UMV instead, personnel work ashore. At the first glance, the STCW Convention is inapplicable.
Unmanned Vessels and the Law
197
According to Ringbom, such a distinct differentiation based on the place of work is contrary to the aim of the convention. [44] The exclusion of remotely operated vessels from the scope of the STCW Convention causes legal uncertainty and promotes – contrarily to the interests of the automation itself - national regulations rather than international harmonization. National legislators would adopt training and qualification schemes for ships flying their flag. Kampantais as well sees the risk of a legal uncertainty arising in a legal vacuum and, hence, prefers the application of the STCW Convention. [45] Karlis fears that the uncertainty prevents operators from investing in research of new technology and hampers innovation. Hence, she deems the application of the STCW Convention necessary to gain the total profit out of UMV operations. [46] Most scholars want to prevent a legal vacuum and therefore argue purpose-oriented. As the uncertainty following from a legal vacuum can be detrimental to the technical advantage, this line of argument is favourable. Following the majority of scholars, the applicability of the STCW Convention and Code is presumed for this analysis. Section A-VIII/4 (1), para. 32, 18, and 24 STCW Code further specify the requirement of physical attendance of the watchkeeping officer as prescribed in Reg. VIII/2 (2.1) STCW Convention. The wording clearly indicates the purpose of the provisions, physical presence on board. Although Reg. I/13 STCW Code allows for exemptions for trials, a permanent deviation from the manning principle cannot be based on this provision. [44] The remote operation of a vessel constitutes therefore a violation of the STCW Convention. [47] The presented literature acknowledges the difficulties arising out of the application of the STCW Convention, however, the authors accept the clear wording of the provision as a limit to the interpretation and call for an amendment. Ringbom assumes that the STCW Convention has to be amended. [44] According to Jokioinen et al., a differentiation should be made between autonomous and remotely operated vessels [48]. According to Art. 12 (1) (a) STCW Convention, a two-third majority of the present member states can adopt amendments to the Convention, provided that one third of the member states of the Convention are present. At least 37 states [49] have to vote for an amendment to fulfil the mentioned requirement. It seems possible to achieve such a majority. The MSC concludes that provisions regarding the “remote operator” and the relationship to the seafarers serving on board should be included in an amendment [50]. To conclude, UNCLOS and SOLAS do not explicitly prevent unmanned shipping, however, clarifying amendments should be made to ensure legal certainty. Substantial amendments are required for the STCW Convention and Code.
4 Look-Out Duty 4.1 Statutory Framework COLREGS is referred to as the “rules of the road” and prescribes the navigational rules ships have to observe to prevent collisions between two or more vessels. Rule 5 prescribes that “every vessel shall at all times maintain a proper look-out by sight and hearing as well as by all available means appropriate in the prevailing circumstances and conditions to make a full appraisal of the situation and the risk of collision.”
198
E. R. Lange
According to Rules 1 (1) and 3 (a), the duty applies to all vessels on the high sea. Ch. VIII of the STCW Convention and Ch. XX of the STCW Code also govern the lookout and prescribe how to arrange the watchkeeping. 4.2 Technical Solutions As of today, the crew does not fulfil the lookout duty merely by sight and hearing. Radar, sound reception systems, Global Positioning System, the automatic identification system, and other technical means assist the crew to monitor the ship’s surroundings. Ch. V/19 (2.3) of SOLAS makes some of the facilities even mandatory for ships of 300 gross tonnage and upwards. These systems transfer information to the ship’s bridge and provide a comprehensive informational basis for the officer in charge to take navigational decisions. When the crew is not present on board, however, the perception of the ship’s vicinity has to be conducted in another way than today. The collected data must not only be transmitted to the mainland but also the technical processing and possibly prioritization of the data to deal with conflicting data or insufficient bandwidth will be necessary. [51] An autonomous ship needs even more sophisticated systems that can take decisions in accordance with COLREGS. Natural scientists and engineers are diligently developing sensor-based systems to enable a full appraisal of the situation on and around the ship. The following section examines from a legal perspective whether the lookout duty can be fulfilled without human presence onboard. 4.3 Status of Opinion Carey assesses first the obligor of the lookout duty. Rule 5 of COLREGS addresses the vessel itself; however, according to Rule 2 (a) either the vessel, the master, the owner, or the crew is responsible for compliance with the COLREGS rules. The use of “or” could impose the obligation solely on the vessel. But, Carey points out that the legislator presumed human presence onboard when enacting COLREGS. The wording “sight and hearing” can be seen in the reference to the human capacity of seeing and hearing. Carey applies this literal interpretation and therefore reasons that at least one person must look and hear onboard the vessel to fulfil the lookout duty. [52] UMV, equipped with sensor and camera systems, providing an even more detailed and reliable picture of the ships surrounding and transmitting the received data to a shore-based office by audio-visual means would not fulfil the lookout duty as no humans are involved. [53] Carey concludes by calling for an amendment or revision of COLREGS to consider the peculiarities of UMV and enable a sensor-based lookout that is needed to use UMV commercially [54]. As Carey, Ntovas et al. refer to the obligor of the lookout duty – the vessel itself – and presume only little tension when the human lookout is replaced by a camera and audio system. [55] This interpretation does not consider the presumption of the legislator and is therefore unlikely to be widely applied. The CMI applies a purpose-based approach and points out that Rule 5 intends to ensure human involvement in the decision-making process. [56] In contrast to Carey, the CMI assumes that shore-based human supervision of the vessel can fulfil this duty. [57] But, to ensure legal certainty, Rule 5 should be clarified. [57] For remotely operated
Unmanned Vessels and the Law
199
vessels the CMI expects only modest amendments, yet autonomous vessels require greater changes [58]. Lutter’s approach is similar to the one by the CMI. She assumes that remotely operated vessels will comply with the lookout duty as far as recordings from the vessel are transmitted to the shore-based control centre. She deems though that unsupervised vessels are in breach of the duty as long as the wording of Rule 5 is not clarified. Eventually, she hints at the technical risk of delay in and loss of communication. In such a case even a remotely operated vessel would be in breach of the duty [59]. Deketelaere bases his reasoning on the objective of the lookout duty, i.e. that information shall be gathered in an organized way. [60] Pritchett extracts the purpose of the lookout duty from two judgements [61] and adds that the lookout shall ensure a comprehensive information basis about the ships surrounding allowing to take navigational decisions in good time. Both authors think that a sensor-based system can serve this purpose and hence do not expect legal problems. [62] The argumentation seems convincing, however, it can be questioned whether the principles of judgements passed in 1928 and 1858 can be used to support the implementation of the latest technology since the judges at that time had no inducement to distinct the human lookout from any technological information collection process. Ringbom and Jokioinen et al. [63] apply an objective-based approach as well and do not assume a violation of COLREGS by a sensor-based lookout as long as the same level of safety is ensured. Ringbom adds a historic argument. When the design of enclosed bridges made it necessary to install sound reception systems to transfer sound signals to the crew, only an amendment of SOLAS by Reg. V/19 (2.1.8) was needed. This indicates that the IMO does not apply a strictly literal understanding of the rules wording and an amendment of SOLAS by the tacit amendment procedure could provide legal certainty for a technical lookout. Nevertheless, Cioto criticizes that in contrast to a sound reception system a sensor-based lookout affects the principle of human involvement in the lookout and does not merely transfer the signal inside the bridge, but rather replaces marines by processing the collected data. Cioto, therefore, rejects this comparison. He adds that prior to implementation it has to be proven, that a sensor system can give a full appraisal of the situation [64]. The presented purpose-based approaches deviate from a strictly literal understanding of Rule 5 but deliver striking arguments for a frictionless implementation of UMV into the current legal framework. However, the prescription of human involvement by Rule 5 makes an amendment at least for autonomous vessels necessary. Pursuant to Art. VI (2) COLREGS a two-third majority of those member states present and voting in the Assembly is required to amend COLREGS. Such a majority is rather unlikely considering other attempts to implement new technology. Adopting a non-binding resolution on interpretation of Rule 5 by willing member states could be a first step to enable UMV operation. However, such propositions of a ‘coalition of the willing’ could hazard the global acceptance of COLREGS. Considering that navigational decisions usually affect several vessels, the uniform application of international regulations is crucial [65].
200
E. R. Lange
To summarize, most of the scholars assume that sensor-based lookout systems can be accommodated with COLREGS. For clarification and legal certainty, an amendment of the current regulation is preferred, even if considered unlikely.
5 Conclusion The analysis of legal aspects affected by UMV has shown that there is still a long way to go to ensure legal certainty for the commercial use of UMV. Scholars, the IMO, the MSC and other stakeholders research on the most effective ways to address UMV. As UMV affect several provisions in different conventions, a new legal instrument governing UMV could be preferable to amendments of each convention. Whether a new convention, a supplementing protocol, or another instrument is suited best has to be further discussed. All legislative procedures take considerable time to reach an internationally prevailing consent or majority. [58] Therefore, the CMI and Ringbom are only guardedly optimistic that the legal implementation of UMV can keep up with the technical development. [66] Ringbom further emphasizes that a new legal framework has to consider UMVspecific issues like risk of delay of information transmission, loss of communication, and cyber threats [44] and that the level of autonomy can vary during the voyage. Adequate flexibility is therefore necessary [67]. The technical research and the legal development concerning UMV are intertwined. The operation of UMV depends on a certain global legal framework. To not prevent technological advantages beneficial for the society for economic, environmental, or social reasons such a framework should be goal-based and open to new technology. The time when remotely operated, autonomous as well as manned vessels sail side by side will probably be the most challenging, as the vessels cannot communicate easily with each other and their operation modes differ significantly. A reliable legal framework for this period is particularly important. Autonomous operation of means of transportation is known from cars, trains, drones, and trucks. Commercial international shipping is, however, more challenging as it affects the law on both the international and national level and is a highly regulated area of law. The task is to create an internationally accepted and uniformly applied legal framework without passing the time-consuming amendment process for international conventions and codes.
Abbreviations CMI Comité Maritime International COLREGS Convention on the International Regulations for Preventing Collisions at Sea IMO International Maritime Organization MASS Maritime Autonomous Surface Ships SOLAS Convention for the Safety of Life at Sea STCW International Convention on the Standards of Training, Certification and Watchkeeping for Seafarers UMV Unmanned maritime vessels UNCLOS United Nations Convention for the Law of the Sea
Unmanned Vessels and the Law
201
References 1. MSC.1/Circ. 1604, Interim Guidelines for MASS Trials, 14 June 2019 2. MSC.1/Circ. 1638, Outcome of the Regulatory Scoping Exercise for the Use of Maritime Autonomous Surface Ships, 3 June 2021 3. IMO Homepage. https://www.imo.org/en/About/Strategy/Pages/Default.aspx. Accessed 5 Jan 2022 4. The UK Maritime Autonomous Systems Regulatory Working Group published “MASS UK Industry Conduct Principles and Code of Practice” as a voluntary code in November 2020. https://www.maritimeuk.org/media-centre/publications/maritime-autonomous-sur face-ships-industry-conduct-principles-code-practice-v4/. Accessed 5 Jan 2022 5. DNVGL Class guideline, DNCGL-CG-0264, Edition September 2018, https://rules.dnv.com/ docs/pdf/DNV/cg/2018-09/dnvgl-cg-0264.pdf. The Lloyd’s register Code. https://www.lr. org/en/unmanned-code/. Accessed 5 Jan 2022 6. Unmanned Systems technology news page. https://www.unmannedsystemstechnology.com/ 2021/07/sea-kit-usv-receives-first-lloyds-register-unmanned-marine-systems-certification/. Accessed 5 Jan 2022 7. Industry Europe. https://industryeurope.com/sectors/transportation/slef-propelled-ship-yarabirkland-completes-maiden-voyage/. Accessed 5 Jan 2022 8. Norwegian University of Science and Technology Homepage, News. https://www.ntnu.edu/ autoferry/news. Accessed 5 Jan 2022 9. Zeabuz homepage, projects. https://zeabuz.com/miliampere/. Accessed 5 Jan 2022 10. Homepage of the Norwegian Maritime Authority, Sjøfartsdirektoratet. https://www. sdir.no/en/news/news-from-the-nma/worlds-first-test-area-for-autonomous-ships-opened/. Accessed 5 Jan 2022 11. SEA-KIT press release. https://www.sea-kit.com/post/press-release-sea-kit-usv-successfu lly-completes-22-days-of-offshore-operation. Accessed 5 Jan 2022 12. FinFerries Homepage. https://www.finferries.fi/en/news/press-releases/finferries-falco-wor lds-first-fully-autonomous-ferry.html. Accessed 5 Jan 2022 13. UNCLOS is ratified by 174, COLREGS by 162, SOLAS 1974 by 167 and STCW 1978 by 166 states as it can be seen at IMO Homepage. This compromises approx. 99.99 % of the world’s total tonnage. https://wwwcdn.imo.org/localresources/en/About/Conventions/Status OfConventions/StatusOfTreaties%20(1).pdf. Accessed 5 Jan 2022 14. Jokioinen, E., et al.: Advanced Autonomous Waterborne Applications (AAWA) Position paper, Remote and Autonomous Ships: The Next Steps, Rolls-Royce, p. 37 (2016); RodriguezDelgado, J.: The legal challenges of unmanned ships in the private maritime law: what law would you change? In: Port, Maritime and Transportation Law Between Legacies of the Past and Modernization (5), 493–523; 498–499 (2018) 15. Lutter, C.: Unmanned shipping. In: Ehlers, P., Paschke, M. (eds.) Maritime Law – Current Developments and Perspectives, Publication on the Occasion of the 35th Anniversary of the Institute for the Law of the Sea and Maritime Law (Hamburg), pp. 229–249, 232. LIT Verlag GmbH & Co. KG, Zürich (2018) 16. Rodriguez-Delgado: pp. 499–501 (2018) 17. Deketelaere, P.: The legal challenge of unmanned vessels. Universiteit Gent (2017); Kampantais, N.: Legal and regulatory implications of the unmanned ship operation. In: 6th International Symposium on Ship Operations, Management & Economics, March 2018, p. 2; Rodriguez-Delgado, p. 502 (2018) 18. Comité Maritime International: CMI International Working Group Position Paper on Unmanned Ships and the International Regulatory Framework, p. 3 (2016). https://comite maritime.org/wp-content/uploads/2018/05/CMI-Position-Paper-on-Unmanned-Ships.pdf. Accessed 5 Jan 2022
202
E. R. Lange
19. CMI: p. 3 (2018); Jokioinen: p. 37 (2016) 20. Part I, Article I (a) of the Convention on the International Maritime Organisation names the purpose of the IMO as “To provide machinery for co-operation among Governments in the field of governmental regulation and practices relating to technical matters of all kinds affecting shipping engaged in international trade” 21. Lutter: p. 233 (2018) 22. BGH Urteil vom 14. 12. 1951 - I ZR 84/51: Unter einem Schiff im Rechtssinne ist jedes schwimmfähige, mit einem Hohlraum versehene Fahrzeug von nicht ganz unbedeutender Größe zu verstehen, dessen Zweckbestimmung es mit sich bringt, daß es auf dem Wasser bewegt wird 23. Sager, H.: § 553 HGB. In: Drescher, I., Fleischer, H., Schmidt, K. (eds.) Münchener Kommentar zum Handelsgesetzbuch, 4th edn., München (2020) 24. Ntovas, A., Tsimplis, M., Veal, R., Quinn, S., Serdy, A.: Liability for operations in unmanned maritime vehicles with differing levels of autonomy Southampton. Institute of Maritime Law, Southampton, p. 42 (2016) 25. Suri, M.: Autonomous vessels as ships – a definition conundrum. In: IOP Conference Series: Materials Science and Engineering, vol. 929, p. 012005 (2020). p. 9 26. IMO Resolution A: 1047(27). Adopted in 2011 27. Ringbom, H.: Regulating autonomous ships – concepts, challenges and precedents. Ocean Dev. Int. Law 50(2–3), 141–169 (2019). p. 3 28. Ringbom: p. 3 (2019) 29. Ringbom: pp. 11–12 (2019) 30. Ringbom: p. 11 (2019) 31. Lutter: p. 235 (2018) 32. Jessen, H.: Autonome Schiffe als zukünftige Herausforderung für das Seevölkerrecht und für den Rechtsrahmen der IMO. In: Verkehrsgerichtstag, D (ed.) Veröffentlichung der auf dem 55. Deutschen Verkehrsgerichtstag vom 25.- bis 27, January 2017 in Goslar gehaltenen Vorträge, Referate und erarbeiteten Empfehlungen 2017, pp. 265–273, 267 (2017) 33. Guilfoyle, D.: Article 94 UNCLOS. In: Proelss, A.: United Nations Convention on the Law of the Sea, a Commentary, pp. 707–713, Marginal no. 12, C. H. Beck oHG, München (2017) 34. Ringbom: p. 17 (2019); Jokioinen: p. 44 (2016); Allianz: Safety and Shipping Review, p. 5 (2020). https://www.agcs.allianz.com/content/dam/onemarketing/agcs/agcs/reports/AGCSSafety-Shipping-Review-2020.pdf. Accessed 5 Jan 2022 35. Pritchett, P: Ghost ships: why the law should embrace unmanned vessel technology. Tulane Maritime Law J. 40(1), 197–225 (2015). p. 204 36. Chircop, A.: Testing International Legal Regimes: The Advent of Automated Commercial Vessels. The German Yearbook on International Law, vol. 60, pp. 109–142 (2017). p. 111 37. Noma, T.: Existing conventions and unmanned ships - need for changes? World Maritime University Dissertations, vol. 527 (2016) 38. Ntovas: p. 49 (2016) 39. Jessen: p. 270 (2017) 40. Pritchett: p. 203 (2015) 41. Jo, M., Kim, Y., Lee, A., Seo, J.: Study on the potential gaps and themes identified by IMO Regulatory Scoping Exercise (RSE) for the use of Maritime Autonomous Surface Ships (MASS). IOP Conf. Ser. Mater. Scie. Eng. 929, 012014 (2020). p. 4 42. Chircop: p. 111 (2017) 43. MSC.1/Circ. 1637, Annex, p. 32 44. Ringbom: p. 19 (2019) 45. Kampantais: p. 3 (2018) 46. Karlis, T.: Maritime law issues related to the operation of unmanned autonomous cargo ships. WMU J. Marit. Affairs 17, 119–128 (2018). pp. 122–123
Unmanned Vessels and the Law
203
47. Ringbom: p. 18 (2019) 48. Jokioinen: p. 40 (2016) 49. 166 states are members to the STCW Convention; 2/3 majority of at least 1/3 present member states amounts to 37 50. MSC.1/Circ. 1637, Annex, p. 82 51. Ringbom: p. 14 (2019) 52. Carey: p. 10 (2017) 53. Carey: p. 11 (2017) 54. Carey: p. 15 (2017) 55. Ntovas: p. 67 (2016) 56. CMI: p. 14 (2018) 57. Ibid 58. CMI: p. 21 (2018) 59. Lutter: p. 237 (2018) 60. Deketelaere: p. 64 (2017) 61. Dahlmer v. Bay State Dredging & Contracting Co., US Court of Appeals for the First Circuit - 26 F.2d 603 (1st Cir. 1928), May 31, 1928 and Chamberlain v. Ward, 62 U.S. 21 How, p. 548 (1858) 62. Pritchett: p. 205 (2015) 63. Jokioinen: p. 46 (2016) 64. Coito: p. 300 (2021) 65. Ringbom: p. 13 (2019) 66. Ringbom: p. 12 (2019) 67. Ringbom: pp. 3–4 (2019)
Low Emission Choices in Freight Transport: Comparing Land and Short Sea Shipping Alternatives Esa Hämäläinen1 , Tommi Inkinen2(B) , and Eunice O. Olaniyi2 1 Brahea Centre, University of Turku, 20014 Turku, Finland
[email protected]
2 Department of Geography and Geology, University of Turku, 20014 Turku, Finland
{tommi.inkinen,eunice.olaniyi}@utu.fi
Abstract. This study compares carbon dioxide (CO2 ) emissions in multimodal transport chains. Given the recent pressure towards responsible logistics, the International Maritime Organization (IMO) has given several resolutions aiming to decrease all emission types in sea transports. The study analyzes different combinations of alternative delivery chains of paper rolls from the production facility to the customer. The results show emissions originating from each transport leg. In the case of land transports, the greenhouse emissions from solely truck-based deliveries are two times higher than in intermodal solutions, i.e. different combinations of truck and rail. Electrified rail transportation lowers the overall emissions significantly. This indicates that a potential for electric trucks, particularly in the case of short distance deliveries. Keywords: Multimodal transport · Short sea shipping · Trucks · Emissions · Clean shipping
1 Introduction Global climate change agreements are trying to mitigate and limit atmospheric emissions, such as carbon dioxide (CO2 ). However, they have had limited impact on the environment thus far. As the situation deteriorates, global logistics operations and freights are still predominantly based on heavy-duty trucks (HDTs) and cargo ships that use nonrenewable fuels (IPCC 2019). Land freight transports produce extensive volumes of emissions such as CO2 ; nitrogen oxides (NOx); black carbon (BC); polycyclic aromatic hydrocarbons (PAH); and particulate matter (PM) 0.5 and 2.5, which have significant health and global climate change impacts (IPCC 2019; Brunila et al. 2020). So far, 27% of Europe’s CO2 emissions originate from the transport sector (EEA 2019). A typical international logistic chain covers land–port–maritime–port–land transport operations and uses multiple value streams from the point of dispatching, to storage, and up to consignees. Most land logistics chains and deliveries use light-duty trucks (LDT) and HDTs. They primarily use fossil fuels and in turn, produce large amounts of emissions that are detrimental to the environment and human health (Kuo et al. 2017). © The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 M. Freitag et al. (Eds.): LDIC 2022, LNLO, pp. 204–216, 2022. https://doi.org/10.1007/978-3-031-05359-7_17
Low Emission Choices in Freight Transport
205
Mandal et al. (2020) explained that this “high mountain of technology hindrance” must be surmounted in order to find alternative solutions for the engine and fuel technology. Moultak et al. (2017) argued that electric-technology vehicles could bring about emissions reductions (covering a whole life cycle) in diesel-dominated freight transportation, and heavy-duty electrical trucks (e-HDT) might provide a good option in the mitigation of CO2 emissions (Sen et al. 2019). A case in point is China, a densely populated country with a vast logistics network plagued with increased combustion engine vehicles. Recently, China has experienced a significant increase in the total number of alternative new engine solutions, such as electrical, plug-in hybrid, fuel cell, and hydrogen fuel-cell vehicles to curb air pollution. The country has taken significant strides towards the utilization of alternative solutions to solve its environmental issues. Fortunately, the European market is catching up with China’s usage of electric-driven busses. According to Transport & Environment (2018) in 2017, the number of electric bus orders in Europe increased from 400 (2016 figures) to over 1 000. Other remarkable development is seen from a spatial planning point of view from a Dutch study (Boogaard et al. 2012) that reveals that when HDTs operations were stopped in five residential areas in the Netherlands (Amsterdam, The Hague, Den Bosch, Tilburg, and Utrecht), environmental pollution decreased substantially. The coming years will certainly witness a shift from the regular fossil niche to mainstream and the beginning of a steep and necessary uptake curve of non-fossil options. It is important to note that without the availability of necessary infrastructure for charging points for the smooth running of electrically powered trucks, low emissions technologies like e-HDTs remain expensive and rare. Even though e-HDT’s operational range is shorter than the average HDTs or LDTs, the reduction of greenhouse gases and regional pollutants could be achieved by adopting cleaner vehicle technologies using a combination of electricity or non-fossil fuels (Kuo et al. 2017). These options indicate possibilities for successful transport policies that can reduce emissions significantly and improve air quality in densely populated areas. European maritime sector came up with stringent rules to promote good sustainable environmental practices. Because of various shipping regulations, there are already different options for ship owners to consider how to reach desired emission reductions. These options include (but are not limited to) the use of scrubber technologies, waste heat recovery systems, alternative fuels, and the use of batteries and solar energy. Recently, many newly built container ships are using liquefied natural gas (LNG) that significantly reduces CO2 emissions when compared to diesel. However, LNG emissions have issues related to methane that makes it a controversial fuel type in overall emission reduction. Environmental regulation is important for eco-innovations as Makkonen and Inkinen (2018) addressed. Their study depicted maritime scrubber – a system that removes excess sulphur from ships – as an example of incremental innovation. It is also a response to stricter emission regulations. So far, emissions have reduced in the EU, although the level is debatable: CO2 emissions have reduced approximately 20% during the last decade (Olaniyi and Viirmäe 2016). SOx and fine particles have reduced to almost 99% and NOx emissions decreased to some 85% (Prause and Olaniyi 2019). The creation of the Sulphur emission control area (SECA) legislation brought about lowered SOx values in the atmosphere dramatically in the EU, especially in the port areas
206
E. Hämäläinen et al.
connoting that international binding legislation has a significant impact on emissions levels (Lähteemäki-Uutela et al. 2019). It emphasizes the significance of policymaking and environmentally driven international cooperation as a key factor in (technological) environmental improvement. Although, Hämäläinen and Inkinen (2019a) explain that many of these technological solutions present only acute (short-term) solutions that may not be applicable in longer periods. In the end, a sustainable transport chain should be through a complete transformation from the existing diesel engine-based technique to low or zero emission techniques (Hämäläinen and Inkinen 2019b). Responsible industrial operations should continuously consider environmental consequences contrasted with economic gains. Particularly, transport emitted pollution and environmental stress should be accounted for, especially, in value and efficiency. Wu et al. (2016) argue that air pollution problems link to all types of vehicle emissions and the success or failure of export-oriented long-distance logistics is dependent on a secured and efficient functioning of the logistics chains. A responsible transport chain begins from several information flows (e.g. finance, transport, and location data). They have to be integrated into secure and reliable (physical) freight transportation. All organizations in the chain should be committed to contemporary requirements and regulations striving to achieve low emissions objectives. The objective of this work is to determine the most suitable land–sea-land transportation combination with the lowest CO2 emissions generation for a Nordic paper company’s product. Using the analyses of different combinations of alternative delivery chains of paper rolls from the production site until the end-users, the study calculates possible CO2 emissions generated by each delivery logistics mix. These combinations include the use of trucks, trains, and ships. The result from the study is intended for policy interpretations towards emissions reduction in the land–sea-land transports. This paper has a following structure: The Sect. 2 shows the methodology and the data analyses of the study. The Sect. 3 highlights the results. The Sect. 4 discusses the results and their implications. The last Sect. 5 concludes the work.
2 Methodology There are three major logistics challenges that are dependent on transportation mode, port operation efficiency, and shipping processes. The first step is the cargo transfer from the warehouse to the port (land process with either trucks or trains). The second step is the sea process (from the port A to port B with ships). The final third step is the destination land process (the final trip to delivery address using trucks or trains and short distance transport options, e.g. e-trucks or other low emission vehicles). This process is clearly visible in our case study: the paper rolls are transported from the warehouse to ports by HDTs. The cargo is stored in the ports for a while, and then loaded into ships. Sea process is done with short sea shipping (SSS). On arrival at the new port, the rolls are stored and transported by trucks or train onwards to the consignees. McKinnon and Piecyk (2010: 2012) recommended in their (widely used) report that “the average CO2 emission factor for road transport operations should be 62 g CO2 ton-km. This value is based on an average load factor of 80% of the maximum vehicle payload and 25% of empty running.” On the other hand, the corresponding figure for
Low Emission Choices in Freight Transport
207
rail transport operations is 22 g per transported freight ton. The value is based on the following components from the updated EcoTransIT (2020): – – – –
The average split between diesel and electric haulage, The average carbon intensity of electrical power sources, The average energy efficiency of the locomotive, Assumptions on average train load factors.
We acknowledge that emission data on freight transport may have significant variations depending on the data collection site and time. There are differences between countries and regions in terms of energy policies, supported means of transport, and technological progression. It is highly difficult to create universal (or even European) emission indicators (or factors). Still, according to McKinnon and Piecyk (2010; 2012), the estimation of the CO2 emissions may be built on two main freight transport operations: energy consumption and the level of transport activity. McKinnon and Piecyk brought to light the energy-based approach for CO2 emissions for all freight transport and the transport activity approach as an indication of transport operation volume. (1) Energy-based approach: The application of emission factors and the conversion of energy values to CO2 are as follows: – Liters (L) of fuel (unit of energy) for trucks, diesel-hauled trains, barges, and ships – Kilowatt-hours (kW/h) for electric rail and pipeline. (2) Activity-based approach: Used if there are no available energy data, it uses carbon footprint as an alternative to estimate transport operations where CO2 /grams is calculated as: Tones transported x average distance traveled x CO2 emissions factor/ton − km (1) Estimation of emissions is based on system boundaries (SB) integrated data, which are differentiated into five levels (Table 1) by the Swedish environmental organization (NTM; Erixon et al. 2003; Updated 2006). These levels are cumulative, starting from the SB1 and going up to SB5 covering all emissions from the transportation sector. This study focuses on SB1, thus calculating the direct carbon emissions from freight transport from the paper mill to two different countries in Europe. We use the total life cycle emission calculations with complete and reliable data storages (data is obtained from the mill and it is continuously updated in daily operations). Electrically powered trucks are classified under SB1 and they produce low or zero emissions in daily use. As a reference, Özsalih et al. (2009) made an empirical analysis with a company level case data (from Unilever) and concluded that intermodal transport (road/water or train/water) is probably greener than pure road transports.
208
E. Hämäläinen et al.
Table 1. The system boundaries in the environmental calculation. Adapted from Erixon et al. (2003; 2006). SB1: Confines to the calculation of emissions from actual transport operation, most of which emanate from vehicle exhaust. In the case of electrified rail freight operations, it includes emissions from the electrical power source SB2: Takes account of the extraction, production, refining, generation, and distribution of energy, using a “well-to-tank” perspective SB3: Includes the servicing and maintenance of vehicles and transport infrastructure SB4: Includes emissions from the manufacturing of the vehicles, construction of transport infrastructure, including their subsequent scrappage and dismantling SB5: Includes emissions associated with the management of transport operations, mainly office functions and activities of staff
3 Results 3.1 Conditions, Factors, and Variables Peripherally located large bulk industrial site requires a well-functioning transportation chain to deliver their products to their primary markets. Usually, annual production volumes may start from 350 000 tons to one million tons of high-quality pulp and paper products. Nearly 100% of this product category are delivered to international and global markets. If the site is located inland (requiring long-distance land transport), as in the case company, the intermodal transport chain includes several phases. It starts from the production site to the port, loading the freight (or containerization), SSS, port operations, and finally inland transportation to customers. The delivery (order sizes) tons may vary according to each buyer, making delivery and route planning a complex problem (Hämäläinen 2011). Estimations have the following assumptions: (1) From industrial site to outward port (40 km): There are two options: truck (diesel) or train (two-way electricity production is estimated) transportation. (2) SSS transport to the destination country (1 800 km): All paper products are inside containers and transported by container ships that comply with the required maritime bunker and emissions regulations. (3) From import port to customers behind various export routes (100 km per 200 000 tons and 300 km per 150 000 tons): Options are either truck or intermodal (rail and truck). Emission factors taken from Hämäläinen (2011) explored how different modes of transport produce CO2 (grams per ton-km). The authors recognize a considerable variation between the calculated factors (even in the same logistic model), and used the following three factors (routes and phases) to create several intermodal combinations: – The emission values are dependent on the engine technology. The selected combinations indicate the largest variety in calculated emission levels for each transport option.
Low Emission Choices in Freight Transport
209
– It is important to recognize that a single large mill (such as the case here) may have thousands of customers in different locations. – The average delivery (i.e. order) may range from a few tons up to hundreds of tons. This again results in hundreds of transports that vary considerably in their lengths. 3.1.1 Case Study: CO2 Emissions Per Ton-Km from the Mill to Port The case mill operates annually for around 8 000 h (corresponding approx. 91% of the production efficiency). Inside the mill, one large paper machine produces approximately 44 tons of paper per hour, and daily production can reach around 956 tons. The mill runs typically around the clock, suggesting that the logistic chain must operate continuously from mill warehouse to export port (intermediate storage); otherwise, the central warehouse would become overstocked without storage for the new incoming and newly produced products. Usually, there are online store orders as all products are manufactured on-demand basis (by customer orders). Table 2 presents CO2 emissions per ton-km from the mill to port with two transport modes. Transportation by diesel truck is the first option, and the second one is electrified freight train. Table 2. Data from the mill to port – production volumes Export of heavy freight (paper rolls)
Production and Volumes
Mill/machine h
8000
Tons/hour (91% efficiency)
43.75
Daily production (round) 955.5 Heavy Duty Truck 60% utilization
59
Volume (ex-mill tons)
350000
Distance from a port (km)
40
We assume (from the company’s document review) that trucks are fully loaded from the mill to the port and that they are only partially loaded from the port back to the mill. Therefore, the assumption is that the trucks run on average approximately with 60% utility. The annual paper transportation from the paper mill to port by diesel trucks produces about 826 tons of CO2 greenhouse emissions. It is further assumed that the electricity used in the trains is produced either by fully renewable natural resources (hydro or solar power). Thus, the direct CO2 footprint remains low (around 0.003 CO2 grams per ton-km) used in Table 3.
210
E. Hämäläinen et al.
3.1.2 Transportation Combinations and Their Total CO2 Emissions We make an assumption that 22 CO2 grams per ton-km emission level are reached. In a comparison of motor and fuel technologies, it is detectable that the total CO2 emission level varies significantly. In other words, diesel trucks are estimated to reach 826 tons, electronic train with a combination of non-renewable and renewable energy production ends up to 308 tons, and finally, renewable electricity production operated trains reach only 0.042 tons. Electrified trains (with renewable electric power) may be considered highly scalable technology with nearly zero usage emissions (comparative figures in Table 3). Table 3. CO2 (tons) production with different transport means. Adapted from McKinnon and Piecyk (2010). Export of heavy freight (paper rolls) Transport g CO2 /ton-km either diesel trucks (60% utility) or Electrified rail freight service from 0.003 g CO2 /ton-km for electricity generated by renewables
Option 1
Option 2
Diesel
Electric Train
59
0.003
Electrified trains (el is produced by a mix of different sources gCO2 /ton-km)
22
Total CO2 emissions (tons) el train renewables and non-renewables
308
Total CO2 emissions (tons) el train renewables
0.042
Total CO2 emissions (tons)
826
308.042
An example is calculated for 1 800 km long SSS container line. It is further assumed that everything is transported to the central market. In real life, the variety of transportation options from a Nordic country to the main (European) markets, is large and both container ship types and distances vary extensively (see Hämäläinen 2011). The transport volumes are very high so that the total CO2 footprint of 350 000 tons of paper transported 1 800 km by SSS rises to 10 080 tons. Overall, 1 800 km distance is considered a typical sea route from a Nordic country to several main European ports like Poland, Germany, or the Netherlands. This is important as the applied distance has a fundamental impact on the CO2 levels in these calculations. The different bunkering fuel cargo ships use are mainly LNG, MGO, methanol, and HFO (with scrubbers by regulation). All these bunkers produce different amounts of CO2 . Table 4 approaches transport emissions through both truck and intermodal options and yields the following interpretations as it presents transportation from import harbor to the customers with the CO2 factor per tonnage Customers are businesses with varying order sizes and delivery times (see Hämäläinen et al. 2017). Two average scenarios are used to include the final transport leg from ports to customers: If all consignments are transported by trucks (59 CO2 grams per ton-km), the total CO2 footprint of deliveries (in the importing country) would be approximately 3
Low Emission Choices in Freight Transport
211
Table 4. Transportation options in the export country from port to end customers. Source: Authors. CO2 (g) factor/ton-km
Truck
Intermodal road/train
59
26
Av. distance from the port
Delivered tons
CO2 tons
CO2 tons
100
200 000
1 180
520
300
150 000
Cumulative total in the port country (CO2 tons)
2 655
1 170
3 835
1 690
835 tons. Second, if deliveries are transported by intermodal transport chain (rail and road), the total CO2 emissions are less than half (1 690 tons) in comparison to truck deliveries only. Truck (diesel) is a two times higher greenhouse gas emitter when compared to intermodal option because CO2 emission factor in rail transportation is considerably lower. Table 5 captures the CO2 emissions from different transportation modes outside and within the importing country. Table 5. Total CO2 tonnages by different transport modes from the paper mill to end-customers. Source: Authors. Transport option
Total CO2 emissions with 350 000 tons
Context
Truck diesel
826
Export country
El. Train 100% renewables
0.042
Export country
Non-renewables + renewables train
308
Export country
Diesel (MGO) ship
10 080
SSS
Truck diesel
3 835
Import country
Intermodal road/ electrified rail
1 690
Import country
Table 5 distinguished the substantive and pertinent question of how to mitigate CO2 emissions produced especially in SSS, and this is because long-distance heavy industry deliveries are often dependent on long sea journeys. This connotes high CO2 grams per ton-km factor.
212
E. Hämäläinen et al.
Table 6 presents different transport means combinations and their total CO2 emissions. There are six choices where the lowest greenhouse footprint (combination 1) is produced by electrified rail service (Renewable electricity production) + SSS + intermodal road and electrified rail service together. Table 6. Summary of the 350 000 tons transport (mill to markets). Source: Authors. Comparing emissions
First
2nd
Third
Total CO2
CO2 (tons)
%
Renew Train + SSS + Intermodal road/rail
0.042
10080
1690
11770
0
0
Ren + non ren El train + SSS + Intermodal road/rail
308
10080
1690
12078
308
2.616
Truck + SSS + Intermodal road/rail
826
10080
1690
12596
826
7.017
Ren + non ren El train + SSS + Truck
308
10080
3835
14223
2453
20.84
Renew Train + SSS + Truck
0.042
10080
3835
13915.042
2145
18.22
Truck + SSS + Truck
826
10080
3835
14741
2971
25.24
The example logistics chain produces 11 770 tons of CO2 emissions. Note that the significant contribution of emissions is by maritime transportation (95–98%) because there is no viable alternative to replace SSS. However, at the same time, while a long sea journey creates a significant amount of CO2 footprint, there are more alternatives available for inland logistics. Calculated combinations in Table 6 do not include dieselonly rail emissions because diesel locomotives are no longer widely used by the case mill.
4 Discussion A plausible choice an export company for decreasing CO2 footprint is to focus on electrified land transportation and the lowest (i.e. the most environmentally friendly) CO2 maritime transport. These options offer a scalable solution to decrease airborne transport emissions. In the nearest future, electrified trucks will gain more popularity in short-distance final deliveries, but currently they are not extensively available in the array of logistics suppliers. When electric trains (operating in countries with low CO2 emission power plants) are in use, or when container ships operate SSS, intermodal transportation will bring clear environmental benefits. However, this is not the case in all situations. When intermodal transport uses Ro-Ro cargo and Ro-Pax ships, lowering carbon emissions might be challenging because the easiest way to reduce emissions from shipping activities without compromising demand is to reduce emitted CO2 (aggregate) per unit. (ICCT 2017). Transport sector needs environmental innovations. This is critical because blue and green innovations can resolve several of the challenges associated with the climate crisis
Low Emission Choices in Freight Transport
213
(OECD/ITF 2016). However, implementation must be economically viable and they require holistic understanding, e.g. of the production of electricity. Nordic countries, such as Sweden and Finland, have a significant production volume of hydrological, wind, and in some cases nuclear power that offers (after plant construction and dismissing fuel transport and power cell/grid production in other countries) nearly CO2 free electricity for rail transport. Yet, some of the generated electricity is produced by coal power or other non-renewable power solutions. Consequently, the generated emissions (in electricity production) are quite high. They correspond approximately one-third of total diesel truck CO2 emissions. Maritime sector has been able to build industry-level structures to achieve emissions reductions for a long time (Anner et al. 2006). Even though water transportation is a relatively simple way to move freight between countries (emissions per transported unit is the lowest among transport modes), globally the maritime transportation sector produces between 2–3% of total CO2 emissions (EU 2019). This is a significant amount of airborne emissions. It also explains why IMO (2009) started already a decade ago numerous policies aiming to lower all maritime emissions. In terms of transport volumes, sea transportation remains the most environmentally friendly way (per unit or capita) to move passengers and freight between geographical locations. However, the downside is that emission particles are spread several miles to coastal areas. Evidently, reduction of emissions to the barest minimum remains the most preferred choice. Conversely, and from the climate point of view, it does matter what transport modes are employed by heavy industries in their long-leg transports, as they are sea transported in any case. Modal shifts from road to rail have essential impacts. Last-mile logistics especially in inner-city regions should be done with electrified trucks, and while this may be expedient, the replacement of the whole truck fleet from diesel technology to electrified ones remains an expensive investment. This results in the fact that industrylevel changes may likely take a long time. Moreover, the transport range (of e-trucks) is still short. The transport chain needs to be carefully planned from an environmental perspective. Lastly, inter-, trans-, and supranational regulatory structures can only be strengthened when unified across sectors and most importantly across countries (Olaniyi and Viirmäe 2016). The current mismatched and expensive cyclicality and capital utilization currently witnessed in the marmite industry, for example, can only be resolved through transnational coordination, which may take years to achieve. In particular, the environmental innovations and blue/green technologies stem from various compliance activities that may establish new key drivers for the future economic growth and social wealth.
5 Summary This study scrutinized the seeming complexity of multimodal transport emissions. The case dealt with the transportation and delivery process of products from a large paper production company to their end customers. The goal was to illustrate with different combinations of transport modes the total levels of generated greenhouse emissions. This is needed to determine the best-combined transport means for the low generation of CO2 .
214
E. Hämäläinen et al.
The result shows that typical transportation of bulky material from the production site to the end-users amounts to close to a million (826) tons of CO2 greenhouse emissions. The bulk of the emissions are coming from shipping. Intercity CO2 emissions from all truck deliveries are also high when compared to rail transports. The paper presented several example calculations to illustrate the emission effect of intermodal transport on emissions. The results indicate that the greenhouse emissions from all truck (diesel) inland deliveries are two times higher than in the case of inland intermodal logistics (i.e. a combination of truck-rail). This is because CO2 emission factors in rail transportation are comparably low when the renewably produced electricity is used. Focusing on electrified land transportation and low CO2 maritime transportation is a conceivable solution to lower emissions and to create a satisfactory low pollution (or even pollution-free) environment. Logistics companies need to work harder in focusing on intermodal transport or a combination of different means of transport for goods deliveries. Policymakers would need to use stringent regulations to enforce successful joint efforts from all actors to achieve this change.
References Anner, M., Greer, I., Hauptmeier, M., Lillie, N., Winchester, N.: The industrial determinants of transnational solidarity: global interunion politics in three sectors. Eur. J. Ind. Relat. 12(1), 7–27 (2006) Boogaard, H., et al.: Impact of low emission zones and local traffic policies on ambient air pollution concentrations. Sci. Total Environ. 435–436, 132–140 (2012). https://doi.org/10.1016/j.scitot env.2012.06.089 Brunila, O.-P., Inkinen, T., Kunnaala-Hyrkki, V., Hämäläinen, E., Ala-Rämi, K.: Black carbon, maritime traffic and the arctic. In: Pongrácz, E., Pavlov, V., Hänninen, N. (eds.) Arctic Marine Sustainability. SPS, pp. 165–177. Springer, Cham (2020). https://doi.org/10.1007/978-3-03028404-6_8 Deloitte: Truck Market 2024: Sustainable Growth in Global Markets (2014). https://www2. deloitte.com/content/dam/Deloitte/tr/Documents/process-and-operations/truck-studie-2014-s. pdf. Accessed 14 June 2019 EcoTransIT: Ecological Transport Information Tool for Worldwide Transports - Methodology and Data (2020). https://www.ecotransit.org/download/EcoTransIT_World_Methodology_D ata_100521.pdf. Accessed 4 May 2020 EEA: Air quality in Europe (2019). https://www.eea.europa.eu/themes/air. Accessed 15 Nov 2021 Transport and Environment: Marketplace, economic, technology, environmental, and policy perspectives for fully electric buses in the EU (2018). https://www.transportenvironment.org/wpcontent/uploads/2021/07/Electric-buses-arrive-on-time-1.pdf. Accessed 15 Nov 2021 Erixon, M., Carlson, R., Pålsson, A.C.: Measuring the environmental impact of products CPM’s experiences of tools, methods, and the provision of information. Industrial Environmental Informatics. Chalmers University of Technology. Swedish Environmental Protection Agency (2003). https://www.naturvardsverket.se/Documents/publikationer/620-5349-3. pdf?pid=2987. Accessed 14 Dec 2019 Erixon, M., Tivander, J., Pålsson, A.C., Carlson, R.: General method for integration of industrial environmental information systems. Chalmers University of Technology (2006) EU: EU shipping’s climate record Maritime CO2 emissions and real-world ship efficiency performance (2019). https://www.transportenvironment.org/sites/te/files/publications/Study-EU_shi ppings_climate_record_20191209_final.pdf. Accessed 6 Jan 2020
Low Emission Choices in Freight Transport
215
Hämäläinen, E.: Economic, geographic characteristics in the Finnish paper industry-a case study. Fennia Int. J. Geogr. 189(2), 63–75 (2011). https://fennia.journal.fi/article/view/4445 Hämäläinen, E., Inkinen, T.: Industrial applications of big data in disruptive innovations supporting environmental reporting. J. Ind. Inf. Integr. 16, 100105 (2019). https://doi.org/10.1016/j.jii. 2019.100105 Hämäläinen, E., Twrdy, E., Inkinen, T.: Cost aggregation in export logistics chain. J. Open Innov. Technol. Mark. Complex. 3(1), 1–14 (2017). https://doi.org/10.1186/s40852-017-0077-9 Hämäläinen, E., Inkinen, T.: Big data in emission producing manufacturing industries–an explorative literature review. In: ISPRS Annals of Photogrammetry, Remote Sensing & Spatial Information Sciences, vol. IV-4-W9, pp. 57–64 (2019a). https://doi.org/10.5194/isprs-annals-IV-4W9-57-2019a Naya Olmer, N., Comer, B., Roy, B., Mao, X., Rutherford, D.: ICCT: The International Council of Clean Transportation. Greenhouse gas emissions from global shipping report, 2013– 2015 (2017). https://theicct.org/sites/default/files/publications/Global-shipping-GHG-emissi ons-2013-2015_ICCT-Report_17102017_vF.pdf. Accessed 15 Nov 2021 IOICA: International Organization of Motor Vehicle Manufacturers (2020). http://www.oica.net/ production-statistics/. Accessed 14 Sept 2019 IPCC: The Intergovernmental Panel on Climate Change. Special Report on climate change, desertification, land degradation, sustainable land management, food security, and greenhouse gas fluxes in terrestrial ecosystems (2019). https://www.ipcc.ch/srccl/ Kuo, S.Y., Lin, P.C., Lu, C.S.: The effects of dynamic capabilities, service capabilities, competitive advantage, and organizational performance in container shipping. Trans. Res. Part A Policy Pract. 95, 356–371 (2017). https://doi.org/10.1016/j.tra.2016.11.015 Lähteenmäki-Uutela, A., et al.: The impacts of the sulphur emission regulation on the sulphur emission abatement innovation system in the Baltic Sea region. Clean Technol. Environ. Policy 21(5), 987–1000 (2019). https://doi.org/10.1007/s10098-019-01684-2 Makkonen, T., Inkinen, T.: Sectoral and technological systems of environmental innovation: the case of marine scrubber systems. J. Clean. Prod. 200, 110–121 (2018). https://doi.org/10.1016/ j.jclepro.2018.07.163 Mandal, J., Goswami, A., Wang, J., Tiwari, M.K.: Optimization of vehicle speed for batches to minimize supply chain cost under uncertain demand. Inf. Sci. 515, 26–43 (2020). https://doi. org/10.1016/j.ins.2019.12.009 McKinnon, A., Piecyk, M.: Measuring and managing Co2 emissions of European Chemical Transport. European Chemical Industry Council, Edinburgh (2010). https://cefic.org/app/upl oads/2018/12/MeasuringAndManagingCO2EmissionOfEuropeanTransport-McKinnon-24. 01.2011-REPORT_TRANSPORT_AND_LOGISTICS.pdf. Accessed 20 Apr 2020 McKinnon, A.C., Piecyk, M.: Setting targets for reducing carbon emissions from logistics: current practice and guiding principles. Carbon Manag. 3, 629–639 (2012) Moultak, M., Lutsey, N., Hall, D.: Transitioning to zero-emission heavy-duty freight vehicles. White Paper. The International Council of Cleaner Transportation (2017) OECD/ITF: Reducing Sulphur Emissions from Ships: The Impact of International Regulation (2016). http://www.itf-oecd.org/reducing-sulphur-emissions-ships-impact-internationalregulation. Accessed 15 Nov 2021 Olaniyi, E.O.: Towards EU 2020: an outlook of SECA regulations implementation in the BSR. Baltic J. Eur. Stud. 7(2), 182–207 (2017). https://doi.org/10.1515/bjes-2017-0016 Olaniyi, E.O., Viirmäe, M.: The economic impact of environmental regulations on a maritime Fuel Production Company. Res. Econ. Bus. Cent. East. Eur. 8(2), 58−84 (2016). https://ideas. repec.org/a/ttu/rebcee/93.html Özsalih, H., de Kok, A.G., Fransoo, J.C.: A methodology for transport buying companies to estimate CO2 emissions in transport: application in Unilever European Logistics. Technische Universiteit Eindhoven (2009)
216
E. Hämäläinen et al.
Prause, G., Olaniyi, E.O.: A compliance cost analysis of the SECA regulation in the Baltic Sea. J. Entrepreneurship Sustain. Issues 6(4), 1907–1921 (2019). https://doi.org/10.9770/jesi.2019. 6.4(26) Buhaug, Ø., et al.: IMO: Second IMO GHG Study. London, UK, April 2009 Sen, B., Ercan, T., Tatari, O., Zheng, Q.P.: Robust Pareto optimal approach to sustainable heavyduty truck fleet composition. Res. Cons. Rec. 146, 502–513 (2019). https://doi.org/10.1016/j. resconrec.2019.03.042 Wu, X., Wu, Y., Zhang, S., Liu, H., Fu, L., Hao, J.: Assessment of vehicle emission programs in China during 1998–2013: achievement, challenges, and implications. Environ. Poll. 214, 556–567 (2016). https://doi.org/10.1016/j.envpol.2016.04.042
Digital Twin Features for the Intelligent Container Reiner Jedermann1(B)
, Walter Lang1
, Martin Geyer2
, and Pramod Mahajan2
1 Institute for Microsensors, -Actuators and -Systems (IMSAS), University Bremen,
Bremen, Germany [email protected] 2 Leibniz Institute for Agricultural Engineering Potsdam-Bornim, Potsdam, Germany
Abstract. The “Intelligent Container” for remote monitoring of refrigerated transports of fresh fruits already implements typical features of digital twins, including remote sensing and modeling of physical and biological objects. This article asks how the Intelligent Container can be extended to make the best use of digital twin concepts. Existing applications in agricultural science focus on offline simulation models that can predict shelf life and the effects of packaging on cooling but cannot integrate real-time data or correct their current estimates according to those data. This update feature is considered a key component of digital twins. The related challenges and algorithms can be best understood from the viewpoint of systems theory and state-space description. Internal properties of real objects can be either directly measurable, hidden, or unobservable, and implementation of the update process should be adapted accordingly. Using ocean transport and banana processing as an example, this paper demonstrates how models can be made “updateable”, in addition to discussing the necessary steps for linking different sub-models over a standardized platform according to the “publish/subscribe” pattern. Keywords: Digital twin · Intelligent Container · State observer · Updatable model · Food logistics · Cold chain
1 Introduction The “Intelligent Container” (IC) was developed over 10 years ago to monitor temperature conditions and consequent changes in the quality of fruit during ocean transportation to improve logistical process planning [1]. Although IC development began long before the term digital twins (DTs) was used publicly, DTs now represent a popular concept for remote management of various complex systems, including shop floors, chemical plants, and wind farms [2]. Agricultural applications of DTs have been reported since 2017 [3]. The IC already implements some typical DT features, including real-time remote sensing and modeling of physical objects. This article’s major focus is using DT concepts and applications to guide future extensions of the IC, which can be considered a proxy for other remote sensing and modeling applications in food logistics that contend with © The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 M. Freitag et al. (Eds.): LDIC 2022, LNLO, pp. 217–228, 2022. https://doi.org/10.1007/978-3-031-05359-7_18
218
R. Jedermann et al.
similar issues. Furthermore, we consider special challenges associated with modeling biological objects. The scientific literature includes several definitions of DTs. According to [4], “a digital twin is a virtual, dynamic model in the virtual world that is fully consistent with its corresponding physical entity in the real world and can simulate its physical counterpart’s characteristics, behavior, life, and performance in a timely fashion.” Alternate definitions are listed in [5], and a more detailed five-dimensional definition can be found in [2]. However, there is still no binding definition of DT, with [6] writing that “each DT is legitimate in its corresponding context, which makes an overarching definition of DT more abstract and thus difficult to comprehend and to imagine”. They further suggest using a taxonomy of seven dimensions or features to classify different DT applications. Their approach indicates that low implementation levels for some features can be sufficient for certain applications and do not provide a basis for rejecting classification as a DT. Nonetheless, it is worth considering whether additional benefits can be achieved by increasing the implementation level. Although [7] provides a more ambitious definition of DTs in agricultural science, this iteration is almost impossible to achieve, even with current state-of-the-art modeling: “As such, this digital replica evolves and reacts hygrothermally and metabolically in a similar way as its physical counterpart––a real fruit or vegetable––but now in-silico and preferably in real time.” The similarity of the twins can be maintained only if changes to the physical object are continuously measured, and the digital counterpart is updated accordingly. [8] raised the question, “how to tell the difference between a model and a digital twin”. Key point is the ‘digital thread’ [8] to update the model based on real-time sensor data. Given the literature mostly neglects the complexity of this update process, its principal difficulties and solutions comprise this paper’s second special focus. The sensor system is mostly realized by Internet-of-Things (IoT) technologies, and the complexity of the DT’s data processing varies largely depending on the target application. According to the taxonomy of [6], the processing can be classified into four stages or goal types. The first three are acquiring data, predicting future changes to the physical object, and conducting virtual experiments. The latter provides the option of testing possible interventions on the digital platform before applying them to the real object. The fourth action implementation stage involves the DT automatically adjusting controls for the physical object, thus closing the feedback loop. Biological objects differ from work-pieces in the manufacturing scenarios in which DTs have traditionally been deployed. Related models often entail high numbers of “hidden” states that cannot be directly measured. Section 2 discusses the consequent challenges and the update process involved in realizing the digital thread. Section 3 considers DT software realization and demonstrates how simulation models can be converted to an updateable form. Having applied this concept to the IC, Sect. 4 summarizes the steps involved in converting a monitoring solution to a DT-compatible platform.
Digital Twin Features for the Intelligent Container
219
1.1 The Intelligent Container The temperature history during cold chain operations substantially impacts the quality of agricultural products, although this is often not visible from the outside. A green banana remains a green banana, whether it was transported at perfect temperature and atmospheric conditions or whether the temperature was a few degrees too warm. However, in the latter case, an unwanted ripening process can commence shortly after the transport’s arrival, making the fruit useless for further commercial processing. By revealing hidden changes in quality to the operator, fruit in a critical state can be prioritized for harbor handling and further processing; in the worst-case scenario, a timely replacement delivery can be organized. The IC system comprises a network of wireless sensors inside the cargo hold that measure temperature and other environmental conditions. The sensor data is processed by a gateway mounted in the container, with the effects of temperature deviations evaluated by a so-called shelf-life or green-life model. The gateway sends either full sensor data or warning messages alone to a cloud server via cellular or satellite networks [9], allowing the user to make decisions based on forward projection of the predicted shelf-life, i.e., to assess whether the consignment will arrive in an acceptable quality state after the expected transport duration. Making the actual quality of agricultural products visible enables stock rotation optimization. Assigning deliveries according to the first-expires first-out approach can reduce losses of highly perishable products by 8% to 14% of the total volume [10]. 1.2 Digital Twins in the Food Supply Chain Modeling has a long tradition in agricultural science, especially for predicting the effect of temperature deviations on shelf-life. [11] provides model parameters for the shelf-life of 60 different fruits and vegetables. Elsewhere, an online tool developed by the FRISBEE project [12] provides shelf-life models for six food products and enables simulation of the effects of different transport modes on total energy consumption and shelf-life, with one model for apples including the effect of air humidity on moisture loss. The FRISBEE tool also enables virtual experiments that can compare the effects of different cooling and transport modes on product quality and carbon footprints. Meanwhile, [13] provides a range of mathematical models for simulating accumulated gas concentrations in the headspace above fresh produce and related shelf-life changes. Other models based on computational fluid dynamics (CFD) simulation analyze spatial airflow and temperature profiles in containers, trucks [14], and inside boxes according to different packing designs [15]. [16] combined CFD modeling for temperature conditions and biological models to create a DT for mango fruits, and [17] used data loggers to collect temperature data from various transports, allowing detection of weak spots in the cooling chain using detailed thermal and quality modeling of the fruit. Although purely offline modeling has value, automated data integration arguably generates more benefits. For [7], the supply chain for agricultural produce needs not merely models but DTs because “each shipment is subject to a unique and unpredictable set of temperature and gas atmosphere conditions”.
220
R. Jedermann et al.
Other DT applications in agriculture have IoT origins. According to [3], although 28 applications have emerged since 2017, most are at a conceptual or prototype stage. Most deployed applications focus on remote sensing, such as monitoring the filling height of feed silos for livestock. Only two other deployed applications include detailed modeling; these use deep learning to monitor animal behavior or proactively prevent tractor malfunctions. Remote sensing solutions for refrigerated containers can be split into “remote container monitoring” systems, which focus on the maintenance and operation state of the cooling engine, and wireless data loggers, which can be placed inside boxes or pallets. The missing link between these two solution types [9] remains an obstacle for implementing DTs.
2 Making Models Updateable Applying multidisciplinary knowledge is among the key enablers of DTs [2]. DTs should provide a holistic model of a real object. Models for fluid dynamics, heat transfer and biochemical processes must be selected, programmed, and linked, with the complexity of modeling representing a challenge of its own. However, even if a complex simulation model is available, one DT-specific challenge remains; the set of models must be continuously “updateable” using real-time sensor data. Despite the importance of this key feature of DTs [8], few articles on DTs describe how the update process can be realized. We suggest three levels for categorizing solutions to implement the update process. First, the update is implemented by simply overwriting obsolete sensor data in the model. This is especially possible if the system state can be described using discrete values, which are measurable without noise, such as the location of a workpiece at a certain machine on a shop floor. A model tracking the workload of each machine simply has to overwrite information at each new RFID scanning event. The second level includes incremental models. Crucial system properties often have an accumulative character, such as a component’s accumulated mechanical stress. In the food logistics context, agricultural products have a high initial shelf-life, with a certain amount of quality lost each day depending on the current deviations from the optimal transport conditions. For computational efficiency, instead of recalculating the whole model, a decrement is recalculated according to the incoming real-time sensor data. Such incremental models are only applicable if the critical indicator––i.e., quality– –has a direct integral relation to a measurable quantity. Otherwise, a more complex update method is necessary. This third level entails methods that estimate system states that are not directly measurable. The general approach can be best understood within the framework of systems theory as discussed in the following section. The update process can be realized as a state-space observer. An overview of related mathematical methods regarding DTs can be found in [5]. Nonetheless, this paper focuses on the consequences from systems theory rather than on mathematical details. 2.1 State-Space Description The systems theory perspective indicates that each dynamic system can be described by a set of time-dependent variables x(t), or, in short, the state vector x. Vectors are
Digital Twin Features for the Intelligent Container
221
demarcated using bold font in the following. The transitions of x can be described using a set of first-order ordinary differential equations [18] with the initial value x0 . The change of states over time ∂x ∂t is given by a function of current state x, the control vector u, and system noise wi . The measurable quantities of the physical object y are given by a linear combination of the system states x and measurement noise wM (Fig. 1). Except for laboratory experiments, typical biological objects have fewer measurable outputs than internal states. Certain internal states are of less interest, and some can be excluded to produce a simplified model. However, others give crucial information regarding the object, including, for example, the degeneration of a biochemical substance, which indicates the loss of fruit quality over time. A simple example model for a fruit box might comprise three states, namely, the box core, surface temperatures, and the remaining concentration of the biochemical substance; the latter is not directly measurable. Vector u contains the system inputs or control variables; in this case, the cooling setpoint and the ambient temperature. The measurement noise wM includes the intrinsic noise of the sensor element, amplifier noise, and quantization noise of the analog-digital converter. The system or process noise wi describes unpredictable time-varying fluctuations in system states with known and unknown causes, including direct thermal noise, time-varying airflow due to turbulence, and random variations in the speed of biochemical reactions.
Fig. 1. State-space description of dynamic systems. Example properties are given in the white legends, and the principle of the update process is in orange.
In the case of a linear system, the differential equations simplify to (1), with matrix A describing the time behavior of the states and their mutual influences. Matrix B quantifies the effects of the inputs on the states. ˙ ∂x (1) = Ax + Bu + wi ∂t The DT model’s first task involves estimating “hidden” states that cannot be directly measured, such as the fruit quality indicator. Next, the model predicts the system’s
222
R. Jedermann et al.
future behavior. If the current state is at least roughly known, the model can predict future changes to x and the expected measurement values y. Due to noise, errors in the estimation of the model parameters A, B and the initial state x0 , predictions increasingly deviate from the real object over time. The update process must correct the estimated system states based on the available information, namely, the measured values y and the known control inputs u (see lower part of Fig. 1). A so-called state observer provides methods for correcting the states based on the error between actual measurement y and prediction yp , with the Kalman filter [18] representing the most common realization of such. A state can be either. • Directly measurable, for example, surface temperatures, in which case, the Kalman filter can be deployed to reduce measurement noise); or • Hidden, for example, core temperatures, which are not directly accessible by a sensor but impact surface temperature, making it feasible to estimate core temperature via long-term observation of the surface. Shelf-life cannot be directly measured. Destructive laboratory methods cannot be applied by real-time surveillance. However, if shelflife losses relate to increased respiration activity or heat production, shelf-life is at least theoretically observable. However, in practice, noise reduces the accuracy of such estimations. • Not observable, if quality changes do not cause temperature changes or other measurable outputs. In such cases, future projections based on the assumed initial value x0 are possible, but no later corrections can be made based on the measurements. Improvements compared with an incremental model are mostly unfeasible. Practical systems often feature non-linear relations, such as the relationship between temperature and shelf-life loss. In such cases, a Hidden Markow model can realize state observation [5]. Nonetheless, the general problem of observability remains, independent of non-linearity and of the model structure, even for black-box or machine-learning models. The DT literature mostly neglects such observability problems [5]. 2.2 Complexity of Biological Systems In moving from initial applications of DTs with pure mechanical objects to agricultural products involving biological systems, model complexity largely increases. Numerous chemical and enzymatic reactions must be considered. Although some of them can be excluded for simplified modeling, real-time measurements cannot access most of these states. If modeling begins at the point of harvest, the initial state x0 underlies variations due to influences such as nutrition, weather conditions during the growing period, mechanical damage during harvest, and the position of the fruit on the tree or plant. However, even fruits with identical harvest conditions demonstrate substantial differences in ripeness state. This biological variance creates a certain degree of modeling and state estimation uncertainty that must be considered during model deployment and networking with subsequent models. For example, our tests identified typical deviation of ±5 days in the shelf-life of bananas despite these bananas being harvested at the same time and transported and stored at the same temperature [19].
Digital Twin Features for the Intelligent Container
223
Furthermore, values of model parameters, such as the matrices A, B, can deviate or even drift over time. The diameter of gaps between pallets in a container affects airflow and, thus, temperature. The distribution of gaps is not known in advance due to careless stowage and non-cuboid pallet shape. Pallets might even move slightly during transportation, changing gap diameters. Our IC project developed a method [20] to describe the heat transfer from supply air to box temperature using a surrogate model featuring only two free time-constant parameters. During the two-week ocean transportation, these two parameters were identified from the temperature data: cooling efficiency and heat generation by ripening processes. Upon arrival, the fruit was exposed to ethylene to initiate the ripening process. Thereafter, the first model parameter was fixed to the identified value, and the second parameter was replaced with a time-dependent state variable that could be estimated using a Kalman filter.
3 Software Structure of Digital Twins The updateable models must be interlinked and fed the correct data. Values for past and current real-time readings and predictions for the future must be combined on the timeline as model inputs. Past model inputs can be logged control inputs to the physical object, direct sensor readings, Kalman filtered sensor data, or simulated outputs of preceding models. Future model inputs must be generated differently, with control inputs mostly assumed to remain at the last known value; furthermore, instead of sensor data, the related quantities should be predicted using the preceding models. The complexity of the required software platform increases with goal type. The action implementation stage includes bi-directional data exchange over the digital thread and complex decision-making based on a tool that tests and compares different possible interventions using virtual experiments. 3.1 Software Components The numerous possible configurations for connecting different sub-models with data streams for control inputs, sensor readings, model outputs and predicted values drives the need for standardized software solutions. Such platforms provide modularity for easily swapping or adding sensors and models. Furthermore, direct point-to-point linking of the models using proprietary interfaces becomes almost unmanageable for complex systems. Event-drive architectures using the “publish/subscribe” pattern are a common method for providing generic interfaces for DT platforms, with [21] suggesting an architecture comprising three layers: • A streaming platform or persistence layer acts as a mediator between sensors and models. Sensors publish their data to so-called topics or message queues. Models subscribe to certain topics to receive notifications of new data becoming available and retrieve information from other models. • The DT is linked to the physical object by the communication and integration layer. Basic components are sensors that capture the process state and mostly wireless
224
R. Jedermann et al.
communication that provides a secure and reliable channel for transmitting process information. • The sub-models are programmed as event-driven services that react to published events such as the availability of new sensor data or sensed distortions of the physical process. Although more complex descriptions of DTs exist, e.g. [2], the basic software components can be described in terms of these three layers. Numerous software platforms for DTs are available, with Source Forge listing more than 40 entries on its “Best Digital Twin Software” list [22] and Dashdevs including 20 commercial solutions on their list [23]. Eclipse IoT projects provide several components associated with implementing a DT [24]. The development of a specific DT can be based on such solutions. For example, communication and integration can mostly be handled by existing IoT solutions and enhanced using project-specific sensor types. Notably, the Message Queuing Telemetry Transport (MQTT) protocol has become one of the most common standards for transmitting real-time sensor data (Fig. 2). Several research projects [21, 25] use Apache Kafka as an open-source streaming platform for the persistence layer that is combined with the Eclipse Hono protocol adapter for MQTT. Commercial IOT soluƟons and wireless sensors T
rH
Product specific sensors Flow
Bio-physical models Event-driven Services Updateable implementaƟon
• Graphics output • ConfiguraƟon for goal type
…
MQTT Gateway
User Interface Services
Specific Model 1
Specific Model 2
Specific Model …
Publish / Subscribe Streaming plaƞorm / Persistence Layer Topics / Message queuing system Long term storage
Fig. 2. Typical DT software architecture. Necessary project-specific extensions to standard software and available IoT solutions are marked in orange.
Although some platforms entail general models, such as time-series analysis and deep learning, agricultural applications require the development of fruit and transportchain-specific models that depends on the user. The system is completed with the introduction of user interfaces for graphical output and control. 3.2 Converting Simulation Models into Event-Driven Services Describing physical objects using a set of sub-models should be as concise and extensive as possible. The first step involves collecting available models and then fine-tuning them in laboratory experiments for the specific product before completing the set of
Digital Twin Features for the Intelligent Container
225
sub-models with the missing elements. In the second step, the observability of system properties should be verified. Models that are only available in an offline form for simulation purposes must be converted into an updateable form in the third step, enabling the model to react to events such as the availability of new real-time data. If crucial system states are unobservable, an incremental model can represent the best solution. Otherwise, it is worthwhile pursuing optimal use of available sensor data using a state observer or a statistical- or machine-learning-based approach. Step four considers the uncertainties of model prediction, especially where states are unobservable. Error propagation for the uncertainty of model inputs must be included and forwarded through the model chain. Notably, sensors are a cost factor, requiring consideration of whether some of the sensors could be replaced by modeling without significantly increasing the prediction’s overall uncertainty. Finally, step five programs interfaces for communication with the streaming platform, which typically follows the “publish/subscribe” pattern. 3.3 From the Intelligent Container to a General Digital Twin Given our IC prototype was designed for demonstration purposes, it was focused on one specific product in one transport chain with a specific stowage scheme. Two example models were tested to predict the effect of temperature variations on the green-life of bananas and to predict box temperature as a function of a coefficient for cooling efficiency and a proportional factor for ripening heat. The aforementioned application of a Kalman filter as observer demonstrates the complexity of linking models in different configurations for the processing of real-time sensor data. Thus far, we have only demonstrated that our embedded processing platform has sufficient resources to run the Kalman filter. The model was only tested by manually combining recorded and predicted data from transport and ripening processes. For real-time deployment, numerous services must be linked in different configurations, including biological models, modules for parameter estimation, and data streams. Data streams must switch between past measurements from the physical object and future predictions from the DT models, with the latter becoming obsolete upon the arrival of new real-time data. A more generalized solution is preferable to the earlier point-to-point linking of modules. Streaming platforms provide more flexibility to exchange modules. For example, if a new model provides a more accurate estimation of the coefficient for cooling efficiency by direct airflow measurement, it can directly publish its prediction to the related topic after disabling a less accurate estimator, and subsequent models do not have to know where the data derives from. The IC demonstrator was mostly based on temperature sensors. The accuracy of prediction can be improved by integrating new sensor types such as 2-dimension airflow sensors and nondispersive infrared ethylene sensors [1]. Such sensors should implement DT-compatible protocols (e.g., MQTT). Accordingly, commercially available and project-specific sensors can be easily incorporated into the IC. Although other models––such as the influence of humidity on shelf-life and water condensation on fruit [26] and the headspace model for the accumulation of humidity and gases [13]––are available in principle, these have not been integrated into the IC.
226
R. Jedermann et al.
The most important task is converting available models into event-driven services as the previous section described. Notably, compliance with DT standard solutions enables the IC to be integrated into a system of systems. A DT for order and transport management can include several sub-systems representing each refrigerated container.
4 Summary The success of DTs in other applications fields drives the motivation to represent perishable products in the cool chain by detailed digital counterparts. Modeling the related biochemical and heat-transfer processes is critical to taking advantage of DT concepts. Furthermore, different research groups have established a focus on either fruit modeling or IoT technologies for real-time monitoring. Neither a binding definition of DT nor practical considerations preclude such applications from being considered a DT because each DT is legitimate in its corresponding context [6]. Instead, it is important to consider how certain applications can benefit from implementing additional DT features. The necessary combination of models with real-time data leads to the question, how the model can be updated by new incoming sensor measurements. Incremental models or simply overwriting obsolete values are often insufficient for updating the DTs of food products. Adopting systems theory with state-space descriptions provides a framework for understanding the problems and opportunities associated with a more elaborate update process. Although unmeasurable quantities of a physical object might be hidden, they are often observable in principle by their effects on other measurable values. The steps to making better use of DT features in the IC entail implementing stateobservers for hidden quantities, converting models to event-driven services, and implementing streaming platforms. These steps are the same for other agricultural processes. The streaming platform can be implemented using open-source solutions or commercial cloud services, with IoT sensor solutions broadly available. DT platforms provide adapters to handle various standard protocols, and only product-specific sensors remain to be added for special measurement tasks. Notably, the lack of models is the key obstacle to increasing the use of DTs in agriculture. Agricultural models and remote sensing applications are generally programmed as proprietary solutions. The models of several research groups have been published as mathematical algorithms or as online tools with a graphical user interface. Digital twins provide the most useful platform for combining and integrating such models following reprogramming as event-driven services. By applying the principles this article describes, available simulation models can be prepared to be part of a DT solution that offers better real-time monitoring and control in the food logistics field.
References 1. Jedermann, R., Lang, W.: 15 years of intelligent container research. In: Freitag, M., Kotzab, H., Megow, N. (eds.) Dynamics in Logistics, pp. 227–247. Springer, Cham (2021). https:// doi.org/10.1007/978-3-030-88662-2_11
Digital Twin Features for the Intelligent Container
227
2. Qi, Q., Tao, F., Hu, T., Anwer, N., Liu, A., Wei, Y., et al.: Enabling technologies and tools for digital twin. J. Manuf. Syst. 58, 3–21 (2021). https://doi.org/10.1016/j.jmsy.2019.10.001 3. Pylianidis, C., Osinga, S., Athanasiadis, I.N.: Introducing digital twins to agriculture. Comput. Electron. Agric. 184, 105942 (2021). https://doi.org/10.1016/j.compag.2020.105942 4. Zhuang, C., Liu, J., Xiong, H.: Digital twin-based smart production management and control framework for the complex product assembly shop-floor. Int. J. Adv. Manuf. Technol. 96(1–4), 1149–1163 (2018). https://doi.org/10.1007/s00170-018-1617-6 5. Cronrath, C., Ekström, L., Lennartson, B.: Formal properties of the digital twin – implications for learning, optimization, and control. In: 2020 IEEE 16th International Conference on Automation Science and Engineering (CASE), pp. 679–684 (2020). https://doi.org/10.1109/ CASE48305.2020.9216822 6. Uhlenkamp, J. F., Hribernik, K., Wellsandt, S., Thoben, K. D.: Digital twin applications: a first systemization of their dimensions. In: 2019 IEEE International Conference on Engineering, Technology and Innovation (ICE/ITMC), pp. 1–8 (2019). https://doi.org/10.1109/ICE.2019. 8792579 7. Defraeye, T., et al.: Digital twins are coming: Will we need them in supply chains of fresh horticultural produce? Trends Food Sci. Technol. 109, 245–258 (2021). https://doi.org/10. 1016/j.tifs.2021.01.025 8. Wright, L., Davidson, S.: How to tell the difference between a model and a digital twin. Adv. Model. Simul. Eng. Sci. 7(1), 1–13 (2020). https://doi.org/10.1186/s40323-020-00147-4 9. Jedermann, R., Praeger, U., Lang, W.: Challenges and opportunities in remote monitoring of perishable products. Food Packag. Shelf Life 14(A), 18–25 (2017). https://doi.org/10.1016/ j.fpsl.2017.08.006 10. Lang, W., Jedermann, R.: What can MEMS do for logistics of food? Intelligent container technologies: a review. IEEE Sens. J. 16(18), 6810–6818 (2016). https://doi.org/10.1109/ JSEN.2016.2576287 11. Tijskens, L.M.M.: Discovering the future: modelling quality matters (Ph.D. Thesis). University of Wageningen (2004). http://library.wur.nl/WebQuery/wurpubs/lang/334193 12. Gwanpua, S.G., et al.: The FRISBEE tool, a software for optimising the trade-off between food quality, energy use, and global warming impact of cold chains. J. Food Eng. 148, 2–12 (2015). https://doi.org/10.1016/j.jfoodeng.2014.06.021 13. Jalali, A., Linke, M., Geyer, M., Mahajan, P.: Integrative programming for simulation of packaging headspace and shelf life of fresh produce. MethodsX 8, 101514 (2021). https:// doi.org/10.1016/j.mex.2021.101514 14. Moureh, J., Flick, D.: Airflow pattern and temperature distribution in a typical refrigerated truck configuration loaded with pallets. Int. J. Refrig. 27(5), 464–474 (2004). https://doi.org/ 10.1016/j.ijrefrig.2004.03.003 15. Ambaw, A., Mukama, M., Opara, U.L.: Analysis of the effects of package design on the rate and uniformity of cooling of stacked pomegranates: numerical and experimental studies. Comput. Electron. Agric. 136, 13–24 (2017). https://doi.org/10.1016/j.compag.2017.02.015 16. Defraeye, T., Tagliavini, G., Wu, W., Prawiranto, K., Schudel, S., Kerisima, M.A., et al.: Digital twins probe into food cooling and biochemical quality changes for reducing losses in refrigerated supply chains. Resour. Conserv. Recycl. 149, 778–794 (2019). https://doi.org/ 10.1016/j.resconrec.2019.06.002 17. Shoji, K., Schudel, S., Onwude, D., Shrivastava, C., Defraeye, T.: Mapping the postharvest life of imported fruits from packhouse to retail stores using physics-based digital twins. Resour. Conserv. Recycl. 176, 105914 (2022). https://doi.org/10.1016/j.resconrec.2021.105914 18. Brown, R.G., Hwang, P.Y.C.: Introduction to Random Signals and Applied Kalman Filtering: with MATLAB Exercises, 4th edn. Wiley, Hoboken (2012)
228
R. Jedermann et al.
19. Jedermann, R., Lloyd, C., Poetsch, T.: Communication techniques and challenges for wireless food quality monitoring. Phil. Trans. R. Soc. A 372(2017), 20130304 (2014). https://doi.org/ 10.1098/rsta.2013.0304 20. Jedermann, R., Lang, W.: Model based estimation of biological heat generation during coldchain transport and processing. In: 3rd IIR International Conference on Sustainability and the Cold Chain, International Institute of Refrigeration (IIR), London, UK (2014) 21. López, C.E.B.: Real-time event-based platform for the development of digital twin applications. Int. J. Adv. Manuf. Technol. 116(3–4), 835–845 (2021). https://doi.org/10.1007/s00 170-021-07490-9 22. SourceForge: Slashdot Media: best digital twin software. https://sourceforge.net/software/dig ital-twin/. Accessed 26 Oct 2021 23. dashdevs: Product owner talks: 20 digital twins solution providers. https://dashdevs.com/ blog/product-owner-talks-20-digital-twins-service-companies/. Accessed 26 Oct 2021 24. Eclipse Foundation Europe GmbH: Eclipse IoT open source projects. https://iot.eclipse.org/ projects/. Accessed 26 Oct 2021 25. Kamath, V., Morgan, J., Ali, M.I.: Industrial IoT and digital twins for a smart factory: an open source toolkit for application design and benchmarking. In: 2020 Global Internet of Things Summit (GIoTS), pp. 1–6 (2020). https://doi.org/10.1109/GIOTS49054.2020.9119497 26. Linke, M., Praeger, U., Mahajan, P.V., Geyer, M.: Water vapour condensation on the surface of bulky fruit: some basics and a simple measurement method. J. Food Eng. 307, 110661 (2021). https://doi.org/10.1016/j.jfoodeng.2021.110661
Transportation Networks and Vehicle Routing
A New Lower Bound for the Static Dial-a-Ride Problem with Ride and Waiting Time Minimization Christian Pfeiffer(B) and Arne Schulz Institute of Operations Management, University of Hamburg, Hamburg, Germany {Christian.Pfeiffer,Arne.Schulz}@uni-hamburg.de
Abstract. The paper focuses on the static dial-a-ride problem with ride and waiting time minimization. This is an important problem setting of significant practical relevance, as several ridesharing providers launched in recent years in large cities. In contrast to the standard dial-a-ride problem, these providers focus on the general public. Therefore, they are amongst others in competition with taxis and private cars, which makes a more customer-oriented objective necessary. We minimize the sum of relative detours of all customers. The paper introduces upper bounds for the arrival times and an initial lower bound for the objective value. Our approach is tested in a computational study with realistic test instances. Keywords: Demand responsive transport Branch-and-bound · Lower bound
1
· Dial-a-Ride ·
Introduction
Private mobility in cities grows more and more important. In many cities, the residents have access to a wide range of motorized mobility options including public transportation, taxis or their private car. Over the past years ridesharing has grown to be a competitive alternative to the previous options. Uber, the biggest company in that area, connects private drivers and potential customers. MOIA—a subsidiary company of the automobile manufacturer Volkswagen—, ioki or CleverShuttle—subsidiary companies of the main German railway company Deutsche Bahn—have different business models to Uber [1,9,12]. They use their own fleet of cars and employed drivers. MOIA, ioki, and CleverShuttle use apps where customers can ask for a ride from a pickup location to a delivery location. The app matches the requested rides and makes an offer the customer can accept or decline. As the prices are lower than those for taxis, this concept works only if customer requests are combined as efficiently as possible. On the one hand, the customers decline tours that have a too long detour and use public transport, their private car or a taxi instead. On the other hand, few shared rides are not cost-effective for the provider. This paper develops a new lower bound for the optimization problem a provider of such a ridesharing service encounters. The problem is known as c The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 M. Freitag et al. (Eds.): LDIC 2022, LNLO, pp. 231–243, 2022. https://doi.org/10.1007/978-3-031-05359-7_19
232
C. Pfeiffer and A. Schulz
the dial-a-ride problem (DARP) in the literature and often optimized to minimize costs. In this paper, we consider an objective function for the DARP that does not aim to minimize costs but instead minimizes the customer inconvenience and was first introduced in Pfeiffer and Schulz [19]. Minimizing customer inconvenience allows the service provider to compete better with other modes of transport. The paper is structured as follows: First, we classify our problem setting into the literature (Sect. 2) and give a detailed problem description (Sect. 3). Then, we present the lower bound in Sect. 4. This is followed by a computational study in Sect. 5. Afterwards, we feature a conclusion in Sect. 6.
2
Literature Review
While customer inconvenience is often included in traditional formulations of the DARP [16], it is often modeled with the help of constraints: time windows and maximum ride time restrictions are usually utilized [2,6,21]. Our formulation forgoes these constraints and relies on the objective function to minimize the customer inconvenience. The DARP was originally proposed by Psaraftis [20], who solves the problem with dynamic programming. Since then there has been a large variety of employed methods to solve the DARP. Cordeau [2] and Ropke et al. [21] use Branch-and-Cut algorithms to solve the DARP exactly. One of the models presented in Ropke et al. [21] is the basis for the model formulation presented in this paper. Gschwind and Irnich [6] and Parragh et al. [18] utilize Branch-andPrice methods. For their Branch-and-Price algorithm, Gschwind and Irnich [6] split the original problem into a master and a subproblem. The master problem selects the tours while the subproblem aims to identify tours that improve the solution. The subproblem is solved with a labeling algorithm that utilizes dynamic time windows to reduce the size. Pfeiffer and Schulz [19] first introduced the DARP variant that we examine in this paper. They present an Adaptive Large Neighborhood Search and a Dynamic Program to solve the problem. The Dynamic Program can only solve instances with a single vehicle while the Adaptive Large Neighborhood Search algorithm can be utilized for multi-tour instances as well. Much of the literature for the DARP considers the transport of elderly, ill or handicapped passengers [4,7,11]. Nevertheless, the DARP has also been investigated in the context of demand responsive transport in urban areas [10,18]. For more in-depth reviews of the DARP literature we refer to Ho et al. [8] and Parragh et al. [16]. A brief overview about papers that minimize customer inconvenience in a DARP is presented in Table 1. The majority of the examined literature considers some form of cost minimization objective as well.
A Lower Bound for the Static DARP
233
Table 1. DARP literature about minimizing customer inconvenience ALNS: Adaptive Large Neighborhood Search DP: Dynamic Program GA: Genetic Algorithm MDLS: Multi-Directional Local Search PR: Path Relinking TS: Tabu Search VND: Variable Neighborhood Descent VNS: Variable Neighborhood Search Authors (year)
Constraints
Algorithm
Sexton and Min. excess ride time, Bodin (1985) [22] min. delivery time deviation
Vehicle capacity, latest delivery time
Benders
Diana and Dessouky (2004) [5]
Min. total distance, min. excess ride time, min. idle time
Max. ride time, vehicle capacity, time windows
Regret insertion
Jorgensen et al. (2007) [11]
Min. total transportation time, min. excess ride time, min. wait time
Vehicle capacity, soft time windows
GA
Parragh et al. (2009) [17]
Min. total distance, min. mean user ride time
Max. ride time, vehicle capacity, time windows
VNS, PR
Paquette et al. (2013) [15]
Min. total distance, min. mean (relative) user ride time, min. time, window deviations
Max. ride time, vehicle capacity, time windows, vehicle incompatibilities
TS
Molenbruch et al. Min. total distance, min. (2017) [13] mean user ride time
Max. ride time, vehicle capacity, time windows, pairing restrictions
MDLS, VND
Pfeiffer and Min. relative ride and Schulz (2021) [19] waiting time
Vehicle capacity
ALNS, DP
3
Objective
Problem Description
Compared to the standard DARP formulation [3], our problem features differences in the objective function as well as some of the constraints (compare [19]). Similar to the standard formulation we consider a number of requests n, each with their own given demand (i.e. the number of passengers in the request). There are K homogeneous vehicles, which all start and end at the same depot, and all have the same capacity Q. For each request i ∈ P there is a pickup (P = {1, ..., n}) with demand qi > 0 and a delivery location i + n ∈ D (D = {n + 1, ..., 2n}) with demand qi+n = −qi which need to be visited by the same vehicle and in the order pickup before delivery. Request i ∈ P has the earliest pickup time ri . Between all pairs of locations there are deterministic travel times tij which fulfill the triangle inequality. Accordingly, ri+n = ri + ti,i+n for all i ∈ P . If a vehicle arrives at a customer’s pickup location before the request time, it has to wait. Every customer needs to be serviced. We use distances and travel times synonymously in the following. Unlike the standard DARP formulation we do not minimize the driven distances. Instead we minimize the customer inconvenience by minimizing the sum of relative detours over all customers. The detour is defined as the sum of waiting and riding times. More formally, the detour is calculated by subtracting the sum
234
C. Pfeiffer and A. Schulz
of the request time and the direct travel time between pickup and delivery location from the arrival time at the delivery location. The detour is then divided by the direct travel time between pickup and delivery location to receive the relative detour. It is weighted by the number of passengers qi of request i ∈ P . By measuring the objective this way, a given nominal detour will have a larger effect on a customer with a short direct route than on one with a longer direct route. This reflects the actual inconvenience that many customers perceive in such a situation [14]. In metropolitan areas, ridesharing providers compete with several other modes of transportation such as public transport, taxis or private cars. As a result, it is paramount for the ridesharing company to be attractive for its customer base. Since many ridesharing providers utilize electric cars, reducing the traveled distance is furthermore of reduced importance, as recharging is cheap. In addition, should autonomous vehicles become more commonplace, driver costs can further be disregarded which reinforces the need to focus on customer inconvenience. Due to the minimization of customer inconvenience in the objective function, we omit maximum ride time restrictions and time windows. As a result, there is always a trivial feasible solution for our problem as long as no customer demand is higher than the vehicle’s capacity. Simply sequencing each delivery location right after its pickup location will never violate the capacity constraint in that case. The journey time of each customer will also be minimized in such a solution, but there might be large waiting times. Our objective will therefore aim to pool different requests together, which increases the journey time in favour of the waiting time, to reduce the overall customer inconvenience.
4
Mixed-Integer Program
In this section, we present a 2-index formulation for our problem setting (compare [19]). We define the problem on a directed graph G = (I, A), where I is the set of nodes and A is the set of arcs. Let I = {0, 2n + 1} ∪ P ∪ D be the set of all locations, where 0 and 2n + 1 represent the depot with r0 = r2n+1 = 0 and q0 = q2n+1 = 0. We use the indices i, j ∈ I. Moreover, let Mij be sufficiently large constants. We use the following variables: Xij = 1 if a vehicle drives from location i directly to location j and 0 otherwise, Qi specifies the number of passengers in the vehicle after leaving location i, and Bi indicates the departure time from location i. We assume that a request time ri is never smaller than the distance from the depot to the pickup node i, i.e. ri ≥ t0,i ∀i ∈ P , which means that visiting only a single customer in a tour will always result in an objective value of 0 for that tour. 4.1
Model Formulation
We need to define the sets S ⊆ I and S. S is the set of all sets S for which 0 ∈ S, 2n + 1 ∈ / S and there is at least one request i that has its delivery node in S but
A Lower Bound for the Static DARP
235
not the pickup node, i.e., S = {S : 0 ∈ S ∧ 2n + 1 ∈ / S ∧ ∃i : (i ∈ / S ∧ n + i ∈ S)}. The model formulation is then: min
i∈P
qi
Bi+n − ri − ti,i+n ti,i+n
(1)
Xij = 1
∀j ∈ P ∪ D
(2)
Xij = 1
∀i ∈ P ∪ D
(3)
i∈I
j∈I
X0j = K
(4)
j∈P
Xij ≤ |S| − 2
∀S ∈ S
(5)
Bi + tij − Mij (1 − Xij ) ≤ Bj
∀i ∈ I, j ∈ P ∪ D
(6)
Qi + qj − Q(1 − Xij ) ≤ Qj Bi ≥ ri
∀i ∈ I, j ∈ P ∪ D ∀i ∈ I
(7) (8)
i,j∈S
max{0, qi } ≤ Qi ≤ min{Q, Q + qi } Xij ∈ {0, 1}
∀i ∈ I ∀i ∈ I, j ∈ I
(9) (10)
The model was previously introduced in Pfeiffer and Schulz [19] and is based on the model in Ropke et al. [21]. Compared to Pfeiffer and Schulz [19] we set the discussed penalty factor for maxima w to 0 and therefore omit it from our model formulation, as this leads to a fair evaluation of our lower bound. A verbal description for the remaining model can be found in Pfeiffer and Schulz [19]. To find violations of Constraints (5), we use the same procedure as Ropke et al. [21]: Each time an integer solution is found in the Branch-and-Bound tree, we solve a maximum flow problem for every customer to determine whether the precedence constraints are met. If they are violated, the corresponding cuts are added and the solution is marked as infeasible. 4.2
Calculation of an Initial Lower Bound
In order to reduce the initial gap, we provide a lower bound for the objective value to the solver. A slightly modified version of this lower bound can also be used to tighten the bounds of the Bi variables (compare Subsect. 4.3). In order to calculate the lower bound, we solve subproblems of our original problem by only considering a subset of customers. We generate all possible subsets of customers with sizes [K + 1, . . . , K + φ] with φ ≥ 1. Each subproblem is solved to optimality and the corresponding objective value is saved. We do not consider smaller subsets because for an instance with n ≤ K the optimal objective value is always 0. The subsets are solved to optimality by going through all possible assignments of customers to tours and then solving the individual tours with
236
C. Pfeiffer and A. Schulz
the help of the dynamic program in Pfeiffer and Schulz [19]. The objective value for each tour is stored in a hash table and reused throughout the procedure to decrease the running time. These individual subproblems each constitute a lower bound for the original instance. However, combining them improves the bound further: We look for a combination of subproblems where no customer appears more than once across the combined subproblems. The maximum combination, i.e. the combination where the sum of the respective objective values is maximal, is then a lower bound for the original problem. This selection problem is solved with CPLEX. When evaluating a subproblem l with the customers Pl ⊂ P , we also utilize the already calculated solutions for all smaller subsets m with customers Pm ⊂ Pl to further reduce the computational effort. The largest optimal objective value of these smaller subsets m constitutes a lower bound for the objective of the larger subset l since adding a customer can never improve the objective value. If a particular assignment of customers to tours for subset l returns an objective value equaling the largest lower bound, the rest of the possible assignments do not need to be considered. We use this procedure also to generate n additional lower bounds where we remove a single customer each. This is accomplished by only combining subproblems where the customer does not appear. These additional n lower bounds are used in Subsect. 4.3 to improve the bounds for Bi . 4.3
Bounding of Variables
In order to tighten the variable bounds for the model formulation, we use the objective value of a feasible initial solution and the individual lower bounds for the problem without customer i from Subsect. 4.2. Let Z U B be the objective value of an initial solution and ZiLB the lower bound for the problem without customer i. Let BiLB and BiU B denote the lower and upper bound of variable Bi . We set them as follows: ⎧ ⎪ ∀i ∈ P ⎨ri LB Bi = ri + ti−n,i ∀i ∈ D (11) ⎪ ⎩ 0 otherwise and BiU B
⎧ (Z U B −Z LB )·t i,i+n i ⎪ + ri ⎨ qi UB = Bi−n + ti−n,i ⎪ ⎩ 0
∀i ∈ P ∀i ∈ D otherwise.
(12)
While the lower bounds in (11) are already included in the model formulation (1)–(10) by (8) and the restriction that i + n is visited after i by the same vehicle, the idea behind the upper bounds in (12) is that we already know that in the optimal solution, Bi of a single customer i can never be so large that the customer i would generate an objective value greater than that of the initial solution. This bound can be improved if we know how much of the objective value
A Lower Bound for the Static DARP
237
all the other customers will generate at minimum, which is the lower bound for the instance without the customer i. The upper bounds in (12) have, to the best of our knowledge, never been used before. Furthermore, BiLB and BiU B allow us to set Mij = BiU B − BjLB + tij
∀i ∈ I, j ∈ P ∪ D.
where applicable, we also adapted the bound tightening constraints from Cordeau [2] and Ropke et al. [21] for the variables Bi and Qi .
5
Computational Study
The Branch-and-Bound algorithm was implemented in C++, using the CPLEX C++ API and CPLEX version 12.9. All tests were run on a single thread of an AMD Ryzen 3990X and with 256 GB of RAM. 5.1
Instances
The instances are generated based on a district in the city of Hamburg, Germany, where a dial-a-ride service is operating. Within the district there are virtual stops which are used by the dial-a-ride provider. We draw the request locations randomly from these virtual stops. The request times are uniformly distributed but will always be large enough for a vehicle to drive straight from the depot to a pickup location and arrive before or at the request time. More precisely, if maxT = maxi∈P t0i is the maximum distance from the depot to a pickup location, the request times are drawn from the range [maxT, maxT + β · n] with β ≥ 0. β controls the temporal concentration of the requests and was set to 100 in our experiments, as this yielded instances that were challenging but not too difficult. Higher values for β tend to make the instances easier, as there is more time between requests on average. For smaller values of β the system’s capacity is exceeded as requests arrive faster than they can be completed. The distances between the locations are not symmetrical as the city layout is taken into account. However, all distances fulfill the triangle inequality. The vehicle capacity was set to 6 passengers, which is common in practice. The demand of each request (i.e. the number of passengers) is determined by an exponential distribution with a λ of 0.9. The value is rounded up. If the resulting value is not in the interval [1, 6], the value is discarded and redrawn until it is in the valid range. The number of requests n was varied between 10, 15, 20, and 30. We varied the number of tours K between 1, 2, and 3 for all three n. We generated 10 instances for every combination, which results in a test bed of 120 instances in total. For the mixed-integer program (MIP) we set the CPLEX parameter MIP emphasis to bestbound which helps in reducing the optimality gap. The other CPLEX settings are set to their defaults unless otherwise noted.
238
C. Pfeiffer and A. Schulz
5.2
Discussion of Results
We performed tests for several different configurations: the MIP without a starting solution and without the preprocessing lower bound procedure (MIP0 ), and the MIP with the starting solution mentioned in Subsect. 4.3 (generated by the Adaptive Large Neigborhood Search (ALNS) in Pfeiffer and Schulz [19]) and with the preprocessing lower bound using different values for the parameter φ as described in Subsect. 4.2 (MIPφ ). φ is varied between 1, 2, 3, and 4. Finally, we ran an extra MIP configuration where we used depth first search and set the CPLEX parameter MIP emphasis to feasibility (MIPDF 4 ). This last configuration is otherwise identical to MIP4 and was used to put more focus on improving the upper bound. With the exception of MIP0 , all other configurations used the improved bounds from (12). Each CPLEX run had an hour of computation time which was exhausted for almost all instances. Only some of the smaller instances did so most often (in 10 of with n = 10 could be solved to optimality. MIPDF 4 the 120 instances). In Table 2, the average upper bounds for all MIP configurations as well as the starting solution by ALNS are displayed. The configuration with the best but even MIPDF cannot improve the starting average upper bound is MIPDF 4 4 solutions much in the one hour run, which indicates that CPLEX struggles with the model formulation. MIP0 performs worst since it does not receive the initial ALNS solution and has difficulties to find feasible solutions for instances with n > 10. None of the feasible results of MIP0 ever beat the initial solution. Configurations MIP1 to MIP4 only vary slightly from one another and only rarely find a better solution than the initial one. MIP1 , for example, can only improve on the initial solution’s upper bound in 2 out of the 120 instances. Table 2. Average upper bounds of all MIP configurations and the intial solution from ALNS. For config MIP0 , – indicates that not all 10 instances were solved to feasibility n
K MIP0 MIP1
MIP2
MIP3
MIP4
MIPDF ALNS 4
10 1
61.30
59.69
59.69
59.69
59.69
59.69
59.69
10 2
21.84
21.80
21.80
21.80
21.80
21.80
21.80
10 3
8.17
8.17
8.17
8.17
8.17
8.17
8.17
15 1
– 113.19 113.19 113.19 113.19 113.19
113.19
15 2
–
48.34
48.34
48.34
48.34
48.34
15 3
–
24.30
24.30
24.30
24.30
24.30
24.30
20 1
– 187.75 187.80 187.75 187.80 187.62
187.80
20 2
–
78.89
78.89
78.89
78.89
78.78
78.89
20 3
–
37.06
37.06
37.06
37.06
37.06
37.06
30 1
– 382.00 382.00 382.00 382.00 381.79
382.00
30 2
– 151.33 151.33 151.34 151.34 150.89
151.34
30 3
–
78.73
78.73
78.73
78.73
78.72
48.34
78.73
A Lower Bound for the Static DARP
239
Table 3 displays the average lower bounds before (LB b ) and after (LB a ) the execution of CPLEX in every configuration, i.e. the lower bound after our lower bound procedure was executed and the lower bound after the CPLEX run. Since we did not execute the lower bound procedure in the configuration MIP0 , the respective “before” column is omitted (and effectively 0 for every entry). Clearly, as φ is increased, the lower bound is increased as well since more combinations are evaluated. For φ ≥ 2, the lower bound improvement procedure can clearly outperform the lower bound that is received when only running CPLEX for an hour (MIP0 ). In instances with few vehicles and many customers, this effect is more pronounced. In the most extreme case, for n = 30 and K = 1 the average lower bound is improved by more than 400% (compare the result of MIP4 ). Additionally, for instances with few vehicles the lower bound preprocessing is very fast as well: For n = 30 and K = 1 the average computation time was less than three minutes when setting φ = 4 (compare Table 4). The difficulty of solving the model with a traditional MIP solver is further reemphasized by the fact that the preprocessing lower bounds are rarely if ever improved in the subsequent CPLEX run where the lower bound was provided as an initial lower bound. Combining this result with the insight that the upper bounds are also only barely improved, it seems that using a heuristic method in combination with the lower bound preprocessing beats a traditional MIP solver in both solution gap as well as time. The computation times for the preprocessing lower bound can be seen in Table 4. For configurations with low φ or instances with low K, the computation times are negligible. For instances with a large number of vehicles, the computaTable 3. Average lower bounds before (LB b ) and after (LB a ) the CPLEX run of different MIP configurations n
K MIP0 MIP1 LB a LB b LB a
MIP2 LB b LB a
MIP3 LB b LB a
MIP4 LB LB a b
MIPDF 4 LB b LB a
10 1
21.83 14.33 17.34 23.05 23.05 30.25 30.25
38.36
38.36
38.36
38.36
10 2
13.02
10 3
7.68
13.39
13.79
13.39
14.42
7.42
5.23
7.66
5.23
7.80
15 1
16.79 21.18 21.18 35.80 35.80 45.84 45.84
56.42
56.42
56.42
56.42
15 2
11.72
9.56 10.16 14.16 14.16 19.29 19.29
22.34
22.34
22.34
22.34
15 3
9.16
5.40
9.20 10.71 10.71
13.30
13.30
13.30
13.30
20 1
19.67 29.00 29.00 48.21 48.21 64.84 64.84
78.35
78.35
78.35
78.35
20 2
12.43 12.27 12.27 18.95 18.95 25.44 25.44
30.40
30.40
30.40
30.40
15.13
15.13
15.13
15.13
20 3
9.05
5.88 11.30
8.74 11.13 12.04 12.76
2.37
3.90
6.71
7.64
8.39
8.97
7.30
4.42
7.59 10.32 10.32 12.58 12.58
30 1
21.15 44.11 44.11 74.55 74.55 96.75 96.75 119.18 119.18 119.18 119.18
30 2
15.18 18.76 18.76 28.07 28.07 36.79 36.79
44.56
44.56
44.56
44.56
30 3
11.31 10.67 10.77 16.43 16.43 20.84 20.84
25.47
25.47
25.47
25.47
240
C. Pfeiffer and A. Schulz
tion times increase substantially when also setting φ = 4. However, they are still well below the allocated time of the MIP while generating higher quality lower bounds. In future research, the computation times could be reduced if the subproblems are selected heuristically instead of evaluating all of them. The same table also shows the average relative improvement of the lower bounds when incrementing φ. Since the relative improvement of the lower bounds decreases with each additional increase of φ, lower values for φ can be worthwhile to employ, as they maintain computation times rarely exceeding a minute. In Fig. 1, the occupancy rate for the various instances is visualized. Understandably, reducing the number of vehicles and increasing the number of requests increases the occupancy rate per vehicle. This increases the profit for the service provider but is unattractive for the customers. A low occupancy rate, however, is economically not sustainable for the provider but beneficial for the customers. It is therefore important to find a suitable number of vehicles that provides an adequate service level (see Table 2) while still being economically viable.
n=10
n=15
0.3
Relative share of total ride time
0.2
0.1
Ride time with (X) passengers 0.0 n=20
0 passengers 1 passenger 2 passengers 3 passengers 4 passengers 5 passengers 6 passengers
n=30
0.3
0.2
0.1
0.0 3
2
1
3
2
1
K
Fig. 1. Development of the occupancy rate when varying K and n for the solutions by MIP4 . Each bar is the relative share of ride time with that amount of customers. For example, for n = 10 and K = 3 the vehicles have two passengers in 20% of the time.
A Lower Bound for the Static DARP
241
Table 4. Average time for the calculation of the preprocessing lower bound in seconds (t) and the average relative improvement of the lower bound (ΔLB ) when incrementing φ n
6
K
t (s) φ=1 φ=2 φ=3 φ=4
ΔLB (%) φ : 1 to 2 φ : 2 to 3 φ : 3 to 4
10 1
0.00
0.01
0.10
0.53
60.90
30.67
26.74
10 2
0.01
0.02
0.10
0.50
47.72
38.33
10.65
10 3
0.01
0.02
0.12
0.57
63.57
13.91
17.97
15 1
0.01
0.04
0.49
4.88
69.90
28.76
23.42
15 2
0.01
0.07
0.57
5.27
47.77
36.46
16.06
15 3
0.04
0.15
0.93
6.30
64.42
20.57
24.93
20 1
0.01
0.09
1.42
20.79
66.85
34.67
20.80
20 2
0.03
0.19
1.91
23.42
53.99
34.27
19.95
20 3
0.06
0.41
3.89
40.88
56.10
22.69
20.83
30 1
0.01
0.25
6.63
159.84 69.05
29.77
23.11
30 2
0.05
0.80
11.67 188.54 49.73
31.16
21.09
30 3
0.20
2.40
40.30 557.24 54.02
26.69
22.31
Conclusion
The paper presents a new initial lower bound for the static dial-a-ride problem with ride and waiting time minimization. The initial lower bound outperformed the lower bound of CPLEX after one hour of computation in almost all of our 120 instances. Furthermore, the initial solutions of the ALNS were rarely beaten by CPLEX after an hour of computation. Therefore, we conclude that using the initial lower bound in combination with a metaheuristic such as ALNS is more beneficial relating to computation time and solution quality than using CPLEX with one hour of computation time. Our research leads to several possible directions for further research. First of all, the computation times for the initial lower bound can be improved by selecting the solved subproblems heuristically. In the same way, φ can be increased without increasing the computation time. Furthermore, the problem formulation can be adapted such that the customers can determine their latest arrival times at the delivery location instead of their earliest departure times (request times). The model determines the departure times instead of the arrival times in this case. This is a realistic problem setting because customers might have important appointments at their destination and want to be sure to be on time. Funding. This project was supported by the Hamburger Beh¨ orde f¨ ur Wissenschaft, Forschung, Gleichstellung und Bezirke (BWFGB; Hamburg authority for science, research, equalization, and districts). No grant number is available.
242
C. Pfeiffer and A. Schulz
Availability of Data. The instances are available under the following link: http:// doi.org/10.25592/uhhfdm.9670. Conflict of Interest. The authors declare that they have no conflict of interest.
References 1. CleverShuttle: Shared rides for a better furture (2022). https://www.clevershuttle. de/en/. Accessed 03 Jan 2022 2. Cordeau, J.F.: A branch-and-cut algorithm for the dial-a-ride problem. Oper. Res. 54(3), 573–586 (2006) 3. Cordeau, J.F., Laporte, G.: The dial-a-ride problem: models and algorithms. Ann. Oper. Res. 153(1), 29–46 (2007) 4. Detti, P., Papalini, F., de Lara, G.Z.M.: A multi-depot dial-a-ride problem with heterogeneous vehicles and compatibility constraints in healthcare. Omega 70, 1– 14 (2017) 5. Diana, M., Dessouky, M.M.: A new regret insertion heuristic for solving large-scale dial-a-ride problems with time windows. Transp. Res. Part B Methodol. 38(6), 539–557 (2004) 6. Gschwind, T., Irnich, S.: Effective handling of dynamic time windows and its applications to solving the dial-a-ride problem. Transp. Sci. 49(2), 335–354 (2015) 7. Heilporn, G., Cordeau, J.F., Laporte, G.: An integer l-shaped algorithm for the dial-a-ride problem with stochastic customer delays. Discret. Appl. Math. 159(9), 883–895 (2011) 8. Ho, S.C., Szeto, W.Y., Kuo, Y.H., Leung, J.M.Y., Petering, M., Tou, T.W.H.: A survey of dial-a-ride problems: literature review and recent developments. Transp. Res. Part B Methodol. 111, 395–421 (2018) 9. ioki GmbH: ioki—inspiring smart mobility (2022). https://ioki.com/en/home/. Accessed 03 Jan 2022 10. Jaw, J.J., Odoni, A.R., Psaraftis, H.N., Wilson, N.H.: A heuristic algorithm for the multi-vehicle advance request dial-a-ride problem with time windows. Transp. Res. Part B Methodol. 20(3), 243–257 (1986) 11. Jorgensen, R.M., Larsen, J., Bergvinsdottir, K.B.: Solving the dial-a-ride problem using genetic algorithms. J. Oper. Res. Soc. 58(10), 1321–1331 (2007) 12. MOIA GmbH: Reunion begins here (2022). https://www.moia.io/en. Accessed 03 Jan 2022 13. Molenbruch, Y., Braekers, K., Caris, A., den Berghe, G.V.: Multi-directional local search for a bi-objective dial-a-ride problem in patient transportation. Comput. Oper. Res. 77, 58–71 (2017) 14. Nie, W.: Waiting: integrating social and psychological perspectives in operations management. Omega 28(6), 611–629 (2000) 15. Paquette, J., Cordeau, J.F., Laporte, G., Pascoal, M.M.: Combining multicriteria analysis and tabu search for dial-a-ride problems. Transp. Res. Part B Methodol. 52, 1–16 (2013) 16. Parragh, S.N., Doerner, K.F., Hartl, R.F.: A survey on pickup and delivery models: part ii: transportation between pickup and delivery locations. J. f¨ ur Betriebswirtschaft 58(2), 81–117 (2008) 17. Parragh, S.N., Doerner, K.F., Hartl, R.F., Gandibleux, X.: A heuristic two-phase solution approach for the multi-objective dial-a-ride problem. Networks 54(4), 227– 242 (2009)
A Lower Bound for the Static DARP
243
18. Parragh, S.N., de Sousa, J.P., Almada-Lobo, B.: The dial-a-ride problem with split requests and profits. Transp. Sci. 49(2), 311–334 (2015) 19. Pfeiffer, C., Schulz, A.: An ALNS algorithm for the static dial-a-ride problem with ride and waiting time minimization. OR Spectr. 44, 87–119 (2022) 20. Psaraftis, H.N.: A dynamic programming solution to the single vehicle many-tomany immediate request dial-a-ride problem. Transp. Sci. 14(2), 130–154 (1980) 21. Ropke, S., Cordeau, J.F., Laporte, G.: Models and branch-and-cut algorithms for pickup and delivery problems with time windows. Networks 49(4), 258–272 (2007) 22. Sexton, T.R., Bodin, L.D.: Optimizing single vehicle many-to-many operations with desired delivery times: I. Scheduling. Transp. Sci. 19(4), 378–410 (1985)
An Auction-Based Multi-Agent System for the Pickup and Delivery Problem with Autonomous Vehicles and Alternative Locations Johan Los1 , Frederik Schulte1(B) , Matthijs T. J. Spaan2 , and Rudy R. Negenborn1 1
2
Department of Maritime and Transport Technology, Delft University of Technology, Mekelweg 2, 2628 CD Delft, The Netherlands {F.Schulte,R.R.Negenborn}@tudelft.nl Department of Software Technology, Delft University of Technology, Van Mourik Broekmanweg 6, 2628 XE Delft, The Netherlands [email protected]
Abstract. The trends of autonomous transportation and mobility on demand in line with large numbers of requests increasingly call for decentralized vehicle routing optimization. Multi-agent systems (MASs) allow to model fully autonomous decentralized decision making, but are rarely considered in current decision support approaches. We propose a multiagent approach in which autonomous vehicles are modeled as independent decision makers that locally interact with auctioneers for transportation orders. The developed MAS finds solutions for a realistic routing problem in which multiple pickup and delivery alternatives are possible per order. Although information sharing is significantly restricted, the MAS results in better solutions than a centralized Adaptive Large Neighborhood Search with full information sharing on large problem instances where computation time is limited. Keywords: Autonomous vehicle routing · Pickup and delivery problem · Alternative locations · Preferences · Multi-agent system Auctions
1
·
Introduction
The overarching trends towards automation and service orientation in transportation (Speranza 2018) go along with a rising need for decentralized decision support of individuals and autonomous vehicles. Autonomous transportation services may no longer depend on (human-controlled) centralized routing, but may autonomously optimize routes on the level of a single vehicle and may therefore even act as independent vehicular entrepreneurs. As Mobility as a Service (MaaS) solutions, such services rely on a digital platform (mobile app or web page) through which the end-users can access all the necessary resources for c The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 M. Freitag et al. (Eds.): LDIC 2022, LNLO, pp. 244–260, 2022. https://doi.org/10.1007/978-3-031-05359-7_20
An Auction-Based MAS for Pickup and Delivery with Alternative Locations
245
their trips (Jittrapirom et al. 2017). The big players in the automobile industry anticipate this development: while Toyota sees itself in the transition from a car manufacturer to a mobility service provider (Buckland and Sano 2018), Volkswagen even envisions a mobility platform on which vehicles would act as autonomous entrepreneurs (Munford 2018). When vehicles act as independent intelligent agents, their coordination and cooperation becomes increasingly significant (Qu et al. 2008), particularly for cooperative routing and traffic management (Zhou et al. 2017). While many cooperative transportation models assume a centralized planning approach with full information sharing (Guajardo and R¨ onnqvist 2015), there are other applications where competition seriously limits information sharing (Feng et al. 2017). While mobility platforms enable horizontal collaboration with multiple advantages, they also can easily turn into a problem of “coopetition”, describing a situation in which logistics service providers are competitors in one market and cooperate in another market (Cruijssen et al. 2007). This raises a need for decentralized control: fully centralized planning requires a full exchange of information, which is not in the partners’ interest when they are competitors in other markets (Cleophas et al. 2019). Moreover, assuming that such platforms may become dominant design in future mobility (Atasoy et al. 2020), there might easily be 100.000 vehicles or more and a respective number of requests—posing a tremendous computational challenge for centralized approaches based on NPhard problems. Embedding agent-based routing models in multi-agent systems (MASs) is one way to explicitly model decentralized optimization with limited information sharing (Los et al. 2020b). This approach differs from combinatorial auctions, as reviewed by Gansterer and Hartl (2018), among others in the fact that each request is evaluated by agents locally and no centrally defined bundles are auctioned. Thus, agents are given a larger degree of freedom. In this work, we develop a multi-agent approach for solving a decentralized Generalized Pickup and Delivery Problem with Preferences (GPDPP) (Los et al. 2018), where customers or operators can specify multiple alternative timelocation combinations for pickup or delivery. Autonomous vehicle agents solve their individual decentralized subproblems based on the requests they receive. Moreover, order agents are responsible for the individual transport orders, that is, they try to find an assignment of the order to a vehicle, which can be understood as an intelligent (algorithmic) contract. We compare the decentralized MAS approach with a centralized single-agent system (SAS) approach assuming full information availability (see Fig. 1). Despite the limited information sharing, the MAS solutions outperform the solutions of the single-agent approach under certain conditions. The results demonstrate that the MAS is particularly useful when computation time is limited and the problem size is large. We introduce the problem in Sect. 2. In Sect. 3, we describe the developed MAS in detail. Then, in Sect. 4, we introduce the centralized SAS to compare the different approaches in Sect. 5. Finally, in Sect. 6, we summarize our findings and give proposals for future research.
246
2
J. Los et al.
Decentralized Generalized Pickup and Delivery Problem with Preferences
In this section, we describe the GPDPP this paper deals with. First, we give a general motivational introduction. Next, we describe the problem from the local vehicles’ and orders’ perspective, and finally, we consider the problem from the global perspective. An extensive formal description is given by Los et al. (2018).
OA1
OA2
OA3 CA
VA1
VA2
VA3
(a)
(b)
Fig. 1. Different computational approaches. In (a), a decentralized multi-agent approach is shown, where all orders and vehicles are represented by order agents and vehicle agents, respectively, that interact locally with each other, without communicating with a central manager. In (b), a centralized single-agent approach is shown, for which all vehicle and order information is sent to a central manager, and there is no direct local communication. OA: order agent; VA: vehicle agent; CA: central agent. The dotted box represents the computational system, outside is the physical system. Information flow is represented by arrows.
2.1
Problem Motivation
Our problem differs in two aspects from classical Pickup and Delivery Problems (PDPs) (Cordeau et al. 2008) or Dial-a-Ride Problems (Molenbruch et al. 2017). First, we consider the problem to be inherently decentralized, that is, vehicles are independent, can attach to or detach from the system at any moment, and might not be willing or able to share all information. Transport requests also might continuously appear, disappear, or change, and require immediate actions. Computing a central solution might be no longer possible in such situations. Second, we consider a realistic problem with alternative locations and preferences. Instead of a single pickup location and a single delivery location per order,
An Auction-Based MAS for Pickup and Delivery with Alternative Locations
247
as is the case in the classical PDP, we allow for multiple alternative pickup and/or delivery locations per order, with their individual time windows. Furthermore, different preference values can be assigned to each location. One of the possible pickup locations and one of the possible delivery locations is chosen and attended by a vehicle (Los et al. 2018). A motivation for this generalization of the PDP is the high absence rate of customers in regular home delivery processes. Next to the (still preferred) option of home delivery between 10:00 and 12:00, for example, a (less appreciated but still acceptable) delivery at the locker box station two streets away might be allowed, with the advantage of a larger time window. Also, different pickup or delivery time windows can be assigned to the same physical location. In this scenario with multiple alternatives, a higher delivery success rate can be achieved, and the transport operator has more flexibility in designing efficient routes. 2.2
Local Problem Definitions
A problem instance consists of a set C of transport orders and a set V of vehicles. In the next sections, we consider the problem from the perspective of the individual orders and vehicles. Order Problem. Each order c ∈ C has a load quantity Qc , a set of possible pickup alternatives Pc , and a set of possible delivery alternatives Dc . A pickup or delivery alternative i is defined by a tuple ni , ei , li , di , pi , where ni is the location where the order can be picked up or delivered, the earliest service start time ei and the latest service start time li determine the time window in which the service can start, the service duration di determines the time that is needed for loading or unloading at ni , and the preference value pi ∈ (0, 1] describes the relative satisfaction for the alternative. We assume that each order has at least one pickup alternative i and one delivery alternative j with pi = pj = 1, meaning that there are no other alternatives preferred over these. An order c ∈ C needs to be served by a vehicle k ∈ V , that is, it needs to be picked up by k as described by one pickup alternative i ∈ Pc and delivered by k as described by one delivery alternative j ∈ Dc , while Mkc + β((1 − pi ) + (1 − pj )) should be minimized. Here, Mkc are the marginal routing costs for vehicle k to include order c into its route, and β is a positive parameter representing the weight of dissatisfaction relative to travel cost. Vehicle Problem. Each vehicle k ∈ V has a capacity Bk , a start location αk and an end location ωk . For each pair of nodes i, j, we denote the travel time and travel costs from location ni to location nj by tij and cij , respectively. A feasible route for a vehicle k ∈ V is a sequence of locations that meets the following requirements: – the vehicle starts its route at αk and stops at ωk ; – all time constraints are respected, that is, if the vehicle serves an alternative i, it arrives at ni between ei and li , it leaves ni not before di after arrival time, and traveling from ni to nj takes a time of at least tij ;
248
J. Los et al.
– all precedence constraints are respected, that is, if the vehicle serves a pickup alternative i and a delivery alternative j belonging to the same order, the arrival at ni takes place before the arrival at nj ; – all capacity constraints are respected, that is, the load of the vehicle may never exceed its capacity Bk throughout its route, but is increased or decreased with the order load quantity at a pickup or delivery location, respectively. A vehicle k ∈ V can transport orders if it keeps a feasible route, and the vehicle agent should minimize the travel costs. Thus, the vehicle agent tries to find combinations of a pickup alternative and a delivery alternative of the new order that can most efficiently be incorporated into its current route, such that the previously agreed on alternatives from already included orders still can be served. Hence, the vehicle agent locally solves multiple instances of a standard single-vehicle PDP (see, e.g., Parragh et al. 2008). 2.3
Local and Global Perspective
From a local perspective, order agents need transportation by a vehicle that realizes a preferred pickup and delivery, but is not too bad in terms of vehicle route costs. Vehicle agents must obtain routes with minimal travel costs. These local objectives contribute to the objective from the global perspective: to find a minimal cost solution. A global solution consists of a set of feasible routes (one for each vehicle k ∈ V ), such that each order c ∈ C is served by one vehicle. Global cost is defined as the sum of travel costs and dissatisfaction costs, where the sum of cij values for i and j such that the edge from ni to nj is part of a vehicle’s route constitutes the travel costs, and all non-preferred alternatives i that are served by a vehicle contribute a term β(1 − pi ) to the dissatisfaction costs.
3
Multi-Agent System
We propose a MAS approach to solve the GPDPP in a decentralized manner, in accordance with the assumption that vehicles and orders can independently attach to a platform. We introduce two types of agents that represent the main stakeholders of the problem: order agents, each responsible for getting one of the orders transported, and vehicle agents, each representing one vehicle. For finding an assignment of orders to vehicles, the order agents and vehicle agents communicate with each other in a multi-agent auction (Wooldridge 2009). Order agents act as auctioneers that offer a transportation task. Vehicle agents act as bidders. Hence, there is no central auctioneer, and no central authority that is aware of the global routing plan, but information is exchanged locally (see Fig. 2). Although the solution quality might be suboptimal from a global perspective, this approach resembles a transportation demand and supply market that allows for a fast response to dynamic events. Our auction mechanism is based on the systems described by M´ahr (2011), Gath (2016), and Los et al. (2020a; 2022; 2020b).
An Auction-Based MAS for Pickup and Delivery with Alternative Locations
3.1
249
General Approach
Order agents try to make a contract with a vehicle agent for a pickup and delivery with high preference values, but are cooperative in the sense that they take vehicle routing costs into account and accept lower preference values if this decreases the routing costs enough. Vehicle agents are responsible for making contracts with order agents, but have the local goal of minimizing the sum of travel costs while keeping a feasible route. Input
Local
Output
Bid p+d option, vehicle, costs
Order properties
Confirmation yes/no, bid
Auctioning system
Cancellation contract
Current contract
Acceptance bid
Contract update costs
Request p+d options, load quantity, auction end time
(a)
Input
Request p+d options, load quantity, auction end time
Local
Output
Vehicle properties
Bid p+d option, vehicle, costs
Cancellation contract
Routing system
Confirmation yes/no, bid
Acceptance bid
Current route plan
Contract update costs
(b)
Fig. 2. Information input and output, as well as local information flows, for an order agent (a) and a vehicle agent (b).
250
J. Los et al.
When they enter the system, order agents send a request for transportation to a well-selected set of vehicle agents. These compute the routing costs for inserting the order into their current route (by solving multiple single-vehicle subproblems, each with a different combination of an alternative pickup and delivery location of the new order, together with the orders already included in their route) and propose different bids. Order agents evaluate the bids from different vehicles and choose the one that is best, based on the routing costs of the bid and their own preferences for the alternative locations. If no changes have occurred meanwhile in the route of the chosen vehicle agent, the order will be inserted into its route. Orders auction themselves again after some time to check if there are better options due to the high dynamics within vehicle routes. In line with Los et al. (2020b), order agents interact only with a well-chosen subset of the vehicles (instead of with all vehicles) to limit the communicational and computational load. Although we might lack some good bids from other vehicles, we expect a better result due to a gain in time. In contrast to other MAS approaches, we introduce two properties that are specific for the GPDPP. First, preference costs are locally considered by the order agent. Second, it is possible for vehicle agents to send multiple bids to an order agent, resulting from different combinations of pickup and delivery alternatives. An example of the general auction procedure is shown in Figs. 3 and 4, and detailed agent algorithm descriptions are given in the next sections. 3.2
Order Agent
An order agent keeps track of the contract of the order, consisting of a transporting vehicle, one pickup and one delivery alternative that are agreed on, as well as the costs for transportation. Initially, there is no contract; the order agent organizes auctions for obtaining and improving a contract. An order agent starts its first auction in the system immediately after its release time. First, it selects a set of vehicle agents to send a request for transportation. As in other approaches, the set of all vehicles in the system can be used, but this can result in an overload of messages and subsequent vehicle computations, although not all of them have a high potential of being useful. For example, consider an order that needs to be picked up and delivered in the northern part of a city, and a vehicle that has only pickups and deliveries in the southern part of the city in its current plan. A match is not likely in this case. Different selection heuristics are possible, based on, e.g., the current vehicle locations, the planned routes, and the occupancy rate of the vehicles. In this paper, we select the vehicles based on the spatiotemporal distance of the different pickup and delivery alternatives of the order to the planned routes of the vehicles. The order agent opens the auction by sending all its possible pickup and delivery locations, the corresponding time windows and service durations, the load quantity, and the time at which the auction will end to all selected vehicles. If an order agent receives a bid (consisting of a pickup alternative i, a delivery alternative j, and the marginal travel costs) from a vehicle agent, it adds its dissatisfaction costs β(2 − pi − pj ) (see Sect. 2) for the specific alternatives to
An Auction-Based MAS for Pickup and Delivery with Alternative Locations Order agent 1 P1 ={5}, D1 ={6,7}
Vehicle agent 1 R1 : α1 –1–2–ω1
Vehicle agent 2 R2 : α2 –3–4–ω2
Request {P1 ={5}, D1 ={6,7}}
Option p:5, d:7 R2 : α2 –3–4–ω2 : 6
Req. {P1 ={5}, D1 ={6,7}} Option p:5, p:6 R1 : α1 –1–2–ω1 : 7
time
Bid {p:5, d:7, cost:5} Bid {p:5, d:6, cost:2}
α1 –5–6–1–2–ω1 α1 –5–1–6–2–ω1 α1 –5–1–2–6–ω1 α1 –1–5–6–2–ω1 α1 –1–5–2–6–ω1 α1 –1–2–5–6–ω1
: : : : : :
15 14 13 10 9 10
Option p:5, d:7 R1 : α1 –1–2–ω1 : 7 α1 –5–7–1–2–ω1 α1 –5–1–7–2–ω1 α1 –5–1–2–7–ω1 α1 –1–5–7–2–ω1 α1 –1–5–2–7–ω1 α1 –1–2–5–7–ω1
251
: : : : : :
16 15 15 12 11 12
α2 –5–7–3–4–ω2 α2 –5–3–7–4–ω2 α2 –5–3–4–7–ω2 α2 –3–5–7–4–ω2 α2 –3–5–4–7–ω2 α2 –3–4–5–7–ω2
: : : : : :
13 12 17 11 13 14
Option p:5, d:6 R2 : α2 –3–4–ω2 : 6 α2 –5–6–3–4–ω2 α2 –5–3–6–4–ω2 α2 –5–3–4–6–ω2 α2 –3–5–6–4–ω2 α2 –3–5–4–6–ω2 α2 –3–4–5–6–ω2
: : : : : :
∞ ∞ ∞ ∞ ∞ ∞
Bid {p:5, d:7, cost:4}
Select best bid
Accept {p:5, d:6}
Set route R1 : α1 –1–5–2–6–ω1
Confirm
Fig. 3. Schematic overview of an auction round in the MAS, with an order agent having one pickup option and two delivery alternatives, and two vehicles with current routes α1 –1–2–ω1 and α2 –3–4–ω2 . The order agent sends a request with its pickup option 5 and delivery alternatives 6 and 7 to both vehicle agents. The vehicle agents each consider the two options, one with delivery alternative 6 and one with delivery alternative 7. They insert the new locations into their current routes and compare the costs (as defined by the graph of Fig. 4) of the different new routes to the cost of their current routes. A bid with the least increase in costs is sent back. Note that insertion of delivery alternative 6 is not feasible for vehicle agent 2; hence, only one bid is sent back. The order agent selects the best bid (consisting of delivery alternative 6 with a cost of 2), and notifies vehicle agent 1 of its acceptance. (The example abstracts from time windows and preferences.)
252
J. Los et al. 6 4 α1
1
4
2
2
1
3
3
2
3
ω1
3
3 1
2
4
5 5 3 α2
1
2
3
3 3
3
4
4
4 3
1
ω2
3
2
3
4
7
Fig. 4. Graph with initial vehicle routes and an order with two possible delivery locations corresponding to the auction process overview of Fig. 3. Edges represent the travel costs between locations.
the bid the the
travel costs to obtain the total costs for the bid. Subsequently, it stores the in a sorted list with increasing bid costs. When the auction time has ended, order agent selects the first bid of its list, compares the costs of that bid to costs of its current contract, if possible, and acts appropriately:
– If the costs of the selected bid are lower than that of the current contract, or there is no current contract, the order agents asks the vehicle agent that proposed the bid to insert the order into its route. If a positive response follows, the order agent updates its current contract, cleans up its bid list and schedules to start a new auction after some time. Furthermore, a message is sent to the vehicle agent of the previous contract (if applicable) to inform this agent that the order can be removed from its route. In case of a negative response of the vehicle agent, the bid has become outdated. In this case, the order agent possibly includes a new bid of the vehicle agent into its bid list, selects the next bid of its bid list and repeats the procedure. – If the costs of the selected bid are not lower than the costs of the current contract, the current contract is still the best option. The agent cleans up its bid list and schedules to start a new auction after some time. – If there is no bid selected (i.e., the bid list was empty) and there is no current contract, the order agent immediately starts a new auction. If vehicle routes have been changed in the meantime, probably it will obtain some bid from the new auction. This is urgent since there is no contract yet.
An Auction-Based MAS for Pickup and Delivery with Alternative Locations
3.3
253
Vehicle Agent
A vehicle agent keeps track of the planned route of the vehicle, along with earliest and latest possible times for each location, and the used vehicle capacity at each trajectory. Initially, the route only consists of the vehicle’s start and end locations. When a vehicle agent receives a request from an order agent, it checks whether the auction has not yet ended. If there is still time, it computes the marginal travel costs for inserting each combination of alternatives into its current route, that is, it solves the single-vehicle PDP multiple times: once for each possible combination of a pickup and a delivery alternative of the new order. If an insertion is possible, a bid consisting of the marginal travel costs (the costs of the new route minus the costs of the current route), the pickup alternative and the delivery alternative is sent to the order agent. Hence, a vehicle agent can return multiple bids based on one request. For quick vehicle computations, we use a fast greedy insertion heuristic instead of solving the local vehicle problem in an exact way. The current sequence of the route will be kept, and feasibility (of time windows and capacities) will be checked for insertion of the new pickup and delivery at all possible positions (see Fig. 5). If a vehicle agent receives the acceptance of a bid from an order agent, it checks whether including the corresponding pickup and delivery alternatives into its route is still possible for the same (or less) costs. If this can be done, the vehicle agent updates its route accordingly and confirms this to the order agent. Otherwise, it sends a negative response to the order agent, together with a new bid for the same pickup and delivery alternatives, if possible. The rationale is that the vehicle still might have a better offer than other vehicles, although the costs might be higher than in the initial bid. Each time a vehicle agent changes its route plans (after insertion or removal of an order), it informs all order agents that are affected by the changes about their new routing costs: for all order agents that have a pickup or delivery directly before or after an inserted or deleted location in the route, the vehicle agent computes what it would gain by removing the pickup and delivery of that order. These actual routing costs will be sent to the corresponding order agents; they update the costs of their contracts, which is useful when they compare bids to their contract in a new auction.
254
J. Los et al.
(a) p1
p3
p2
d2
d1
p2
d2
d1
d2
d1
d3
(b) p1
p3
d3
(c) p1
p2
p3
d3
(d) p1
p2
d2
d1
p3
d3
(e) p1
p2
d2
d1
p3
d3
Fig. 5. Different insertion possibilities for a new pickup (p3 ) and delivery (d3 ) into a vehicle’s route consisting of two orders. The greedy heuristic keeps the sequence of the current route (p1 , p2 , d2 , d1 ). The number of possible routes to check is quadratic in the number of pickups and deliveries.
An Auction-Based MAS for Pickup and Delivery with Alternative Locations
4
255
Single-Agent System
To measure the performance of our MAS, we compare with a solution that is computed in a centralized way. For such a SAS, it is assumed that all order and vehicle information is available at one central place. We use an Adaptive Large Neighborhood Search (ALNS) algorithm (Ropke and Pisinger 2006), adapted to the situation with multiple locations and preferences, as SAS. First, an initial solution is computed by a greedy heuristic. Then, ALNS iteratively looks for improvements of the current solution by changing parts of it. In each iteration, some orders are removed from the current solution, and reinserted again into the remaining routing plan. Different heuristics for removal and reinsertion can be used; they are selected based on their performance in previous iterations. For details, see Algorithm 1 and Ropke and Pisinger (2006). An indication of the quality of ALNS applied to the GPDPP is given by Los et al. (2018). Algorithm 1: Adaptive Large Neighborhood Search Input: feasible solution x Initialize best solution xb ← x Initialize weights w while stop criterion not met do Select destroy and repair heuristics d and r based on w xt ← r(d(x)) if accept(x, xt ) then x ← xt end if costs(xt ) 0, system has an H∞ noise attenuation performance index γ if it is stochastically stable for all nonzero w(k). Here e is the error value calculated from estimation signal and target signal in system. (2) e22 < γw22 In order to estimate the corresponding signal state, the following full-order filter Gf used to estimate the robust normalized boundary condition is designed in a limited-dimensional probability space. ˆθ(k) y(k) x ˆ(k + 1) = Aˆ1,θ(k) x ˆ(k) + Aˆ2,θ(k) x ˆ(k − dk ) + B ˆθ y(k) ˆ(k) + E zˆ(k) = Fˆθ x k
k
(3a) (3b)
ˆθ(k) , Fˆθ(k) , E ˆθ(k) are the filter system parameter matrixes. zˆ(k) Aˆ1,θ(k) , Aˆ2,θ(k) , B refers to the measurement target signal output. Sojourn interval is defined between the adjacent instant as sn kn+1 − kn . The kn indicates the time instant sample at the nth jump with the initial condition k0 = 0. Combine system G and filter Gf , we obtain the state estimation error system Ge as follows: ˜ x ˜(k + 1) = A˜1 x ˜(k) + A˜2 x ˜(k − dk ) + Bw(k) ˜ e(k) = F˜ x ˜(k) + Gw(k)
(4a) (4b)
where x ˜(k) = [xT (k), x ˆT (k)]T , e(k) = z(k) − zˆ(k). And A1,θ(k) A2,θ(k) 0 0 Gθ(k) ˜ ˜ ˜ A1 = ˆ , A2 = ,B = ˆ Bθ(k) Cθ(k) Aˆ1,θ(k) Bθ(k) Dθ(k) 0 Aˆ2,θ(k) ˆθ(k) Cθ(k) , −Fˆθ(k) ], G ˜ = [−E ˆθ(k) Dθ(k) ] F˜ = [Hθ (k) − E Next, to abstract the state from cargo throughput process, we need to give the following propositions which are useful to comprehend the quantization state in filter.
330
B. Ren et al.
Proposition 1. The cargo situation can be quantified into level set according to the busy degree of cargo throughout capacity. Remark 2. In this paper, we quantify the cargo situation according to the degree of busy. Considering the mean value of cargo throughout capacity, one typical quantization way is to quantize the level by designing a fuzzy membership function, such as the Gaussian distribution. Proposition 2. The modal θ(k) of port system is jumped between the quantization levels of cargo situation, and it follows the semi-Markov process. Proposition 3. The state x(k) in semi-Markov jump system is composed by a cargo imported process and a delayed state quay crane handled. Remark 3. System state x(k) and delayed state x(k − dk ) represent different realistic meaning. We assume that the quay crane is delay dk affected after it was launched. Meanwhile, the cargo amount state x(k) involving with the imported queue process can be substituted by Poisson distribution in application. Besides, we consider the state value as an increment formal, which is convenient to build the s-MJLs.
3
Method
In this section, in order to develop the objective of the H∞ filtering synthesis method, a new H∞ performance criterion for the filtering error system (4a) and filter design formation is presented as follows. Lemma 1. [17] If there exists matrices M , N T and L with appropriate dimension that are satisfied with M T ≥0 ∗ N then the following inequality holds T a M T −L a −2a Lb ≤ b ∗ N b T
Theorem 1. The filter error system Ge is stable and with H∞ performance norm bound γ > 0, if there exists positive matrixes N, S, R, Q, P and state delay dk satisfying the upper boundary condition that 0 < dk < d¯k , such that the following LMIs(linear matrix inequalities) ⎛ ⎞ ¯ −P 0 P A¯1 P A¯2 PB ⎜ ∗ −I F˜ ⎟ ˜ 0 G ⎜ ⎟ ⎜ ∗ ∗ S ⎟0 ∗ Q
> 0,
331
(6)
I I ¯= 0 , ¯2 = 0 , B , A ˜ A˜1 + A˜2 − I −I A˜2 B W 0 S = −P + d¯k M + 0 d¯k (Q + R)
Proof. We construct the Lyapunov-Krasovskii functional V (k, θ(k)) = V1 (k, θ(k)) + V2 (k, θ(k)) + V3 (k, θ(k)) where
(7)
V1 (k, θ(k)) = x(k)T P x(k) V2 (k, θ(k)) =
−1 k−1
βsT [R + Q]βs
j=d¯k s=k+j
V3 (k, θ(k)) =
k−1
(8)
x(k)T W x(k)
s=k−dk
Consider the positive-definite matrices P, Q, R and we define increment between different state β = x ˜(k + 1) − x ˜(k). Then we calculate ΔVk V (k + 1) − V (k) = ΔV1 + ΔV2 + ΔV3 . It can be shown that ΔV1 = x(k + 1)T P x(k + 1) − x(k)T P x(k) T k−1 k−1
x(k) x(k) = (A¯1 P A1 − P ) β(s)A¯2 A2 β(s) + β(k) β(k) s=k−d¯k s=k−d¯k ¯ T P A¯1 x(k) + w(k)T B ¯ T P Bw(k) ¯ + 2w(k)T B β(k) k−1 k−1
x(k) T ¯ T P A¯2 β(s) − 2 A¯T1 P A¯2 β(s) −2 w(k)T B β(k)
s=k−d¯k
x(k) ΔV2 = β(k) T x(k) ΔV3 = β(k)
s=k−d¯k
k−1 k−1
x(k) 0 0 −{ δ(s)}T (Q + R){ δ(s)} ¯ β(k) 0 dk (Q + R) s=k−dk s=k−dk W 0 x(k) − x(k − dk )T W x(k − dk ) 0 0 β(k) (9)
T
332
B. Ren et al.
Using Lemma 1, we obtain k−1
−2
¯T P A ¯2 β(s) ≤ w(k)T B
s=k−d¯k
k−1
−2
s=k−d¯k
k−1
s=k−d¯k
x(k) β(k)
T
w(k) β(s)
T
¯2 ¯T P A Z S−B ∗ Q
w(k) β(s)
⎤T ⎤ ⎡ x(k) x(k) ¯ ¯ P A M N − A 1 2 ¯T P A ¯2 β(s) ≤ ⎣ β(k) ⎦ ⎣ β(k) ⎦ A 1 ∗ R β(s) β(s) s=k−d¯k k−1
⎡
(10) Substituted above inequalities to (7), then we have Jk = E
∞
[eT (k)e(k) − γ 2 wT (k)w(k)]
k=0
[eT (k)e(k) − γ 2 wT (k)w(k) + ΔVk ] ⎤⎞ ⎛⎡ T ⎤ ⎡ A¯1 F¯ T G S + F¯ T F¯ N ¯ +⎣ ⎦⎠ φ ∗ −W S = φT ⎝⎣ A¯T2 ⎦ P[A¯1 , A¯2 , A] T T 2 ¯ ¯ A ∗ ∗ G G + dk Z − γ I (11) k−1 where φ = [x(k)T , β T , { s=k−dk δ(s)}T , w(k)T ]T . ≤E
Remark 4. Note that under the discrete-time situation, this delay-independent stability condition is obtained when the dk > 1. Theorem 2. Consider the error filter system Ge with state delay upper bound d¯k , if there exist positive define matrices P11 , P22 , Q11 , Q22 and matrices P12 , Q12 −1 with appropriate dimensions. R = ς11 > 0, ω > 0 such that Θ11 Θ12 3, for this would already include 99% of the possible actual values underlying a given measurement. We propose choosing an α between 0.5 and 1.5, but this is up to the designer. Second, along with the first observation, α should be chosen depending on the constraint technique. A smaller α may suffice for a tolerant technique, while a higher α might be more appropriate for a restrictive technique. Third, defining bi ∈ {−1, 0, 1} is a compromise as it has to be ensured that an algorithm employing this technique will terminate. More true to the cause would be bi ∈ [−1, 1]. However, we achieve identical results when gj is continuous in the relevant neighbourhood. With this technique mainly being developed for experimental setups, this is the case for most situations. Fourth, domain experts must supply the evolutionary algorithm expert with the necessary information on standard deviations of the measurements. This necessity of configuration motivates our next contribution, which the following subsection explains. 3.2
Configurable Constraint Handling
Especially in real-world applications, domain knowledge is significant for designing the evolutionary algorithm. Designers make some decisions initially, e.g., the encoding or the choice of mutation and recombination operators. Others, however, may be subject to change when new variables are introduced or information about the included variables changes. When faced with a dynamic environment, a changing infrastructure or a developing research system, these changes are likely to happen more often than less often. Plump et al. propose a data description language for providing information on data to the evolutionary algorithm without having to change the principal design of this algorithm [9]. They use
354
C. Plump et al.
experiment example ( ” x ” ) { measuring method measurement ( ” x ” ) {
}
}
quotient d e s c r i p t o r ” v a l u e 0” ( ” x . 0 ” ) i n § c o n s t r a i n t ( ” x . 0 ” >= 0 . 0 , ” s t r i c t ” ) § standardDeviation ( 1 . 5 ) ;
[ ”µm” ]
quotient d e s c r i p t o r ” v a l u e 1” ( ” x . 1 ” ) i n § c o n s t r a i n t ( ” x . 1 ” >= 0 . 0 , ” s t r i c t ” ) § standardDeviation ( 0 . 8 ) ;
[ ”µm” ]
§ c o n s t r a i n t ( ” x . 1 ” / ” x . 2 ” >= ” x . 0 ” , ” vague ” ) § c o n s t r a i n t ( ” x . 1 ” >= ” x . 0 ” , ” n o i s e ” )
Listing 1.1. Exemplary data description file
this to include dependency information on the search space dimensions to modify mutation and recombination operators. We enhance this data description language to include information about constraints on all search space dimensions. Listing 1.1 shows an example of this, where one can see the included boundary and constraint information. However, as mentioned above, it might make sense to treat different constraints with different techniques depending on the domain. Which technique to choose for which constraint requires the expertise of both experts (domain and evolutionary algorithm). The domain expert needs to assess the strictness and assurance of a given constraint—whether, for example, it is defined by a physical law or only a known relationship that may be somewhat vague. On the other hand, the evolutionary algorithm expert needs to define the constrainthandling technique to use. We introduce a separation of concern in this case: In the data description file, the domain expert can label each constraint with a type. Additionally, he can supply the standard deviation necessary for noisesensitive constraint handling. The evolutionary algorithm expert can define the constraint-handling technique for every type defined by the domain expert in a configuration file for the evolutionary algorithm. Listing 1.2 shows one possible corresponding configuration file for a given data description file.
4
Implementation
We implemented the proposed constraint handling with Jenetics [13] to evaluate our proposed approach. Jenetics is a genetic algorithm, evolutionary algorithm, genetic programming, and multi-objective optimisation framework. The implementation uses Xtext [7] for parsing data description language files. We generate the configured constraint handling mechanisms based on the domain information and the evolutionary algorithm’s configuration. Since the constraint handling mechanisms differ, we used two of Jenetic’s extension points to implement the strategies. For eliminating individuals, we use Jenetic’s constraint mechanism. This mechanism allows a client application to check all individuals of a generation. If the constraint implementation marks an individual as invalid, Jenetis
Choosing the Right Technique for the Right Restriction
355
{ ” algorithm ” : { ” c o n s t r a i n t −h a n d l i n g ” : { ”strict”: { ” c a l c u l a t i o n ” : { ”name” : ” normal ” } , ” h a n d l i n g ” : { ”name” : ” k i l l A t B i r t h ” } }, ” vague ” : { ” c a l c u l a t i o n ” : { ”name” : ” normal ” } , ” h a n d l i n g ” : { ”name” : ” m a l u s F o r F i t n e s s ” , ” s m o o t h i n g ” : 1 . 0 } }, ” noise ” : { ” c a l c u l a t i o n ” : { ”name” : ” s t a n d a r d −d e v i a t i o n ” , ” alpha ” : 1 . 0 } , ” h a n d l i n g ” : { ”name” : ” k i l l A t B i r t h ” , ” s m o o t h i n g ” : 1 . 0 } } }, ... } ... }
Listing 1.2. Excerpt of the evolutionary algorithm configuration
eliminates the individual and produces a new one. For the malus-based constraint handling, we decorated the fitness function. The decorator pattern allows us to use the default fitness calculation and modify the result based on the constraint evaluation and the configuration.
5
Evaluation
To gain insight into the effects of our presented approach, we carried out a thorough evaluation. With this evaluation, we plan to investigate the following research questions: Research Question 1. Which influence does the adjustability of constraint handling techniques have on the conformance of individuals to the given constraints? Research Question 2. Is the above mentioned influence, if present, dependent on the choice of encoding: real-valued or bit encoding? 5.1
Setup of Evaluation
We use four standard benchmark functions for our evaluation as fitness functions: the Ackley-Function, the Rastrigin Function, the Rosenbrock Function, and the Weighted Sphere Function. These four functions have been chosen according to the classification attributes of functions vital for evolutionary algorithms: Separability and Modality. Each function represents one combination of these two attributes. All four functions were evaluated with ten dimensions as well as two dimensions. Additionally, to achieve some shift of the functions, we did not simply minimise towards zero but minimised the distance to the given target and varied that.
356
C. Plump et al.
Furthermore, to effectively answer Research Question 2 we used both encodings: A 64-bit floating-point real-valued encoding and a bit encoding with 14 decimal places. We chose an elite selector combined with a tournament selector for parent and offspring selection. For the real-valued encoding, we chose a Gaussian Mutator and a Line Recombinator and for the bit encoding a Swap Mutator and a single crossover as recombination. These operator combinations had proven the most effective for our given situation in previous work. We carefully chose data description files for the 2D and 10D case with standard deviations and constraints. We varied the assigned type of constraint between strict, vague, noise to symbolise the domain expert’s preferred type of constraint-handling. As constraint-handling techniques, we implemented the kill-at-birth approach as representative for the restrictive methods and the penalty approach as representative for tolerant methods. For both the kill-at-birth and the penalty approach, we added the option of noise-sensitivity with a configurable degree α. ackley
rastrigin
rosenbrock
weighted Sphere
1000
750
bit
Number of violating individuals
500
250
noise−handling
0
kill 1000
malus
750
double
500
250
0
all−
t t t t d d e d e e d e e e e e ried ried ried ried noisall−stricll−vagull−varie ue−va noisall−stricll−vagull−varie ue−va noisall−stricll−vagull−varieue−va noisall−stricll−vagull−varie ue−va all− all− all− a a a a a tVag a tVag a tVag a tVag stric stric stric stric
Fig. 1. Number of constraint-violating individuals for constraint x0 ≥ 0.0 in a twodimensional search space with x0 representing the first dimension. The columns show the results for the different benchmark functions, the rows differentiate the chosen encoding. For each configuration, there is one boxplot for the bit-encoding and one for the double encoding.
We defined five different configurations: all-strict (all constraints are to be handled with a strict technique), all-vague (all constraints are to be handled with a vague technique), all-noise (all constraints are to be handled noise-sensitive), and all-varied (all constraint types are used) and strict-vague-varied (only strict and vague types occur).
Choosing the Right Technique for the Right Restriction
357
All in all, we had 8000 different configurations. Each was run 20 times for statistical reasons and with 100 generations and 1000 individuals. 5.2
Results
Figure 1 presents an overview of the acquired data. The boxplots show the distribution of the number of individuals in the last generation, which violates this configuration’s first constraint. The description noise-handling refers to the chosen constraint-handling technique for constraints of type noise (kill for the restrictive one, and malus for the tolerant one). The first observation is the high variance for configurations with only type vague (all-vague) constraints and, of course, all-noise with noise-handling malus. Second, both configurations with variations and the all-strict configuration have much lower variance and, in most cases, a smaller median, i.e., they have fewer constraint-violating individuals in more cases. Furthermore, the smallest results are achieved either by all-varied, when employing malus as noise-handling and strictVague-varied. Figure 2 shows the same overview for the second constraint. Please note, that for varied configurations the first constraint was of type strict and the second of type vague or noise, resp.. We see again a higher variance for the all-vague case and occasionally also in the all-varied configuration with malus noise-handling. However,
ackley
rastrigin
rosenbrock
weighted Sphere
1000
750
bit
Number of violating individuals
500
250
noise−handling
0
kill 1000
malus
750
double
500
250
0
all−
t t t t d e d e d e d e e e e e ried ried ried ried noisall−stricll−vagull−varieue−va noisall−stricll−vagull−varie ue−va noisall−stricll−vagull−varie ue−va noisall−stricll−vagull−varie ue−va all− all− all− a a a a a tVag a tVag a tVag a tVag stric stric stric stric
Fig. 2. Number of constraint-violating individuals for constraint x1 ≥ x0 in a twodimensional search space, x0 , x1 representing the first and second dimension, resp. The columns show the results for the different benchmark functions, the rows differentiate the chosen encoding. For each configuration, there is one boxplot for the bit-encoding and one for the double encoding.
358
C. Plump et al.
with the exception of these outliers, all-varied outperforms the other configurations. For both constraints, all-strict seems to perform better when employed with a double (real-valued)encoding instead of a bit-encoding. The difference in the results for both constraints stems partially from the number of variables in the constraints and additionally, because the second constraint receives the vague or noise technique in the varied configurations. 5.3
Discussion
With the above-made observations, we can discuss our research questions from above. Research Question 2 has already been answered by the last sentence above: It seems to have an influence, especially with constraints of type strict. This may be caused by the high variability of the bit encoding, as a single bit flip can have a huge impact on the offspring, whereas the Gaussian Mutator (employed for double encoding) instead produces offsprings in the neighbourhood. Therefore, double encoding is more respective of restrictive techniques. Research Question 1 asked for the influence of the configuration-possibility: For both situations, the all-varied case outperform or equals the others. Even—and this is interesting—when the constraint handling for the specific constraint is equal. However, the difference in the other constraint seems to have an effect through the recombination. This is particularly interesting, and we would like to further investigate this in the future.
6
Conclusion and Further Research
Evolutionary algorithms are beneficial as a tool for optimisation when the optimisation task has many local optima or the function is very complex such that direct approaches like Hillclimbing fail. This increased complexity often occurs when trying to “invert” machine-learned predictions. These situations are often precisely those that need a considerable amount of domain knowledge. We proposed using this domain knowledge to choose the proper constraint-handling technique for a given restriction, allowing different techniques to occur for different constraints. Furthermore, to cope with the fact that experimental setups are usually subject to noise, we defined noise-sensitive constrained-handling techniques based on the standard deviation of the present noise. We expanded our framework for evolutionary algorithms to check for noise sensitivity and extended a domain-driven data description language to adopt these changes. We evaluated our approach by comparing evolutionary algorithms where every constraint received the same treatment to those adapted based on the type of constraint. We found our approach to have the desired effect and intend to include more constraint-handling techniques into our framework for future work.
Choosing the Right Technique for the Right Restriction
359
References 1. Coello Coello, C.A.: Theoretical and numerical constraint-handling techniques used with evolutionary algorithms: a survey of the state of the art. Comput. Methods Appl. Mech. Eng. 191(11), 1245–1287 (2002). https://doi.org/10. 1016/S0045-7825(01)00323-1. https://www.sciencedirect.com/science/article/pii/ S0045782501003231 2. Deb, K.: Multiobjective Optimization Using Evolutionary Algorithms. Wiley, New York (2001) 3. Deb, K., Goel, T.: A hybrid multi-objective evolutionary approach to engineering shape design. In: Zitzler, E., Thiele, L., Deb, K., Coello Coello, C.A., Corne, D. (eds.) EMO 2001. LNCS, vol. 1993, pp. 385–399. Springer, Heidelberg (2001). https://doi.org/10.1007/3-540-44719-9 27 4. Drechsler, R., Eggersgl¨ uß, S., Ellendt, N., Huhn, S., M¨ adler, L.: Exploring superior structural materials using multi-objective optimization and formal techniques. In: International Symposium on Embedded Computing and System Design (ISED), pp. 13–17 (2016) 5. Drechsler, R., Huhn, S., Plump, C.: Combining machine learning and formal techniques for small data applications - a framework to explore new structural materials. In: 23rd Euromicro Conference on Digital System Design, DSD 2020, Kranj, Slovenia, 26–28 August 2020, pp. 518–525. IEEE (2020). https://doi.org/10.1109/ DSD51259.2020.00087 6. Ellendt, N., M¨ adler, L.: High-throughput exploration of evolutionary structural materials. HTM J. Heat Treat. Mater. 73, 3–12 (2018). https://doi.org/10.3139/ 105.110345 7. Eysholdt, M., Behrens, H.: Xtext: implement your language faster than the quick and dirty way. In: Proceedings of the ACM International Conference Companion on Object Oriented Programming Systems Languages and Applications Companion, OOPSLA 2010, pp. 307–309. Association for Computing Machinery, New York (2010). https://doi.org/10.1145/1869542.1869625 8. Michalewicz, Z., Schoenauer, M.: Evolutionary algorithms for constrained parameter optimization problems. Evol. Comput. 4(1), 1–32 (1996). https://doi.org/10. 1162/evco.1996.4.1.1 9. Plump, C., Berger, B.J., Drechsler, R.: Domain-driven correlation-aware recombination and mutation operators for complex real-world applications. In: IEEE Congress on Evolutionary Computation, CEC 2021, Krak´ ow, Poland, 28 June– 1 July 2021, pp. 540–548. IEEE (2021). https://doi.org/10.1109/CEC45853.2021. 9504931 10. Riche, R.L., Knopf-Lenoir, C., Haftka, R.T.: A segregated genetic algorithm for constrained structural optimization. In: Proceedings of the 6th International Conference on Genetic Algorithms, pp. 558–565. Morgan Kaufmann Publishers Inc., San Francisco (1995) 11. Steinbacher, M., et al.: Descriptors for high throughput in structural materials development. High Throughput 8(4) (2019). https://doi.org/10.3390/ht8040022 12. Weicker, K.: Evolution¨ are Algorithmen. Vieweg+Teubner Verlag (2015). https:// doi.org/10.1007/978-3-658-09958-9 13. Wilhelmst¨ otter, F.: Jenetics (2021). https://jenetics.io
Production Planning and Scheduling
Managing Complexity in Variant-Oriented Manufacturing: A System Dynamics Approach Phillip Kießner1(B)
and H. Niles Perera2,3
1 Chair of Enterprise Logistics, Technische Universität Dortmund, Leonhard-Euler-Straße 5,
Dortmund, Germany [email protected] 2 Center for Supply Chain, Operations and Logistics Optimization, University of Moratuwa, Katubedda 10400, Sri Lanka [email protected] 3 Professor H.Y. Ranjit Perera Institute for Applied Research, Nugegoda 10250, Sri Lanka [email protected]
Abstract. This paper proposes a System Dynamics (SD) approach to support decision-making to manage variety induced complexity. Offering product variety leads to increasing internal complexity, which results in higher inventory and increasing setup processes. Managing the trade-off between marketing-, logisticsand product management complicates the decision process in offering sufficient variety to the market. This leads to numerousness of stock keeping units (SKUs), all of which are required to maintain various key performance indicators and inventory levels. Managing this variety induced complexity to optimize the overall business success requires an understanding of its System Dynamics behavior and interrelation. The reviewed literature reveals that existing metrics do not capture the necessary dynamic system behavior sufficiently to measure the impact of long-term strategies. The proposed model combines System Dynamics and the portfolio-fitness index (PFI) metric to capture the required dynamic system behavior. Applying scenarios to a national electronics company through a case study demonstrates the ability to manage complexity using System Dynamics and the PFI metric. The outcome of this research is a System Dynamics model that can manage variety induced complexity by offering scenario analysis to support decision-making. The findings in the case study suggest that reducing the complexity does not automatically lead to competitive advantages. Understanding the dynamic behavior of complexity impacts forms a basis for decision-making. Thus, the model’s findings provides insights to manage complexity in the most efficient manner. Keywords: System Dynamics · Complexity · Operations research · Product variety · Supply chain complexity
1 Introduction Companies face increasing customer requirements, decreasing product life cycles and dynamics of growing markets [9]. Therefore, companies are compelled to opt for growing product portfolios, which leads to higher inventory levels through product variety and © The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 M. Freitag et al. (Eds.): LDIC 2022, LNLO, pp. 363–375, 2022. https://doi.org/10.1007/978-3-031-05359-7_29
364
P. Kießner and H. N. Perera
increasing complexity [24]. A great challenge facing companies is how to deal with the requested product variety while maintaining optimal performance. To ensure sales and differentiation potential against the competition, increasing complexity needs to be taken into consideration. The presence of a large number of SKUs in an organization is termed as numerousness. Existing literature has shown how this leads to complexity creating challenges for organizations to meet optimal performance [33]. This situation reinforces the conflict of goals between the marketing, manufacturing, and logistics management functions of the organization. Marketing management aspires to allow high levels of flexibility for customers through its product configuration possibilities, while the logistics management aims to maintain low inventory levels at high product availability [21]. At the same time, manufacturing seeks to reduce both costs and time in the production process. Accomplishing all performance goals simultaneously is impossible and this presents a trade-off. The use of metrics informs the status at a particular time aggregated down to a single number to simplify decision-making [25]. The challenge associated with managing complexity is understanding both the relationship and impacts occurring due to system behavior over time [15, 34]. Simulation and optimization methods can support decision-making. However, existing methods aiming to manage complexity do not fully capture the dynamic impact of product variant complexity to logistics, manufacturing, and sales. A System Dynamics (SD) approach allows a scenario-based analysis to measure and manage complexity caused by product variety. This paper proposes a SD approach to help managers to understand the interrelation between necessary complexity levels for sales and its impact on logistics and manufacturing. The model optimizes the understanding of the systems behavior over time (BOT) and enables scenario building to provide guidance for decision makers. Existing literature focused on the impact of complexity and inventory management separately [28, 32]. Although there are case studies on reducing complexity through SKU tail management projects, etc. the literature focusing on integrating complexity and inventory management is scarce [4, 33]. This paper’s value consists of combining the impacts to support the decision process while considering the systems perspective. The paper is structured as follows. Section 2 reviews the relevant literature presenting complexity measurement approaches and metrics. Section 3 explains the methodology implementing SD. Followed by Sect. 4 demonstrating the simulation and analysis of the SD model. Section 5 and 6 closes with the results and the conclusion, respectively.
2 Literature Review This section reviews the relevant literature under three parts. The first part gives an overview of complexity measurement approaches in the related field of product variety. Part two reviews a contribution of complexity metrics, to indicate the current status of options available to measure complexity and its decision-making potential. The third and last part considers several SD approaches with the focus on complexity applied in different industries and their major findings.
Managing Complexity in Variant-Oriented Manufacturing
365
2.1 Complexity Measurement Approaches In this paper complexity is defined as the system characteristic that is dependent on the number of system elements, including the interrelations of elements, and its resulting number of possible system states [2, 6]. Complex systems interact with multiple actors and are influenced by time delays between cause and effect, which lead to feedback and dynamic system behavior [12, 36, 37]. In this context, dynamic complexity is characterized by the system behavior over time. To cope with necessary internal (company) and external (market) complexities, the product varieties build an interrelation between necessary market complexity and its consequence on internal complexity [18, 29]. The most recent research is reviewed from the portfolio complexity perspective. Related literature indicates different perspectives and is subdivided into the following areas: 1) managing complexity of manufacturing and assembly processes due to product variety, and 2) integral approaches in managing complexity. A great deal of research on complexity has focused on the field of manufacturing and its performance optimization. Results have shown that variety leads to an increasing complexity in assembly and supply processes, which impacts supply chain configuration and inventory control [1, 7, 11, 16, 22, 30]. For this reason, Hu et al. [16] aim for robust performance by proposing a unified measure of product variety-induced manufacturing complexity for assembly systems and supply chains. Reis [30] focuses on variety-based complexity. By using a case study, the findings present a negative correlation between a manufacturing system performance and its complexity. Managing complexity due to operational strategy, Frizelle [11] presents a mathematical approach to measure complexity in manufacturing. As a result, the method allows alternative courses of action leading to the development of an operational strategy. This review presents various approaches to measure complexity by discussing findings of interrelation and impacts of product variety focusing on manufacturing. However, the main focus of the reviewed literature is on static complexity. The problem of dynamic system behavior has been underreported and scarcely understood. Reviewing general variety-induced complexity methods, Abdelkafi [1] focuses on mass production. The developed framework combines the perspective of the company, supply chain and customers. The method proposes a comprehensive key metrics system leading to variety management strategies. Investigating the dimensions of external and internal complexity, Marti [22] provides guidelines of action on how to balance it. The model and its metrics aim to optimize the product’s architecture, leading to increasing customer benefits and causing less internal complexity. The model results are displayed in a complexity matrix. Providing variety leads to internal complexity and costs. Presented approaches [1, 22] focus on measuring complexity to deduce management strategies or balancing internal and external complexity. This enables managers to estimate the costs of introducing variety. Martin & Ishii [23] propose the concept of Design for Variety (DFV). The model holds three indices: commonality, differentiation point, and setup costs. Presenting information about handling external complexity by product variety and its impact to internal complexity represented by product lines is important. Using a multi-agent based approach, Becker et al. [7] represent saleable product variants based on market mechanisms which suit real customers’ requirements.
366
P. Kießner and H. N. Perera
Outlining the approaches, existing methods have the ability to deal with complexity. Nevertheless, the methods do not capture dynamic complexity over time. Also, the interrelation between marketing-, logistics- and manufacturing management dealing with product variety is partially investigated, leaving a gap in the contemporary understanding [1, 7]. 2.2 Overview Complexity Metrics Literature recommends using measures such as the number of variants and product models to evaluate complexity levels [5]. Strategies in variant management include considering modularity and (parts) commonality measurements of components and products. Covering metrics are presented in the design for the variety method. The commonality index is being used to investigate the percentage of standardized parts that are being reused for other product models [23]. The differentiation point index (DI) indicates differentiation occurring in the process. The lower the number, the better for the product structure. Identifying costs in the context of variety induced complexity, the setup cost index (SI) measures the percentage cost contribution of setups to the total costs for the product [23]. The measures reflect the problem dealing with variety induced complexity and its impact on companies, such as decreasing costs due to standardization and setup costs in the manufacturing environment. However, metrics do not capture the external complexity presented by the customers and its effects for the product portfolio. In order to offer variety, companies have to increase external complexity while simultaneously trying to decrease their internal complexity to be more efficient. The Portfolio-FitnessIndex (PFI) [31] represents the ratio of sold product variants to offered product variants (1), to allow a measurement of the needed external complexity. Voffered represents the offered variants over a time horizon to the market. VSold shows the number of sold variants over a time horizon. The aim of the metric is a low PFI value. PFI = 1 −
VSold Voffered
0 ≤ PFI ≤ 1
(1)
High PFI values indicate that companies should implement policies to decrease complexity. One strategy to lower the PFI could be to limit the product configuration possibilities. Using the PFI as a single metric, the time period of measurement of sold variants defines the result quality. For instance, seasonal variation is not taken into consideration. To cope with complexity in dynamic business environments, companies need appropriate tools. Reviewed metrics enables a comfortable application to measure complexity and its impacts. Despite this, a dynamic system behavior and its diversification over time are non-informative, through single metrics. 2.3 System Dynamics Approaches System Dynamics is a simulation and modelling approach to solve complex system problems with qualitative and quantitative methods combining feedback control theory [10, 35]. Table 1 outlines publications relating to SD with a focus on product variety and their findings in different industries. The reviewed literature presents several models and frameworks aiming to explain the systems behavior in optimizing business performance.
Managing Complexity in Variant-Oriented Manufacturing
367
Table 1. System dynamics literature focused on product variety and complexity Author(s)
Year
Source
Key findings
Reference
Zafer Do˘gruyol, Samet Güner
2021
Journal of the Faculty of Engineering and Architecture of Gazi University
Increasing product variety increases the design load and has a positive impact on productivity and quality
[14]
Moeen Smmak Jalali, 2020 Fatmei Ghomi, Massoud Rabbani
Journal of Industrial Engineering & Management Systems
Product variety required [17] to produce Make To Order & Make To Stock products
Mazieh Mehrjoo, Zbigniew J Pasek
2014
CIRP Conference on Manufacturing Systems
SD model showed a trade-off between cost and revenue when variety increases in the fashion industry
[24]
Daniel Kasperek, Sebastian Maisenbacher, Maik Maurer
2014
International DSM Conference
SD framework to analyze the dynamic behavior of engineering design processes
[19]
Joakim Stock
2009
Doctoral Thesis in Production Engineering
SD model to assess the influence of product variety on performance in steel production
[38]
Adolfo Cresp Marquez, Carol Blanchar
2008
International Journal of proposes a SD Production Economics simulation model to assess and compare portfolios of contracts with suppliers
[8]
Analyzing the literature in detail reveals that dealing with variety-induced complexity through the product portfolio and its dynamic interrelation to marketing-, logisticsand product management is not adequately investigated. Existing complexity management focuses on single metrics offering a comfortable application and potential for quick decision-making [22, 23, 31]. Regarding the dynamic system behavior and its interrelation, SD is a capable method to improve the decision-making process of managing complexity.
368
P. Kießner and H. N. Perera
3 Methodology Designed to identify cause-effect dependencies of systems and connected complexity management, SD is a popular and validated simulation method for decision-making [13]. It captures essential characteristics of real-world management containing nonlinear behavior, delay, and feedback [26, 35, 39, 40]. Creating a SD model is based on a five-step modelling development process [3, 13, 35]. Figure 1 displays the design process.
Identifying the problem
Conceptualization (causal loop diagram)
Formalization (stock and flow diagram)
Analysis of the model (Validation)
Evaluation of the model (Selection of policies)
Fig. 1. System dynamics design process [35]
The first step of SD modelling is to identify the problem. Managing complexity in terms of offering product variety is a challenge for companies due to nonlinear system behavior and its impact on internal complexity. Finding interdependencies of external and internal complexity presented by the product portfolio enhances the understanding of system behavior. Understanding the problem statement allows one to define system boundaries and model construction. The model is derived from literature and practice to ascertain the system’s elements and boundaries [24, 35, 38]. The main problem for the company in the case study is the spread of the demand linked to product variants, with the consequence of an expanded portfolio. One business goal is to offer a wide product portfolio (flexibility) at a lower complexity level. Flexibility is defined in the paper as the ability to react to the customer’s needs within a short lead time with a wide product portfolio. Finding the right ratio between flexibility and complexity is the main challenge for the company. In the case study, two questions were investigated: 1. How can the dynamic complexity level be displayed? 2. What is a suitable complexity level when considering flexibility and complexity? After the problem is identified, a causal loop diagram is needed. Examining the relationship between the variables and the developments of feedback and delays is the vital activity in the conceptualization phase. In SD, the feedback structure is represented by causal loop diagrams (CLD) [35]. These feedback loops present the system structure that captures the dynamic system behavior. The qualitative perspective of CLDs generates information relating to system behavior by applying the syntax (R: reinforcing loop, B: balancing loop) [20, 35]. Reinforcing loops suggest that two interrelated activities are interlaced and grow in tandem over time while balancing loops indicate that two interrelated activities balance each other out over time [35]. The variables were discussed and approved in a workshop with team members of the logistics, manufacturing, and sales
Managing Complexity in Variant-Oriented Manufacturing
369
departments of the case company. CLD indicates that the product variety influences the demand, inventory and production capacity. In particular, the demand indicates a balancing and reinforcing impact on the system behavior controlled by the product variety. For this reason, the PFI metric measures the system’s complexity level. For logistics and marketing, the key metrics were represented by inventory and sales since a limitation of the product portfolio affects both inventory and sales at the same time. Thus, the key metrics represent a superior view on the impact of complexity for logistics and sales. After the CLD is designed, the stock and flow diagram (SFD) is needed to investigate the model’s performance (Fig. 2). The “stock” represents the state of the “flow”, and the latter is changed by decisions [3]. Equations simulating the system’s behavior, are needed to consider interactions between levels, rates, auxiliaries, and constants [24, 35, 38]. In the fourth step, the validation of the model is executed by different tests (structure and behavior). The last step allows analyzing the impact of variables on the system behavior by various scenarios.
4 Simulation and Analysis of the Model The practical application of the method is demonstrated by a numerical example from a national electronics company. The medium-sized company is located in Germany with a distribution network spanning across the globe. In the case study, a product family of electrical switches consisting of 17 variants was analyzed. The SFD is simulated using the Vensim (version 8.0.9) software (Fig. 2). The settings of the model are represented by the time boundaries of the model by one year and its time unit one week. The external complexity in the model is represented by consumer demand due to different product variants. Consumer’s wish to have a large variety of products leads companies to react to the external complexity. The manufacturing process represents the internal complexity trying to respond to external complexity. This generates the system’s behavior aiming to offer flexibility to customers by providing a variety of SKUs to satisfy demand. To understand the impact of variant induced complexity to logistics and marketing, inventory and sales were identified as the key metrics. Reviewed literature indicates that balancing internal and external complexity provides an efficient way to manage complexity. The PFI metric represents the external complexity level in terms of providing variety leading to internal complexity [7, 22, 23, 30]. A basic scenario demonstrates the status quo of the system. In contrast to other approaches, the PFI will not decrease to a specific value based on expert interviews. In fact, in a second scenario the portfolio is reduced through a limitation of the offered product variants and its demand. This process aims to analyze the impact of the portfolio limitation to the complexity level represented by the PFI and its effects to the key measurements. To decrease the SKU portfolio, the product variants have been analyzed and sorted by its selling times in an ABC-analysis aiming to eliminate low performing product variants. The advantage of this proceeding is that the complexity level will be decreased based on portfolio performance without trying to gain a PFI value defined by experts. For the basic scenario, 21 variants were analyzed. After the ABC-analysis, four variants were identified as low performing variants and were eliminated from the portfolio. In the limitation scenario 17 variants persist, which need to be investigated for their impacts on
370
P. Kießner and H. N. Perera Difference inventory level Numbe of sold variants Number of variants
Range of coverage ZK in days
Stock value ZK
B
Delivery time
Range of coverage ZK
Mean price ZK
Order
Target stock level ZK
PFI Incoming goods rate ZK
Inventory ZK
Outgoing goods rate ZK
Safety stock ZK
Price FG
Sales
Safety stock in days
Delivery relaibility
ATP Backlog
Required semi-finished parts
Inventory level
Mean price FG R Stock value FG
Demand Safety stock FG
Incoming goods rate
Inventory Finish Goods (FG)
Outgoing goods rate Target stock level FG
R
Inventory deviation
Capacity utilization Manufacturing efficiency
Range of coverage FG
B
Order quantity rate
Production quantity
BOM coefficent Capacity
Finish good rate
Fig. 2. Stock and flow diagram of the manufacturing process
the complexity level, inventory, and sales metrics. The demand of the different product variants and scenarios is loaded into the SD model by the variable “demand” and delivers additional dynamic behavior into the system. The model validation is built as the model passes model structure and behavior tests. Tests of the model’s structure are performed as follows: stock and flow maps and its equations are analyzed by direct inspection and compared with the knowledge of the structure of a production process. Dimensional consistency is checked by Vensim’s unit check feature. Tests for the model behavior are performed to test behavior anomalies and extreme policies. In case of an abnormal behavior of the model, its structure has been examined to identify errors and correct them. Several policies of extreme conditions were modeled to represent dynamic consequences. The model was revised appropriately in the presence of abnormal behavior [27]. The model’s accuracy has been verified after comparing simulated and actual data at the inventory level. This result was validated by the experts of the company.
5 Results After the formulation process and the simulation, key metrics for logistics and sales are displayed in Fig. 3. Surprisingly, the analysis shows that the reduction of the portfolio through the ABC-analysis influences the complexity level marginally. At the same time, it is observed that a ratio between the minimal and maximal values (0–0,9) of the PFI exist. This clearly shows that an evaluation of the complexity level based on a single value without time response and expert opinions is not meaningful. This indicates the
Managing Complexity in Variant-Oriented Manufacturing
371
importance of consideration of the impacts of complexity regarding time. It can also be seen that a high complexity level is not necessarily negative for the company if high values do not persist over long-term.
Fig. 3. Overview of key metrics over time
The impact of portfolio reduction on logistics and sales is reflected through a decreasing inventory level and lower sales (Table 2). The greatest difference between the basic and the limitation scenario is seen in the inventory level in week 49 (Fig. 3). The scenarios diverge at the end of the reviewed period. This effect is due to the change in demand. While demand is higher in the basic scenario, the inventory level is increased due to the ordering policy. In the limitation scenario, the demand and the ordered quantity are lower, with the consequence that in successive periods with high demand inventory is subject to the risk of running out and only increasing again in the following periods (cf. week 49 and week 51). Table 2. Average values for key metrics Metrics
Unit
Basic scenario
Limitation scenario
Difference
Inventory level
Piece
2302
2082
220
Sales
Euro
980
832
148
PFI
Dml
0,528
0,508
0,02
The evaluation demonstrates that the complexity level in the basic scenario is sufficient for the analyzed product family, so that (1) a high degree of flexibility is ensured
372
P. Kießner and H. N. Perera
and (2) the effects of complexity on logistics and sales are minimal. In summary, a reduction in the portfolio has a positive effect on the complexity level, but the effects on inventory and sales are minor. The model’s result indicates an average complexity level for the product family between 0.528 and 0.508.
6 Conclusion This paper presents a System Dynamics model that investigates long-term system behavior of managing complexity generated through variety (numerousness) of products through a case study of a German electronics manufacturer. The portfolio-fitness index is introduced and implemented as a dynamic metric to measure the complexity. Applying the PFI to a SD system offers the ability to measure and control complexity by understanding the impacts of dealing with complexity. The model is simulated using data from the case company applying several scenarios. A limitation scenario to decrease the portfolio investigates the impact on the system behavior and the trade-off between marketing and logistics- and product management functions. Comparing the simulation results, the limitation scenario indicates positive effects in logistics and the complexity level compared to sales. The major finding in the case study shows that identifying an optimal ratio between the complexity level, logistics and sales depends on the dynamic system behavior and the impact of complexity strategies (i.e., reducing the product variety). Findings also indicate that a continuous decrease in complexity strategy does not lead to a better ratio between complexity and flexibility. The paper has twofold implications: (1) The SD model is proposed as a decision support tool to analyze long-term impacts of complexity strategies for manufacturing companies. Through the proposed model, it is possible to foresee the impact of a portfolio reduction on the inventory level, sales and the dynamic complexity levels measured by the PFI. (2) The simulation model can help to understand system behavior through scenario analysis. The simulation results offer support for decision makers in the most efficient manner. The paper’s originality is presented through the results of the simulation model. Results highlight that quantitative measurements of complexity over time support the decision-making process compared to the current handling through expert opinions. The work is unique as it identifies the impact of continuous decrease in complexity strategy and its effects on the company’s levels of complexity and flexibility. The research has some limitations, such as the average time for delivery based on a single sourcing strategy. Additionally, the model does not include replacement products, which may generate additional inventory. Effects such as product cannibalization during the process of reducing SKUs were also out of scope. Future research can be directed to examine aspect of dealing with increasing complexity due to new product variants in the portfolio. Effects such as product cannibalization in the presence of multiple SKUs within the same product family, as well as how product rationing/reduction can impact cannibalization would be areas worthy of future research. Findings of the research can lead to a rethink about complexity strategies with the goal to find relations between high flexibility and the impact on logistics and sales to generate competitive advantages. Systems thinking would allow manufacturing firms to manage their strategies effectively to find an ideal balance between competing interests of business functions within the firm.
Managing Complexity in Variant-Oriented Manufacturing
373
References 1. Abdelkafi, N.: Variety Induced Complexity In Mass Customization. Concepts and Management. Operations and Technology Management, vol. 7. Erich Schmidt, Berlin (2008) 2. Ashby, W.R.: An Introduction to Cybernetics, 1 edn. Wiley, New York (1956). https://doi. org/10.5962/bhl.title.5851 3. Bala, B.K., Arshad, F.M., Noh, K.M.: System Dynamics. Springer, Singapore (2017). https:// doi.org/10.1007/978-981-10-2045-2 4. Balachandra, K., Perera, H.N., Thibbotuwawa, A.: Human factor in forecasting and behavioral inventory decisions: a system dynamics perspective. In: Freitag, M., Haasis, H.-D., Kotzab, H., Pannek, J. (eds.) LDIC 2020. LNL, pp. 516–526. Springer, Cham (2020). https://doi.org/ 10.1007/978-3-030-44783-0_48 5. Blecker, T., Abdelkafi, N.: Using suitable key metrics for the management of complexity in variety-rich environments. In: Himpel, F., Kaluza, B., Wittmann, J., (eds.) Spektrum des Produktions- und Innovationsmanagements, Bd 44. Gabler, Wiesbaden, pp. 97–106 (2008). https://doi.org/10.1007/978-3-8350-5583-4_8 6. Blecker, T., Abdelmalek, N.: Innovative Process Optimization Methods in Logistics. Emerging Trends, Concepts and Technologies. Operations and Technology Management, vol. 13. Schmidt, Berlin (2010) 7. Blecker, T., Abdelkafi, N., Kreutler, G., Friedrich, G.: Dynamic multi-agent based variety formation and steering in mass customization. In: Camp, O., Haferlach, T., Rastetter, J. (eds.) Enterprise Information Systems VI, pp. 116–126. Kluwer Academic, Dordrecht (2006). https://doi.org/10.1007/1-4020-3675-2 8. Crespo Marquez, A., Blanchar, C.: The procurement of strategic parts. Analysis of a portfolio of contracts with suppliers using a system dynamics simulation model. Int. J. Prod. Econ. 88(1), 29–49 (2008). https://doi.org/10.1016/S0925-5273(03)00177-4 9. Desai, P., Kekre, S., Radhakrishnan, S., Srinivasan, K.: Product differentiation and commonality in design: balancing revenue and cost drivers. Manag. Sci. 47(1), 37–51 (2001). https:// doi.org/10.1287/mnsc.47.1.37.10672 10. Feng, Y.: System dynamics modeling for supply chain information sharing. Phys. Procedia 25, 1463–1469 (2012). https://doi.org/10.1016/j.phpro.2012.03.263 11. Frizelle, G., Woodcock, E.: Measuring complexity as an aid to developing operational strategy. Int. J. Oper. Prod. Manag. 15, 26–39 (1994). https://doi.org/10.1108/01443579510083640 12. Groesser, S.N.: Co-evolution of legal and voluntary standards: development of energy efficiency in Swiss residential building codes. Technol. Forecast. Soc. Change 87, 1–16 (2014). https://doi.org/10.1016/j.techfore.2014.05.014 13. Grösser, S.N.: Complexity management and system dynamics thinking. In: Grösser, S.N., Reyes-Lecuona, A., Granholm, G. (eds.) Dynamics of Long-Life Assets. From Technology Adaptation to Upgrading the Business Model, vol. 10, pp. 69–92. Springer, Cham (2017). https://doi.org/10.1007/978-3-319-45438-2_5 14. Guner, S., Do˘gruyol, Z.: Ürün çe¸sitlili˘ginin tasarım, tedarik ve üretim süreçlerine etkisinin sistem dinami˘gi yakla¸sımı ile incelenmesi. Gazi Üniversitesi Mühendislik-Mimarlık Fakültesi Dergisi (2021). https://doi.org/10.17341/gazimmfd.731788 15. Harrison, J.R., Lin, Z., Carroll, G.R., Carley, K.M.: Simulation modeling in organizational and management research. AMR 32(4), 1229–1245 (2007). https://doi.org/10.5465/amr.2007. 26586485 16. Hu, S.J. Zhu, X., Wang, H., Koren, Y.: Product variety and manufacturing complexity in assembly systems and supply chains. CIRP Ann. 57(1), 45–48 (2008). https://doi.org/10. 1016/j.cirp.2008.03.138
374
P. Kießner and H. N. Perera
17. Jalali, M.S., Ghomi, S.M.T.F., Rabbani, M.: A system dynamics approach towards analysis of hybrid make-to-stock/make-to-order production systems, 143–163 (2020). https://doi.org/ 10.7232/iems.2020.19.1.143 18. Jiao, J., Zhang, L., Zhang, Y., Pokharel, S.: Association rule mining for product and process variety mapping. Int. J. Comput. Integr. Manuf. 21(1), 111–124 (2008). https://doi.org/10. 1080/09511920601182209 19. Kasperek, D., Maisenbacher, S., Maurer, M.: Structure-based compilation of system dynamics models for assessing engineering design process behavior. In: Marle, F., Jankovic, M., Maurer, M., Schmidt, DM., Lindemann, U., (eds.) Risk and change Management in Complex Systems: Proceedings of the 16th International DSM Conference, 2–4 July 2014, Paris, France, pp. 232– 242 (2014). https://doi.org/10.3139/9781569904923.023 20. Kunc, M.: System Dynamics. Palgrave Macmillan UK, London (2018). https://doi.org/10. 1057/978-1-349-95257-1 21. Lambert, D.M.: Supply Chain Management: Processes, Partnerships, Performance. Supply Chain Management Institute (2008) 22. Marti, M.: Complexity Management. Optimizing Product Architecture of Industrial Products. Deutscher Universitäts-Verlag, Wiesbaden (2007). https://doi.org/10.1007/978-3-83505435-6 23. Martin, M.V., Ishii, K.: Design for variety: a methodology for understanding the costs of product proliferation. In: ASME Design Engineering Technical Conference Proceedings (1996) 24. Mehrjoo, M., Pasek, Z.J.: Impact of product variety on supply chain in fast fashion apparel industry. Procedia CIRP 17, 296–301 (2014). https://doi.org/10.1016/j.procir.2014.01.082 25. Mintz, O., Bart, Y., Lenk, P., Reibstein, D.: Drowning in metrics: how managers select and trade-off metrics for making marketing budgetary decisions. SSRN J. (2019). https://doi.org/ 10.2139/ssrn.3502600 26. Morecroft, J.D.: Strategic Modelling and Business Dynamics. A Feedback Systems Approach. Wiley, Chichester (2007) 27. Olivares-Aguila, J., ElMaraghy, W.: System dynamics modelling for supply chain disruptions. Int. J. Prod. Res. 59(6), 1757–1775 (2021). https://doi.org/10.1080/00207543.2020.1725171 28. Perera, H.N., Fahimnia, B., Tokar, T.: Inventory and ordering decisions: a systematic review on research driven through behavioral experiments. Int. J. Oper. Prod. Manag. 40(7/8), 997–1039 (2020). https://doi.org/10.1108/IJOPM-05-2019-0339 29. Ramdas, K., Fisher, M., Ulrich, K.: Managing variety for assembled products: modeling component systems sharing. M&SOM 5(2), 142–156 (2003). https://doi.org/10.1287/msom. 5.2.142.16073 30. Reiss, M.: Komplexitätsmanagement als Grundlage wandlungsfähiger Produktionssysteme. Industrie Management 27(3), 77–81, (2011) 31. Rennekamp, M.: Methode zur Bewertung des Komplexitätsgrades von Unternehmen, 1. Aufl. Produktionssystematik, 8. Apprimus-Verl., Aachen (2013) 32. Santos, V., Sampaio, M., Alliprandini, D.H.: The impact of product variety on fill rate, inventory and sales performance in the consumer goods industry. J. Manuf. Technol. Manag. 31(7), 1481–1505 (2020). https://doi.org/10.1108/JMTM-06-2019-0213 33. Sheffi, Y.: The Power of Resilience. How the Best Companies Manage the Unexpected. The MIT Press, Cambridge (2015). https://doi.org/10.7551/mitpess/9780262029797.001.0001 34. Skaržauskien˙e, A.: Managing complexity: systems thinking as a catalyst of the organization performance. Meas. Bus. Excell. 14(4), 49–64 (2010). https://doi.org/10.1108/136830410110 93758 35. Sterman, J.: Business Dynamics. Systems Thinking and Modeling for a Complex World With. TMHE IE OVERRUNS. McGraw-Hill Education, Europe (2002)
Managing Complexity in Variant-Oriented Manufacturing
375
36. Sterman, J.D.: Learning in and about complex systems. Syst. Dyn. Rev. 10(2–3), 291–330 (1994). https://doi.org/10.1002/sdr.4260100214 37. Sterman, J.D.: All models are wrong: reflections on becoming a systems scientist. Syst. Dyn. Rev. 18(4), 501–531 (2002). https://doi.org/10.1002/sdr.261 38. Storck, J.: Strategic and operational capabilities in steel production: product variety and performance. KTH (2009) 39. Warren, K.: Strategic Management Dynamics. Wiley, Chichester (2008) 40. Jayalath, M.M., Perera, H.N.: Mapping post-harvest waste in perishable supply chains through system dynamics: a Sri Lankan case study. J. Agric. Sci. Sri Lanka 16(3), 526–543 (2021). https://doi.org/10.4038/jas.v16i03.9477
Dynamic Lot Size Optimization with Reinforcement Learning Thomas Voss(B) , Christopher Bode, and Jens Heger Leuphana University, Universit¨ atsallee 1, 21335 L¨ uneburg, Germany [email protected]
Abstract. Production planning and control has a great influence on the economic efficiency and logistical performance of a company. In this context, this article gives an insight into the use of simulation as a virtual model of a filling machine in the process industry. Furthermore, it shows the possibilities of a reinforcement learning (RL) approach for dynamic lot sizing. The contribution indicates a possible implementation in an ERP system and shows how a decision support tool can support the planner to save up to 5% of costs compared to a human planner and a heuristic approach proposed by Groff. Keywords: Simulation
1
· Reinforcement learning · Lot sizing
Introduction
For every manufacturing company, it is a challenge to find an optimal balance between flexibility and efficiency in changing markets. The high product variety and the requirements for the shortest possible delivery times pose new challenges for production planning and control (PPC). The choice of the production lot size marks a trade-off between low order placement costs in production and low inventory costs in the warehouse. Machine learning (ML) as a driver of artificial intelligence (AI) has made more and more applications for simplifying private life [21]. However, a study by the German Federal Ministry of Economics shows that in 2019, only 17 500 of the companies surveyed were using AI, which represented about 6% of all companies surveyed [3]. The reasons are manifold: besides lack of understanding, lack of confidence in the methods, lack of practical implementation concepts, the industry also lacks skilled workers in that area. Due to the discovered capabilities of ML in complex environments, the potential of lot sizing with the help of reinforcement learning (RL) is considered in this contribution. The contribution is based on a real world use case in the process industry and presents a possible exemplary implementation.
2
Challenges in Batch Size Planning
The trade-off in lot sizing is the positioning between low order placement costs for large lot sizes and low inventory costs for small lot sizes. Large lot sizes lead c The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 M. Freitag et al. (Eds.): LDIC 2022, LNLO, pp. 376–385, 2022. https://doi.org/10.1007/978-3-031-05359-7_30
Dynamic Lot Size Optimization
377
to higher safety stock levels and result in higher inventory costs, therefore, meeting the customer demand without delivery delay. Besides the environmentally induced fluctuation in demand for different products, the increase in product variants due to more individual customer needs results in smaller batch sizes being manufactured on multi-purpose lines, which causes changeover costs. The following hybrid manufacturing system, which produces discrete products from liquid raw material, indicates why setup costs are a decisive factor. The manufacturing flow (see Fig. 1) is simplified as follows: (1) mixing and buffering intermediate products, (2) filling the bottles, and (3) stacking, palletizing, and storing discrete products. In this use case, the technical equipment must be cleaned of the preliminary product before being able to process the next intermediate product. Additionally, changing to different bottle sizes and different labels cause additional changeover effort, which leads to additional setup costs. From interviews, it was concluded that the filling machine for process step (2) is the hold-up resource. For that reason, it is critical to determine the correct quantity of the specific raw material to be filled (lot sizes), and using the correct bottle sizes, with the appropriate labeling (product).
Fig. 1. The three-step production process
3
Methodology for the Calculation of a Lot Size
Based on the uncertainty of customer behavior, it is important to break down static structures in order to be able to react flexibly. Since lot sizing can be viewed as a dynamic, multicriteria decision-making process, flexibility is critical [2]. Technological innovations and increasing digitalization in particular offer new opportunities in this regard. In this scenario, the use of real-time data with high quality and smooth integration can have a positive impact on the decisionmaking and control process of PPC [14]. There are already multiple approaches in the literature to calculate the lot size, ranging from static models calculating the optimal lot sizes with mathematical equations to simulation-based optimization [11] and ML methods [6,16]. It should be pointed out that multiple dynamic approaches have been presented [6,7,10,17], still, none of the aforementioned approaches utilize the advantages provided by the usage of reinforcement learning. Additionally, it should be noted that known approach such as Andlers-Lot Size, being deterministic models, are not fully suitable for the process industry,
378
T. Voss et al.
which imposes particular challenges or restrictions. In the case of simple deterministic models, the lack of capacity constraints, the lack of sequence-dependent setup costs or, the premise of static demand are deficiencies [12]. For that reason, the one-stage single product model developed by Wagner and Whitin [8] to calculate dynamic lot sizes is considered in this contribution in addition to the human planner and the RL-agent presented later. The model can be solved using exact as well as heuristic methods, given that the optimal solution requires tremendous amounts of computational power. The proposed heuristic approach can be considered as a sequential method, therefore it can be applied in a static and a dynamic way. Generally, the lot size for each time interval is raised until a threshold is reached and the calculations are terminated. This detail is utilized by Groff, assuming that the cost minimum for the lot sizes is given once setup and storage costs are equal [5]. To be more precise, the requested amount of products is added to the lot size until the storage cost is larger than the setup cost and the restriction is invalid [4]. After all, given the single product model and the missing capacity constraints, the resulting lot sizes and production programs are rarely feasible and therefore cannot be transferred to a real production process. Last, dynamic approaches utilizing ML methods are applicable to all PPC tasks that involve predictions [13]. Experiments with RL for the dynamic adjustment of lot sizes have already shown promising results in smaller scenarios [19,20]. Still, the use of ML-based methods lacks practical implementation concepts. In this contribution, the two approaches will be compared to a human planner baseline, and advantages, as well as drawbacks, will be evaluated. In addition to the example of RL for the dynamic selection and adjustment of lot sizes, the implementation of the same in a real system is also presented in this paper. The development of a method for the dynamic selection of lot sizes with reinforcement learning in the process industry considering different performance indicators and the above-mentioned deficits in the context of the process industry is aimed at. The results of previous contributions are to be confirmed in the context of the application example with a real use case of 35 products.
4
Dynamic Adaptation with Reinforcement Learning
Reinforcement learning stands alongside supervised and unsupervised learning in the classification. Based on interaction with its environment and observations of change, an agent independently learns action strategies to maximize a received reward. An agent can choose its actions depending on the situation, which is highly dependent on the current state of the system. Depending on the action performed, the system state changes and the agent receives a reward and penalty for its interaction. Problems solved with RL are characterized by the lack of analytical solutions, but with the knowledge of an environmental model. The very high number of possible states and the lack of knowledge about the rewards of certain state-action pairs justify the approach that the best possible action
Dynamic Lot Size Optimization
379
is approximated probability-based with neural networks (NN). With NNs in the RL approach, reasonable estimates can be made for unknown situations even in complex problems [18]. In this use case, the different elements of the agent are linked to the lot size calculation. The observations that describe the state of the system can be, for example, the existing stocks or open orders. The reward describes a variable that defines the goal of the training (e.g., reducing inventory costs). The action space describes all possible actions that are possible given the observations (e.g. which product is filled in which lot size). The action trigger describes when exactly an action is performed (for example, when the filling machine requires a new order).
5
Scenario and Realization
In the application considered in this contribution, the manufacturing process is divided into three process steps (see Fig. 1). The mixing of the bulk products in seven different agitated tanks with five storage tanks as buffer storage is the first step. Connected to this via several product lines is the filling machine. The product is then portioned onto appropriate pallets according to the defined order size. After palletizing, the products are subjected to quality control. Based on three common bulk products, the filled products can contain the same liquid but different filling quantities, called format in the following. In addition, different labels are possible for products with the same liquid (bulk) and the same filling quantity (format). Between two bulk products, the time-consuming cleaning of the lines in the filling machine is necessary. A format change, as well as a label change, is penalized with an additional setup time. In the simulation, a mix of 35 different products from 3 bulk goods with 2 formats are considered as a representative subset for the product portfolio. Table 1 below describes the setup times for bulk, format, and label changeover. The values have been tracked for the use case at the partner’s plant. Table 1. Duration of setup based on task Task
Duration [h]
Change of label
1
Cleaning between two different bulks 5 and 8 Change of Format
10
Figure 2 shows the structure of the approach. Based on the historical data and the parameters of the production system, a simulation model is created that can be used to train the agent. For this purpose, the model is exported and imported into the software-as-a-service platform of Pathmind [1] which handles the training in this contribution. Pathmind uses state-of-the-art training strategies such
380
T. Voss et al.
as proximal policy optimization [15] in combination with population-based training [9], in order to provide the best agent. The action strategy created during the training can be deployed offline for evaluation as well as online in the production environment. The offline use in combination with the simulation model can be used to validate and verify the action strategy. If the actions of the strategy are plausible and perform as desired in the evaluation, it is made available online for decision support (in this case SAP) and is accessible via a REST API.
Fig. 2. All elements needed for the implementation [19]
6
Observation and Action Space
The existing planner makes decisions based on the specifications of the Material Resource Planning (MRP) run and organizational rules. The primary goal of the staff is to have very few setups by adhering to a set of rules for format and bulk change. If possible, a format should be run on a filling machine for a certain number of weeks, e.g. format A for 4 weeks and format B for 3 weeks. The format rule is adjusted approximately once a year if there are no major changes in demand. For certain fill levels of tanks, there are additional restrictions regarding the change of bulk products. Except for the organizational rules, the agent has all the observations that the human and the Enterprise Resource Planning (ERP) system also has (see Table 2). During the process analysis, it was found that certain observations were considered by the planner for adjusting the lot sizes, which the regular MRP run does not consider. These were also provided to the agent. Provided these observations, the agent then makes its decision. In this scenario, the agent could always make a new decision when the filling system had finished the current job. Thus, the time intervals between actions were flexible. The invalid bulk switches, when a tank has not yet been emptied, are realized at the agent by action constraints, called action masking. For better understanding, the observations, the actions and the evaluation of the change in the system are explained in more detail below. As mentioned above, the approach developed by Groff only considers the requested amount of products, neglecting all other observations. The different observations have been discussed with process experts, related to the use case.
Dynamic Lot Size Optimization
381
Table 2. observations available to the different planners Observation
Human Groff RL-agent
Requested amount of products
x
Stocked amount of products
x
Reach of stock
x
x x
x
x
Amount of products being work in process x
x
Last product manufactured
x
x
Last bulk manufactured
x
x
Last lot size
x
x
Capacity of the bottleneck
x
x
Amount of bluk product available
x
x
Possible next bulk
x
x
Handmade rules for changing the format
x
Given the RL approach, due to coding restrictions, a combination of dynamically choosing product and lot size was not doable. For that, a fixed lot size and a setup time of zero for the same format and bulk between orders of the product campaigns have been considered, so the lot size can thus be implicitly decided when the agent selects only the same product. Table 3 shows a comparison of the action spaces. The human can specify larger values for the lot size and thus have to make fewer decisions. The human can also choose the unscheduled time as desired. The calculation of the reward is the sum of defined performance indicators (see Table 4) whose progression is observed over the simulation. The individual factors can be weighted for training to focus on certain performance indicators and to punish certain behaviors. For the best reward function in this case, over 40 different reward functions were tested. To calculate the reward, the indicators are compared before and after the action. Table 3. Available actions to the planners Decision Human
Groff
RL-agent
Product One out of 35 different products Single product or undefined slack time at a time
One out of 35 different products or four hours slack time
Lot size Chosen freely from five up to maximum tank size
Fixed lot size being 1/5 of the tank
Choose freely without restrictions
382
T. Voss et al. Table 4. Aspects of the reward function available
Elements
Description
Referenceframe
Cost
Cost for work in process Cost for products in stock Setup cost Transportation cost Cost for keeping the machine empty
Per Per Per Per Per
Revenue
Revenue per Pallet
Additional rewards Reward for keeping the product sequence Reward for keeping the bulk sequence Reward for keeping the format sequence
Hour Hour Setup Pallet Hour
Per pallet Per Setup Per Setup Per Setup
Penalty
For not delivered products being requested Once in 24 h
Other
Service level in %
Every 15 min
Within one simulation run, the agent could test 1 500 actions and thus collect state-action pairs with a corresponding score. During the 1.5-h training, 250 iterations with 8 simulation runs each were performed per reward function, resulting in a total number of 3 million state-action pairs with corresponding rewards. This database forms the basis for the agent’s knowledge.
7
Evaluation of the Different Policies
In the following, the impact of batch size on the performance of the system based on a historical program, the approaches proposed by Groff and the newly developed RL approach are compared. The sales and the actual production plan of the last seven months is considered as the reference value for the new strategies. The amount of products requested is available for the simulation model on a weekly basis. For comparison purposes, the historical production program was simulated over a defined period of time and the resulting performance was standardized to 1 as a benchmark. Groff and RL can thus be compared in performance with percentage change against the existing metrics (see Table 5). For the performance indicators storage, setup, transport and, work in process cost, lower values are better. For the revenue as well as the service level, larger values indicate better performance. The historic program, developed by a human planner, provides predefined lot sizes that are used to meet demand. Analyzing the historical production schedule, the ordered demand was fulfilled by a factor of 1.4, reaching a service level for “on time delivery” over 90%. As previously mentioned, in some cases, the lot sizes are manually adjusted by the human planner in order to comply with organizational regulations. The comparison below will show, that the strategy to produce one format and one bulk product for as long as possible will result in a high amount of stock.
Dynamic Lot Size Optimization
383
Table 5. Comparison of selected performance indicators the different policies Performance
Human Groff RL-agent
Storage cost
1
0.61
0.31
Setup cost
1
1.42
1.36
Transportation cost 1
0.69
0.68
WIP cost
1
1
1
Service level
1
1.2
0.78
Revenue
1
1.01 0.99
The comparison shows that regardless of the neglected capacity constraints, the storage cost is reduced to 2/3 compared to the human planer, using the Groff heuristic. The setup cost is increase compared to the human planer. After all, the service level is increased since all customer orders are fulfilled, given the nature of the proposed approach. To conclude, as the revenue stays the same and the Groff heuristic is able to decrease stock at the cost of more setup, there is a slight increase in reward and therefore profit by around 4%. The same observations hold for the RL-agent. The agent learns to satisfy the demand in terms of quantity by determining the sequence and batch size of production orders. A direct comparison of the considered performance indicators shows that the agent generates a comparable profit to the historical program. Furthermore, it shows that the agent significantly reduces inventory in the warehouse (to 1/3 of the original cost). Just like the Groff approach, the agent causes an increase in setup costs, in this case up to 36% more than the historical program. As a result of the low amount of products in inventory, the service level drops below the required 90% mark. With the reduction of inventory, fewer pallets were produced, which results in lower intra-plant transportation costs. In the end, it can be concluded that the agent receives around 5% more total reward. In monetary terms, this can save up to 5% of the total costs in the use case, based on storage, setup and, transportation costs. The potential could be validated with lower initial inventories, which then resulted in 4% savings. After all, this comparison shows, that both strategies are reducing the amount of products in stock and accepting the higher costs for setup operations to reduce overall cost. It can be concluded that the RL-agent did learn a feasible strategy. Additionally, and in contrast to the approach provided by Groff, due to its nature, the RL-approach is much more dynamic and can adjust to change without recalculating all lot sizes, providing the biggest advantage over the other approaches.
8
Implementation in the ERP System
It can be summarized that a dynamic lot sizing with reinforcement learning can achieve comparable results to the human planner. The simulation of hybrid
384
T. Voss et al.
production in the process industry under consideration of setup, process, and storage cost enables the agent to learn a profitable action strategy. Since Pathmind’s action strategy is available both in the cloud and locally as a microservice deployed via REST API, POST requests can be used to query actions in the real production environment. Thus, based on the description of the original planning procedure, several options for implementation are possible. All of them are based on the top half of Fig. 2, emphasizing the online aspects. The action strategy of reinforcement learning can be used at this point as a decision support system to provide the planner with assistance in dynamic lot size selection. For this purpose, the observations, as described above, are transmitted via the interface, which then outputs a suggestion for the current production order on the machine. The planner can then accept or reject this recommendation. This is the simplest solution, as it can be used in parallel independently of the existing system. Furthermore, this implementation makes it possible to get to know the technology and to build up trust in the system. The next step is then the implementation in the MRP run to obtain a fully automated solution. A detailed description of the implementation will be omitted here.
9
Outlook
The scaling of this use case to several production plants can be built up by a multi-agent scenario with common observation, action, and action space. An extended simulation model is mandatory for this. Furthermore, future action decisions are also conceivable in ’tuple’ decisions (product and lot size). Further investigations should consider alternative action spaces and constraints to improve logistic performance indicators. The confidence thus built in comparable AI support systems could create practical scenarios in which humans and machines work hand in hand.
References 1. pathmind.com 2. Azadnia, A.H., Saman, M.Z.M., Wong, K.Y.: Sustainable supplier selection and order lot-sizing: an integrated multi-objective decision-making process. Int. J. Prod. Res. 53(2), 383–408 (2015). https://doi.org/10.1080/00207543.2014.935827 3. Bundesministerium f¨ ur Wirtschaft: Einsatz von k¨ unstlicher intelligenz in der deutschen wirtschaft: Stand der ki-nutzung im jahr 2019 (2020) 4. Domschke, W., Scholl, A., Voß, S.: Produktionsplanung: Ablauforganisatorische Aspekte, 2, u ¨ berarb. und erw. aufl. edn. Springer-Lehrbuch. Springer, Berlin (1997) 5. Groff, G.K.: A lot-sizing rule for time-phased component demand. Prod. Inventory Manag. 1, 47–53 (1979) 6. Hachicha, W.: A simulation metamodelling based neural networks for lot-sizing problem in MTO sector. Int. J. Simul. Model. 10(4), 191–203 (2011). https://doi. org/10.2507/IJSIMM10(4)3.188 7. Hakeem-Ur-Rehman, Wan, G., Zhan, Y.: Multi-level, multi-stage lot-sizing and scheduling in the flexible flow shop with demand information updating. Int. Trans. Oper. Res. 28(4), 2191–2217 (2021). https://doi.org/10.1111/itor.12645
Dynamic Lot Size Optimization
385
8. Wagner, H.M., Whitin, T.M.: Dynamic version of the economic lot size model. Manag. Sci. 5(1), 89–96 (1958). http://www.jstor.org/stable/2626974 9. Jaderberg, M., et al.: Population based training of neural networks (2017) 10. Liu, R., Wang, S., Pi, Y., Qin, Q.: An effective heuristic with evolutionary algorithm for the coordinated capacitated dynamic lot-size and delivery problem. Comput. Ind. Eng. 153, 107051 (2021). https://doi.org/10.1016/j.cie.2020.107051 11. Maier, J.T., Voß, T., Heger, J., Schmidt, M.: Simulation based optimization of lot sizes for opposing logistic objectives. In: Ameri, F., Stecke, K.E., von Cieminski, G., Kiritsis, D. (eds.) APMS 2019. IAICT, vol. 567, pp. 171–179. Springer, Cham (2019). https://doi.org/10.1007/978-3-030-29996-5 20 12. M¨ unzberg, B.: Multikriterielle Losgr¨ oßenbildung: Zugl.: Hannover, Univ., Diss., 2013, Berichte aus dem IFA, vol. 2013,5. PZH-Verl., Garbsen (2013) 13. Schmidt, M., Maier, J.T., Grothkopp, M.: Eine bibliometrische analyse: Produktionsplanung und -steuerung und maschinelles lernen. wt Werkstattstechnik online 110(4), 220–225 (2020) 14. Schuh, G., Reuter, C., Prote, J.P., Brambring, F., Ays, J.: Increasing data integrity for improving decision making in production planning and control. CIRP Ann. 66(1), 425–428 (2017). https://doi.org/10.1016/j.cirp.2017.04.003 15. Schulman, J., Wolski, F., Dhariwal, P., Radford, A., Klimov, O.: Proximal policy optimization algorithms (2017) 16. S ¸ enyi˘ git, E., D¨ u˘ genci, M., Aydin, M.E., Zeydan, M.: Heuristic-based neural networks for stochastic dynamic lot sizing problem. Appl. Soft Comput. 13(3), 1332– 1339 (2013). https://doi.org/10.1016/j.asoc.2012.02.026 17. Slama, I., Ben-Ammar, O., Dolgui, A., Masmoudi, F.: Genetic algorithm and Monte Carlo simulation for a stochastic capacitated disassembly lot-sizing problem under random lead times. Comput. Ind. Eng. 159, 107468 (2021). https://doi.org/10. 1016/j.cie.2021.107468 18. Sutton, R.S., Barto, A.G.: Reinforcement Learning: An Introduction. MIT Press, Cambridge (2018) 19. Voß, T., Rokoss, A., Maier, J.T., Schmidt, M., Heger, J.: Outperformed by a computer? - comparing human decisions to reinforcement learning agents, assigning lot sizes in a learning factory. SSRN Electron. J. (2021). https://doi.org/10.2139/ ssrn.3859196 20. Wang, J., Li, X., Zhu, X.: Intelligent dynamic control of stochastic economic lot scheduling by agent-based reinforcement learning. Int. J. Prod. Res. 50(16), 4381– 4395 (2012). https://doi.org/10.1080/00207543.2011.592158 21. Welsch, A., Eitle, V., Buxmann, P.: Maschinelles lernen. HMD Praxis der Wirtschaftsinformatik 55(2), 366–382 (2018). https://doi.org/10.1365/s40702018-0404-z
Integrated Simulation-Based Optimization Approach for Production Scheduling: A Use Case Application in a Machining Process Ícaro Romolo Sousa Agostino1(B) , Mauricio Randolfo Flores da Silva1 Enzo Morosini Frazzon1 , and Luciana Amaral Stradioto Neto1,2
,
1 Industrial and Systems Engineering Department, Federal University of Santa Catarina,
Florianópolis, SC 88040-900, Brazil {icaro.agostino,mauricio.randolfo, luciana.stradioto}@posgrad.ufsc.br, [email protected] 2 Olsen, Palhoça, SC 88133-510, Brazil
Abstract. The scheduling of complex manufacturing systems requires the integration of production, inventory, and maintenance to obtain a robustness control and performance. In this context, this paper proposes a data-driven simulationbased optimization approach for production scheduling to optimize job shop sequencing decisions considering inventory availability and machine breakdowns. The approach consists in the implementation of a combination of genetic algorithm with discrete-event simulation model aiming to optimize the selection sets of dispatching rules for groups of machines in the job shop to improve the manufacturing system performance. The approach was evaluated in a machining process, part of a job system of a Brazilian industry of the dental and medical sector, achieving better performance for the average lead time with the improvement of 10.27% compared to the current strategy of the company. Keywords: Manufacturing systems · Optimization · Discrete-event simulation
1 Introduction The production line of manufacturing systems is composed of resources to transform raw material in final products, featuring a system subject to uncertainty, which implies a relevance in the definition of an adaptive and optimized production scheduling [1]. In order to define the optimal scheduling, the initial processing time, the priority of each job, which machine to process or when to perform maintenance in each machine [2] is essential to determine key performance indicators (KPIs) to control the production [3, 4]. However, organizing an optimized scheduling can be challenging to manufacturing systems, since it requires the use of heuristic methods [5]. To solve this problem, [6] suggests the use of dispatching rules associated with KPIs. Additionally, in order to deal with disruptions during the processing time, such as machine breakdown and raw material availability [3], rescheduling is highly efficient to © The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 M. Freitag et al. (Eds.): LDIC 2022, LNLO, pp. 386–395, 2022. https://doi.org/10.1007/978-3-031-05359-7_31
Integrated Simulation-Based Optimization Approach for Production Scheduling
387
quickly adapt the schedule and control the production system. In this context, another challenge related to production scheduling is inventory unavailability, characterized by [7] as a crucial disruption to the production. Although inventory availability has directly impact on production, many research study this problem separated from the production schedule [8, 9]. Moreover, [10] explains that the scheduling of complex manufacturing systems requires the integration of production, inventory, and maintenance in real-time to obtain an optimized scheduling. In the literature, simulation-based optimization method has been used in many research to deal with this problem [11, 12]. [13] describes the simulation-based optimization as a reactive approach triggered by disruptive events, like machine breakdowns or new jobs arrival, in which a set of dispatching rules is applied to each individual machine to optimize the KPIs of the system. Therefore, the method is efficient to dynamic scenarios, subject to disruptions and that demand real-time control. In this context, this paper proposes a data-driven simulation-based optimization approach for production scheduling, using simulation-based optimization to optimize job shop sequencing decisions considering disruptions, as inventory availability and machine breakdown. Therefore, the main contribution of this paper was to propose and evaluate in a real use case a simulation-based optimization method to optimally define the manufacturing system scheduling based on the best combination of dispatching rules, that considers not only the aspects of production control, but also dynamic machine failures and available inventory levels. The remainder of this paper is structured as follows: in Sect. 2 a related work is presented to support the research and presents the specific contribution of this paper. Section 3 describes the proposed approach in detail. In Sect. 4 the approach is evaluated by a test case of a job shop production environment. The paper closes with a conclusion and an outlook.
2 Related Work and Contribution In the literature, different methods have been developed to design adaptive and optimized production scheduling for manufacturing systems. [14] used a mathematical program approach to construct a dynamic scheduling of production, but the researchers concluded that a powerful heuristic is required to obtain more efficient results. [5] presented a framework combining mathematical model and heuristic approach to design and plan the schedule of supply chains, however the method is not tested in a real case, which makes the efficiency of the method uncertain. The production scheduling and maintenance planning is explored by Pan et al. [15] in a scenario considering a single machine to evaluate the use of sensors and prognostic technologies with the objective of minimizing the maximum tardiness, but without considering the complexity of a real manufacturing system. On one hand, [11] solve the optimization schedule with discrete event simulation, in which the production schedule is simulated using probability distributions to analyze perturbations and time uncertainties, contrasting the developed model with traditional models that use dispatching rules to optimize production schedule. On the other hand, [12] considered an adaptive simulation-based optimization (aSBO) based on dispatching rules to optimize the production schedule, however the method only considers
388
Í. R. Sousa Agostino et al.
machine breakdowns and demand update as disruptions, missing the impact of inventory availability in the system. In parallel, Huang and Süer developed a model with dispatching rules being selected by a genetic algorithm to simulate a job shop scheduling scenario, in which the main objective is minimize makespan, average flow time, and minimum and total tardiness [16]. This approach is improved by [10], since the authors adopted a genetic algorithm to perform the simulation-based optimization model in a manufacturing system aiming to determine the most suitable dispatching rules for each individual machine, from real-time monitoring of system conditions. Agostino et al. [13] studied an approach to automatically synchronize the planning and control of production according to the current state of the shop floor. The proposed approach utilizes Digital Twin to optimize the use of dispatching rules for each machine in the system every time a stochastic event is identified, allowing a constant updating in the process. [17] applied the simulation-based optimization method based on a genetic algorithm and obtained a significant reduction in the number of late orders in a scenario of integrated scheduling of production and transport.
3 Data-Driven Simulation-Based Optimization 3.1 Simulation-Based Optimization Approach The approach considered in this study is based on the method presented in Frazzon et al. [10] and the integration of inventory availability to the production control. Thus, our implementation consists in adopting a simulation-based optimization approach (SBO) to optimize the selection of dispatching rules for each machine in the job shop. Besides, our model consists in two parts, in which the first represents the job shop environment, including the real scenario characteristics, as available resources, products demand, dynamic order arrival, and the processing time and operations required for producing each item. In the second part, a genetic algorithm is implemented aiming to artificially evolve the results of the system from a restricted population, which means that each combination of dispatching rules to attend the system is considered as a viable solution and each iteration will evolve the results proposing new possible solutions for the production schedule [18]. Additionally, predictive maintenance is considered as a job with high priority, while the inventory level is considered as a decision parameter for the system with stochastic behavior, in which the production of an item only starts if there is enough raw material to finish the production of that order. The occurrence of inventory disruptions due the non-availability of raw material was defined based on the historical data of the use case analyzed. Regarding maintenance, we also considered original data from resources used on the simulated production line to include predictive maintenance in our model, while historical data of mean time between failures and mean time to repair were used to input machine breakdown probability distribution in the model. The output of our approach is an optimized set of most suitable dispatching rules for each resource according to the current status of the system, in which the set is mainly focused on optimizing the selected KPI. The Fig. 1 shows the complete proposed approach.
Integrated Simulation-Based Optimization Approach for Production Scheduling
389
Fig. 1. Proposed approach adapted from [10]
4 Use Case The use case considered in this paper was applied in a machining process part of a job system of a Brazilian industry of the dental and medical sector, named Olsen. The scenario consists of 23 products and 11 machine resources, in which each product has a specific trajectory in the production line, and some products skip resources in the production line. Hence, the number of operations per product is between two and nine. Regarding the resources, besides the machines used directly to process the jobs, we
390
Í. R. Sousa Agostino et al.
also included the resources from manual inspection and maintenance in the scenario. Additionally, the processing time are stochastics for each product. Setup times and time to recover after a failure are included in all machines considered in the system. Table presents the list of resources included in the scenario (Table 1). Table 1. Resources list Resources
Quantity
Resources
Computer numerical control lathe
2
Milling machine
Quantity 1
Ergomatic lathe
2
Inspection
1
Mechanical lathe
1
Operators
12
Bench drills
4
Electric maintenance
2
Band saw
1
Mechanic maintenance
2
The demand is monthly, in which we considered 18 months of production with a daily production time of 528 min. The demand is converted to jobs that represents the grouping of items from the same product to be processed. Thus, during the simulation time, we have about 820 jobs arriving in the system to produce 133,570 individual products. The monthly demand is presented in detail in Fig. 2. We considered that each order arrives in the system in the first day of each month and have until the end of the month to be delivered, this approach is similar to the real company scenario nowadays.
Fig. 2. Monthly historical demand
Moreover, the machine breakdowns and inventory are also included in our scenario. On one hand the model considers periodic maintenance for each machine in frequency that can be daily, weekly, monthly, quarterly, semimanual, or yearly, and corrective maintenance, implemented with probability distribution according to historical data from each resource. In average 2965 periodic maintenance and 50 failure events are executed in each simulation run for the 11 machines in the system over the experiment. On the other hand, inventory data is implemented considering the raw material required for products and the delivery frequency of each raw material. Thus, before starting the
Integrated Simulation-Based Optimization Approach for Production Scheduling
391
production of a job, the model checks the available amount of raw material and in case of nom availability the job waits for new arriving material to be produced. This behaviour was also modelled using probabilistic distributions according to historical data from the company. 4.1 Experiment Description The simulated scenario was implemented using Simmer: Discrete-Event Simulation package for R language [19]. The experiment was executed in a stochastic environment, which includes variations in production, inventory and maintenance processing based on the historical data provided by the company. The optimization engine used a genetic algorithm from the package GA for R language [20]. The optimization was defined with the following parameters based on preoptimization tests: A population of 30 solutions for each generation, with a mutation rate of 40% and crossover rate of 80%. As termination criteria was defined 2 rules: (i) after 30 consecutive generations without any improvement in the best fitness value; (ii) after the maximum of 100 generations. If one of this two criteria is reached the GA algorithm stops and the best solution is chosen. During the optimization, each individual solution is simulated 10 times to evaluate the quality of the solutions considering the stochastic aspects of the simulated environment. The approach was applied to find an optimized set of dispatching rules considering dynamic effects related to stochastic behaviors in production, the non-availability of raw material in the inventory, and machine failures. The optimal solution should deal in an integrated way with all these aspects to optimize the KPI’s of the production system. The final solution is composed by a set of dispatching rules for each group of machines, being able to dynamically react to the stochastics events over the production execution. For this work we selected a collection of six dispatching rules to apply in this use case experiment, based on the consulted literature [10, 21–23]: (i) Fist in first out (FIFO); (ii) Earliest due date (EDD); (iii) Shortest Processing Time (SPT); (iv) Slack remaining time (SLK); (v) Critical ratio (CR); (vi) and Slack per remaining operation (SL/OPN). At the end of the experiment, the performance of the simulation-based optimization (SBO) approach was evaluated by means of two main key performance indicators (KPI): (i) the total number of tardy jobs; and (ii) the mean lead time. For each KPI a separated optimization experiment was conducted over 18 months of operation based on the demand data provided by the company. To evaluate the quality of the proposed SBO approach the results was compared against five benchmark scenarios, that used a static set with the same dispatching rule for each group of machines and also, the current strategy of the company that uses a fixed production schedule based on FIFO approach for production execution. So, our optimization is for the entire period of 18 months, and can be readjusted when events of interest occur, such as unavailability of stocks, machinery breakdown or changes in production plans. The implemented SBO model and all the experiments was developed only using open sourcing technology. The entire implementation was conducted using R language 4.0.4 with the packages ‘GA’ [20] for optimization with genetic algorithm and ‘Simmer’ [19] for discrete event simulation. The code is free available for reproducibility purpose at https://github.com/icaroagostino/SBO_LDIC2022.
392
Í. R. Sousa Agostino et al.
The computer used to perform the experiments has the processor Intel i7-9750H, with 6-core 12-threads. The method was implemented and executed with multi-core parallelization using the features of the ‘GA’ package that provides this built-in execution feature to improve the method performance. 4.2 Results and Analysis The SBO approach proposed was tested focused on optimize the due date adherence in terms of number of tardy jobs and the average lead time of the production lines. This two KPIs are the most important parameters for the company that we considered in this research. However, for future applications and analysis it is possible to include different optimization KPIs due the flexibility of the proposed method. In the Fig. 3 we can see the comparison of the results for the mean lead time in minutes over the experiment considering 18 months of operations, and the evolution of the SBO approach over the generations.
Fig. 3. (i) Lead time results comparison; (ii) Lead time optimization evolution
The SBO approach proposed by this paper achieves a significantly improvement compared in the performance compared with the current strategy of the company (fixed schedule). The difference was 10.27% better than the current strategy and 37.98% better than the worst set of fixed (CR) rules for this scenario. The method was able to find the optimal solution after 84 generations using the stop criteria of 30 generations without improvements. In the same direction, Fig. 4 shows the comparison of the results for the total number of tardy jobs over the experiment considering 18 months of operations.
Fig. 4. (i) Tardy jobs results comparison; (ii) Tardy jobs optimization evolution
Integrated Simulation-Based Optimization Approach for Production Scheduling
393
For the total number of tardy jobs, the improvement of the SBO approach proposed over the other approaches was less significant. However, the method was capable to overcome the current strategy improving the performance reducing the number of tardy jobs in 3.17%. Compared with the less efficient approach for this KPI (SL), the proposed method was 3.94% better. The summary of all the results is showed in the Table 2 with all the improvements compared with the current status of the company (fixed schedule) achieved in the experiments conducted. Table 2. Performance comparison Mean lead time (minutes)/improvement
Number of tardy jobs/improvement
SBO (optimized)
1559.76
(10.27%)
SBO (optimized)
Fixed schedule
1738.21
–
CR
125
(0.79%)
SL
1927.56
(−10.89%)
SL/OPN
125
(0.79%)
EDD
1946.53
(−11.98%)
EDD
126
(0.00%)
122
(3.17%)
SPT
1994.53
(−14.75%)
Fixed schedule
126
–
SL/OPN
2105.77
(−21.15%)
SPT
126
(0.00%)
CR
2152.15
(−23.81%)
SL
127
(−0.79%)
The SBO approach proposed in this research was able to deal with stochastic events such as non-availability of raw material, and multiple machine breakdowns. Being able to find in time manager better strategies to schedule the production over time. The simulation-based optimization run achieved the best solution in 84 generations for the lead time KPI, totaling 25200 rounds of simulations (10 replications, 30 individuals in the solution population in 84 generations) to improve the performance of the manufacturing system. The total computation time to determine the final solution for the experiment was approximately 1.9 h, with each iteration requiring about 1.35 min using the parallelization execution of the method.
5 Conclusion and Outlook The main contribution of this paper was to propose and evaluate the method to optimize a highly dynamic production system considering variations in production times and demand, combined with stochastic availability of raw material and machine failures. In the best of our knowledge the similar approaches founded in the literature usually considered only production aspects when apply optimization techniques to the scheduling process, and in some cases combine with a secondary aspect as inventory management [7] or maintenance planning. This paper expanded the work of [10] and [13] by considering a solution that integrates information from production, maintenance, and inventory at the same time to find an optimized solution for production schedule and control. Besides, more complex simulation models, more complex prioritization policies, variations in GA parameters
394
Í. R. Sousa Agostino et al.
and changes to scenario settings, as the production rescheduling policy, may be tested to evaluate the robustness of the model in future research. Furthermore, we intend to expand our method in the future to be able to dynamically interact with the scheduling overtime considering the digital twin concept of modeling. Acknowledgement. This study was financed in part by the Coordenação de Aperfeiçoamento de Pessoal de Nível Superior - Brasil (CAPES) - Finance Code 001 and by CAPES under reference number 99999.006033/2015-06, in the scope of the BRAGECRIM program. Also, we would like to thank the Olsen managers for providing the data for the use case of this study.
References ´ c, A.: Predictive scheduling as a part of intelligent job scheduling 1. Sobaszek, Ł, Gola, A., Swi´ system. Adv. Intell. Syst. Comput. 637, 358–367 (2018) 2. Valledor, P., Gomez, A., Priore, P., Puente, J.: Solving multi-objective rescheduling problems in dynamic permutation flow shop environment with disruptions. Int. J. Prod. Res. 56, 6363– 6377 (2018) 3. Schuh, G., Reuter, C., Prote, J.P., Brambring, F., Ays, J.: Increasing data integrity for improving decision making in production planning and control. CIRP Ann. Manuf. Technol. 66, 425–428 (2017) 4. Nyhuis, P., Mayer, J.: Modelling the influence of setup optimized sequencing on lateness and productivity behaviour of workstations. CIRP Ann. Manuf. Technol. 66, 421–424 (2017) 5. Israel, E.F., Frazzon, E.M., Cordes, A.K., Hellingrath, B., Lopes, A.A.: Operational supply chain planning method for integrating spare parts supply chains and intelligent maintenance systems. IFAC-PapersOnLine 50(1), 12428–12433 (2017) 6. Kuck, M., Eike, B., Freitag, M., Hildebrandt, T., Frazzon, E.M.: Towards adaptive simulation based optimization to select individual. In: Winter Simulation Conference (WSC), pp. 3852– 3863. WSC, Las Vegas (2017) 7. Takeda-Berger, S.L., Zanella, R.M., Frazzon, E.M.: Towards a data-driven predictive-reactive production scheduling approach based on inventory availability. IFAC-PapersOnLine 52, 1343–1348 (2019) 8. Mohammaditabar, D., Hassan, G.S., Obrien, C.: Inventory control system design by integrating inventory classification and policy selection. Int. J. Prod. Econ. 140, 655–659 (2012) 9. Muckstadt, J.A.: Analysis and Algorithms for Service Parts Supply Chains. Springer, New York (2005). https://doi.org/10.1007/b138879 10. Frazzon, E.M., Kück, M., Freitag, M.: Data-driven production control for complex and dynamic manufacturing systems. CIRP Ann. 67, 515–518 (2018) 11. Vieira, G.E., Frazzon, E.M.: Searching for production robustness through simulation-based scheduling optimization. In: Freitag, M., Haasis, H.D., Kotzab, H., Pannek, J. (eds.) LDIC 2020. LNL, pp. 351–362. Springer, Cham (2020). https://doi.org/10.1007/978-3-030-447830_34 12. Pimentel, R., Santos, P.P.P., Carreirão Danielli, A.M., Frazzon, E.M., Pires, M.C.: Towards an adaptive simulation-based optimization framework for the production scheduling of digital industries. In: Freitag, M., Kotzab, H., Pannek, J. (eds.) LDIC 2018. LNL, pp. 257–263. Springer, Cham (2018). https://doi.org/10.1007/978-3-319-74225-0_35
Integrated Simulation-Based Optimization Approach for Production Scheduling
395
13. Agostino, Í.R.S., Broda, E., Frazzon, E.M., Freitag, M.: Using a digital twin for production planning and control in industry 4.0. In: Sokolov, B., Ivanov, D., Dolgui, A. (eds.) Scheduling in Industry 4.0 and Cloud Manufacturing. International Series in Operations Research & Management Science, vol. 289, pp. 39–60. Springer, Cham (2020). https://doi.org/10.1007/ 978-3-030-43177-8_3 14. Scholz-Reiter, B., Novaes, A.G.N., Makuschewitz, T., Frazzon, E.M.: Dynamic scheduling of production and inter-facilities logistic systems. In: Kreowski, H.J., Scholz-Reiter, B., Thoben, KD. (eds.) Dynamics in Logistics, pp. 443–453. Springer, Heidelberg (2011). https://doi.org/ 10.1007/978-3-642-11996-5_40 15. Pan, E., Liao, W., Xi, L.: A joint model of production scheduling and predictive maintenance for minimizing job tardiness. Int. J. Adv. Manuf. Technol. 60, 1049–1061 (2012) 16. Huang, J., Süer, G.A.: A dispatching rule-based genetic algorithm for multi-objective job shop scheduling using fuzzy satisfaction levels. Comput. Ind. Eng. 86, 29–42 (2015) 17. Frazzon, E.M., Albrecht, A., Pires, M., Israel, E., Kück, M., Freitag, M.: Hybrid approach for the integrated scheduling of production and transport processes along supply chains. Int. J. Prod. Res. 56(5), 2019–2035 (2017) 18. Jimenez, J., González-Neira, E., Zambrano-Rey, G.: An adaptive genetic algorithm for a dynamic single-machine scheduling problem. Manag. Sci. Lett. 8(5), 1117–1132 (2018) 19. Ucar, I., Smeets, B., Azcorra, A.: Simmer: discrete-event simulation for R. J. Stat. Softw. 90(2), 1–30 (2019) 20. Scrucca, L.: GA: a package for genetic algorithms in R. J. Stat. Softw. 53(4), 1–37 (2013) 21. Rolf, B., Reggelin, T., Nahhas, A., Lang, S., Müller, M.: Assigning dispatching rules using a genetic algorithm to solve a hybrid flow shop scheduling problem. Procedia Manuf. 42, 442–449 (2020) 22. Nguyen, S., Mei, Y., Xue, B., Zhang, M.: A hybrid genetic programming algorithm for automated design of dispatching rules. Evol. Comput. 27(3), 467–496 (2019) 23. Freitag, M., Hildebrandt, T.: Automatic design of scheduling rules for complex manufacturing systems by multi-objective simulation-based optimization. CIRP Ann. 65, 433–436 (2016)
Scheduling Workforce in Decentrally Controlled Production Systems: A Literature Review Julia Schwemmer(B) , Mathias K¨ uhn, Michael V¨ olker, and Thorsten Schmidt Chair of Logistics Engineering, Institute of Material Handling and Industrial Engineering, TU Dresden, Dresden, Germany [email protected] https://tu-dresden.de/ing/maschinenwesen/itla/tl
Abstract. Decentral production control plays a crucial role within the paradigm of Industry 4.0. Due to fast and flexible decisions on allocation and sequencing, there is no baseline schedule in advance. Moreover, the fourth industrial revolution modifies the organizational structures in the area of human resources, too. Despite changed tasks, the human is still a key factor with a coordinating, controlling and directing function— but without knowing the exact time of requirement. The workers are not available 24 h a day but are provided individually via personnel schedules. Creating a personnel schedule for the changed tasks without an overall baseline schedule becomes a crux of efficient staff deployment in the vision of Industry 4.0. This article presents the current state of this research aspect and derives a challenge for future research. Keywords: Workforce scheduling production control · Industry 4.0
1
· Timetabling · Decentral
Introduction
According to Ernst et al., “Personnel scheduling, or rostering, is the process of constructing work timetables for its staff so that an organisation can satisfy the demand for its goods or services” [1, p. 3]. The field of personnel scheduling is very broad and continues to grow. Its importance and timeless relevance is illustrated by a large number of publications (e.g. see [1–3]). For example, in a well-established literature review for this topic of Van den Bergh et al. in 2013 [2] the authors investigated almost 300 contributions, only selecting a representative, not a completed set of papers. Since its publication, it has been cited about 450 times (by the end of July 2021), according to the database Scopus, ¨ listing peer-reviewed research papers. Ozder et al. [3] have compiled an overview of different existing scheduling categories, including staff scheduling (Fig. 1). Human resource planning is not a static field of research. Its priorities and influencing factors are changing. One particular and actual example is the growing importance of satisfying the needs of the employees [2,4]. This includes the development of flexitime-models to advance the self-determination of employees. c The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 M. Freitag et al. (Eds.): LDIC 2022, LNLO, pp. 396–408, 2022. https://doi.org/10.1007/978-3-031-05359-7_32
Scheduling Workforce in Decentrally Controlled Production Systems
397
Fig. 1. Categories of scheduling (based on [3])
Scheduling staff is a topic also triggered by economic considerations [2]. On the one hand personnel is a key factor that contributes significantly to the success of a company. On the other hand it is also one of the strongest cost drivers. Therefore, the organization of staff and the labor costs are at the center of investigation of optimization procedures [2]. Of course, the goal is not to blindly cut personnel resources. Instead, the company should strive for perfect synchronization of personnel requirements and supply through efficient organization. The first step in creating an efficient staffing schedule is to identify the exact requirements [1]. This includes the knowledge of when exactly which type of employee (e.g. which qualification) is needed for which period of time. Based on the knowledge of demand, shifts can be determined and employees can be assigned individually to tasks or task bundles. Since people are not machines, the human resources department has to observe many regulations during the scheduling process (e.g. regulations of labor law, collective labor agreement or employment contract). Thus the personnel scheduling problem is highly constrained and very complex [1]. The derived problem classes for research considerations are usually classified as NP hard [3,5]. This contribution is structured as follows: Sect. 2 provides some background information on the specified research topic, the process of our literature review and the research questions. Section 3 shows the results of the state-of-the-artstudy and determines an existing research gap in more detail. The paper closes with a short conclusion in Sect. 4.
2
Background Information of Investigation
The following two subsections provide background information on the specified research topic and the performed process of literature review.
398
2.1
J. Schwemmer et al.
Workforce Scheduling Dilemma in Industry 4.0
The concept of Industry 4.0 brings many technical and organizational changes. This does not stop at human resource planning and scheduling or at the nature of employee task either, as we have already described in [6]. However, a large part of the research community agrees that humans will continue to play a central role in the smart factory [7,8]. The deserted factory is an utopian vision, which is not necessarily to be striven for. In this contribution, we particularly focus on machine-dominated manufacturing environments. We do not refer to a mainly manually dominated project manufacturing like complex assembly processes, as described in [9]. Nevertheless, especially in machine-dominated production environments, the nature of the tasks for the production workers will change [7,8]. Within cyber-physical production systems (CPPS) the human will mainly have coordinating, controlling and directing tasks [7,8]. On the one hand, progressive automation and digitalization will omit simple tasks that are usually characterized by continuous time-requirements. Consequently, steady deployment continuity of manual tasks during the production process will decrease. On the other hand, in CPPS, the complexity of the production system and of the tasks themselves will lead to high qualification requirements [8]. These described perspectives define a future scenario with more specific qualification classes and less consistent periods of working time. In addition to the change of tasks, the fourth industrial revolution will have influence on the organizational level of workforce scheduling, too. A base element of Industry 4.0 is the decentralization of production control [10]. Therefore, decisions on sequence and allocation of resources and orders will take place at lowest shop floor level. This will enable a high degree of flexibility and very rapid reactivity to process disturbances in production control. Decisions without long lead times will create a real-time control (as commonly called). Thus, there will be no (detailed) basic schedule for the production system. It will not be known (at least not in detail), which operation will be scheduled on which machine at what time. At this point, the dilemma of workforce scheduling in Industry 4.0 arises. In contrast to the production schedule, the staff schedules have to be determined some days or weeks in advance. In order to create an efficient workforce timetable (as described in Sect. 1), the information of the requirements from the production should already be available. However, with decentralized control, this information is not available, due to the missing basic production schedule. This prevents efficient resource planning of the workforce. In addition to technical and organizational changes, there is a transformation in the attitude of work. This tendency is sometimes also referred to as Work 4.0 [4] or New Work [11]. Work-life balance is increasingly coming to the fore [4]. This means that the balance between working time and free time gains higher relevance to employees. The main focus is no longer on career and salary. In contrast, compatibility of work and family or hobbies increases in importance. Companies must adapt to the changing wishes of their employees in order to be able to recruit and retain skilled personnel. Especially in highly specialized areas, a scarcity of skilled workers cannot be excluded in the future [7].
Scheduling Workforce in Decentrally Controlled Production Systems
2.2
399
Systematic Literature Review Process
A systematic literature review consists of several steps. For the generation of this contribution we used, for example, the guideline given in [12]. Fettke [12] divides the review procedure into the following five steps: Problem formulation, literature search, literature evaluation, interpretation, and presentation. Firstly, the problem formulation includes the research questions for the investigation. The main focus is to compile the current status of existing contributions to the field of tactical to operative workforce planning for decentrally controlled production systems. Above all, the essential part is the workforce requirements planning and individual deployment roster. The exact background to the problem statement is examined in detail in Sect. 2.1. The derived main research question is thus: Is there any method or algorithm dealing with the dilemma of workforce scheduling in decentrally production systems, as described in the Sect. 2.1? The included aim is to investigate whether there are existing problem formulations or a research gap—and to specify it. Our study focuses on the basic model (i.e. problem formulation) and not explicitly on an applied solution method in the existing literature. In this sense, we evaluate whether this problem type is already formulated in research or how it is formulated. Secondly, the goal of the literature search is to find suitable literature within the questionnaire. We mainly applied the following search phrases, which we combined in different ways: Personnel or workforce, scheduling or timetabling or planning, decentral production control or Industry 4.0 or adaptive manufacturing. The search process was mainly based on the two search engines Scopus and Google Scholar. Moreover, on a match, we looked at the reference list as well as at similar contributions proposed by the publishing medium. Thirdly, the literature evaluation part systematizes the found matches of the literature search and sorts it with regard to its relevance. Van den Bergh et al. [2] distinguish between managerial and technical papers in their literature review. According to [2], managerial contributions do not contain any mathematical formulation of the problem or any algorithmic description of the solution method. We prefer to name the separation as quantitative and qualitative. For our main classification part of the literature review in Sect. 3.2 we will focus on contributions containing mathematical or algorithmic formulations, i.e. quantitative methods. If there are interesting approaches in qualitative/text-based papers, we will discuss them in the text. Due to the starting point of Industry 4.0 described in Sect. 3.1, we do not consider literature before 2011. We also restricted the application area to the field of manufacturing because of the broadness of the personnel-scheduling field. Most areas of service systems (see Fig. 1) are not relevant due to the continuous and temporally known demands. We do not consider these any further. In addition, we exclude the problem class of resource constrained project scheduling projects (RCPSP) as it does not fit to the machine-dominated manufacturing system as described in Sect. 2.1. Fourthly, there is the step of analysis and interpretation. The purpose is to examine and evaluate the sorted papers against the background of the problem formulation. In this contribution, we do not give a complete listing of all
400
J. Schwemmer et al.
available literature. However, we give a representative set of the respective problems studied, to determine the state of the art. We used four previous literature reviews about the general topic of personnel planning to guide our investigation: [1–3,13]. In addition, there is a literature review of job shop scheduling in Industry 4.0, but it is not dealing with personnel planning: [14]. Finally, in the phase of presentation, the results of the study are to be presented to the public. With this in mind, this contribution shares some gained insights (see especially Sect. 3).
3
State of the Art
Section 3.1 gives a brief historical insight, followed by the results of literature review and the located research gap in Sects. 3.2 and 3.3. 3.1
Selected Relevant Trends in the Last Years
Market conditions are subject to constant change [10]. On the one hand, there is the trend towards individualization with small batch sizes [10]. On the other hand, there is a global competition with cost pressure and shortening delivery times [10]. One answer to these trends is the development of CPPS—with the changed role of the production workforce. As described, highly specialized employees with short work assignments will play a central role in cost optimization. Therefore, there are some up-coming questions: Will the classic view of assigning and primarily optimizing orders or resources (resources in unspecific form or in sense of machine [15] or, e.g., also seen in textbooks like [16]) be sufficient? Will the derivation of workforce schedule from the production schedule be possible and the subsequent second level optimization of workforce sufficient? Furthermore, the increasing flexibility of working time will play an important role. Turning away from rigid shift times to flexible working hours can be triggered from both point of views, the employer’s and the employee’s point of view. From the employee’s perspective, it is the on-going change in attitude to work (Work 4.0). This does not only affect people with office jobs but also employees in production. From the employer’s perspective, it is the promising possibility of perfect short- to medium-term synchronization of demand and supply. The synchronization of the demand of workforce is not a new topic for the research community. Already at the turn of 2000, there are a lot of contributions that deal with flexitime-models (e.g. see [15]). These approaches include for example working hours with seasonal accounts. Even skill-based allocations are taken into account [15]. Topics like individual efficiency and individual preferences of task assignments came up (e.g. see [17,18]), too. However, the workforce is still often just a second level resource at optimization process, like in [17]. The idea of Industry 4.0 emerged later in history. In 2011, Kagermann, Lukas and Wahlster introduced the “fourth industrial revolution” at the Hannover Fair [19]. From 2012 onwards, the German government also pursued the strategy of Industry 4.0 [10].
Scheduling Workforce in Decentrally Controlled Production Systems
3.2
401
Classification of Contributions
A targeted view of the state of the art for the described problem will be presented in a nutshell. Table 1 shows the problem classes sighted based on the papers found in connection with the research questions. In particular, there are the classes of employee-timetabling, Job Shop Scheduling Problem (JSP/JSSP) & Flow Shop Scheduling Problem (FSP/FSSP) & Flexible JSP (FJSP/FJSSP) and Line Balancing Problems. In addition, there are some contributions without assigning their model to an already commonly known (c.k.) problem class. The class of employee-timetabling does not typically take into account scheduling or assignments of machines. Table 1 is not a complete list of all sub-problems of each class but the specifications we found in the context of the described dilemma. Based on the considered problem case from Sect. 2.1, we can derive the following characterizations for the investigation of the available literature: – Problem Classification: For research considerations, real problems are categorized. This also facilitates the allocation and recognition of research studies. Each class has its own characteristics and attributes. Usually there are benchmark instances to test the efficiency of algorithms. In some contributions, the classification is missing. Thus, we add a short substitution like “Assignment of tasks within a shift” or “skill development”. Table 1. Sighted problem classes Class
Specification
Ref.
Employee-timetabling (no machines)
- General Employee Scheduling Problem (GES) - Shift Minimization Personal Task Scheduling Problem (SMPTSP) - General Task-based Shift Generation Problem (GTSGP)
e.g. [20] e.g. [21, 22] e.g. [23]
JSP, FJSP and FSP
-
Dual-Resource Constrained Job Shop Problem (DRCJSP) e.g. [24] Dual-Resource Constrained Flexible Job Shop Problem (DRCFJSP) e.g. [25–27] Dual-Resource Constrained Assembly Flexible Job Shops Problem e.g. [28] JSP with Daily Shift Scheduling Problem (JSP-DJSSP) e.g. [29] Job Shop with Skilled Operators (JSSO) e.g. [30] Worker-Constrained FJSP (WFJSP) e.g. [5] WFJSP with Sequence-Dependent Setup Times (WSFJSP) e.g. [31] Flexible Job Shop Scheduling with Worker flexibility (FJSPW) e.g. [32] Integrated Employee Timetabling and Job-Shop Scheduling Problem (IET-JSSP) e.g. [33] Job Shop Scheduling Problem with HR-Flexibility e.g. [34] Dynamic Hybrid Flow Shop Problem e.g. [35]
Line balancing problem - Assembly Line Worker Assignment and Balancing Problem (ALWABP) - Multi-Manned Assembly Line Balancing Problem (MALBP)
e.g. [36, 37] e.g. [38]
No assign. of (c.k.) class
e.g. e.g. e.g. e.g. e.g.
-
Problem including skill development/crosstraining Problem including rostering and skill development Stochastic General Manpower Allocation Problem (SGMAP) Task-based worker assignment within a shift but without rostering Scheduling decisions for multi-agent control
[39, 40] [17] [41] [42, 43] [44]
402
J. Schwemmer et al.
– Decentral Production Control: As described earlier, the decentral production control is a central key factor of the scenario considered. This feature indicates whether the method or algorithm takes this into account. – Dual-Resource Constrained: We have borrowed the term from the JSPclassification. It indicates the consideration of man and machine as separate resources that may be used in parallel. The workforce is a different type of resource with different conditions than a machine. Both types of resource are subject to the scheduling process to reach an overall optimization. – Short Discrete Worker Requirement Operation: As described in Sect. 2, in the scenario under consideration, there may be short manual operation tasks that arise discretely in time. These manual operation tasks may be shorter than the associated machine tasks. The worker is only called to the machine at short selective intervals. The machine task is not the shortest element to schedule. Examples are inserting and removing parts, interim inspections, troubleshooting or collaborative subtasks. This feature will be only considered if the attribute “Dual-Resource Constrained” already includes the scheduling of the human resource. However, this attribute does not mean or include multi-machine operation. – Individual, Heterogeneous Qualification Profiles: Each worker has an individual, specific set of skills. Consequently, not every worker can perform each task. Qualifications can be single or multiple assigned to the workers. – “Shiftless” Working Hours: The scheduled working times do not have to follow classic shift models, like three-shift-system with early, late and night shifts. The working hours can be set independent (but with restrictions on working hours). The scheduled time of each employee may vary. Several papers in our review process do not contain shift-table-generation or rostering at all. – Integration of Working Time Preference of the Employee: The employee can contribute to the arrangement of his or her working hours. They can indicate when they would like to work: Buzzword “Work 4.0”. The method considers these wishes. This refers primarily to the time-of-day scheduling (for further information see [2]). – Stochastic Influencing Variables: This feature denotes whether the production scenario is deterministic or stochastic. In the vision described above, stochastic manual process times or production disruptions are to be expected. Other stochastic influencing variables are conceivable. Generally, stochastic scenarios significantly complicate the solution finding process and strongly increase the computational effort. Table 2 provides the information on the described attributes for the selected contributions. There were numerous promising matches from the title in the second phase of review, but classified as irrelevant on closer inspection in the third and fourth phase. An example of these contributions is [45] due to the wrong time horizon. As described, we even do not focus on contributions with qualitative methods for the detailed categorization, as for example [46,47]. However we would like to discuss the (qualitative) contribution of Bauer [47] as it also deals with some highlights of the topic. Bauer [47] deals with the
Scheduling Workforce in Decentrally Controlled Production Systems
403
Table 2. Classification of literature Ref. Author
Year Class
Dec. Prod. Ctrl.
DRC Short Man. Task
Het. Skills
Shiftl. Hours
Time Pref.
(Part.) Stoch.
[20] [21] [22] [23] [24] [25] [26] [27] [28] [29] [30] [5] [31] [32] [35] [33] [34] [36] [37] [38] [39] [40] [17] [41] [42] [43] [44]
2020 2021 2012 2019 2020 2020 2019 2016 2014 2017 2014 2021 2019 2019 2021 2014 2016 2014 2012 2015 2016 2021 2014 2014 2021 2019 2016
No No No No No No No No Yes No No No No No Yes No No No No No No No No No No No Yes
Noa Noa Noa Noa Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes No Yes Noa Yes Yes Yes Yes
Yes Yes Yes Yes Yes Yes No Yes No No Yes Yes No Yes Yes Yes Yes Yes No No No Yes Yes Yes Yes Yes Yes
Yes No No No Nob No Nob Nob Nob Yes Nob Nob No Nob Nob No Nob Nob No Nob Nob Nob Yes Nob Nob Nob Nob
No No No No No No No No No No No No No No No No No No No No No No No No No No No
No No No No Yes Yes Yes No No Yes Yes No No No Yes No Yes Yes No No No Yes No Yes No No Yes
a b
Kletzandera and Musliu Nurmi and Kyng¨ as Krishnamoorthy et al. Kyng¨ as et al. Th¨ urer et al. Andrade-Pineda et al. Th¨ urer et al. Zheng and Wang Sammarco et al. Hur Agnetis et al. M¨ uller and Kress Kress et al. Gong et al. Zhang et al. Guyon et al. Lin et al. Borba and Ritt Moreira et al. Roshani and Giglio Denkena et al. Altendorfer et al. Attia et al. Egilmez et al. Wikarek and Sitek Liu and Liu Qu et al.
GES SMPTSP SMPTSP GTSGP DRCJSP DRCFJSP DRCFJSP DRC-FJSSP DRC-A-FJSP JSP-DSSP JSSO WFJSP WSFJSP FJSPW Dynamic hybrid FSP IET-JSSP JSSP with HR flex. ALWABP ALWABP MAL BP Skill dev. Skill dev. Rost. + skill dev. SGMAP Task assign. Task assign. Scheduling decisions
– – – – No No No No No No No No No No No No No No No No Yes No – Yes No No No
Only personnel No shift-table-generation/rostering
synchronization of personnel requirement and supply within CPPS as well as the timetabling for the workforce in times of Work 4.0. The most significant difference is that he only focuses on compensatory measures, like additional shifts or shift extensions. These compensatory working hours are flexibly divided among the workers in a group agreement, after selecting possible workers based on working time regulations. The additional shift also has fixed shift times for the workers. Moreover, a rigid shift time model covers the entire basic supply of workforce. In addition, the determination of workforce requirement within a decentrally controlled production system is not a key part in [47]. For the authors it was not possible to see whether stochastic influences are assumed. 3.3
Some Gained Insights
Table 1 already shows that there are various problem classes with subcategories under consideration. Looking for the topic of rostering, the class of “EmployeeTimetabling” comes across as one of the first matches. However, this class usually deals only with the scheduling of the human resource. The assignment of
404
J. Schwemmer et al.
machines is no (primary) goal. Therefore, there is no overall optimization. In contrast, the field of JSP/FSP focuses on the scheduling of machines, as they are generally machine-dominated problem classes. Thus, humans usually do not play any further role—at least as long as there are no further specified sub-problems. With the specification of “DRC”, both resources (man and machines) are subject of the scheduling process. In this area there are numerous representatives [24]. Nevertheless, within DRC-JSP, rostering plays a subordinate role and is usually not treated. Within the literature, rostering is also hardly noticed in the other subclasses of the JSP/FSP. For the selected line balancing problems, the same pattern arises. The focus is primarily on the assignment of tasks to employees during the shift. The shift creation itself is not included in the analysis. During the evaluation process it is noticeable that there are only few contributions that deal with shift-table-generation and are not based on rigid shift hours (attribute “Shiftl. Hours” in Table 2). In addition—or maybe consequently, time preferences of individual workers are not considered in any of the studies classified in Table 2 (attribute “Time Pref.”). However, especially regarding to Work 4.0 or New Work, this should be a central topic in employee scheduling. Regardless of our other criteria, this area seems to offer a lot of potential for investigation. Heterogeneous skill sets are taken into account in almost all investigated problems and hardly represent any further need for a new research direction. Nevertheless, almost all explored literature also demonstrates: Within production environments of highly specialized workforce with heterogeneous skill sets, the scheduling of individual workers in time generally has a significant influence to the objective function (see especially [39]). Conducting the evaluation, there is the central question: How many contributions are concerned with the topic of rostering in decentrally controlled production systems? Just referring to our classification in Table 2, there are only three papers which apply to the corresponding attributes “Dec. Prod. Ctrl.” and “DRC”. By adding the subject of rostering, i.e. filtering with these three selection criteria, there will be no publication left. Until now, we were not able to find a quantitative contribution dealing with the dilemma we have described. Table 2 reveals that research gap. To the best of our knowledge, no author already formulated a model for the described problem of the scheduling dilemma of workforce for Industry 4.0. Consequently, there is no problem formulation regarding our research topic. Nevertheless, individual deployment roster based on workforce requirement planning in the environment of decentrally controlled production systems is a promising future research direction. We will pursue this research challenge by further work. The derived classification attributes give an initial impression of what has to be taken into account in order to solve the considered problem. Accordingly, the derived classification attributes can also be applied for the generation of new corresponding problem instances. The difference in application would be that these will not only specify criteria to be fulfilled, but also to what extent they are fulfilled. The following examples illustrate this:
Scheduling Workforce in Decentrally Controlled Production Systems
405
– Decentralized Production Control: Of what type is the control? E.g., agent based or priority rule based. – Individual, Heterogeneous Qualification Profiles: Are there special effects? E.g., the inclusion of learning curves or fatigue.
4
Conclusion
After a short introduction to the topic of personnel scheduling, we briefly explained the focus and the procedure for our state-of-the-art-study. We have limited the literature review to the application area of manufacturing and quantitative contributions since 2011. Section 3 presents the main results. There are several research projects that deal with the technical collaboration of human and machine (e.g. see [48]). However, the organizational aspects of flexible shifting workforce in CPPS have hardly been considered. To the best of our knowledge, there is no contribution dealing with the regarded subject of the workforce planning dilemma in Industry 4.0. We derived some necessary attributes of the described problem case to categorize existing literature. However, none of the investigated publications considers all of the derived aspects. Therefore, we have revealed a research gap in the context of workforce scheduling for decentrally controlled production systems. We will continue to pursue this research direction by future work as the dilemma is worth to be investigated. Further research has to discuss the exact modeling with possible solution paths, classifications as well as potential benchmarks. Further considerations for the literature review can extend the classification to solution method and objective function values. In addition, it would be possible to look even more closely to manually varying processing times in the above context, like learning and unlearning phenomena, effects of fatigue or similar. For future state-of-the-art-studies, it may be also interesting to examine material supply. Against the same background of decentralized production control, a similar problem for material supply may arise as for workforce scheduling. Acknowledgement. We would especially like to thank the German Research Foundation/Deutsche Forschungsgemeinschaft (DFG), which is funding our project with the title “A simulation-based and flexi-time applying prediction model for scheduling personnel deployment times in the production planning process of cyber-physical systems” (project-id: 439188616).
References 1. Ernst, A., Jiang, H., Krishnamoorthy, M., Sier, D.: Staff scheduling and rostering: a review of applications, methods and models. Eur. J. Oper. Res. 153(1), 3–27 (2004). https://doi.org/10.1016/S0377-2217(03)00095-X. ISSN:03772217 2. Van den Bergh, J., Beli¨en, J., De Bruecker, P., Demeulemeester, E., De Boeck, L.: Personnel scheduling: a literature review. Eur. J. Oper. Res. 226(3), 367–385 (2013). https://doi.org/10.1016/j.ejor.2012.11.029. ISSN:03772217
406
J. Schwemmer et al.
¨ ¨ 3. Ozder, E.H., Ozcan, E., Eren, T.: A systematic literature review for personnel scheduling problems. : Int. J. Inf. Technol. Decis. Mak. 19(06), 1695–1735 (2020). https://doi.org/10.1142/S0219622020300050. ISSN:0219-6220, 1793-6845 4. German Federal Institute for Occupational Safety and Health. Flexible Arbeitszeit¨ modelle - Uberblick und Umsetzung. OCLC: 1197143325. 44149 Dortmund (2019). ISBN:978-3-88261-230-1 5. M¨ uller, D., Kress, D.: Filter-and-fan approaches for scheduling flexible job shops under workforce constraints. Int. J. Prod. Res. 1–23 (2021). https://doi.org/10. 1080/00207543.2021.1937745. ISSN:0020-7543, 1366-588X 6. Schwemmer, J., Schmidt, T., V¨ olker, M.: A new simulation-based approach to schedule personnel deployment times in decentrally controlled production systems. In: SIMUL2020: The Twelfth International Conference on Advances in System Simulation, pp. 19–23 (2020) 7. Ittermann, P., Niehaus, J., Hirsch-Kreinsen, H.: Arbeiten in der Industrie 4.0: trendbestimmungen und arbeitspolitische Handlungsfelder. In: Study der HansB¨ ockler-Stiftung, vol. 308, p. 90 (2015) 8. Spath, D., Ganschar, O. (eds.): Organisation, F.-I. f¨ ur Arbeitswirtschaft und: Produktionsarbeit der Zukunft - Industrie 4.0: Studie. Fraunhofer-Verl, Stuttgart (2013). ISBN:978-3-8396-0570-7 9. K¨ uhn, M., V¨ olker, M., Schmidt, T.: An algorithm for efficient generation of customized priority rules for production control in project manufacturing with stochastic job processing times. Algorithms 13(12), p. 337 (2020). https://doi.org/10. 3390/a13120337. ISSN:1999-4893 10. German Federal Ministry of Education and Research. Industrie 4.0/Innovationen f¨ ur die Produktion von Morgen. Technical Report 11055 Berlin, p. 172 (2017) 11. Bergmann, F.: New Work, New Culture: Work we Want and a Culture that Strengthens us. OCLC: on1049797502. Zero Books, Winchester; Washington (2019). ISBN:978-1-78904-064-7 12. Fettke, P.: State-of-the-art des state-of-the-art: Eine Untersuchung der Forschungsmethode review innerhalb der Wirtschaftsinformatik. Wirtschaftsinformatik 48(4), 257 (2006). https://doi.org/10.1007/s11576-006-0057-3. ISSN:09376429, 1861-8936 13. De Bruecker, P., Van den Bergh, J., Beli¨en, J., Demeulemeester, E.: Workforce planning incorporating skills: state of the art. Eur. J. Oper. Res. 243(1), 1–16 (2015). https://doi.org/10.1016/j.ejor.2014.10.038. ISSN:03772217 14. Zhang, J., Ding, G., Zou, Y., Qin, S., Fu, J.: Review of job shop scheduling research and its new perspectives under Industry 4.0. J. Intell. Manuf. 30(4), 1809–1830 (2019). https://doi.org/10.1007/s10845-017-1350-2. ISSN:0956-5515, 1572-8145 15. Grabot, B., Letouzey, A.: Short-term manpower management in manufacturing systems: new requirements and DSS prototyping. Comput. Indust. 43, 11–29 (2000) 16. Steven, M.: Produktionsmanagement, 70565. W. Kohlhammer Verlag, Stuttgart (2014). ISBN:978-3-17-028861-4 17. Attia, E.-A., Duquenne, P., Le-Lann, J.-M.: Considering skills evolutions in multi-skilled workforce allocation with flexible working hours. Int. J. Prod. Res. 52(15), 4548–4573 (2014). https://doi.org/10.1080/00207543.2013.877613. ISSN:0020-7543, 1366-588X 18. Sabar, M., Montreuil, B., Frayret, J.-M.: Competency and preference based personnel scheduling in large assembly lines. Int. J. Comput. Integrat. Manuf. 21(4), 468– 479 (2008). https://doi.org/10.1080/09511920701574842. ISSN:0951-192X, 13623052
Scheduling Workforce in Decentrally Controlled Production Systems
407
19. Kagermann, H., Lukas, W.-D., Wahlster, W.: Industrie 4.0: Mit dem Internet der Dinge auf dem Weg zur 4. industriellen Revolution. In: VDINachrichten (2011) 20. Kletzander, L., Musliu, N.: Solving the general employee scheduling problem. Comput. Oper. Res. 113, 104794 (2020). https://doi.org/10.1016/j.cor.2019.104794. ISSN:03050548 21. Nurmi, K., Kyng¨ as, N.: A successful three-phase metaheuristic for the shift minimization personal task scheduling problem. In: Repoussis, P.P. (ed.) Advances in Operations Research, vol. 2021, pp. 1–12 (2021). https://doi.org/10.1155/2021/ 8876990. ISSN:1687-9155, 1687-9147 22. Krishnamoorthy, M., Ernst, A., Baatar, D.: Algorithms for large scale shift minimisation personnel task scheduling problems. Eur. J. Oper. Res. 219(1), 34–48 (2012). https://doi.org/10.1016/j.ejor.2011.11.034. ISSN:03772217 23. Kyng¨ as, N., Nurmi, K., Goossens, D.: The general task-based shift generation problem: formulation and benchmarks. In: Proceedings of the 9th Multidisciplinary International Conference on Scheduling: Theory and Applications, Ningbo, pp. 301–319 (2019) 24. Th¨ urer, M., Zhang, H., Stevenson, M., Costa, F., Ma, L.: Worker assignment in dual resource constrained assembly job shops with worker heterogeneity: an assessment by simulation. Int. J. Prod. Res. 58(20), 6336–6349 (2020). https://doi.org/10. 1080/00207543.2019.1677963. ISSN:0020-7543, 1366-588X 25. Andrade-Pineda, J.L., Canca, D., Gonzalez, R.P.L., Calle, M.: Scheduling a dualresource flexible job shop with makespan and due date-related criteria. Ann. Oper. Res. 291(1–2), 5–35 (2020). https://doi.org/10.1007/s10479-019-03196-0. ISSN:0254-5330, 1572-9338 26. Th¨ urer, M., Stevenson, M., Renna, P.: Workload control in dual-resource constrained high-variety shops: an assessment by simulation. Int. J. Prod. Res. 57(3), 931–947 (2019). https://doi.org/10.1080/00207543.2018.1497313. ISSN:0020-7543, 1366-588X 27. Zheng, X.-l., Wang, L.: A knowledge-guided fruit fly optimization algorithm for dual resource constrained flexible job-shop scheduling problem. Int. J. Prod. Res. 54(18), 5554–5566 (2016). https://doi.org/10.1080/00207543.2016.1170226. ISSN:0020-7543, 1366-588X 28. Sammarco, M., Fruggiero, F., Neumann, W., Lambiase, A.: Agent-based modelling of movement rules in DRC systems for volume flexibility: human factors and technical performance. Int. J. Prod. Res. 52(3), 633–650 (2014). https://doi.org/10. 1080/00207543.2013.807952. ISSN:0020-7543, 1366-588X 29. Hur, Y.: Optimization models for manufacturing and personnel scheduling. PhD thesis. The University of Texas at Austin (2017) 30. Agnetis, A., Murgia, G., Sbrilli, S.: A job shop scheduling problem with human operators in handicraft production. Int. J. Prod. Res. 52(13), 3820–3831 (2014). https://doi.org/10.1080/00207543.2013.831220. ISSN:0020-7543, 1366-588X 31. Kress, D., M¨ uller, D., Nossack, J.: A worker constrained flexible job shop scheduling problem with sequence-dependent setup times. OR Spectrum 41(1), 179–217 (2019). https://doi.org/10.1007/s00291-018-0537-z. ISSN:0171-6468, 1436-6304 32. Gong, G., Chiong, R., Deng, Q., Gong, X.: A hybrid artificial bee colony algorithm for flexible job shop scheduling with worker flexibility. Int. J. Prod. Res. 58(14), 4406–4420 (2020). https://doi.org/10.1080/00207543.2019.1653504. ISSN:0020-7543, 1366-588X 33. Guyon, O., Lemaire, P., Pinson, R., Rivreau, D.: Solving an integrated jobshop problem with human resource constraints. Ann. Oper. Res. 213(1), 147–171 (2014). https://doi.org/10.1007/s10479-012-1132-3. ISSN:0254-5330, 1572-9338
408
J. Schwemmer et al.
34. Lin, R., Zhou, G., Liu, A., Lu, H., Li, T.: Impact of personnel flexibility on job shop scheduling. Sci. Program. 2016, 1–8 (2016). https://doi.org/10.1155/2016/ 3403058. ISSN:1058-9244, 1875-919X 35. Zhang, X., Sun, X., Liu, Y., Zhang, C., Zhao, L., Shen, W.: An agent- based approach for dynamic scheduling in hybrid flow shops. In: 2021 IEEE 24th International Conference on Computer Supported Cooperative Work in Design (CSCWD), Dalian, pp. 24–30. IEEE (2021). https://doi.org/10.1109/CSCWD49262.2021. 9437879. ISBN:978-1-72816-597-4 36. Borba, L., Ritt, M.: A heuristic and a branch-and-bound algorithm for the assembly line worker assignment and balancing problem. Comput. Oper. Res. 45, 87–96 (2014). https://doi.org/10.1016/j.cor.2013.12.002. ISSN:03050548 37. Moreira, M.C.O., Ritt, M., Costa, A.M., Chaves, A.A.: Simple heuristics for the assembly line worker assignment and balancing problem. J. Heurist. 18, 505–524 (2012). https://doi.org/10.1007/s10732-012-9195-5 38. Roshani, A., Giglio, D.: A simulated annealing approach for multi-manned assembly line balancing problem type II. IFAC-PapersOnLine 48(3), 2299–2304 (2015). https://doi.org/10.1016/j.ifacol.2015.06.430. ISSN:24058963 39. Denkena, B., Dittrich, M.A., Winter, F., Wagener, C.: Simulation-based planning and evaluation of personnel scheduling in knowledge-intensive production systems. Prod. Eng. 10(4–5), 489–496 (2016). https://doi.org/10.1007/s11740-016-0693-4. ISSN:0944-6524, 1863-7353 40. Altendorfer, K., Schober, A., Karder, J., Beham, A.: Service level improvement due to worker cross training with stochastic worker absence. Int. J. Prod. Res. 59(14), 4416–4433 (2021). https://doi.org/10.1080/00207543.2020.1764126. ISSN:0020-7543, 1366-588X 41. Egilmez, G., Erenay, B., S¨ uer, G.A.: Stochastic skill-based manpower allocation in a cellular manufacturing system. J. Manuf. Syst. 33(4), 578–588 (2014). https:// doi.org/10.1016/j.jmsy.2014.05.005. ISSN:02786125 42. Wikarek, J., Sitek, P.: Proactive and reactive approach to employee competence configuration problem in planning and scheduling processes. Appl. Intell. (2021). https://doi.org/10.1007/s10489-021-02594-x. ISSN:0924-669X, 1573-7497 43. Liu, M., Liu, X.: Satisfaction-driven bi-objective multi-skill workforce scheduling problem. IFAC-PapersOnLine 52(13), 229–234 (2019). https://doi.org/10.1016/j. ifacol.2019.11.134. ISSN:24058963 44. Qu, S., Wang, J., Govil, S., Leckie, J.O.: Optimized adaptive scheduling of a manufacturing process system with multi-skill workforce and multiple machine types: an ontology-based, multi-agent reinforcement learning approach. Procedia CIRP 57, 55–60 (2016). https://doi.org/10.1016/j.procir.2016.11.011. ISSN:22128271 45. H¨ ammerle, M.: Methode zur strategischen Dimensionierung der Personalflexibilit¨ at in der Produktion: Wirkungsbewertung von Instrumenten zur Flexibilisierung der Personalkapazit¨ at im volatilen Marktumfeld. In: Schriftenreihe zu Arbeitswissenschaft und Technologiemanagement, vol. 26. Fraunhofer Verlag, Stuttgart (2015). ISBN:978-3-8396-0899-9 46. Stock, T.: Ein Verfahren zur Personalplanung und -steuerung und Restrukturierung der Aufbauorganisation f¨ ur eine bedarfsorientierte und wandlungsf¨ ahige Produktion. In: Stuttgarter Beitr¨ age zur Produktionsforschung 9. Fraunhofer-Verl, Stuttgart (2013). ISBN:978-3-8396-0532-5 47. Bauer, W., (ed.): Selbstorganisierte Kapazit¨ atsflexibilit¨ at in Cyber-Physical- Systems: Abschlussbericht. Fraunhofer Verl, Stuttgart (2015). ISBN:978-3-8396-0932-3 48. Fraunhofer-Institut f¨ ur Arbeitswirtschaft und Organisation Stuttgart. My-CPS. www.mycpstoolbox.de. Accessed 28 Aug 2021
Maintenance 4.0: A Literature Review and SWOT Analysis Danilo Ribamar Sá Ribeiro1(B) , Lúcio Galvão Mendes1 , Fernando Antônio Forcellini1,2 , and Enzo Morosini Frazzon1 1 Graduate Program in Production Engineering, Federal University of Santa Catarina,
Florianopolis, SC 88040-900, Brazil [email protected], [email protected], {fernando.forcellini,enzo.frazzon}@ufsc.br 2 Department of Mechanical Engineering, Federal University of Santa Catarina, Florianopolis, SC 88040-900, Brazil
Abstract. Maintenance 4.0 (M4.0) is described as an innovative and optimized maintenance strategy, integrating existing practices with technologies from Industry 4.0. These technologies have been impacting and transforming several production processes and also maintenance management. It is possible to extract increasingly accurate data about the assets through sensors and other technologies in real-time. However, companies still struggle during the Maintenance 4.0 implementation, becoming an open field for research. Based on this, we developed a consistent literature review, followed by a SWOT analysis (strengths, weaknesses, opportunities, and threats) of Maintenance 4.0 implementation. The synthesis of the results was divided into bibliometric, thematic, and SWOT analyses. The bibliometric analysis showed the continuous growth of the theme in recent years. In the content analysis, the perspectives were grouped into (i) conceptual, theoretical studies, (ii) empirical studies/applications. In the SWOT analysis, we demonstrated the main strengths to achieve the benefits of implementing Maintenance 4.0. The weaknesses and threats identified may challenge some organizations. The implications of this study are in the organization of a comprehensive body of knowledge on Maintenance 4.0. From a practical point of view, it is hoped that this research can be used as a guideline, providing decision support to maintenance managers. Keywords: Maintenance 4.0 · Literature review · SWOT analysis
1 Introduction The fourth industrial revolution, better known by the popular term “Industry 4.0” (I4.0), originally started in Germany in 2011, has gained much attention by promoting the idea of a fully integrated and efficient industry [27]. In this scenario, companies and departments will become more cohesive by automating the value chain [58]. Taking advantage of digital technological innovations, the evolution towards I4.0 occurs at all levels, namely in the maintenance function, which is understood as a support process for © The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 M. Freitag et al. (Eds.): LDIC 2022, LNLO, pp. 409–422, 2022. https://doi.org/10.1007/978-3-031-05359-7_33
410
D. R. S. Ribeiro et al.
production [46]. Maintenance ensures the proper functioning of machines and equipment and, consequently, the entire production line [61]. In this sense, this function is essential since machine failures can negatively affect production (delays or rework) [59]. Within these actions, the use of modern maintenance approaches such as Maintenance 4.0 (M4.0) stands out as one of the predominant topics of this era [30, 59]. Although researchers have envisioned many benefits derived from M4.0, companies still struggle with its practical implementation [64], and few studies have specifically addressed the integration of I4.0 and maintenance [34, 59]. Also, M4.0 still requires much research to have a broad impact as uncertainties about its potential contributions [33]. Therefore, this paper presents an original contribution by investigating the research landscape in M4.0. We assessed the studies found in the scientific literature, in terms of strengths, weaknesses, opportunities, and threats (SWOT), according to their potential to contribute to a satisfactory situation for the I4.0’s context.
2 Theoretical Background 2.1 Maintenance 4.0 Maintenance has evolved from reactive (Maintenance 1.0) to preventive (Maintenance 2.0) and then to condition-based (Maintenance 3.0), the current predictive and prescriptive approach (M4.0) [30, 45, 55]. [30] state that M4.0 is a set of techniques for monitoring the current state of machines to predict failures through the use of real-time automatic analysis and supervised or unsupervised machine learning. It employs sensors to collect accurate data describing manufacturing, equipment condition, and overall operating status. [11] define M4.0 as an organizational design for managing plant maintenance in environments with pervasive digital technologies. Furthermore, they devised M4.0 in four underlying and interrelated dimensions: (i) data-driven decision making, (ii) human capital resources, (iii) internal integration, and (iv) external integration [11, 59]. I4.0 technologies provide the basis for implementing innovative maintenance strategies and optimizing existing practices. The key technologies involved are data collection and analysis technologies such as the Internet of Things (IoT), cloud computing, predictive analytics (such as fuzzy logic, neural networks, evolutionary network algorithms, machine learning, probabilistic reasoning), and equipment repair technologies [30, 59]. According to [45], the success of M4.0 is associated with designing products with specifications that include these technologies. [33] point out that M4.0 can make an essential contribution to the maintenance pattern of the future and should be researched thoroughly considering the characteristics of the industrial field in which it will be deployed. It can be applied in many fields, such as automotive, airlines, transportation, ports, and oil and gas [3, 49].
3 Research Methodology The method used in this research was conceptual-theoretical, based on a systematic literature review (SLR). According to [20], this is a systematic, explicit, and reproducible
Maintenance 4.0: A Literature Review and SWOT Analysis
411
method for identifying the existing body of completed and recorded work produced by researchers. The search strategy was formulated for searching and mapping the articles starting from requirements relevant to the research. In light of this, we used the following search term: “Maintenance 4.0” OR “Smart maintenance” OR “Intelligent maintenance”. Despite not being identical concepts, the terms chosen have great similarities [55]. The search was applied to the titles, abstracts, and keywords of the articles. This SLR relies on articles collected in the following databases: Scopus™ (Elsevier) and Web of Science Core Collection™ (WoS), to provide respectively a broad spectrum [63] and interdisciplinary perspective [19]. These scientific databases were chosen because they are the most widespread for industrial engineering [1]. The search was performed in November 2021. We found 658 publications, of which 215 were from Scopus and 443 were from WoS. In addition, the inclusion and exclusion criteria are presented in Table 1. For this, it followed the Preferred Reporting Items for Systematic Review and Meta-Analysis (PRISMA) approach. Table 1. Inclusion and exclusion criteria. Criteria
Explanation
Inclusion
The research efforts related to studies that focused on M4.0 implementation, its benefits, disadvantages, and opportunities. Document type: peer-reviewed articles; Language: English.
Exclusion
It has only the title, abstract, and keywords in English, but not the full text. It does not have the full text to be considered. It is not an academic article and is not aligned with M4.0. It is used only as an example/ part of research direction and future perspective/ a quoted expression/ in keywords and references.
Source: Adapted from [41].
Figure 1 presents the methodological steps used in this study. The classification taxonomy consists of the following stages: (i) identification of the research problem; (ii) the search in the journal bases; (iii) selection of articles that contain the research term in the title, or the abstract, or the keywords; (iv) the removal of duplicate articles; (v) there was no refinement by date, as this criterion would reduce the scope of the research; (vi) composition of the bibliographic portfolio and analyses [2, 19, 20, 23]. The final selection totaled 38 papers with alignment to the theme addressed. The synthesis of the results was divided into bibliometric, thematic, and SWOT analyses. For the bibliometric analysis, we used the software R 3.5.2 [53], supported by the package “Bibliometrix” 2.1.2 [7]. This software was chosen because it has the most extensive techniques and tools compared to others with the same functionality [42]. Then, the second step is a thematic analysis, a common technique for qualitative data [17], which comprises the organization of the studies into guiding axes for the research [43]. At last, a SWOT analysis was conducted. This tool is characterized by the analysis of strengths (s), weaknesses (w), opportunities (o), and threats (t) [29]. Recently, it has been used to provide comprehensive synopsis in several areas [13, 50, 51, 62]. Its use in this research
412
D. R. S. Ribeiro et al.
was because it is considered a robust approach to summarize the results [26] and allows to present all the information collected in the literature review also with a visual approach that is useful to better frame the different instruments [50].
Fig. 1. Methodological research flow process.
4 Results and Discussion 4.1 Bibliometric Analysis Table 2 shows the temporal evolution of publications based on the final bibliographic portfolio. This analysis allows us to highlight the growing interest in research on M4.0, considering the selection criteria adopted. It is noted that the year 2020 has the most significant number of publications in the final portfolio. Table 2. Publication temporal evolution. Year
2016
2017
2018
2019
2020
2021
Papers
1
2
2
5
18
10
There are four journals in the final portfolio, which have more than one paper. “International Journal of Production Economics” has three papers. “International Journal of Productivity and Performance Management”, “Journal of Quality in Maintenance Engineering,” and “Computers in Industry” each have three papers.
Maintenance 4.0: A Literature Review and SWOT Analysis
413
Figure 2a identifies the most recurrent terms in the final portfolio, i.e., keywords cooccurrence network. For this, we based on [28, 52]. The central term “maintenance” is connected with the terms “Industry 4.0” and “Decision making”, indicating the direction of the reviewed papers.
Fig. 2. (a) Keywords co-occurrence network; (b) Thematic map.
Figure 2b shows a thematic map of the area, relating the density and centrality of the terms most evident in the titles of the analyzed publications. The themes in the upper-right quadrant represent the motor themes, well-developed and essential areas for structuring a search field. The base themes are necessary for developing the field in the lower-right quadrant, being more generic and transversal. The themes in the upperleft quadrant represent specialized themes. In the lower-left quadrant are the emerging themes [2]. 4.2 Content Analysis This section presents the main results of the analysis of the contributions of the articles in the final bibliographic portfolio classified by perspectives. The bibliographic classification of the portfolio considered two perspectives: (i) conceptual, theoretical, and model studies; (ii) empirical/applications studies. Table 3 presents the studies classified according to perspectives. Table 3. Classification of the resulting bibliographic portfolio. Classification
Authors
(i) conceptual, theoretical studies
[4, 5, 16, 18, 21, 30, 35, 38, 44, 45, 54, 56, 57, 60, 61]
(ii) empirical studies/applications
[3, 6, 8–12, 14, 15, 22, 24, 31–33, 36, 37, 39, 40, 47–49, 55, 59]
414
D. R. S. Ribeiro et al.
In the following subsections, each perspective was analyzed to highlight the recent advances in the area and trends observed in the studies from a scientific and practical point of view. Conceptual, Theoretical Studies [5] presented a concept for digitalized maintenance (DM), maps the conceptualized DM to maintenance problems in industries, and highlighted challenges that might be faced when realizing this concept. [45] presented an M4.0 model, disruptive maintenance based on new I4.0 technologies. Although the authors do not apply the model, they state that M4.0 is an undeniable trend. [30] describe M4.0 from the perspective of the challenges of the fourth industrial revolution and the economic, environmental, and social challenges of sustainable development. [44] reviewed some Machine Learning techniques for predictive maintenance to highlight the advantages of their application to improve situational intelligence, performance, and reliability. Based on a literature review, [38] developed a framework to assist in querying Integrated Electronic Technical Publications (IETP) in and out of the aviation maintenance hangar. [61] proposed integration of Lean tools and I4.0 technologies based on a three-step methodology, which included technological and industrial mapping, 25 points of synergy were identified. Maintenance, specifically TPM, was one of the areas addressed in this research. [35] analyzed and categorized 170 maintenance IPs to support IP selection targeting the anticipated effects of M4.0. To ensure dynamic, robust, and sustainable production systems through the implementation of M4.0. [18] reviewed applications of Digital Twins (DT) for maintenance. DT replicates a physical asset, process, or system used for control and decision-making [25]. As a result, the authors provide an agenda of future researches for the industrial sector. [16] presented a comprehensive overview of current predictive maintenance (PdM) issues, providing a deeper understanding of the dynamic maintenance paradigm. As a result, they found that its current development primarily involves remote status tracking of monitored equipment. [4] proposed a methodology for calculating overall equipment effectiveness (OEE) in the context of the I4.0 environment, based on Nakajim’s OEE indicator. The proposed methodology identifies weak points in the manufacturing process, which corrective measures can eliminate. [60] presented the evolution of intelligent decision support for maintenance practice from Condition-based Maintenance (CBM) then prognostic use to the M4.0 paradigm and the introduction of IoT and Cloud-enabled solutions. [56] reviewed the M4.0’s challenges, focusing on inspections. The authors also provide a conceptual framework for autonomous inspection and maintenance practices for linear assets using autonomous robots and data from different sources. [21] mapped out the vision of how maintenance could be integrated into this new change in industrial factories. The authors state that digitalization is the critical element for the future success of factories, as all networked machines must be controlled, and all digital data must be stored, processed, and analyzed in a meaningful way. [54] described a proactive approach to M4.0 and intelligent logistics. Then, the authors showed a view of the changes needed in the process area according to I4.0. To develop systems that support M4.0, [57]
Maintenance 4.0: A Literature Review and SWOT Analysis
415
mentioned that self-maintenance machines might be a better option. In this context, these authors discussed the concept, needs, implementation scenarios, and self-maintenance issues. Empirical Studies/Applications [59] examined the integration of I4.0 and Total Productive Maintenance (TPM) technologies through interviews in four large manufacturing companies (metal, automotive, electronics, and machinery). As a result, they provide a conceptual model of TPM based on the Diffusion of Innovation Theory (DIT). [36] proposed a cyclical process to support M4.0 implementation that includes six steps: benchmarking, defining clear objectives, setting strategic priority, planning key activities, lifting implementation, and follow-up. They analyzed the process in three industrial cases (discrete manufacturing company, process industry, and infrastructure and traffic services company). [55] performed a qualitative and quantitative investigation of the scientific literature to clarify the relationship among the different maintenance concepts, including M4.0. The authors conducted a case study involving eight manufacturing companies. [48] developed a reference architecture design for M4.0 system deployment in the era of I4.0. The case study was conducted at a company in the oil and gas sector. [47] presented a novel simulation model that enables estimating the lifetime benefits of an industrial asset when an M4.0 management system is utilized. [31] identified maturity dimensions for M4.0 of built assets that can be measured. The maturity dimensions are validated for university corporate facilities management organizations. [3] proposed an M4.0 web platform and its main KPIs: Availability, which relates to mean time between failures (MTBF) and mean time to repair (MTTR); Production Rate and Energy Consumption, OEE, and Overall Performance Effectiveness (OPE). [22] provided a framework to design a suitable 4.0 ecosystem for integrating maintenance activities with design and data collection solutions. To validate it, they conducted a case study based on an aircraft assembly line. [40] gave special attention to the open challenges regarding robustness, scalability, and security that have arisen through adopting the IoT concept in the industry, resulting in a scalable system for fleet monitoring and visualization. They validated through an M4.0 use case, leading to an open test and research platform for accelerated bearing life testing. [39] applied predictive analytics, big data processes, and tools to design CBM plans for train axle bearings to increase PdM intervals and train reliability. [37] discussed the usability and generic aspect of Lean applied to M4.0, where they are validated using data from three different asset management assessment projects in organizations with different types of production. [33] examined a diesel generator maintenance operation to investigate the potential benefits of M4.0 in proportion to human reliability aspects. For the first part of a series of papers, [11] developed a conceptual definition of M4.0 and its four underlying dimensions (data-driven decision making, human capital resources, internal integration, and external integration). In the second paper, [12] presented an empirically based research agenda that reflects the heterogeneity in industrial adoption and performance of M4.0. The methods used in both publications [11, 12] were focus groups and interviews with experts from different companies. Following these publications, [10] developed a psychometric instrument that measures the four dimensions of M4.0.
416
D. R. S. Ribeiro et al.
[9] presented an intelligent approach using unsupervised Machine Learning techniques for pre-processing and analyzing PdM to get qualified and structured data. To this end, they demonstrate its validity through a case study (an engine component line). [8] presented the application of Augmented Reality (AR) integrated with Computerized Maintenance Management Software to demonstrate the potential of this integration to optimize the maintenance procedure pipeline to increase the profitability and competitive advantage of an industrial company. [6] have proposed an M4.0 mathematical framework for power generators. The proposed approach is advanced through three main concepts: Markov Chains, Fuzzy Logic, and Maintenance Effort Degree. The authors use a case study to justify the effectiveness of the proposal. [32] developed a framework that presents the necessary resources and their connection when an organization wants to implement M4.0 and its outcomes. They conduct an exploratory case study that includes interviews at a digital rail maintenance development company and its primary customer, the traffic agency. [24] investigated existing machine criticality assessment and identified components of the criticality assessment tool to increase productivity. To do this, they adopted a multiple case study with multinational manufacturing companies. To implement M4.0, [15] built a prototype of the cyber-physical system (CPS) on a cell manufacturing system consisting of several additional manufacturing machines and a 3D Scanner. The method used to build the prototype was Rapid Application Development (RAD). [49] described a floating data concept for an industrial internet application for efficient road maintenance. This concept is demonstrated in a case study of a stretch of road. The resulting improvement in winter maintenance will promote reliable, security, and sustainable transport of goods and people. [14] provided a framework for an M4.0 decision support system. Applications using big data analytics and a case study specific to the electric utility industry are detailed. 4.3 SWOT Analysis Table 4 highlights the main results found in the studied publications by a SWOT (strengths, weaknesses, opportunities, and threats) analysis of M4.0. Many factors influence (facilitate or inhibit) the implementation of M4.0, [11] such as culture, leadership, technology investments, and IT security. However, implementation can be elevated by using strengths, opportunities, and potential obstacles [36]. Based on [62], the following classification was adopted for the SWOT dimensions: benefits obtained by implementing M4.0 (Strengths); potential benefits that can be achieved by implementing M4.0 (Opportunities); difficulties and/or barriers in the implementation of M4.0 (Weaknesses); potential failures/problems caused by the implementation of M4.0 (Threats). The benefits perceived in the strengths section, in general, are already known and studied in articles in the literature. The M4.0’s benefits are related to economic, social, and environmental challenges. In the opportunities dimension, future research can focus on analyzing the potential advantages obtained in some areas, such as circular economy and sustainability, and employee engagement. Moreover, in the weaknesses and threats dimensions, barriers, difficulties, and problems encountered in implementing M4.0. These dimensions open a range of possibilities for future researches.
Maintenance 4.0: A Literature Review and SWOT Analysis
417
Table 4. SWOT analysis results. Strengths
Weakness
- Improves economic efficiency [3, 5, 30, 44, 45]; maximizes production profits and minimize all costs and losses, including assets [16, 21, 24, 56] - Reduces maintenance time and inventory [30, 44, 47, 59, 61] - Extended equipment life, improvement in availability [3, 5, 45, 47, 48, 56] - Adjusts real-time data of machine status and predict future failures in addition to enabling data-driven decision making [3, 10, 15, 16, 21, 31, 44, 54, 59–61] - Facilitates knowledge management of maintenance activities and past events [12, 31, 56, 59] - Decreases spare parts and lubricant usage [30] - Improves environmental and worker safety [5, 30, 45] - Minimizes end of life [30] - Provides improvements to accident mitigation [5] - Allows experimentability, i.e., can be tried without full implementation [59]
- Complexity, complex or challenging innovation to use and understand [9, 36, 45, 59] - Difficulties in demonstrating financial benefits (observability) [59] - The need to incorporate maintenance professionals specialized in these new areas of knowledge [5, 31, 45, 59] -The challenging establishment of the level that this state variable of equipment should have to trigger a maintenance intervention [45] - A selection of corrective/reactive, preventive, and condition-based maintenance may not be sufficient for the development of M4.0 [36] - Requires researches to have a broad impact [21, 33] - Cost of adopting new ICTs [59] and necessity of high capital investment [5] - ICTs and related technologies have a short economic life and may become obsolete and replaced by newer generation systems [31] - Difficulties in assessing data quality within a reasonable period due to high dimensionality and short time of data [5, 9] - Required technology can be prohibitive for medium and small companies [5]
Opportunities
Threats
- Redesign of maintenance policies and role of workers to support technological innovations [45, 59] - Develop new technologies that affect maintenance engineering. It presents a unique opportunity to make a disruptive evolution of M4.0 [45] - Contribute to a circular and sustainable economy [30, 45] - Develop new organizational capabilities [21, 31] - Increase employee engagement, enabling a mix of top-down and bottom-up approaches [31, 36] - Build cost-benefit integrated models that include impact across asset management [16] - Open challenges for mining of data streams provided by sensors [60] - Capture and store data streams from production machinery and the audit trails of decision-making within Semantic technologies [60]
- Maintenance organizations that are unable to adapt can drag the entire company to its end [45] - Unplanned deployments can result in high costs to the organization [61] - Even with preventive and periodic maintenance schedules, machine failures are not completely under control [61] - Do not capture IoT reliability for M4.0 applications, various concepts (such as cybersecurity, reliability, resiliency, among others.) must be integrated into M4.0 modeling [16] - Regulation and standards liability issues, data ownership, intellectual property, and safety and security [5] - Implementing M4.0 without an effective predictive horizon of the associated PdM and effective frequency of opportunistic maintenance intervals does not guarantee the gain of its lifetime benefits [47]
418
D. R. S. Ribeiro et al.
5 Conclusion This paper performed a literature review and SWOT analysis about M4.0. The bibliometric analysis showed the importance of the subject today and the interest in the theme by quality journals and conferences. The content analysis allowed us to organize the thematic addressed in guiding axes of the research, according to the similarities and trends found. Moreover, the SWOT analysis demonstrated the strengths that allow the achievement of the benefits provided by the implementation of M4.0 and consequently improve the principal operational indicators: security, reliability, availability, and cost [3, 30, 45]. In contrast, the weaknesses and threats identified may challenge some organizations. The implementation of M4.0 involves many other aspects and impacts various sectors involved in maintenance, logistics [39, 49, 54], occupational health, safety and environment, design (the use of intelligent components can lead to different reliability allocation solutions) [18], high management [16]. This study shows that there is still much to be done to provide the knowledge needed in terms of M4.0 implementation. In this sense, this research presents itself as a helpful starting point for researchers, practitioners, and stakeholders when addressing this topic. However, the authors point out that there are limitations. The limitations are as follows: (i) the keywords employed for the literature search phase may limit the results found and included in this study, (ii) only two databases were consulted, (iii) the literature sample contains only journals articles published in English, i.e., research results or practices in other languages were not reported in this study, and (iv) content analysis may generate interpretation bias. The following research agenda is suggested: it is recommended to use other keywords related to the theme that were not used in this research to cover more publications. In addition, the opportunities noted in the SWOT analysis can be new investigations, as we have mentioned. New research in these contexts can support the achievement of these benefits. Also, we recommend investigating how to mitigate the threats presented in the SWOT analysis. Acknowledgments. The authors acknowledge the financial support of Coordination for the Improvement of Higher Education Personnel (Brazil) through the CAPES-DS scholarship – under reference numbers 88882.438684/2019–01 and 88887.647046/2021–00.
References 1. Acerbi, F., Taisch, M.: A literature review on circular economy adoption in the manufacturing sector. J. Clean. Prod. 273, 123086 (2020). https://doi.org/10.1016/j.jclepro.2020.123086 2. Agostino, Í.R.S., Ristow, C., Frazzon, E.M., Taboada Rodriguez, C.M.: Perspectives on the application of internet of things in logistics. In: Freitag, M., Haasis, H.-D., Kotzab, H., Pannek, J. (eds.) LDIC 2020. LNL, pp. 387–397. Springer, Cham (2020). https://doi.org/10.1007/9783-030-44783-0_37 3. Aksa, K., Aitouche, S., Bentoumi, H., Sersa, I.: Developing a web platform for the management of the predictive maintenance in smart factories. Wirel. Pers. Commun. 119(2), 1469–1497 (2021). https://doi.org/10.1007/s11277-021-08290-w
Maintenance 4.0: A Literature Review and SWOT Analysis
419
4. Aleš, Z., Pavl˚u, J., Legát, V., Mošna, F., Jurˇca, V.: Methodology of overall equipment effectiveness calculation in the context of Industry 4.0 environment. Ekspolatacja i Niezawodn Maint Reliab 21, 411–418 (2019). https://doi.org/10.17531/ein.2019.3.7 5. Algabroun, H., Bokrantz, J., Al-Najjar, B., Skoogh, A.: Development of digitalised maintenance – a concept. J. Qual. Maint. Eng. (2020). https://doi.org/10.1108/JQME-04-20190039 6. Alvarez-Alvarado, M.S., Jayaweera, D.: Operational risk assessment with smart maintenance of power generators. Int. J. Electr. Power Energy Syst. 117, 105671 (2020). https://doi.org/ 10.1016/j.ijepes.2019.105671 7. Aria, M., Cuccurullo, C.: Bibliometrix : An R-tool for comprehensive science mapping analysis. J. Inform. 11, 959–975 (2017). https://doi.org/10.1016/j.joi.2017.08.007 8. Ariansyah, D., Rosa, F., Colombo, G.: Smart maintenance: a wearable augmented reality application integrated with CMMS to minimize unscheduled downtime. Comput. Aided Des. Appl. 17, 740–751 (2019). https://doi.org/10.14733/cadaps.2020.740-751 9. Bekar, E.T., Nyqvist, P., Skoogh, A.: An intelligent approach for data pre-processing and analysis in predictive maintenance with an industrial case study. Adv. Mech. Eng. 12, 168781402091920 (2020). https://doi.org/10.1177/1687814020919207 10. Bokrantz, J., Skoogh, A., Berlin, C., Stahre, J.: Smart maintenance: instrument development, content validation and an empirical pilot. Int. J. Oper. Prod. Manag. 40, 481–506 (2020). https://doi.org/10.1108/IJOPM-11-2019-0746 11. Bokrantz, J., Skoogh, A., Berlin, C., Wuest, T., Stahre, J.: Smart maintenance: an empirically grounded conceptualization. Int. J. Prod. Econ. 223, 107534 (2020). https://doi.org/10.1016/ j.ijpe.2019.107534 12. Bokrantz, J., Skoogh, A., Berlin, C., Wuest, T., Stahre, J.: Smart maintenance: a research agenda for industrial maintenance management. Int. J. Prod. Econ. 224, 107547 (2020). https://doi.org/10.1016/j.ijpe.2019.107547 13. Bonfante, M.C., Raspini, J.P., Fernandes, I.B., Fernandes, S., Campos, L.M.S., Alarcon, O.E.: Achieving sustainable development goals in rare earth magnets production: A review on state of the art and SWOT analysis. Renew. Sustain. Energy Rev. 137, 110616 (2021). https://doi. org/10.1016/j.rser.2020.110616 14. Bumblauskas, D., Gemmill, D., Igou, A., Anzengruber, J.: Smart Maintenance Decision Support Systems (SMDSS) based on corporate big data analytics. Expert Syst. Appl. 90, 303–317 (2017). https://doi.org/10.1016/j.eswa.2017.08.025 15. Cahyati, S., Achdianto, S: A prototyping of additive manufacturing cell in cyber physical system for maintenance 4.0 preparation. Int. J. Adv. Sci .Technol. 29, 575–584 (2020) 16. Compare, M., Baraldi, P., Zio, E.: Challenges to IoT-enabled predictive maintenance for Industry 4.0. IEEE Internet Things J. 7, 4585–4597 (2020). https://doi.org/10.1109/JIOT. 2019.2957029 17. Dhir, A., Talwar, S., Kaur, P., Malibari, A.: Food waste in hospitality and food services: a systematic literature review and framework development approach. J. Clean. Prod. 270, 122861 (2020). https://doi.org/10.1016/j.jclepro.2020.122861 18. Errandonea, I., Beltrán, S., Arrizabalaga, S.: Digital Twin for maintenance: a literature review. Comput. Ind. 123, 103316 (2020). https://doi.org/10.1016/j.compind.2020.103316 19. Ferenhof, H.A., Vignochi, L., Selig, P.M., Lezana, Á.G.R., Campos, L.M.S.: Environmental management systems in small and medium-sized enterprises: an analysis and systematic review. J. Clean. Prod. 74, 44–53 (2014). https://doi.org/10.1016/j.jclepro.2014.03.027 20. Fink, A.: Conducting Research Literature Reviews: From the Internet to Paper. Sage Publications (2019) 21. Fusko, M., Rakyta, M., Krajcovic, M., Dulina, L., Gaso, M., Grznar, P.: Basics of designing maintenance processes in Industry 4.0. MM Sci. J. 2018, 2252–2259 (2018). https://doi.org/ 10.17973/MMSJ.2018_03_2017104
420
D. R. S. Ribeiro et al.
22. Giacotto, A, Costa Marques, H., Pereira Barreto, E.A., Martinetti, A.: The need for ecosystem 4.0 to support maintenance 4.0: an aviation assembly line case. Appl. Sci. 11, 3333 (2021). https://doi.org/10.3390/app11083333 23. Gimenez, I.N.K., Morosini Frazzon, E., Uriona, M.: Big data and business analytics in the supply chain: a review of the literature. IEEE Lat. Am. Trans. 13, 3382–3391 (2015). https:// doi.org/10.1109/TLA.2015.7387245 24. Gopalakrishnan, M., Skoogh, A., Salonen, A., Asp, M.: Machine criticality assessment for productivity improvement. Int. J. Product Perform. Manag. 68, 858–878 (2019). https://doi. org/10.1108/IJPPM-03-2018-0091 25. Grieves, M., Vickers, J.: Digital Twin: mitigating unpredictable, undesirable emergent behavior in complex systems. In: Kahlen, F.-J., Flumerfelt, S., Alves, A. (eds.) Transdisciplinary Perspectives on Complex Systems, pp. 85–113. Springer, Cham (2017). https://doi.org/10. 1007/978-3-319-38756-7_4 26. Helms, M.M., Nixon, J.: Exploring SWOT analysis – where are we now? J. Strateg. Manag. 3, 215–251 (2010). https://doi.org/10.1108/17554251011064837 27. Hofmann, E., Rüsch, M.: Industry 4.0 and the current status as well as future prospects on logistics. Comput. Ind. 89, 23–34 (2017). https://doi.org/10.1016/j.compind.2017.04.002 28. Huang, J.-J., Tzeng, G.-H., Ong, C.-S.: Multidimensional data in multidimensional scaling using the analytic network process. Pattern Recognit. Lett. 26, 755–767 (2005). https://doi. org/10.1016/j.patrec.2004.09.027 29. Humphrey, A.: SWOT Analysis for Management Consulting, 1st edn. SRI Alumni Newsletter (2005) 30. Jasiulewicz-Kaczmarek, M., Legutko, S., Kluk, P.: Maintenance 4.0 technologies–new opportunities for sustainability driven maintenance. Manag. Prod. Eng. Rev. 11 (2020). https://doi. org/10.24425/mper.2020.133730 31. Johannes, K., Theodorus Voordijk, J., Marias Adriaanse, A., Aranda-Mena, G.: Identifying maturity dimensions for smart maintenance management of constructed assets: a multiple case study. J. Constr. Eng. Manag. 147, 05021007 (2021). https://doi.org/10.1061/(ASCE)CO. 1943-7862.0002112 32. Johansson, N., Roth, E., Reim, W.: Smart and sustainable emaintenance: capabilities for digitalization of maintenance. Sustainability 11, 3553 (2019). https://doi.org/10.3390/su1113 3553 33. Kandemir, C., Celik, M.: A human reliability assessment of marine auxiliary machinery maintenance operations under ship PMS and maintenance 4.0 concepts. Cogn. Technol. Work 22(3), 473–487 (2019). https://doi.org/10.1007/s10111-019-00590-3 34. Kolberg, D.: Knobloch J, Zühlke D: Towards a lean automation interface for workstations. Int. J. Prod. Res. 55, 2845–2856 (2017). https://doi.org/10.1080/00207543.2016.1223384 35. Lundgren, C., Bokrantz, J.: Skoogh A: Performance indicators for measuring the effects of Smart Maintenance. Int. J. Product Perform Manag. 70, 1291–1316 (2021). https://doi.org/ 10.1108/IJPPM-03-2019-0129 36. Lundgren, C., Bokrantz, J., Skoogh, A.: A strategy development process for Smart Maintenance implementation. J. Manuf. Technol. Manag. 32, 142–166 (2021). https://doi.org/10. 1108/JMTM-06-2020-0222 37. Maier, H.T., Schmiedbauer, O., Biedermann, H.: Validation of a lean smart maintenance maturity model. The. Glas. 14, 296–302 (2020). https://doi.org/10.31803/tg-20200706131623 38. Marques, H.C., Giacotto, A., Scussiatto, CE, Abrahão FTM: Semantic voice search in IETP: filling the gap for maintenance 4.0. J. Qual. Maint. Eng. 27, 500–516 (2021). https://doi.org/ 10.1108/JQME-05-2020-0038 39. Márquez, A.C., de la Fuente Carmona, A., Marcos, J.A., Navarro, J.: Designing CBM plans, based on predictive analytics and big data tools, for train wheel bearings. Comput. Ind. 122, 103292 (2020). https://doi.org/10.1016/j.compind.2020.103292
Maintenance 4.0: A Literature Review and SWOT Analysis
421
40. Moens, P., et al.: Scalable fleet monitoring and visualization for smart machine maintenance and industrial IoT applications. Sensors 20, 4308 (2020). https://doi.org/10.3390/s20154308 41. Moher, D., Liberati, A., Tetzlaff, J., Altman, D.G.: Preferred reporting items for systematic reviews and meta-analyses: the PRISMA statement. PLoS Med 6, e1000097 (2009). https:// doi.org/10.1371/journal.pmed.1000097 42. Moral-Muñoz, J.A., Herrera-Viedma, E., Santisteban-Espejo, A., Cobo, M.J.: Software tools for conducting bibliometric analysis in science: an up-to-date review. Prof. Inf. 29 (2020). https://doi.org/10.3145/epi.2020.ene.03 43. Munaro, M.R., Tavares, S.F., Bragança, L.: Towards circular and more sustainable buildings: a systematic literature review on the circular economy in the built environment. J. Clean. Prod. 260, 121134 (2020). https://doi.org/10.1016/j.jclepro.2020.121134 44. Nacchia, M., Fruggiero, F., Lambiase, A., Bruton, K.: A Systematic mapping of the advancing use of machine learning techniques for predictive maintenance in the manufacturing sector. Appl. Sci. 11, 2546 (2021). https://doi.org/10.3390/app11062546 45. Navas, M.A., Sancho, C., Carpio, J.: Disruptive maintenance engineering 4.0. Int. J. Qual. Reliab. Manag. 37, 853–871 (2020). https://doi.org/10.1108/IJQRM-09-2019-0304 46. Nissoul H, El Harrouti T, Serrou D, Abouabdellah A: Impact of maintenance 4.0 on the performance of the production function: application to the automotive industry in Morocco. In: 2020 5th International Conference on Logistics Operations Management (GOL). IEEE, pp. 1–5 (2020). https://doi.org/10.1109/GOL49479.2020.9314737 47. Nordal, H., El-Thalji, I.: Lifetime benefit analysis of intelligent maintenance: simulation modeling approach and industrial case study. Appl. Sci. 11, 3487 (2021). https://doi.org/10. 3390/app11083487 48. Nordal, H., El-Thalji, I.: Modeling a predictive maintenance management architecture to meet industry 4.0 requirements: a case study. Syst. Eng. 24, 34–50 (2021). https://doi.org/10.1002/ sys.21565 49. Odelius, J., Famurewa, S.M., Forslöf, L., Casselgren, J., Konttaniemi, H.: Industrial internet applications for efficient road winter maintenance. J. Qual. Maint. Eng. 23, 355–367 (2017). https://doi.org/10.1108/JQME-11-2016-0071 50. Olivieri, M., Andreoli, M., Vergamini, D., Bartolini, F.: Innovative contract solutions for the provision of agri-environmental climatic public goods: a literature review. Sustainability 13, 6936 (2021). https://doi.org/10.3390/su13126936 51. Paes, L.A.B., Bezerra, B.S., Deus, R.M., Jugend, D., Battistelle, R.A.G.: Organic solid waste management in a circular economy perspective – a systematic review and SWOT analysis. J. Clean. Prod. 239, 118086 (2019). https://doi.org/10.1016/j.jclepro.2019.118086 52. Prell, C.: Social Network Analysis: History. Sage, Theory and Methodology (2012) 53. R Core Team, R: A Language and Environment for Statistical Computing (2018) ˇ Zrnic, N.: Proactive approach to smart 54. Rakyta, M., Fusko, M., Hercko, J., Závodská, L, maintenance and logistics as a auxiliary and service processes in a company. Istraz i Proj za privredu 14, 433–442 (2016). https://doi.org/10.5937/jaes14-11664 55. Roda, I., Macchi, M.: Maintenance concepts evolution: a comparative review towards advanced maintenance conceptualization. Comput. Ind. 133, 103531 (2021). https://doi.org/ 10.1016/j.compind.2021.103531 56. Seneviratne, D., Ciani, L., Catelani, M., Galar, D.: Smart maintenance and inspection of linear assets: an industry 4.0 approach. ACTA IMEKO 7, 50 (2018). https://doi.org/10.21014/acta_i meko.v7i1.519 57. Singh, S., Galar, D., Baglee, D., Björling, S.-E.: Self-maintenance techniques: a smart approach towards self-maintenance system. Int. J. Syst. Assur. Eng. Manag. 5(1), 75–83 (2013). https://doi.org/10.1007/s13198-013-0200-7 58. Stock, T., Seliger, G.: Opportunities of sustainable manufacturing in Industry 4.0. Procedia CIRP 40, 536–541 (2016). https://doi.org/10.1016/j.procir.2016.01.129
422
D. R. S. Ribeiro et al.
59. Tortorella, G.L., Fogliatto, F.S., Cauchick-Miguel, P.A., Kurnia, S., Jurburg, D.: Integration of Industry 4.0 technologies into total productive maintenance practices. Int. J. Prod. Econ. 240, 108224 (2021). https://doi.org/10.1016/j.ijpe.2021.108224 60. Turner, C.J., Emmanouilidis, C., Tomiyama, T., Tiwari, A., Roy, R.: Intelligent decision support for maintenance: an overview and future trends. Int. J. Comput. Integr. Manuf. 32, 936–959 (2019). https://doi.org/10.1080/0951192X.2019.1667033 61. Valamede, L.S., Akkari, A.C.S.: Lean 4.0: a new holistic approach for the integration of lean manufacturing tools and digital technologies. Int. J. Math. Eng. Manag. Sci. 5, 851–868 (2020). https://doi.org/10.33889/IJMEMS.2020.5.5.066 62. Wijesooriya, N., Brambilla, A.: Bridging biophilic design and environmentally sustainable design: a critical review. J. Clean. Prod. 283, 124591 (2021). https://doi.org/10.1016/j.jclepro. 2020.124591 63. Zemigala, M.: Tendencies in research on sustainable development in management sciences. J. Clean. Prod. 218, 796–809 (2019). https://doi.org/10.1016/j.jclepro.2019.02.009 64. Zonta, T., da Costa, C.A., da Rosa Righi, R., de Lima, M.J., da Trindade, E.S., Li, G.P.: Predictive maintenance in the Industry 4.0: a systematic literature review. Comput. Ind. Eng. 150, 106889 (2020). https://doi.org/10.1016/j.cie.2020.106889
Using Supervised Learning to Predict Process Steps for Process Planning of Third-Party Logistics Marius Veigt1(B) , Lennart Steinbacher1 , and Michael Freitag1,2 1 BIBA - Bremer Institut für Produktion und Logistik GmbH, University of Bremen,
Hochschulring 20, 28359 Bremen, Germany {vei,stb,fre}@biba.uni-bremen.de 2 Faculty of Production Engineering, University of Bremen, Badgasteiner Straße 1, 28359 Bremen, Germany
Abstract. There is intense competitive pressure in the third-party logistics industry. As a result, logistics providers have to respond to tenders quickly and with convincing concepts. This article shows how logistics process planning in tender management can be accelerated using methods of supervised learning. Under the premise that similar processes from past projects can be transferred and adapted to a new project, an assistance system suggests appropriate process steps in the form of MTM (methods-time measurement) codes to the planner using N-Gram analysis and a decision tree. This procedure accelerates the process planning and can lead to an increase in the quality of the planned logistics processes. Keywords: Contract logistics · Third-party logistics · Process planning · Supervised learning · n-gram · Decision tree
1 Introduction 1.1 Motivation The outsourcing of operations outside the company’s core competencies is common practice [1]. This trend has continued to the present day, particularly in third-party logistics, evidenced by the constant growth in the market volume of the third-party logistics industry [2]. Nevertheless, the contract logistics industry is also under pressure to innovate, and the players have to adapt to new technical possibilities to remain competitive [3]. The high individualization of the services is characteristic of third-party logistics [4]. The logistics provider must consider the specific conditions of the customer’s location as well as the product- and customer-specific requirements. Hence, the logistics provider must investigate and analyze the tender and the not-described requirements to plan and calculate a proper logistics concept for the customer. Experienced and highly qualified employees are necessary for this task [5]. However, due to fluctuating demand for logistics services, it is unavoidable that less experienced planners have to take on this task as © The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 M. Freitag et al. (Eds.): LDIC 2022, LNLO, pp. 423–434, 2022. https://doi.org/10.1007/978-3-031-05359-7_34
424
M. Veigt et al.
well. This increases the required time and reduces the quality of the logistics concept. Thus, it reduces the chances of winning the tender. Spiegel et al. have achieved initial successes with an approach of modularization and standardization of logistics processes [6]. Testing and evaluating this approach revealed that even extensive process modules require adjustments for specific tenders. These adjustments imply that standardization is not the only approach for process optimization in the planning and calculation process. Recent research has analyzed the planning and calculation process in the tender management of third-party logistics and identified the demand and potential for support by an artificial intelligence (AI) based assistance system [7]. The assistance system should support the planners in accelerating the planning process and decrease individual quality fluctuations of the planned concept. This paper presents initial results on how AI in the form of supervised learning can support process planning in third-party logistics. 1.2 Problem Description The planning process of a logistics system in contract logistics is comparable to the procedure in factory planning [7]. Figure 1 illustrates the planning process structure schematically according to VDI Guideline 5200, which is used to classify the present research work. Veigt et al. identified one potential for AI support in phase 2: extracting data from the tender; and three potentials in phase 3: process planning, creating an ideal layout, and creating variants for a real layout [7]. These results reflect the findings of other research work in this domain. For example, Teucke et al. investigated how assistance systems can support planning tasks in logistics. As a result, Teucke et al. identified: data gathering and preparation (phase 2), generation of solution variants, early evaluation of solution variants (phase 3), and implementation support of the selected planning variant (phase 4–6) [8]. This contribution examines the support provided for process planning on phase 3.
Phase 1 setting of objectives
Phase 2 establishment of the product basis
Phase 3 concept planning
Phase 4 detailed planning
project management
Phase 5 preparation for realization
Phase 6 monitoring of realization
Phase 7 ramp-up support
project closeout
Fig. 1. Phase model of the factory planning process according to VDI Guideline 5200 [9]
The core competence of the planner is to design an efficient logistics process from the requirements formulated in phase 2. A piece of basic but unknown information is the processing time a planner has to calculate for a specific process step. The correctly planned process time is an essential information. On the one hand, planning too-short process times means that not all costs are sufficiently considered. This underestimation would lead to a loss in the project. On the other hand, planning too-long process times would lead to an expensive offer that is not competitive.
Using Supervised Learning to Predict Process Steps
425
To plan process times accurately, logistics planners use the methods-time measurement (MTM). MTM is used to plan manual work processes. When using MTM, all movements performed by humans are referred to as certain basic movements (e.g., reaching, grasping) supplemented by motion elements (e.g., walking, bending, visual control). Work times were empirically determined for these motion elements. In addition, each motion element has a uniform code so that the planners can identify the motion elements based on the codes and combine them into processes [10]. However, this method is very time-consuming and requires experiential knowledge. In this phase of planning, there is a demand for an AI application, which generates suggestions of an MTM code composition on the data basis of previous, similar planning. In this way, inexperienced planners could benefit from the knowledge gained from previous planning, and the work of all planners could be accelerated [7]. 1.3 Research Approach The research hypothesis is that analogies to a current planning process can be found in the data set of past planning processes using supervised learning. In other words: A supervised learning algorithm can predict the subsequent process step or a sequence of process steps appropriate to a new process to be planned by analyzing a data set of in the past planned processes. To identify the process steps, the unique MTM codes will be used. Two questions arise: 1. Which is the first process step in form of an MTM code in a process? 2. Which process step in form of an MTM code is the next one? The question can be posed either as a sequencing problem or as a classification problem. In this contribution, both approaches will be applied and compared. The sequencing problem is known from writing messages on a smartphone. Depending on the letters or words typed, the messenger completes the word or makes suggestions for the following word. Over time, the messenger learns the characteristics of the user and adapts its suggestions. In classification, a data set is assigned to a group of identical or similar data sets based on features. The first question is which features should be taken into account for the classification. Then, based on these features, a decision has to be made as to which class a specific data set belongs. This class represents the MTM code searched for with the highest probability. Both approaches are tested using an industrial data set, and the results are presented and discussed in the following.
2 Methods 2.1 Data Basis The planning data of 13 third-party logistics projects are available for the research work. The data set consists of six different customers from the automotive industry and the
426
M. Veigt et al.
construction industry. The services offered in the available data can be divided into the main processes of goods receipt, storage, retrieval, picking, sequencing, repacking, pre-assembly, goods dispatch, as well as transport, empties, and unique processes. The 13 projects involved the description of in total 440 processes. A process refers to the handling of a product or a load carrier in one main process. For example, the reception of a product represents a process. Putting the same product into storage represents a new process. The reception of another product represents a new process, and so on. The processes contain 260 different MTM codes and 13,637 MTM codes in total. The shortest process contains just one MTM code, the longest 155 MTM codes. The average is about 30 MTM codes per process. Each project was planned in a spreadsheet program as a separate file, so the data had to be merged into one data set during pre-processing. In this data set, each row represents a process step with the associated MTM code. In other columns, the available features customer, location, industry, main process title, process description, used equipment, used load carrier, and assigned employee group are added. In addition, individual notations were standardized, e.g., “goods receipt”, “GR” and “incoming goods” were uniformly written as “goods receipt”. It must be taken into account that the features used equipment, used load carrier, and assigned employee group are sometimes empty. In some cases, this information was not known at the time of planning (e.g., type of load carrier); in other cases, it is not relevant for all process steps (e.g., not every process step requires equipment such as a forklift or a scanner). 2.2 Sequence Analysis with N-Gram Analysis The N-Gram analysis is a widely used method to learn a sequence [11, 12]. N-grams are the result of splitting a text into fragments. First, the text is split, and N sequential fragments are combined as N-grams. The fragments can be letters, words, and similar. Subsequently, the probability with which another gram follows an N-gram is calculated based on existing texts. Consequently, N-gram analysis is used to answer how probably a letter or word fragment will be followed by a particular letter or word [13]. In this research, one MTM code was used as a gram. For the available data set, the probability of which MTM code or sequence of several MTM codes follows a particular MTM code was calculated. On the one hand, the entire data set was used to calculate the probabilities. On the other hand, the data set was limited to main processes. For example, to calculate the probabilities of an MTM code successor in the main process goods receipt, the data set was limited to goods receipt main processes. This limitation is based on the hypothesis that the probability of suggesting the correct successor MTM code can be increased by limiting the data set to similar processes. 2.3 Classification with Decision Trees Decision trees (DT) are a widely used method for classification, and one of the most influential data mining algorithms in the research community identified by the IEEE.
Using Supervised Learning to Predict Process Steps
427
International Conference on Data Mining [14]. Due to DT often generating good results [15], they are a good starting point for searching for a suitable classification method. A DT represents decision rules, and it is used to classify data sets based on specific features (independent variables) [16]. In order to identify the relevant features in the present case, workshops with experts were conducted. The experts determined which factors influence the selection of MTM codes during the process planning. Table 1 lists the results. The comparison with the features available for supervised learning in the data set, see Table 1, reveals a part of the knowledge that is implicitly held by the planners but not available for the application of supervised learning. The features main process, industry, assigned employee group, and the previous MTM codes were used as additional features to compensate for the missing features. In addition, the categorical textual features were converted into numeric values. A simple dictionary was used to number the categories from 0 to N; see also Table 1. The features are nominally scaled. Table 1. Used and missed features for the decision tree
428
M. Veigt et al.
The subsequent MTM code was used as the label (dependent variable), which the DT should predict. A total of 260 different MTM codes were used. Figure 2 displays the distribution of the dependent variable. The MTM code “KA” was used most frequently (approx. 10.5% of the cases). In contrast, the MTM code “PBBB” was used only once (approx. 0.007% of the cases).
Fig. 2. Distribution of the dependent variable (label) in the complete data set.
The supervised learning approach is to classify an input vector shown in Table 1 column “converted features” to predict the following process step in the form of the MTM code. Thus, the planners’ experiential knowledge is to be transformed into a formal rule. In this case, it is essential to note that not all vectors are entirely filled with data. A DT was selected as the supervised learning method due to its ability to deal with missing data and expected good results [15]. In addition, DT are used to represent decision rules. Especially the ID3 algorithm is suitable to create a DT for the application with categorical data on a nominal scale [17]. Consequently, DT are a good fit for the presented task. For training and testing the DT, the ID3 algorithm in the software RapidMiner version 9.10 was used. The DT was generated and evaluated with a minimum size of split of four and a minimum leaf size of two. The small leaf size of two can lead to overfitting of the DT. However, the question of how many times a planner should create a process step sequence before the supervised learning method should learn this sequence led to the following result. A process step sequence created only once may be highly customized or faulty. As soon as a process step sequence has been created twice, the supervised learning method should learn it and suggest it the third time. To take this result into account, the leaf size was set to two. Cross-validation with ten folds was conducted to prevent overfitting of the DT.
3 Results 3.1 Results of N-Gram-Sequence Analysis Firstly, the question of which MTM code should be the first in a process is considered. The data set contains 440 processes. 42 different MTM codes were used as the first
Using Supervised Learning to Predict Process Steps
429
process step. Table 2 shows the probability of the MTM codes most often used in the first process step. Thus, the correct MTM code is suggested with a probability of 19% if the most frequently used MTM code IALW is always suggested. Table 2. Frequency of the most used MTM codes in the entire process Probability
MTM code
Description
19%
IALW
Reading information per word
18%
KA
Walking per Meter
16%
SFISF
Driving internally per meter with forklift truck (stable / empty)
If the analysis is limited to a main process (e.g., goods receipt), the frequency of the MTM codes used in the first position changes. The example of “goods receipt” demonstrates this. 65 goods receipt processes were analyzed. In this example, 12 MTM codes were used in the first process step. Table 3 shows the probability of the most frequently used MTM codes as the first in the goods receipt process. The probability of suggesting the correct MTM code increases to 27%. Consequently, the probability increases by eight percentage points compared to the result shown in Table 2. This result supports the hypothesis that a limitation of the data analysis to similar main processes is appropriate. Table 3. Frequency of the most used MTM codes in the goods receipt processes Probability
MTM code
Description
27%
KA
Walking per meter
14%
AZA
Getting on and off the driver’s seat
12%
SFISF
Driving internally per meter with forklift truck (stable/empty)
In order to predict the subsequent MTM code, the N-gram analysis is performed with N = 1 to N = 5 for the entire data set and the limitation to the goods receipt processes. The entire data set consists of 260 different MTM codes. Thus, the probability of correctly predicting an MTM code is about 0.4% if no analysis is conducted. The goods receipt processes contain 104 different MTM codes. This reduced quantity results in a probability of approx. 1% of suggesting the correct MTM code without any analysis. Table 4 presents the results of the N-Gram analysis. For each MTM code or MTM code sequence, the most probable subsequent MTM code was determined, and a weighted average value was calculated from these probabilities.
430
M. Veigt et al.
Table 4. Average weighted probability of correctly predicting the subsequent process step in the form of an MTM code. Data set
without N-Gram-Analysis
1-g
2-g
3-g
4-g
5-g
Entire processes
0.4%
17%
52%
76%
85%
90%
Goods receipt processes
1.0%
42%
75%
88%
96%
99%
Table 4 shows that the probability of predicting the correct subsequent MTM code increases as N increases. For example, for a sequence of five MTM codes (5-g), the correct subsequent MTM code is predicted in 90% of the cases. If only goods receipt processes are considered, the probability of predicting the correct subsequent MTM code in a goods receipt process can be increased to 99% (5-g). This result confirms the hypothesis that the N-gram analysis should be limited to similar processes. 3.2 Results of Decision Tree Classification The classification with DT was applied to the entire data set. It must be considered that only 440 samples were available for the question regarding the first MTM code. It is because the first MTM code occurs only once in each of the 440 processes. For the question about the subsequent MTM code, 13,197 samples were available. Based on the features shown in Table 1, the DT predicts the first MTM code in 70% of the cases correctly (accuracy). The DT received one to five previous process steps in the form of MTM codes as additional features. Table 5 shows the prediction results of the DT depending on the number of previous process steps. Similar to the N-Gram analysis, the probability of correctly predicting the subsequent MTM code increases with the number of predecessors. Table 5. Results of correctly predicting the subsequent process step in the form of an MTM code using a DT. Evaluated is the accuracy of the prediction.
Entire processes
One Predecessor
Two Predecessor
Three Predecessor
Four Predecessor
Five Predecessor
51.3%
70.8%
74.3%
76.3%
76.5%
Discussion Figure 3 illustrates the compared results of the N-gram analysis and the DT. For the prediction of the first process steps, 440 samples were available. For the prediction of the subsequent process steps, more than 13,000 samples were available. The N-gram analysis indicates probabilities of less than 30% for the first process step. It implies that there is a risk of over 70% of selecting the wrong MTM code. A wrong selection would lead to a “wrong course setting” because the prediction of subsequent
Using Supervised Learning to Predict Process Steps
431
Probability for the correct following MTM code
MTM codes would base on the wrong first MTM code. The decision tree predicts the first process step in 70% of the cases correctly. This value is also too low for an automated determination of the first process step, particularly under the aspect of “wrong course setting”. However, both methods offer the possibility to suggest a prioritized suggestion of probable MTM codes to the planner, as shown in Fig. 4. The planner should be responsible for the final selection and confirmation of the first MTM code. Nevertheless, the focused suggestions speed up the planner’s work since the planner no longer searches for the MTM codes.
96%
100% 80%
88% 75%
70%
60%
51%
71% 52%
40% 27%
42%
19%
17%
20% 0% First process step
1-Gram
85%
99% 90%
76% 74%
76%
77%
N-Gram-Analysis Overall N-Gram-Analysis focussed goods receipt Decision Tree
2-Gram 3-Gram 4-Gram subsequent procces step
5-Gram
Fig. 3. Comparison of N-Gram analysis and decision tree classification considering the previous process steps.
If the previous MTM codes can be considered, the probability of predicting the correct subsequent MTM code increases. One reason is that MTM codes often contain device-specific information. For example, if an MTM code for a forklift truck is used for a process step, it is probably the case that the subsequent MTM code will also be an MTM code for a forklift truck. In addition, the increasing probability at higher grams indicates that there are similar processes in the data set. An explanation for this is that the experienced planners fall back on best-practice processes and use these for new projects. This use of best-practice processes is also highlighted by the limitation to similar processes of goods receipt. A further limitation of the N-gram analysis, e.g., only processes of the same customer or the same location, does not appear reasonable. This comprehensive limitation would exclude similar processes that are performed at other customers or locations from the analysis. In contrast, the DT has the advantage that it can consider and evaluate all features without rigid exclusion. Due to this, the prediction probability of the DT hardly increases from the three previous process steps. While the N-gram analysis also provides higher probabilities with increasing consideration of the previous process steps.
432
M. Veigt et al.
However, the high probabilities in the N-Gram analysis with N ≥ 3 also indicate the risk that the planned processes will be equalized. This equalization would be useful in the meaning of standardization, but it would possibly not fit the customer’s requirements. Accordingly, a combination of both methods is conceivable. An N-gram analysis with low N (e.g. N = 1 or N = 2) provides a list of prioritized MTM codes. In addition, a DT suggests a concrete MTM code. Figure 4 shows such an application for the assistance system mentioned above. On the right side, the N-Gram analysis with N = 1 and limited to the goods receipt processes recommends various probable MTM codes with a short general description. The DT suggests a concrete MTM code in the blue box. The figure is only a mock-up, not a final design.
Fig. 4. Mock-up of the assistance system
In order to increase the probabilities effectively and adequately, the missing influencing factors (see Table 1) should be captured in the assistance system and evaluated as additional features by the DT or another classification method. Further research should investigate if other classification methods achieve better results than DT. Another improvement possibility concerns pre-processing. In the test, a dictionary converted the categorical features. Instead, the embedding procedure can be used to bring more meaning to the numerical notation of the features by generating an embedding vector. In the embedded vector, the similarity of categorical feature specifications is described numerically by several criteria. This conversion can enhance the DT results. However, no matter how high the probability of predicting correctly a subsequent MTM code can be increased, it must be considered that the predictions are always generated based on historical data. A new tender may be outside the valid range, e.g., it may require new processes that do not exist in the existing data set. In order to automatically include the contents of a new tender, a text analysis of the tender is necessary. This aspect addresses the research demand of extracting data from the tender identified in [7].
Using Supervised Learning to Predict Process Steps
433
4 Conclusion MTM code prediction by using N-gram analysis and decision trees can accelerate process planning during tender management in the third-party logistics industry. This contribution presents the first results of how these methods can recommend the first process step as well as the subsequent process steps in the form of an MTM code. The final decision of the correct MTM code is up to the planner. Nevertheless, the prioritized suggestion of probable MTM codes accelerates the planning process. A further acceleration of planning would result from predicting sub-processes in the form of MTM code sequences. There is a demand for further research on suggesting an MTM code sequence. Another demand for further research is the evaluation of other classification methods and the embedding of categorical features. The suggestions made by the AI methods support the standardization of the planned logistics processes since the AI methods identify and suggest appropriate process steps in the form of MTM codes from the existing processes. The AI can process more significant amounts of data than a human can and suggest to the planner which MTM codes his colleagues used in similar situations. If not only the planning is documented in the data set, but also a feedback loop is created in practice so that planning errors are corrected, and best practice processes from logistics operations are documented in the data set, the planning quality would also be increased. The results presented in this contribution are based on supervised learning methods and, thus, are past-oriented. However, since a new tender cannot necessarily be planned with the knowledge of previous projects, there is a demand for further research, especially in two aspects. On the one hand, in evaluating the tender, e.g., with methods of Natural Text Processing, to make the information contained in the tender available to both the planner and the AI in a fast and targeted manner. On the other hand, in the use of reinforcement learning. This approach promises to plan new logistics processes individually into the future without being limited to the data set of past planning. It is conceivable to use this approach for process planning and the development of an ideal layout, and the generation of planning variants in real layout planning. Acknowledgment. This work results from the research project “INSERT- AI-based assistance system for concept planning in production and logistics”, funded by the BAB - Bremer Aufbaubank under the reference number FUE0626B and with funds from the European Regional Development Fund (ERDF).
References 1. Prahalad, C.K., Hamel, G.: The core competence of the corporation. In: Hahn, D., Taylor, B. (eds.) Strategische Unternehmungsplanung—Strategische Unternehmungsführung, pp. 275– 292. Springer, Heidelberg (2006). https://doi.org/10.1007/3-540-30763-X_14 2. Doll, A., Friebel, D., Rückriegel, M., et al.: Global Logistics Markets (2014). https://www. rolandberger.com/publications/publication_pdf/roland_berger_global_logistics_markets_2. pdf. Accessed 04 Aug 2021 3. Hofmann, E., Osterwalder, F.: Third-party logistics providers in the digital age: towards a new competitive arena? Logistics 1, 9 (2017). https://doi.org/10.3390/logistics1020009
434
M. Veigt et al.
4. Large, R.O.: Partner-specific adaptations, performance, satisfaction, and loyalty in third-party logistics relationships. Logist. Res. 3(1), 37–47 (2011). https://doi.org/10.1007/s12159-0110047-8 5. Straube, F., Ouyeder, O., Siegmann, J., et al.: Partner-specific adaptations, performance, satisfaction, and loyalty. In: Blecker, T., Jahn, C., Kersten, W., et al. (eds.) Maritime Logistics in the Global Economy: Current Trends and Approaches, vol. 1, pp. 37–47. Aufl. Eul, Lohmar, (2011). Third-Party Logistics Relationships 6. Spiegel, T., Siegmann, J., Durach, C.F.: Flexible development and calculation of contract logistics services (2014). https://doi.org/10.5281/ZENODO.1091196 7. Veigt, M., Steinbacher, L., Freitag, M.: Planungsassistenz in der Kontraktlogistik: Ein Konzept zur KI-basierten Planungsunterstützung innerhalb einer digitalen Plattform. Industrie 4.0 Management 37(5), 11–15 (2021) 8. Teucke, M., Veigt, M., Engbers, H. et al.: Kontinuierliche Planung von Logistikflächen/Continuous planning of logistic areas within factories. wt Werkstattstechnik online 110(4), 195–200 (2020). https://doi.org/10.37544/1436-4980-2020-04-29 9. Gesellschaft Produktionstechnik: Fabrikplanung: Planungsvorgehen, Ausg. deutsch/englisch. Fabrikplanung, = Factory planning/VDI-Gesellschaft Produktionstechnik (ADB), Fachausschuss Fabrikplanung; Blatt 1. Beuth, Berlin (2011) 10. Karger, D.W., Bayha, F.H.: Engineered Work Measurement: the Principles, Techniques, and Data of Methods-Time Measurement, Background and Foundations of Work Measurement and Methods-Time Measurement, Plus Other Related Material, 1st edn. vol. 4. Industrial Press, New York (1987) 11. Kešelj, V., Peng, F., Cercone, N., et al.: N-gram-based author profiles for authorship attribution. In: Proceedings of the Conference Pacific Association for Computational Linguistics: PACLING 2003, Halifax, pp. 255–264 (2003) 12. Burdukiewicz, M., Sobczyk, P., Rödiger, S., et al.: Amyloidogenic motifs revealed by n-gram analysis. Sci. Rep. 7, 12961 (2017). https://doi.org/10.1038/s41598-017-13210-9 13. Ziegelmayer, D.: Character n-gram-based sentiment analysis. Köln, University, Zugl., Dissertion. Informatik. Verl. Dr. Hut, München (2014) 14. Wu, X., Kumar, V., Ross Quinlan, J., et al.: Top 10 algorithms in data mining. Knowl. Inf. Syst. 14, 1–37 (2008). https://doi.org/10.1007/s10115-007-0114-2 15. Settouti, N., Bechar, M.E.A., Chikh, M.A.: Statistical Comparisons of the Top 10 Algorithms in Data Mining for Classi cation Task. IJIMAI 4, 46 (2016). https://doi.org/10.9781/ijimai. 2016.419 16. Rokach, L, Maimon, O.Z.: Data mining with decision trees: theory and applications, 2nd edn. Series in Machine Perception and Artificial Intelligence, vol. 81. World Scientific, New Jersey (2015) 17. Quinlan, J.R.: C4.5: Programs for Machine Learning, 1st edn. Elsevier, Aufl, p. 1 (2014)
From Linear to Circular Packaging: Enablers and Challenges in the Fashion Industry Sarah Pfoser(B)
, Katharina Herman, Andrea Massimiani, Patrick Brandtner , and Oliver Schauer
University of Applied Sciences Upper Austria, Wehrgrabengasse 1-3, 4400 Steyr, Austria [email protected]
Abstract. Packaging is a logistics function which generates significant amounts of waste. The fashion industry is the business area with the highest share of ecommerce shipments and is therefore responsible for substantial amounts of packaging waste. A shift from linear to circular packaging models is therefore highly desirable, especially in the fashion industry. However, although many fashion retailers recognized the importance of sustainability, hardly any company establishes circular packaging solutions. This paper aims to identify the challenges which prevent fashion retailers from introducing circular packaging. Additionally it will be shown which enablers drive the development of circular packaging. Expert interviews were conducted with companies all along the circular packaging value chain in the fashion industry. A PESTEL analysis shows the Political, Economic, Social, Technological, Environmental, and Legal factors which currently inhibit circular packaging. Three types of enablers where found to be relevant, namely profit, consumer and politics. The findings provide valuable input to develop circular business models promoting the use of circular packaging in the future. Keywords: Circular economy · Sustainable packaging · e-commerce · Fashion industry · PESTEL analysis
1 Introduction Given the recent political, governmental, and consumer’s pressure towards sustainable business practices, circularity has moved to the top of the agenda of many firms. Ecommerce causes a lot of packaging waste Sustainable packaging using reusable boxes and bags is a major concern in the e-commerce business. The software-based trend management tool TRENDONE forms the basis for trend identification and analysis for the concept of sustainable packaging. The macro trends sustainable packaging, zero waste, circular economy and conscious consumer are filtered to the defined megatrend Sustainability. The most significant micro trends imported by the tool are refillable packaging, use of reusable packaging [1], antimicrobial packaging [2], edible films [3], the use of biodegradable and renewable bioplastics [4] and post-consumer recycled resin (PCR) [5]. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 M. Freitag et al. (Eds.): LDIC 2022, LNLO, pp. 435–445, 2022. https://doi.org/10.1007/978-3-031-05359-7_35
436
S. Pfoser et al.
The findings of the trend monitoring software TRENDONE show that the macrotrend Sustainable Packaging including the crucial micro-trends are moving into the focus of manufacturers and consumers. For companies, this finding primarily forms the basis for reducing future uncertainties and serves as a support for strategic decision-making for the introduction of new business models or business innovations. Fashion firms and retailers started to integrate sustainability in their corporate communication. Evidence can be found on the firm’s websites, sustainability reports, social media channels as well as the packaging itself. Empirical evidence shows, that companies which are capable of implementing such goals and rethink business models positively influence growth opportunities, as well as a well-designed business model shall lead to a competitive advantage [6, 7]. Especially packaging is an area where a lot of waste occurs and sustainability is difficult to achieve. Circular Packaging has been incorporated in the sustainability agenda of firms in the e-commerce fashion industry, but there are only scattered pilot projects in practice. Meanwhile, studies prove the enormous economic advances of circular business models, although the question why current systems are still dominantly linear have opened a whole new area of research. In order to allow a transition to the Circular Economy, systemic changes are needed. Companies are required to adopt their business models [8]. However, the application of circular business models is still in its infancy. It requires businesses to enter markets with unstructured settings, little is known about underlying logics, and therefore, approaches remain rather experimental. The importance of circular practices is well understood, however, to realise economic viability and environmental impacts from innovating towards a circular business model requires the understanding of value creation within its context [9, 10]. While linear models follow a “take-make-waste” approach, the key principle of circular economy is said to create and capture value by keeping materials in the loop. A structural literature review explains the need for re-evaluating the elements of a business model and its actors in a circular context [10]. The value creation within a model as such is still poorly understood. The aforementioned challenges result in the need for further research on the resources for constructing a successful circular business model as well as how the actors function within a model as such. It is important to understand what impedes and facilitates the occurrence of a circular packaging model in the fashion industry. This paper therefore aims to answer the following two research questions: RQ1: Which challenges inhibit firms to implement circular packaging models in the fashion industry? RQ2: Which enablers ensure that circular packaging models create and deliver economic, social and environmental value?
2 Methodology The underlying paper focuses on flexible plastics packaging in the e-commerce fashion industry. Flexible packaging refers to “packaging whose shape can be easily changed” [11]. An example for flexible packaging are the polybags (plastic bags) which are typically used to ship textiles in the fashion industry. The value chain of such packaging usually consists of raw material suppliers, brand owners, brand retailers, collectors, sorters and recyclers [12]. The authors included a Logistics and Postal Service Provider as an enabler for Circularity.
From Linear to Circular Packaging
437
In order to assure qualitative results in primary research, the authors put special emphasis on interviewing experts along the value chain. The selection of experts with their background information can be found in the following Table 1. The authors conducted ten interviews with industry experts between March 2021 and May 2021. The focus of the sample was Europe, in order to settle the differences in law, politics and culture caused by geographic location [13]. Interviews were conducted via video call due to the Covid19 restrictions. The interviews lasted 30–40 min on average. Prior to the interviews, desk research was performed on the interview partners, the companies they represent and the sustainability strategies they communicate. Interviewees were invited via mail, receiving an interview guideline. To account for the explorative design of the study, some of the questions were customized for each participant of the value-chain [14]. Table 1. Overview of interviewees. #
Organization
Position
1
Fashion retailer
Senior director transport & logistics
2
Raw material supplier
Innovation project management
3
Machine manufacturer plastics recycling
Technology research manager chemical recycling
4
Fashion retailer
Head of sustainability
5
Fashion retailer
Assistant store manager
6
Logistics and postal service provider
Head of projects and innovation
7
Distributor of polymers and rubber chemicals
Operational business leader
8
Supplier of recycling machines
Sustainability and marketing manager
9
Fashion retailer
Sustainability and innovation manager
10
Machine manufacturer
Vice President plasticising systems & recycling
The 10 interviews were transcribed with the use of the software otter.oi. Later, the data analysis was conducted with the software MAXQDA. MAXQDA is a well-established software for data analysis. With the help of this instrument, the researcher was able to establish a coding system to structure the data. Coding is a method for analysing qualitative data in order to gain deeper insights. Codes are established and assigned to segments of the data in order to systematically analyze qualitative content [15]. Compared to manual human analysis, automating the analysis reduces bias and allows for more consistent and more accurate outcomes. Following an exploratory research design, qualitative data analysis has been conducted to analyze the interview results. Qualitative data analysis is used in research to emphasizes the experiences, opinions, behaviours, and social contexts of research participants. After choosing the appropriate codes, commonalities were analysed and mapped. These differences and commonalities are presented in the following Discussion Section of this paper.
3 Discussion In the following, we will show the challenges and enablers for circular packaging which emerged from the results of the expert interviews. Based on intertextual and interpretive analysis from interview transcripts, this section shows data by means of power quotes in order to support our findings [16, 17].
438
S. Pfoser et al.
3.1 Challenges for Circular Packaging Models In order to illustrate the challenges for circular packaging models, a PESTEL analysis was conducted. PESTEL analysis is a strategic framework used to evaluate the external forces which impact an organization by breaking down these external forces into Political, Economic, Social, Technological, Environmental, and Legal factors [18]. In our case, the challenges for circular packaging models that were mentioned during the expert interviews are classified according to the PESTEL model. Table 2 shows the challenges communicated by interviewees and assigned to the factors of PESTEL. Additionally, the concerned players within the value chain are marked in Table 2. We were not able interview collectors and sorters in this study, therefore these two parts of the value chain are not represented in Table 2. Table 2. PESTEL analysis of challenges for circular packaging models. Material-supplier
Recycler
Retailer
Logistics
Political Changes in taxation causing uncertainties about costs and pricing
x
Environmental Lack of ecological data
x
Social Consumer’s misconception about plastics
x
x
Collection, sorting and bringing the material back to the circle
x
x
Behavioural change of consumers
x
x
Uncertanties about skills needed
x
Compromises in branding
x
x
Technological Quality and consistency of recyclates
x
Customer’s adaption of specifications and products
x
Economic Missing short-term profitability for convincing decision-makers (externally and internally) High prices of recyclates compared to virgin plastics
x x
Low value of polybags
x
Legal Time pressure from legislators Missing uniform standards in the industry
x x
A lack of standards and regulations is mentioned among the upstream portions. The rules are said to be unclear and therefore, uncertainties exist in regard to the costs
From Linear to Circular Packaging
439
and prices. The challenges for the downstream portion consist of a lack of data from the suppliers, a lack of infrastructure for the collection processes and misconceptions from the consumer. The recycler explains a common misconception of plastics by the consumer. Due to the bad reputation of single-use plastics, this is passed on to the perception of multi-use plastic. Furthermore, raw material suppliers explain that their customers are used to over specified products which requires a change in mindset. (…) the challenge that many of our customers, they are still used to the. specifications they have used for the current materials they are. receiving, and they simply insist on receiving materials with the same. quality level. Meaning we have to work closely together with the. customer, to come to specifications and to products which are fulfilling. the purpose but are not over specified. (Expert #1). Further, fashion retailers explain the difficulty of aligning sustainability with branding. Actions are taken to explain to the consumer, that more sustainability can be reached when reducing the number of colors on the packaging for example. We even need to involve the consumer in accepting those challenges. Take colour for example. Colours are very important for branding but do. not always go along well with sustainability. Water, resources and. chemicals are wasted for colours. So, we try to use fewer colour now. The same applies to materials and fabrics. (Expert #9). The logistics partner explains difficulty in its position, when trying to convince stakeholders internally or externally of a sustainable innovation if the profitability is not visible. I find it really hard for really new and innovative ideas to. sometimes look beyond the profitability question. Because sometimes. new and groundbreaking ideas are not really profitable in the. beginning, it might be turning out to be something that is indeed a. gamechanger but in the beginning you have to make iterations just to. make it work. And this is something that especially from an investing. point of view is not always easy to convince your own stakeholders. inside of the company or even customers outside to bring new products. out on the market. (Expert #5). Reflecting on the challenges in the application of Circular Business Models, the machine manufacturer refers to experience in Refurbishing and Reselling used machinery. The expert refers to a lack of an appropriate legal framework.
440
S. Pfoser et al.
Customers expect the same standards from used machinery as from new. machinery. This includes expectations on warrantee. This is not possible. with the current laws. (Expert 10 2021). Furthermore, differences could be identified on an international level. Experts explain that the standards on ecological and social factors vary among countries. High standards in the European Union facilitate innovations in the field of sustainable development. However, China was given as an example for low standards in ecological and social standards, resulting in low prices. International firms that try to implement higher quality and higher prices in countries as such, face the challenge of justifying higher prices to customers. 3.2 Enablers for Circular Packaging Models In order to overcome existing barriers to circular packaging and promote more sustainable packaging practices, key enablers for circular packaging were identified. Three main types enablers where found to be relevant, namely profit, consumer and politics. These enablers refer to the economical, social and political dimension of the PESTEL analysis. Economics of circular innovations are found to be a major concern for those companies willing to implement them. Firstly, stakeholders must be convinced, such as investors internally as well as those changes must be accepted by the end-consumer, as the user of such innovations. This does not only involve a change in behavior but might lead to higher prices. Circular Innovations are often costly in the supply and this could affect the prices for the end-consumer. However, the question whether those costs should be passed on to the consumer have led to a range of opinions. It was emphasized by the experts that consumers in the fashion industry are very price sensitive. While some consumers might pay a premium, a major part is expected not to accept higher costs for sustainable packaging. Expert 5 explains that the consumers are used to free products and services related to the shipping of goods in fashion e-commerce. Further, it is said, that the awareness for the waste problem in this industry is not big enough for the consumer to accept higher prices. So why would they pay an additional price to have something reusable. if they don’t have something they benefit themselves for reusable. packaging. (Expert 5#). The higher prices of recyclates in circular packaging is also highlighted. While one expert explains these costs must be passed on to the consumer, brands claim that creative solutions can be applied now to justify the higher value of the product, which could also be another benefit besides sustainability, such as convenience. On the other side, politics can cause major change in this aspect. One expert refers to the plastic tax. The plastic tax was implemented in Europe in January 2021 and requires companies to pay for each ton of newly used plastics. By recycling plastics, these costs can be avoided.
From Linear to Circular Packaging
441
So for Packaging for fashion for example, if we can develop a solution, recycled materials this kind of tax does not need to be paid, which means. – as long as our recyclate is not 800 euro per ton more expensive than. the fossil based material we can still bring a cost benefit, also to the end user. (Expert #1). In line with previous findings, experts agree that Circular Business Model Innovations require partnerships with other stakeholders and cannot happen on a macro-level. Partnerships with other participants along the value chain and NGOs are mentioned. Empirical Study has shown that the consumer is seen as a key challenge and key enabler at the same time. There are just so many misconceptions about materials used for packaging. An example for that is biodegradability. According to Expert #10, consumers think that biodegradable plastics is better, so this is the driver for our customer to creating a need. They want to use biodegradable plastics, but they don’t know that this is in fact less sustainable. The expert explains that biodegradable plastics can only degrade in certain industrial conditions at a certain temperature. Similar examples were provided on the perception of glass in the packaging industry. Many customers choose glass over plastics, as it is perceived as more sustainable by the end consumer. As explained by the machine manufacture, the most sustainable packaging for beverages would be the use of PET bottles or lightweight PET bottles with a worldwide standard. We consider ourselves as a customer-centric company. However, when it. comes to sustainability, focusing on the customers and consumers can be. a step back. Sometimes their educational level on sustainability is a misconception. (Expert #10). Furthermore, NGOs are seen as a key enabler for communicating the correct values to the consumer. Interactions with consumers have shown that the consumer is overwhelmed with content and it is hard to know whom to believe. Companies rely on NGOs for communicating trustworthy and objective content to consumers. Due to the focus on plastics packaging in this research, experts referred to NGOs, such as the Ellen MacArthur foundation and the plastics bank. Furthermore, manufacturers and recyclers explain that the implementation of certain innovations requires trade-offs for brands as well as consumers. For example, recyclability of packaging limits the opportunities for the use of colors and branding on the packaging. Taken further, in order for brands to accept this tradeoff, the consumer’s acceptance is required.
4 Conclusion The challenges, outlined in the previous chapter, are based on political, economic and consumer grounds. The ecological advantages of circular business model innovation are obvious, if not the motivator for participants of the Circular Economy. However, the key barrier is the economical aspect. This matter of fact was highlighted by several previous studies, e.g. [19] or [20]. The triple bottom line seems to be the key challenge that
442
S. Pfoser et al.
companies are trying to overcome. Experts expect more openness towards alternative solutions. Mindset does not seem to pose a barrier for retailers in the fashion industry, nor the material suppliers, machine suppliers and recyclers. However, the logistics company seems to be in the position to convince stakeholders of sustainable innovations. In regard to a lack of resources, money appears to be the key resource needed. Interestingly, technology does not seem to be a challenge in the development of circular packaging. The technology is available. As discussed, innovations as such are often not implemented due to the lack of appropriate returns in profit. Further, external relations were seen as a key enabler. Experts referred to co-operations along the value chain. Moving down the value chain the dependency on suppliers increases and the importance of the consumer’s acceptance increases. The upstream of the value chain, however, is confronted with the economics of innovations. Customers are used to lower prices, however the prices of recycled materials compared to virgin material remains a key barrier. It has become clear that Circular Business Model Innovation can only be reached on a macroeconomic level. Innovations are found to be customer centric. However, the ecological improvements are demanded from the upper part of the value chain. Therefore, a transparent communication is needed. Furthermore, the consumer acts as a bottleneck for closed loop solutions, as behavioural change and acceptance of business model innovations is required. This is in line with Meherishi et al. [21] who claim for future research on education, awareness and skill development on sustainable packaging practices to develop consumers’ mindset. Further, politics may act as an important role for developing governmental policies, incentives and taxes, trade restrictions, labour regulations and environmental laws. This was previously also outlined by Kawecka and Cholewa-Wójcik [12]. Furthermore, companies explained to innovate solely in the European Union, as the regulations abroad make it hard to compete with competitors’ low prices. The need for worldwide regulations is explained.
5 Managerial Implications, Further Research and Limitations 5.1 Managerial Implications Resulting from the findings of this paper, the authors suggest key activities to be taken by managers to allow for shared value creation with circular e-commerce packaging. All participants along the value chain play a vital role in the transition to a Circular Economy. Firstly, it is suggested to consider the “Design for Recycling” principle, in which products are designed for material recovery, sustainable use, collection and sorting. Resultingly, cooperation with other participants along the value chain is suggested. The increase in the consumer’s acceptance is relevant for increasing the number of packaging recycled and thus, increases the profitability in “closing the loop”. Participants in the downstream value chain are in the position to change the consumer’s behaviour in Circular Business Models. The direct influence on the attitude, perceived value and habits of the consumer have been identified as a potential for improvement. Additionally, the clarification of many misconceptions on plastics shall be placed on the agenda of participants throughout the industry. However, actions are taken already,
From Linear to Circular Packaging
443
transparency along the value chain and cooperation with NGOs are advised in order to increase the objectivity of information communicated to the consumer. On the other hand, upstream and midstream participants are in the position to increase the demand for sustainable innovations by providing information on the ecology of material, parts, products and processes. LCA was identified as a useful tool to measure environmental aspects. In order to mitigate the risk of comparability, the use of international reporting standards is suggested. Finding, that technology can act as a key enabler for Circular Business Model shall encourage managers to invest resources in the development of tools that facilitate the communication of economic, social and ecologic data along the value chain. Overall, many uncertainties on the consumer’s acceptance of Circular Business Models exist. More case studies on testing CBMs are suggested to improve the understanding of this problem in a preliminary stage. A general openness towards changes that come with Circular Innovations is recommended, such as the adaption of specifications in the production process. 5.2 Implications for Further Research The importance of the consumer in circular business model has become obvious. Further research is suggested on the consumer behaviour. The perceptions of consumers on materials could help brands to potentially clarify misconceptions and provide more transparency where needed. It was found that companies shall educate consumers on the sustainability of packaging. Further research on which content is needed is suggested. Furthermore, the collection process for keeping materials in the loop disrupts existing processes in the e-commerce. The relationship of consumers with those models opens a new area of research. Future research on how to incentivize consumers is suggested. Lastly, politics and standardizations within the industry have been identified as an enabler and accelerator for circular business models. Development on national as well as international level are expected. Further research on those topics is recommended. For the purpose of taking the development of Circular Packaging in e-commerce further, the following research agenda has been developed (Table 3). Table 3. Implications for further research. Area of research
Specific problem
Politics
Development of governmental policies, incentives and taxes, trade restrictions, labor regulations, environmental laws
Consumer
Development of Content for educating consumers on the sustainability of materials and processes
Development of international standards for circular business models
Development of methods for incentivicing the consumer for take-back systems
444
S. Pfoser et al.
5.3 Limitations Due to the limited geographical scope, number of experts interviewed, and focus on the fashion industry, the extent to which this work can be generalised to other contexts with different institutions remains unclear and needs further evaluation. This study is purely qualitative and therefore not able to deliver a fully objective and rational PESTEL model [22]. For further research, we suggest that methods such as multi criteria decisionmaking models are needed to analyze the relations and interactions between the identified PESTEL factors.
References 1. Coelho, P.M., Corona, B., Ten Klooster, R., Worrell, E.: Sustainability of reusable packaging– current situation and trends. Resour. Conserv. Recycl. 6, 100037 (2020) 2. Zhong, Y., Godwin, P., Jin, Y., Xiao, H.: Biodegradable polymers and green-based antimicrobial packaging materials: a mini-review. Adv. Ind. Eng. Polymer Res. 3(1), 27–35 (2020) 3. Jeevahan, J.J., et al.: Scaling up difficulties and commercial aspects of edible films for food packaging: a review. Trends Food Sci. Technol. 100, 210–222 (2020) 4. Bhagwat, G., et al.: Benchmarking bioplastics: A natural step towards a sustainable future. J. Polym. Environ. 1–21 (2020) 5. Cecon, V.S., Da Silva, P.F., Vorst, K.L., Curtzwiler, G.W.: The effect of post-consumer recycled polyethylene (PCRPE) on the properties of polyethylene blends of different densities. Polym. Degrad. Stab. 190, 109627 (2021) 6. Salvador, R., Murillo, B., Freire, F., Anthony, H., Cassiano, P., Antonio F.: Circular economy strategies on business modelling: identifying the greates influecnes. J. Clean. Prod. 299 (2021) 7. Wandosell, G., María, P., Alfredo, A., Raúl, B.: Green packaging from consumer and business perspectives. Sustainability 13 (2021) 8. Uusitalo, T., Maria, A.: The concept of value in circular business models. ISPIM (2018) 9. Julia, L.K.N.: A circular business model mapping tool for creating value from prolonged product lifetime and closed material loops. J. Clean. Prod. 197(1), 185–194 (2018) 10. Tom, L., Joakim, W., Vinit, P.: A definition and theoretical review of the circular economy, value creation, and sustainable business models: where are we now and where should research move in the future? Sustainability 10 (2018) 11. Roberge, D.: Flexible packaging materials buying guide. https://www.industrialpackaging. com/blog/flexiblepackaging-materials-buying-guide. Accessed 11 May 2021 12. Kawecka, A., Cholewa-Wójcik, A.: Challenges for the packaging industry in the circular economy. Zeszyty Naukowe Instytutu Gospodarki Surowcami Mineralnymi Polskiej Akademii Nauk 109, 5–16 (2019) 13. Prasad, P.: Crafting Qualitative Research: Beyond Positivist Traditions. Routledge, New York (2017) 14. Stake, R.E.: The Art of Case Study Research. Sage, Thousand Oaks (1995) 15. Stake, R.E.: Qualitative Research: Studying How Things Work. Guilford Press, New York (2010) 16. Pratt, M.G.: From the editors: for the lack of a boilerplate: tips on writing up (and reviewing) qualitative research. Acad. Manag. J. 52(5), 856–862 (2009) 17. Fawcett, S.E., Waller, M.A., Miller, J.W., Schwieterman, M.A., Hazen, B.T., Over-street, R.E.: A trail guide to publishing success: tips on writing influential conceptual, qualitative, and survey research. J. Bus. Logist. 35(1), 1–16 (2014)
From Linear to Circular Packaging
445
18. Rambaree, K., Sundström, A., Wang, Z., Wright, S.A.I.: Qualitative stakeholder analysis for a Swedish regional biogas development: a thematic network approach. Sustainability 13 (2021) 19. Säilä, A.: Challenges and opportunities of packaging in a circular economy. Agro Food Ind. Hi-Tech 29(6), 41–44 (2018) 20. Bening, C.R., Pruess, J.T., Blum, N.U.: Towards a circular plastics economy: interacting barriers and contested solutions for flexible packaging recycling. J. Clean. Prod. 302, 126966 (2021) 21. Meherishi, L., Narayana, S.A., Ranjani, K.S.: Sustainable packaging for supply chain management in the circular economy: a review. J. Clean. Prod. 237, 117582 (2019) 22. Yüksel, I.: Developing a multi-criteria decision making model for PESTEL analysis. Int. J. Bus. Manag. 7(24), 52–66 (2012)
Socio-technical Systems
Assessing Driver Fatigue During Urban Traffic Congestion Using ECG Method Nizami Gyulyev1
, Andrii Galkin1(B) , Tibor Schlosser2 and Oleksii Lobashov1
, Silvia Capayova2
,
1 O. M. Beketov National University of Urban Economy, Kharkiv, Ukraine
[email protected]
2 Slovak University of Technology, Bratislava, Slovakia
{tibor.schlosser,silvia.capayova}@stuba.sk
Abstract. The research paper analyses the level of stress and functional state of the drivers in urban traffic congestion. Therefore, the primary objective of this research is to describe patterns to assess fatigue of the driver during urban traffic congestion. The Electrocardiography (ECG) data is used to assess fatigue of the driver. The model comprising of influence of traffic congestion on the functional state of the average driver, allows us to predict changes to the driver’s state depending on the age, the duration of the traffic congestion and initial state prior to congestion. The value of the initial functional state affects the driver’s functional state during his/her stay in a traffic congestion in different ways. The rising of tension during staying in traffic jam is 10–12% after 7–10 min. The research uses system analysis for data analysis; electrophysiological methods in determining the functional state of the driver and mathematical statistics methods were used during the development of model for analysis of the functional state of the driver. Keywords: ECG · Driver · Age · Driving duration · Functional state
1 Introduction Vehicles are continuously subjected to dynamically changing traffic condition, especially during the peak congestion periods. It adversely affects the mental and psychological condition of drivers and prolonged exposure can lead to deterioration of their functional capabilities. The efficiency of a driver is largely determined by the working conditions he or she is subjected to and is comprised of harmony among the road users (drivers), machines (vehicles) and the surrounding (road environment) (Maji and Tyagi 2018; Chebanyuk et al. 2020). The road environment related traffic condition can be determined using traffic engineering parameters including during urban congestions when traffic volume exceeds highway capacity (Issa 2016). Poor mass transit infrastructure and high traffic density, especially during the morning and evening peak hours further accentuate the traffic congestion (Burko et al. 2020).
© The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 M. Freitag et al. (Eds.): LDIC 2022, LNLO, pp. 449–461, 2022. https://doi.org/10.1007/978-3-031-05359-7_36
450
N. Gyulyev et al.
The driver is always limited by the time of reaction (Gyulyev et al. 2018). Especially, when driving in urban area and congested streets. Often, in these cases, only very fast and accurate actions can prevent a traffic accident (González-Hernández et al. 2020). The driver’s fatigue is ensured not only by high professional training, but also by individual behaviour of the driver. The frequency of starts and stops of movement, staying in congestion have different influence on various drivers’ behaviour and their state (Afanasieva and Galkin 2018). A driving simulator study exploring the effect of different mental models on ADAS system effectiveness presented in (Rossi et al. 2020) Studying the drivers and their influence on functional state of the driver will increase its reliability during the traffic congestions (Gyulyev et al. 2020). Continues, traffic congestion restricts driver freedom thereby, inducing negative emotions and considerably worsening the psycho-physiological performance of the driver, rising fatigue of their activities (Zhang et al. 2016; Gyulyev et al. 2019). So, the fatigue of driver activities consists of the functional and emotional state of the driver, vehicle and road environment (Grabarek 2018). The fatigue of a driver and their functional and emotional state can be measured by observing their neural activity (Gyulyev et al. 2021). The management of road transport is primarily aimed at drawing up and implementing reliable plans for the delivery of goods by road transport from suppliers to consumers. The problem of routing is important in the field of logistics. Route planning is a measure that companies can take to reduce the cost of their activities. For optimal planning of the parameters of the technological process of freight transportation, where the driver is an integral part of this process. Prior studies have emphasized upon the need for studying human factors and their impact on transportation systems (Wang et al. 2015). Capabilities of a person while driving should be considered for traffic operational analysis of transportation infrastructure (Kulbashna et al. 2020). The analysis showed that, despite a lot of work done in this area today, there are very few case study researches among them. At the same time, theoretical research is limited to the field of application and does not provide a complete picture of the research (especially those relating to the environment and the driver). In this case, the accuracy of abstract models that do not take into account the driver’s behavior can be questioned in the real conditions of their application. The second drawback is, the absence of the human factor as such, when designing a route or describing it in models as constants. In contrast to the existing approach to planning transportation process, the driver should considering for route organizing. The driver and vehicle make up due to inextricable link which performed transportation process. Therefore, the primary objective of this research is to described patterns to assess fatigue of the driver during urban traffic congestion. Freight vehicles has been considered as vehicle and congested urban roads as the road environment. The Electrocardiography (ECG) data is used to assess fatigue of the driver. Overall research has been structured as discussed further: 1) Introduction; 2) Method of assessing driver fatigue; 3) Results, which describe an experiment and processing the obtain results; 4) Conclusions.
Assessing Driver Fatigue During Urban Traffic Congestion Using ECG Method
451
2 Method of Assessing Driver Fatigue The fatigue is a complex construct comprising of durability, maintainability and safety (Klyatis 2015). One of the indicators of system fatigue is the probability of failure-free operation, which is the probability of Mean Time Between Failures (Fiondella and Xing 2015). Trouble-free operation from the driver’s perspective within a «Human – technology – environment» system, depends upon individual factors: age, experience, gender, nervous system state and initial state (Gyulyev 2012; Galkin et al. 2018), technology: engine power, ergonomic characteristics of the vehicle etc. (Galkin et al. 2019) and environment: number of lanes, duration of traffic congestion (Eboli et al. 2016). The specifics of the functioning of the vehicle traffic system is the interaction of technical and human factors. The complexity of managing such a system is the need to ensure a kind of balance in the state of each element (subsystem). Otherwise, there are problems (failures) in the functioning of individual subsystems and this affects the functioning of the system as a whole. The functional state of a person can be assessed using an electroencephalogram (EEG), galvanic skin response (GSR), oculography, ECG, and other methods having their own benefits and limitations. The complexity of these methods in processing data is the delimitation and allocation of single individual signals from the information flow that correspond to individual events on the road, which may neglect the results (Prasolenko et al. 2019a). The electrocardiogram has been most informative method for measurement and analysis (Prasolenko et al. 2019b). The designations of the heart rate variability variables are given with regard to the published recommendations of the European Society of Cardiology and the North American Society for Electrophysiology (Heart Rate Variability. Standards of Measurement 1996). Among the variables presented, the complex index of the activity of regulatory systems (IRSA) that was proposed, proceeding from the tasks of space medicine, in the early 1980s occupies an important place. The researches use an integral index for assessing the functional state of the driver – IRSA, which reflects driver’s general reaction against environmental factors and forms the basis of this research work. The integral index of the heart rate variability allowing assesses different degrees of stress of the regulatory systems. The methodology is based on the theory of positive correlation between ‘unevenness of cardio-intervals’ and ‘fatigue level’ (Baevskii 2002). Taking into account the concept of the cardiovascular system as an indicator of adaptive activity of the whole organism, we should first turn to the analysis of changes in heart rate - the universal response of the body in response to any load - physical or emotional. Information about how the body has reached a certain level of activity is encoded in a sequence of cardio intervals. The sequence of cardiointervals of the electrocardiogram is encoded information about the processes that occur not in the heart, but in different parts of the control system: nerve plexuses; endocrine glands; nerve centers located in the brain tissue. The structure of cardiomyopathy can be judged by the state of the mechanisms of physiological regulation. This is an integrated assessment of human fatigue. This technique is based on the theory of directly proportional dependence of non-uniformity of cardio
452
N. Gyulyev et al.
intervals and the level of fatigue, which in the literature is usually called an indicator of the activity of regulatory systems of the body. It is evaluation in conventional units is made with special algorithm (Vaevsky 1984) that based on statistical indicators, histogram indicators and data of spectral analysis of cardiointervals. The level of fatigue allows to differentiate different degrees of tension of regulatory systems and to assess the adaptive capacity of the organism. It is calculated by an algorithm that takes into account five criteria: Pc = |A| + |B| + |C| + |D| + |E| ,
(1)
where A – the total effect of regulation (in terms of mathematical expectation); B – function of automatism (by average square deviation, by variational magnitude and by coefficient of variation); C – vegetative homeostasis (according to a set of indicators: variation scale, amplitude of the mode, index of tension of regulatory systems); D – stability of regulation (by the coefficient of variation); E – activity of subcortical nerve centers (it is determined on the basis of relative capacities of respiratory waves and waves of the first and second order, distinguishing the states of the expressed and moderate increase of activity of subcortical nerve centers). Fatigue levels are measured through the conventional scale having units between 1 to 10. Table 1 provides the value of fatigue level, and the functional states of the driver that can be diagnosed, collected according to Baevskii (2002). Table 1. Characteristics of the various functional states of the driver Functional state
Description
Indicator value, c.u.
Optimal tension level for the regulatory system
Necessary to maintain an active Pc = 1–2 balance between the driver and the environment
Moderate tension level for the regulatory system
Adapting to the environment requires additional functional reserves. Such state arises due to emotional stress or in the event of adverse environmental factors
Pc = 3–4
Expression of tension for the regulatory system
Active mobilization of protective mechanisms, in particular, increased activity of the sympathetic-adrenal system and the pituitary-adrenal system
Pc = 4–6
The state of overstrain of the regulatory systems
The insufficiency of protective and Pc = 6–7 adaptive mechanisms, their inability to provide an adequate response of the organism to the influence of environmental factors (continued)
Assessing Driver Fatigue During Urban Traffic Congestion Using ECG Method
453
Table 1. (continued) Functional state
Description
Indicator value, c.u.
Exhaustion level (asthenization) for This state characterized by the Pc = 7–8 the regulatory system insufficiency of the defensive–adaptation mechanisms, their inability to provide an adequate reaction of the body to the effect of environmental factors. Here, excessive activation of the regulatory systems is not reinforced by the corresponding functional reserves. Activity of the control mechanisms are reduced (insufficiency of the regulation mechanisms) Failure of the adaptation mechanisms
Specific pathological deviations dominate and the ability of the adaptive mechanisms, self-regulation is partially or completely disturbed
Pc = 8–10
According to the stress and its degree of intensity within the regulatory system, diagnosis and the functional state is given (Baevskii 2002). Overstrain of regulatory systems can lead towards failure to adapt, leading to unwanted changes in the functioning of the body systems, as a consequence there will be decrease in reliability of the driver. The state of the regulatory mechanisms of the human body should not reach the level of excessive fatigue, over-stress and failure to adapt.
3 Results 3.1 Experiment Description Driver’s psycho-physiological features and behavior significantly affect the safety of the transport system. For the regression model developed for the pilot study, drivers belonging to all age groups were involved. The experiment required fixing electrocardiogram (Cardiosens) touch points on the driver’s body (KhAI-MEDIKA – equipment for functional diagnostics, 2016 (see Fig. 1). Data was collected while entering the traffic congestion, during their stay in the and at the moment of exiting it. The method of attaching electrodes and the appearance of the program of processing the results are shown at Fig. 2. The age groups of drivers were distributed as follows: 20–30 years old - 21 drivers; 30–40 years old - 24; 40–50 years old - 18; 50–60 - 16; for 60 years - 11. The average driving experience for the first group was 4 years, for the second - 12 years, for the third - 18 years, for the fourth - 22 years and for drivers over 60 years old - 25 years. Female drivers accounted for 20%:
454
N. Gyulyev et al.
Fig. 1. Electrode attachment technique
Fig. 2. Appearance of the Cardiosens software
The software allows you to decipher the data and assess the driver’s level of fatigue. For example, according to Fig. 3. The 2nd level of driver fatigue is 4 cu. units, which corresponds to the norm. The experiment was conducted 90 times to bring it within the permissible error limit and to measure the variability (coefficient of variation). This represent number of drivers (samples) accurately reflects the properties of the general population. The consistency of empirical and theoretical distribution was estimated using Pearson’s fitting criterion for confidence probability. The results of the research showed that the driver of the car during intensive urban traffic makes 400–500 stops, presses the clutch up to 2000 times during 2 h. On the 1st km of the road the driver of the car performs 19,5 operations per minute in average. Works in conditions of an imposed pace and time deficit. These features in the driver’s activity arise when driving a car at high speeds, in a dense traffic flow and in the case of critical traffic situations. Figure 3 provides the experiment’s outcome for selective drivers, regarding evaluation of the functional state. It was found that functional state of the driver changed depending on his/her individual-typological properties. However, the pattern remains consistent with varied amplitude. Driver’s level of tension differs individually during urban traffic congestion. The higher the figure at the time of entering the congestion, higher is the level of tiredness for the driver and higher is the probability of making inadequate decisions. For the driver 2, the functional state in the first hour of driving increases slightly, then somewhat decreases and rises again by the end of second hour. The functional state of the driver 3 rises for one hours, remains stable for another half hour and then rises again. Driver 4 practically does not feel the tension, the functional state is the lowest and it attributed to his confusion and delay in decision making. Various reactions of drivers in the urban traffic congestion caused by next factors: age, temperament, experience, type of car and so on. The results of drivers’ condition during traffic congestion are mapped in Fig. 4.
455
IRSA, c.u.
Assessing Driver Fatigue During Urban Traffic Congestion Using ECG Method
Driving duration, hours
Fig. 3. Change of functional state of drivers during urban driving (inter-peak periods): – driver – driver 2 (aged 50); – driver 3 (aged 60); – driver 4 (aged 30) 1 (aged 20);
Congestion
IRSA, c.u .
Congestion
Duration of movement, min
Fig. 4. Change of functional state of drivers during urban driving (rush hour): – driver 1 – driver 2 (aged 45); – driver 3 (aged 58); – driver 4 (aged 65) (aged 33);
The functional state of driver increases during traffic congestion and then keeps of increasing (Fig. 4). After the first traffic congestion, the functional state of the drivers 1, 2, 3 somewhat stay worst compared to the state of the driver 4. This explains via individual behaviour of the drivers, heavy traffic and prolong staying in congestion.
456
N. Gyulyev et al.
During the second congestion period, the functional state of driver 4 improves marginally, while the functional state of other drivers deteriorates. This can be attributed to the fact that the functional state of driver 4 exceeds the level of tension. During the second congestion period, functional state of drivers 2, 3 and 4 deteriorates, while the functional state of driver 1 improves. A sharp change in the parameters of the traffic flow (an empty road) affects the values in this case. Comparing the results of Fig. 3 and Fig. 4, we can say that drivers become tired quicker during traffic congestion. The change of IRSA per unit time is much faster than in the case of normal traffic flow. The driver’s age has an inverse relationship with fatigue levels during traffic congestion. 3.2 Data Processing Nonlinear regression model was chosen for the analysis the obtained data. The regression model of assessing fatigue of driver during urban connections was estimated:
Рск
0,018 Вв
1,278 Т C
0, 41
(2)
0,291 Рсп ,
where, Рск– the functional state of the driver when driving outside traffic congestion, conventional units (c.u.); Bv – age of the driver in years; TC – duration of the traffic congestion in minutes; Pcn – the functional state when entering the traffic congestion, c.u. Regression model was developed using the variables: driver’s age, duration of driving, number of lanes on the road, level of comfort inside the vehicle, duration of stay in the urban traffic congestion and functional state of driver prior to traffic congestion. The results and the variation range of the model parameters are shown in Table 2 and Table 3. Table 2. Characteristics of the model for changing functional state of the driver in the traffic congestion Factors
Symbol, dimension
Measurement bordering
Coefficient
Standard error
Student’s t-test
The activity index of regulatory systems before entering the traffic congestion
Pcn , c.u.
2,2–5,8
0,291
0,066
4,36
2,0
During the traffic congestion
T c , min
2–20
1,278
0,122
10,48
2,0
Driver’s age
Bv , years
19–67
0,018
0,006
2,98
2,0
Actual
Calculated
Assessing Driver Fatigue During Urban Traffic Congestion Using ECG Method
457
As evidenced in Table 2 and Table 3 it is clear that only three factors were significant. Their significance is evidenced by the higher calculated value of Student’s criterion and absence of zero in the confidence intervals of the model coefficients. The multiple correlation coefficient of the model was 0,96, and the average approximation error was 19,2%. Based on the results we can conclude that functional state of the driver changes during traffic congestion. Table 3. Confidence intervals of model coefficients Factors
Lower bound
Upper bound
The activity index of regulatory systems before entering traffic congestion
0,006
0,031
During the traffic congestion
1,039
0,172
Driver’s age
0,160
1,518
Based on the model outcome we can state that driver’s functional state increases while staying in traffic congestion. The most significant factor influence on the driver’s functional state in urban traffic congestion is its duration and initial functional state of driver while entering the congestion. To a lesser degree, but significantly, the driver’s fatigue level is affected by driver’s age. The nature of sensitivity of the final state of the driver is shown in Fig. 4. In Fig. 5 and Fig. 6 has a smooth curvature devoid of any sudden crests or falls (increase and fall of the functional state), this is due to the fact that the graphs show the emotional levels of an average driver and are devoid of any momentary shocks. Experiments show that there are changes (differences) in the functional state, depending on the age and the initial values of the fatigue level while entering the traffic congestion. Therefore, we are of the opinion that further research is required to account for personal and other individual characteristics of the drivers, and might include, categorizing them based on temperament. The model - influence of traffic congestion on the functional state of the average driver, allows us to predict changes in their state, depending on the age and the duration of the traffic congestion. The value of the initial functional state affects the driver’s functional state during his stay in traffic congestion differently. Therefore, if the driver entering into traffic congestion with an initial states of 5–6 c.u., then within the first 3–7 min his functional state improves by 10–12%. These tendencies are inherent to drivers of all ages. The most significant impact of the duration of the traffic congestion is observed on drivers older drivers. During the analysis of experimental data, it was found that in some cases, the values of the level of fatigue at the exit of the congestion were less than the indication of the level of fatigue at the entrance to the traffic jam. This indicates that some drivers are not deteriorating in traffic jams. To clarify the obtained patterns, there is a need to develop appropriate models for changing the functional state of drivers with different temperaments. However, in the vast majority of drivers the functional condition deteriorates,
458
N. Gyulyev et al.
Fig. 5. Change of the final functional state (Pck ) – Y of the average driver age 20 years depending on the starting level (Pcp ) – Z; and the duration of the urban traffic congestion (T c ) – X
Fig. 6. Change of the final functional state (Pck ) – Y of the average driver age 60 years depending on the starting level (Pcp ) – Z; and the duration of the urban traffic congestion (T c ) – X
which leads to an increase in reaction time and a decrease in the level of safety of the city’s transport system.
4 Conclusions The driver is the most significant link in the «human-machine-environment» system. The reliability of this system depends on 70–80% of his actions. Therefore, while improving
Assessing Driver Fatigue During Urban Traffic Congestion Using ECG Method
459
the design of cars or roads and while creating new technology, psycho-physiological and personal behaviour and human capabilities should always be considered. Harmonization of human and technology is essential for a unified system. While addressing these issues, it is detrimental that which capabilities are appropriate to leave for people, and which ones should be performed by automatic devices. Consequently, both the human activity in its form and content, and the automation policy with respect to different types of technical systems will depend substantially on the allocation of tasks and functions. The allocation of functions between a person and machine is usually carried out according to the principle of prevailing opportunities. The driver of the vehicle is the operator of a complex «man – technology – environment» system. His common task is to receive and process incoming information, to make decisions and to perform control actions. The performance of the operator of any control system depends on the following factors: the characteristics of the incoming information (information flow density, the strength of signals, their duration, the spatial location of the information source, perception ability); the conditions of activity (information flow uniformity, information overload or lack of information); individual characteristics of the operator (psycho-physiological and personal behaviour, resistance to the effects of negative external factors and obstacles, the training and driving experience and age) the operator’s condition (fatigue, illness, mental agitation or depression). Models of influence of congestion on a functional condition of the driver are developed. The driver’s fatigue level increases while in a traffic jam. The reaction of drivers with different types of nervous system is significantly different. Thus, the driver’s fatigue level and the initial state have the most significant effect on the driver’s fatigue level. The results of the work can be used by city authorities and transport companies in the development and organization of the route system in order to minimize the time spent in traffic jams for drivers and thus prevent the excess of fatigue and reaction time. Managers of passenger enterprises can use the results of research to adjust the modes of work and rest for drivers when working on routes by adjusting downtime at the end points. Also, the results of research can be used in professional selection and training of drivers by relevant organizations to develop recommendations in the activity regularities are received. In different traffic jams of different duration, the state and reaction time change, which affect the probability of an accident. At the same time, the identification of patterns of speed changes for different parameters of the route, taking into account the human factor, is today an unsolved problem. It is noted that the use of mathematical methods in the study of the driver as the main link of the system «human-machine-environment» is of paramount importance. When organizing traffic in cities, it is necessary to consideration the results of research, which largely depends on road safety. Further research should be conducted to determine changes in the level of fatigue of drivers of different temperaments.
References Afanasieva, I., Galkin, A.: Assessing the information flows and established their effects on the results of driver’s activity. Arch. Transp. 45(1), 7–23 (2018). https://doi.org/10.5604/01.3001. 0012.0938
460
N. Gyulyev et al.
Vaevsky, R.M.: Mathematical Analysis of Changes in Heart Rate During Stress. Nauka, Moscow (1984) Baevskii, R.M.: Analysis of heart rate variability in space medicine. Hum. Physiol. 28(2), 202–213 (2002) Burko, D., Lobashov, O., Prasolenko, O., Gyulyev, N., Kumar, C.: Establishing patterns of the urban transport flows functioning on urban network parameters. Transp. Res. Procedia 48, 793–800 (2020) Chebanyuk, K., et al.: Pedestrians influence on the traffic flow parameters and road safety indicators at the pedestrian crossing. Transp. Rese. Procedia 45, 858–865 (2020) Eboli, L., Mazzulla, G., Pungillo, G.: Combining speed and acceleration to define car users’ safe or unsafe driving behaviour. Transp. Res. Part C Emerg. Technol. 68, 113–125 (2016) Fiondella, L., Xing, L.: Discrete and continuous reliability models for systems with identically distributed correlated components. Reliab. Eng. Syst. Saf. 133, 1–10 (2015) Galkin, A., Davidich, N., Filina-Dawidowicz, L., Davidich, Y.: Improving the safety of urban freight deliveries by organization of the transportation process considering driver’s state. Transp. Res. Procedia 39, 54–63 (2019) Galkin, A., et al.: Modelling truck’s transportation speed on the route considering driver’s state. Transp. Res. Procedia 30, 207–215 (2018) Grabarek, I.: Projektowanie ergonomiczne s´rodków transportu miejskiego. Oficyna Wydawnicza Politechniki Warszawskiej, Warszawa (2018) Gyulyev, N.: Factor analysis of influence of traffic jams on the state of drivers of different temperaments. East. Eur. J. Enterp. Technol. 5(4), 29–32 (2012). http://journals.uran.ua/eejet/article/ view/4524. (in Russian). Access Gyulyev, N., et al.: Development of models for assessing a driver’s failurefree operation in a transportation system under conditions of traffic congestion. East. Eur. J. Enterp. Technol. 1(3), 24–38 (2020) Gyulyev, N., Lobashov, O., Dulfan, S., Kush, Y., Zagrebelko, Y.: Regularities of changing the functional state of the driver in the transport system in case of traffic congestion. Munic. Econ. Cities 3(163), 187–193 (2021). https://khg.kname.edu.ua/index.php/khg/article/view/5802 Gyulyev, N.U., Lobashov, O.O., Shkabara, O.V., Doroshenko, A.A.: About influence of the driver’s temperament on the time of his reaction in the road jam. Munic. Econ. Cities (140), 86–90 (2018). https://khg.kname.edu.ua/index.php/khg/article/view/5148 Gyulyev, N., Prasolenko, O., Litomin, E., Zinchenko, D.: Study of the influence of road congestion on the fatigue level of a sanguine driver. Technol. Audit Prod. 6/2(50), 31–35 (2019) González-Hernández, B., et al.: The driver’s visual perception research to analyze pedestrian safety at twilight. Transp. Res. Procedia 45, 827–834 (2020) Heart Rate Variability: Standards of Measurement, Physiological Interpretation and Clinical Use, Circulation, 93 (1996) Issa, Y.: Effect of driver’s personal characteristics on traffic accidents in Tabuk city in Saudi Arabia. J. Transp. Lit. 10(3), 25–29 (2016) KhAI-MEDIKA - oborudovanie dlya funktsionalnoy diagnostiki (2016). http://www.khai-med ica.com Klyatis, L.: Introduction to successful predicting of product performance (Reliability, durability, safety, quality, recalls, profit, life cycle cost, and others) (No. 2015-01-0487). SAE Technical Paper (2015) Kulbashna, N., Lynnyk, I., Galkin, A., Prasolenko, O., Dulfan, S.: Safety assessment of adjacent roads sections via maximum entropy driver’s perception field. Commun. Sci. Lett. Univ. Zilina 22(4), 182–190 (2020) Prasolenko, O., et al.: Creating safer routing for urban freight transportation. Transp. Res. Procedia 39, 417–427 (2019a)
Assessing Driver Fatigue During Urban Traffic Congestion Using ECG Method
461
Prasolenko, O., Lobashov, O., Bugayov, I., Gyulyev, N., Filina-Dawidowicz, L.: Designing the conditions of road traffic in the cities taking into account the human factor. In: 6th International Conference on Models and Technologies for Intelligent Transportation Systems, IEEE Advancing Technology for Humanity, p. 8883381 (2019b) Lobanov, E.M.: Proektirovanie dorog i organizatsiya dvizheniya s uchetom psihofiziologii voditelya. Moscow, Transport, 311 (1980). (In Russian) Lieberman, E., Rathi, A.K.: Traffic simulation. Traffic flow theory (1997) Maji, A., Tyagi, A.: Speed prediction models for car and sports utility vehicle at locations along four-lane median divided horizontal curves. J. Mod. Transp. 26(4), 278–284 (2018). https:// doi.org/10.1007/s40534-018-0162-1 Rossi, R., et al.: A driving simulator study exploring the effect of different mental models on ADAS system effectiveness. In: De Paolis, L., Bourdot, P. (eds.) AVR 2020. LNCS, vol. 12242, pp. 102–113. Springer, Cham (2020). https://doi.org/10.1007/978-3-030-58465-8_7 Wang, J., Wu, J., Li, Y.: The driving safety field based on driver–vehicle–road interactions. IEEE Trans. Intell. Transp. Syst. 16(4), 2203–2214 (2015) Zhang, H., Wu, C., Yan, X., Qiu, T.Z.: The effect of fatigue driving on car following behaviour. Transp. Res. F Traffic Psychol. Behav. 43, 80–89 (2016)
Technology Review for Guiding Persons in Airports and Other Hubs Axel B¨orold1(B) , Eike Broda2 , Nicolas Jathe1 , Dirk Schweers2 , Tobias Sprodowski1 , Waldemar Zeitler1 , and Michael Freitag1,2 1
2
BIBA - Bremer Institut f¨ ur Produktion und Logistik GmbH at the University of Bremen, Hochschulring 20, 28359 Bremen, Germany [email protected] Faculty of Production Engineering, University of Bremen, Badgasteiner Straße 1, 28359 Bremen, Germany
Abstract. The COVID-19 pandemic brought public life to a near standstill. Precautionary practices, such as face masks and safe distance, were established to protect people. In addition, various tracking methods were developed to detect possible contacts. In this paper, we review suitable technologies to indicate a solution for a people guidance system, which actively prevents these contacts by suggesting routes through large areas (e.g. airport terminals or train stations). By tracking the people and using destination information, e.g. from tickets, the system should be capable of calculating routes and visualise the suggestion to each person individually. Keywords: Pandemic measures assistance · Distributed control
1
· Human tracking · People
Introduction
Epidemic situations, like the COVID-19 pandemic in the early 2020s, present new challenges for operators of infrastructure with highly attended places like airports, train stations or shopping malls. While it is in public and economic interest to continue the operation, people visiting and working in these places need to be protected. The two consistent safety elements during the COVID-19 pandemic were the usage of face masks and maintaining safe distances between people. Regarding the safety distance, this was the only possible practice in previous pandemics [29]. Different software and hardware solutions were developed to ensure safety measures. In most cases, reactive systems were implemented: Amazon for example developed a system where they measure the distance between people via human tracking [30]. The result is presented on a monitor and signalising an alarm if employees undermine the minimum distance. To become aware of an violation, participants must actively look at the monitor. Other used systems are based on wearables which measure the distance with Bluetooth Low Energy: These systems are used for example at Ford in Plymouth [27] or are sold by c The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 M. Freitag et al. (Eds.): LDIC 2022, LNLO, pp. 462–473, 2022. https://doi.org/10.1007/978-3-031-05359-7_37
Dynamics in Logistics
463
Software AG together with an IoT device from Cumulocity [29]. A drawback is that as further assistance devices are required, which have to be handed out and picked up again, this might be inappropriate for open space with a high number of visiting people due to occurring long waiting queues. All these methods have the disadvantage of being passive systems, i.e. they only control violations that have already occurred and do not provide an active solution how to prevent safety distance violations before they occur. The basic idea of our system is based on the idea of prevention: Each person’s destination is identified and a goal is chosen a priori. An individual routing is provided actively to each person that ensures safety distances to other people in the supervised space. Therefore, different technologies are combined for recognising and tracking people, identifying their destination, planning an individual, feasible route and visualising this planned route. The necessary components and the basic structure of this system are shown in Fig. 1. The basic concept and the requirements are described in [37], while this paper covers the technologies needed to implement such a support system. Input
Calculation
Output
Routing
Routing instructions (via HCI)
Destination
Human Recognition
Fig. 1. Components of the system
In the following, we describe the state of the art of human recognition, routing and human computer interaction (HCI). Section 3 presents the chosen technologies based on the state of the art solutions, with the guidance of people as the main focus point. The paper closes with our conclusion and an outlook.
2
State of the Art
As shown in the previous section, our system consists of four components. The first input is the destination of the individual person, e.g. the gate at an airport or the track at a train station. This information is most accurate if it is provided by the visitor, e.g. by using a smartphone app. Another option is to combine data from different sources, e.g. people are scanning their boarding pass at an airport to enter the security area. At the same moment, this person is recognised by the system and the travel destination and gate, provided by airport systems, can be assigned to this person.
464
2.1
A. B¨ orold et al.
Human Recognition
This subsection addresses different ways to detect and track people as well as identify their movement behaviour. These tasks can be realised by using different data sources and combining them. Due to our given scenario, we focus on data from external video streams and device sensors in mobile devices, such as the built-in accelerometers. The movement behaviour of individuals can be determined by tracking positions over time. When using mobile devices, this can be done with one-dimensional convolutional neural networks (CNNs) as presented in [21] and [46]. CNNs can operate on the x, y, and z acceleration data represented as a one-dimensional vector with the sensor’s values in each direction. The accelerometer provides data based on the direction of motion of the device, which can be interpreted as the direction of motion of the user. This approach can be combined with person detection from a live stream to provide real-time detection of human movement behaviour. In this case, a pose-based CNN (P-CNN), as presented in [6], benefited the overall system for detecting people and their movement behaviour. The P-CNN provides further estimates of the direction of motion of individuals by analysing body parts and inferring the current pose of the detected individuals. The combination of these approaches can estimate the movement behaviour of individuals and additionally identify individuals in a video surveillance environment. Person detection and tracking can also be achieved by using solely external video data. This might be necessary when motion data from mobile devices is not available. Person detection is based on individual frames, while tracking is achieved by analysing the positions of detected persons over time. Currently, the most advanced recognition algorithms are based on CNNs, and recently there has also been active research on transformational models for classification tasks [10]. In the near future, they may also be good candidates for recognition tasks and could be used as replacements instead of CNN-based algorithms. CNN-based detection networks can be divided into two classes, the one-stage and the twostage detectors: Single-stage detectors capture the people in the incoming image in a single computation run, are usually faster, and the computation time is more consistent across different scenes. The drawback is lesser accuracy than of two-stage detectors. Among the best-known algorithms in this category, which are still considered as benchmarks for visual person recognition, are based on YOLO (You Only Look Once) [5,32–34]. Other well-known single-stage algorithms include SSD (Single Shot Detector) [24], RetinaNet [22], RefineDet [47], Efficientnet [41], MobileNets [17], and ExtremeNet [48]. In contrast, two-stage detectors identify regions of interest in the first pass with little computational effort, which are then classified in the second stage. Current two-stage detectors include Fast R-CNN [12], Faster R-CNN [35], HyperNet [20], and R-FCN [7]. Since the detection algorithm is a classification of given first stage proposals, the accuracy is usually higher than that of single-stage detectors. The drawback is that the computational costs depend strongly on the number of proposals and thus on the scene.
Dynamics in Logistics
2.2
465
Routing Components
In this subsection we give an outline of how the information from the human recognition is used to obtain a feasible and reliable solution for people to maintain safety distances and access their target in an appropriate timespan. The routing component is divided into a global and a local component: While the first one is supposed to calculate an abstract route from the current position to the individual destinations expressed in waypoints, the latter one tracks and guides between the obtained waypoints while ensuring distance measures to other people. For the global routing component a geographical Open Street Map (OSM) projection could be used to depict the underlying infrastructure [28]. The global routing component obtains the current status of the passenger by one of the inputs and calculates an abstract to the chosen destination of the passenger, which might be obtained from a smartphone application or the gate information of the flight ticket. Here, shortest path algorithms (e.g. Dijkstra algorithm [9] or A algorithm [13]) could be applied on the underlying OSM map. All people have individual goals/targets where to be navigated, but have to ensure distance measures to each other. The problem how people should be routed in such an infrastructure could be categorised as a path planning problem, which yields a distributed setting due to the natural distinction of any passenger as an individual agent embedded in a multi-agent system [31]. Hence, such a setting allows to utilise routing algorithms and methods from control theory for path planning. In large infrastructures, a high number of people has to be considered. Thus, the algorithms and methods should utilise a moderate computational effort and be scalable. Centralised systems and methods might not match short computation times and complexity rises with a high number of people. Hence, distributed or decentralised systems, capable of controlling high numbers of agents, were proposed in the last decades. In the field of robotics, Distributed Model Predictive Control (DMPC) gained a lot of attention due to its capability to include hard constraints and the ability to achieve real-time capabilities [8,18,44]. To meet the requirements of scalability, real-time and coordination estimation, the local routing component needs to incorporate a Multi-Agent System, which handles each person as an agent and calculates a path from waypoint to waypoint based on an information exchange, which has to be established between the agents. Hence, the scenario could be classified as a distributed noncooperative scenario [39] utilising a distributed model predictive controller [18]. Each agent is equipped with a controller, which acts on the available information. Based on the available controllers, the distributed scenario is achieving a suboptimal solution, which is at best the Nash equilibrium due to the unavailability of information about the goals of the other agents. But still, this procedure may achieve an acceptable solution and ensures the constraints (here safety measures). An optimal control problem (OCP) can be formulated, with the inclusion of other passenger positions based on the tracking component as constraints. The OCP is formulated to utilise a cost function, which penalises the distance to
466
A. B¨ orold et al.
the next waypoint from the global route and the deviation from the reference route. The necessary positions of other people are received and incorporated as distance constraints in the OCP. To equip the OCP with an underlying movement model, linearised pedestrian motion models could be used to additionally reduce the computational burden [3]. Then, by incorporating the predictions of the other persons as constraints, the OCP is solved over a fixed prediction horizon. An optimal path with respect to the distance measures is obtained and communicated to the other agents, which enables them to incorporate the predicted path as distance constraints in their OCP. The whole procedure repeats when the next update of the passenger position arrives, which closes the control loop. An overview of the necessary information to create the constraints is depicted in Fig. 2.
Fig. 2. Information for the optimisation problem
Considering the communication effort, for each time instant the whole predicted path has to be communicated to all other agents, which increases the communication load between the agents. Therefore, different approaches were proposed to lower the communication burden for the coordination. The first approach is based on event-triggering [11], where a threshold for a deviation of a system state is defined, such that the next path prediction is communicated when
Dynamics in Logistics
467
the system reaches this threshold. This method is mostly suitable for systems with a slowly changing dynamic, while other approaches use the quantisation of states to reduce the burden for networks with limited bandwidth [23]. With the assumption that people follow (mostly) the instructions, the evolving path predictions provide the potential of similarity in two successive time instants as the people should be close to the next predicted state: Hence, the communicated prediction can be reduced by sending only updates of the recent changes [38]. 2.3
Human Computer Interaction (HCI)
In the previous subsections we evaluated the applicability of human recognition and routing components. In this part we look upon various interfaces which can be used to communicate the routing results to a person. Since interaction is situation-dependent, there are different interfaces possible to stimulate the visual, tactile or acoustic senses efficiently from the outside. An intelligent system has to decide on its own which interface is most appropriate. The interaction with such an interface is restricted to the limited cognitive resources of humans [45]. The study in [2] with ambient intelligence systems states that intelligent systems may cause competition between human cognitive resources and the users will face a high cognitive workload. For a human-centred system to be accepted by humans different stimuli have to be well defined. Furthermore, not all interfaces are useful in a crowded environment, as can be seen in the following overview. Tactile perception for interaction is only possible via mobile devices carried by humans. If these are personal devices, a connection to the system, e.g. through an application, is needed. Alternatively, additional non-private devices are handed out at the entrance and are collected at the exit. Acoustic stimuli can be emitted via loudspeakers and headphones. In case of the headphones, a likewise integration to the system is needed, which means either a private device or a borrowed one, followed by the same procedure as with the mobile devices. Speakers are useful in general situations such as announcements or alerts. In everyday situations, they are less suitable due to the nondirectional dissemination of information and the difficulty of unambiguous routing. Directional loudspeakers or sound showers can negate this up to a certain extent. They show their strength in places where the distance to the listener does not change, such as navigation announcements in cars, or in places where people automatically move into the sound range, such as in front of an artwork in a museum. However, for moving targets, a supplementary actuator for changing the direction of the speakers is needed, or a sound array has to be installed, which changes the sound direction by superimposition. So far this has not been tested with moving persons, but there is a study of microphone arrays moving in a static sound field which observed a Doppler effect [15]. Therefore, it is likely that these effects will also occur in our scenario, with static or moving objects in moving sound fields. A possibility to compensate this still needs to be found. The best output for the spatial-orienting task is the visual sense. It is much better than acoustic outputs for spatial locations. This is due to the calculation of the localisation from the auditory sources using acoustic indications such as
468
A. B¨ orold et al.
time and level differences between the ears and spectral indications introduced by the head and the pinnae [36]. Public as well as private displays can be used to present visual information. In addition to spatial information, this can also be of an informative nature. To present the routing information in a visual way an augmented reality (AR) system can be used, which projects the information on the ground. This has a positive side effect in enhancing public awareness of other people, because it can be seen by everyone [26]. The visual aids projected in this way can be presented in a reduced form. Therefore, this is easier to be understood without special knowledge and could be used to control the pace of the people [40].
3
Selection of Technologies
Based on the previously presented state of the art, we will select suitable technologies to be used in the proposed system. For the human recognition we propose using two-stage detection algorithms and, if available, in combination with mobile phone data. The reason is that two-stage detectors are usually more accurate than single-stage detectors. The computation time can be reduced because in a tracking application, not every detection of the first stage needs to be checked by the second classification stage in each image. Therefore, depending on the implementation, two-stage algorithms are more suitable for tracking. Furthermore, it is possible to evaluate descriptions of individual persons in the latent space and thus improve the recognisability of individual persons from different camera poses. These considerations lead to what is known as ”multiple object tracking” [14]. The tracking itself can be done by analysing the positions through one-dimensional CNNs. In this network, possible data from mobile devices can also be included to further improve robustness. Considering the global routing component, a map of the infrastructure has to be provided to carry out the routing. In this regard, Open Street Map (OSM) gained large interest in the last years, which allows the customisation of maps for own purposes [28]. As a global router component, various router projects are available for routing based on OSM. Allowing for using customised maps, the OSRM routing engine is a well-maintained and platform-independent project [25]. Furthermore, routing instructions and turning directions are provided, which could be used for one-ways, e.g. separate sections for departure and arrival in airport terminals. Other routers as Routino [4] or OpenTripPlanner [42] are restricted to certain countries or cities. Alternatives as GraphHopper [1] could also be considered, but might be limited considering computational effort due to the usage of a Java Virtual Machine and missing turning restrictions. For the local routing component, divided into a distributed multi-agent-system, the problem formulation as a (local) nonlinear Optimal Control Problem (OCP) requires a nonlinear solver with the ability to handle inequality constraints and computational time-efficient platform-independent frameworks. For optimal control problems, several software packages are fulfilling these requirements, such as ACADO [16], IPopt [43] or NLOpt [19]. As the latter allows for a derivative-free
Dynamics in Logistics
469
formulation of the problem, which could be a necessary requirement regarding quantisation methods [38] and provides therefore most flexibility for the chosen underlying model, this could be a modest compromise regarding real-time capabilities, which are definitely considered by ACADO. Input
Destination App-based, Scan of ticket
Human Recognition Camera-based, CNN
Calculation
Routing Distributed planning (OSM map & DMPC)
Output
Routing instructions via HCI Projection, App-based
Fig. 3. Components of the system
We propose a projection-based AR system as the system’s output for directional information, as this output can be seen by most people without distracting others and further increasing the noise level in a crowded environment. By using projections, people who are not registered in the system can also receive information. These projections can also be used for personal information when information density is low. For additional information, the existing infrastructure can be used, such as interactive monitors, which are primarily used for advertising, or if one is registered, this information can also be pushed directly to one’s smartphone. In summary, for the components of the proposed system, in Fig. 3 the chosen technologies are shown.
4
Conclusion and Outlook
In this paper, we have presented different technologies for the recognition of persons, for HCIs, and the routing of persons to design a people guidance system, which could be applied in airports and other hubs, especially during a pandemic. With this guidance system, the basis for a new opportunity of guiding people through unknown and crowded locations while ensuring safety distances, has been formed. This results in a digital twin of the infrastructure with tracking and guiding of people, who are part of the infrastructure parallelly in real-time. With this digital twin, it is possible to support the people by means of visual,
470
A. B¨ orold et al.
haptic, or auditive signals accordingly. The digital twin additionally allows to check if a person follows the proposed route and to adapt his route if necessary. If a person is not following the proposed route, other people might be re-routed to ensure the safety distances. This support system enables additional usages: At first, if the path and its corresponding time are stored for each person, the measurements of distances between people is also possible at a later time instant. This is especially interesting for the follow up of diseases like COVID-19 to identify people who have been near to an infected person. This way, people are also included who did not use a contact tracing app, which tracks contacts with other people via Bluetooth. Second, on a higher level, mapping the positions of people on a heat map over a real map of the building could help to identify bottlenecks and areas of high usage to derive necessary cleaning intervals of such areas in pandemic situations. To this end, this system allows for a more precise contact tracing. A check-insystem, that registers people to be inside a certain location at a certain time only, would indicate all people inside a large area as infected. Hence, this system could support an efficient and scalable possibility to contain infection risks.
References 1. Graphhopper. (2021). https://github.com/graphhopper/graphhopper/ 2. Bakker, S., Niemantsverdriet, K.: The interaction-attention continuum: considering various levels of human attention in interaction design. Int. J. Des. 10(2), 1–14 (2016) 3. Batkovic, I., Zanonl, M., Lubbe, N., Falcone, P.: A computationally efficient model for pedestrian motion prediction. In: 2018 European Control Conference, ECC 2018, pp. 374–379 (2018). https://doi.org/10.23919/ECC.2018.8550300 4. Bishop, A.M.: Routino: router for openstreetmap data (2021). http://www.routino. org/ 5. Bochkovskiy, A., Wang, C.Y., Liao, H.Y.M.: Yolov4: optimal speed and accuracy of object detection. arXiv preprint arXiv:2004.10934 (2020) 6. Ch´eron, G., Laptev, I., Schmid, C.: P-CNN: pose-based CNN features for action recognition. In: Proceedings of the IEEE International Conference on Computer Vision, pp. 3218–3226 (2015). https://doi.org/10.1109/ICCV.2015.368 7. Dai, J., Li, Y., He, K., Sun, J.: R-FCN: object detection via region-based fully convolutional networks. In: Advances in Neural Information Processing Systems, pp. 379–387 (2016) 8. Defoort, M., Kokosy, A., Floquet, T., Perruquetti, W., Palos, J.: Motion planning for cooperative unicycle-type mobile robots with limited sensing ranges: a distributed receding horizon approach. Robot. Autonom. Syst. 57(11), 1094–1106 (2009). https://doi.org/10.1016/j.robot.2009.07.004 9. Dijkstra, E.W.: A note on two problems in connexion with graphs. Numer. Math. 1, 269–271 (1959). https://doi.org/10.1007/BF01386390 10. Dosovitskiy, A., et al.: An image is worth 16x16 words: transformers for image recognition at scale. arXiv preprint arXiv:2010.11929 (2020) 11. Eqtami, A., Dimarogonas, D.V., Kyriakopoulos, K.J.: Event-triggered strategies for decentralized model predictive controllers. IFAC Proc. 44(1), 10068–10073 (2011). https://doi.org/10.3182/20110828-6-IT-1002.03540
Dynamics in Logistics
471
12. Girshick, R.: Fast R-CNN. In: Proceedings of the IEEE International Conference on Computer Vision, pp. 1440–1448 (2015) 13. Hart, P.E., Nilsson, N.J., Raphael, B.: A formal basis for the heuristic determination of minimum cost paths. IEEE Trans. Syst. Sci. Cybern. 4(2), 100–107 (1968). https://doi.org/10.1109/TSSC.1968.300136 14. Ho, K., Keuper, J., Keuper, M.: Unsupervised multiple person tracking using autoencoder-based lifted multicuts. arXiv preprint arXiv:2002.01192 (2020) 15. Hong, J.Y., He, J., Lam, B., Gupta, R., Gan, W.S.: Spatial audio for soundscape design: recording and reproduction. Appl. Sci. 7, 627 (2017). https://doi.org/ 10.3390/APP7060627. https://www.mdpi.com/2076-3417/7/6/627/htm, https:// www.mdpi.com/2076-3417/7/6/627 16. Houska, B., Ferreau, H., Diehl, M.: ACADO toolkit - an open source framework for automatic control and dynamic optimization. Opt. Cont. Appl. Methods 32(3), 298–312 (2011). https://doi.org/10.1002/oca.939 17. Howard, A.G., et al.: Mobilenets: efficient convolutional neural networks for mobile vision applications. arXiv preprint arXiv:1704.04861 (2017) 18. Jia, D., Krogh, B.H.: Distributed model predictive control. In: Proceedings of the American Control Conference, IEEE, Arlington, pp. 2767–2772 (2001). https:// doi.org/10.1109/ACC.2001.946306 19. Johnson, S.G.: The NLopt nonlinear-optimization package (2004). http://ab-initio. mit.edu/nlopt 20. Kong, T., Yao, A., Chen, Y., Sun, F.: Hypernet: towards accurate region proposal generation and joint object detection. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp. 845–853 (2016) 21. Lee, S.M., Yoon, S.M., Cho, H.: Human activity recognition from accelerometer data using convolutional neural network. In: 2017 IEEE International Conference on Big Data and Smart Computing (bigcomp), pp. 131–134. IEEE (2017). https:// doi.org/10.1109/BIGCOMP.2017.7881728 22. Lin, T.Y., Goyal, P., Girshick, R., He, K., Doll´ ar, P.: Focal loss for dense object detection. In: Proceedings of the IEEE International Conference on Computer Vision, pp. 2980–2988 (2017) 23. Liu, S., Xie, L., Quevedo, D.E.: Event-triggered quantized communication based distributed convex optimization. IEEE Trans. Control Netw. Syst. 5870(99), 1–11 (2016). https://doi.org/10.1109/TCNS.2016.2585305 24. Liu, W., et al.: SSD: single shot multiBox detector. In: Leibe, B., Matas, J., Sebe, N., Welling, M. (eds.) ECCV 2016. LNCS, vol. 9905, pp. 21–37. Springer, Cham (2016). https://doi.org/10.1007/978-3-319-46448-0 2 25. Luxen, D., Vetter, C.: Real-time routing with OpenStreetMap data. In: Proceedings of the 19th ACM SIGSPATIAL International Conference on Advances in Geographic Information Systems, ACM, Chicago (GIS 2011), pp. 513–516 (2011). https://doi.org/10.1145/2093973.2094062 26. Monastero, B., McGookin, D.K.: Traces: Studying a Public Reactive FloorProjection of Walking Trajectories to Support Social Awareness, pp. 1–13. Association for Computing Machinery, New York (2018). https://doi.org/10.1145/ 3173574.3174061 27. Naughton, K.: Ford Tests Buzzing Wristbands to Create Safe Work Distances (2020). https://www.bloomberg.com/news/articles/2020-04-15/ford-testsbuzzing-distancing-wristbands-to-keep-workers-apart 28. OpenStreetMap Contributors. Planet dump (2017). https://planet.osm.org, https://www.openstreetmap.org
472
A. B¨ orold et al.
29. Polychronaki, M., Patrikakis, C.: Social distancing: technology to the rescue. In: Kavoura, A., Havlovic, S., Totskaya, N. (eds.) Strategic Innovative Marketing and Tourism in the COVID-19 Era. SPBE, pp. 95–101. Springer, Cham (2021). https:// doi.org/10.1007/978-3-030-66154-0 11 30. Porter, B.: The company’s latest innovation provides real-time social distancing feedback and we plan to open source the technology. In: AboutAmazon, p. 1 (2020) 31. Purwin, O., D’Andrea, R., Lee, J.W.: Theory and implementation of path planning by negotiation for decentralized agents. Robot. Autonom. Syst. 56(5), 422–436 (2008). https://doi.org/10.1016/j.robot.2007.09.020 32. Redmon, J., Farhadi, A.: Yolo9000: better, faster, stronger. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp. 7263–7271 (2017). https://doi.org/10.1109/CVPR.2017.690 33. Redmon, J., Farhadi, A.: Yolov3: an incremental improvement. arXiv preprint arXiv:1804.02767 (2018) 34. Redmon, J., Divvala, S., Girshick, R., Farhadi, A.: You only look once: unified, real-time object detection. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp. 779–788 (2016). https://doi.org/10.1109/ CVPR.2016.91 35. Ren, S., He, K., Girshick, R., Sun, J.: Faster R-CNN: towards real-time object detection with region proposal networks. Adv. Neural Inf. Process. Syst. 28, 91–99 (2015) 36. Roberts, K.L., Summerfield, A.Q., Hall, D.A.: Presentation modality influences behavioral measures of alerting, orienting, and executive control. J. Int. Neuropsychol. Soc. 12(4), 485–492 (2006). https://doi.org/10.1017/S1355617706060620 37. Rohde, A.K., Pupkes, B., Mortensen Ernits, R., Keiser, D., L¨ utjen, M., Freitag, M.: Challenges and approaches of non-pharmaceutical interventions for airport operations during pandemic situations. In: Dynamics in Logistics, pp. 1–10 (2022) 38. Sprodowski, T., et al.: Differential communication with distributed MPC based on occupancy grid. Inf. Sci. 453, 426–441 (2018). https://doi.org/10.1016/j.ins.2018. 04.034 39. Stewart, B.T., Venkat, A.N., Rawlings, J.B., Wright, S.J., Pannocchia, G.: Cooperative distributed model predictive control. Syst. Contr. Lett. 59(8), 460–469 (2010). https://doi.org/10.1016/j.sysconle.2010.06.005 40. Takahashi, I., Oki, M., Bourreau, B., Kitahara, I., Suzuki, K.: FUTUREGYM: a gymnasium with interactive floor projection for children with special needs. Int. J. Child-Comput. Interact. 15, 37–47 (2018). https://doi.org/10.1016/j.ijcci.2017. 12.002 41. Tan, M., Le, Q.: Efficientnet: rethinking model scaling for convolutional neural networks. In: International Conference on Machine Learning, PMLR, pp. 6105– 6114 (2019) 42. Trimet. Open trip planner (2021). https://github.com/opentripplanner/ OpenTripPlanner 43. W¨ achter, A., Biegler, L.T.: On the implementation of an interior-point filter linesearch algorithm for large-scale nonlinear programming. Math. Program. 106(1), 25–57 (2006). https://doi.org/10.1007/s10107-004-0559-y 44. Xi, Y.G., Li, D.W., Lin, S.: Model predictive control - status and challenges. Acta Autom. Sin. 39(3), 222–236 (2013). https://doi.org/10.1016/S18741029(13)60024-5 45. Xu, W., Dainoff, M.J., Ge, L., Gao, Z.: From Human-Computer Interaction to Human-AI Interaction: New Challenges and Opportunities for Enabling HumanCentered AI (2021). arxiv:2105.05424
Dynamics in Logistics
473
46. Zeng, M., et al.: Convolutional neural networks for human activity recognition using mobile sensors. In: 6th International Conference on Mobile Computing, Applications and Services, pp. 197–205 (2014). https://doi.org/10.4108/icst. mobicase.2014.257786 47. Zhang, S., Wen, L., Bian, X., Lei, Z., Li, S.Z.: Single-shot refinement neural network for object detection. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp. 4203–4212 (2018) 48. Zhou, X., Zhuo, J., Krahenbuhl, P.: Bottom-up object detection by grouping extreme and center points. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. 850–859 (2019)
The Impact of the COVID-19 Pandemic on E-commerce Consumers’ Pro-environmental Behavior Simona Koleva and Stanislav Chankov(B) Jacobs University Bremen, Campus Ring 1, 28759 Bremen, Germany [email protected]
Abstract. The advent of COVID-19 led to a surge in online shopping. As a result of the health crisis, companies reduced their priorities on environmental issues. However, consumers’ concern for sustainability is on the rise. Thus, the purpose of this paper is to examine the impact of the COVD-19 pandemic on e-commerce consumers’ pro-environmental behavior. Accordingly, we conduct an online survey exploring consumers’ online shopping frequency and engagement in environmentally friendly practices before, during and after the COVD-19 pandemic. Applying the Wilcoxon test to compare these three stages, we are able to investigate the shift in e-commerce consumer pro-environmental behavior triggered by COVID-19. The results indicate that the shopping frequency has increased substantially since the start of the pandemic, but will drop down after the end of the pandemic. Moreover, the COVID-19 pandemic was detrimental to consumers’ pro-environmental behavior: during the pandemic consumers showed a tendency towards less environmentally friendly behavior but they have strong intentions to adopt more eco-friendly practices after the pandemic ends. Keywords: Sustainability · Last-mile delivery · Online shopping
1 Introduction The onset of the coronavirus (COVID-19) has significantly impacted consumer behavior, and the shift towards online shopping is expected to remain even after the pandemic is over [1]. In countries, where lockdowns were imposed, most physical stores were forced to close leaving no option to consumers other than online shopping [2]. E-commerce has been growing drastically, but the advent of COVID-19 led to a surge in online shopping worldwide [3]. Due to the high infection risk when going out, the demand for residential deliveries in 2020 grew at an unseen rate [4]. For example, the consumers’ spending on the largest online shopping holiday in the world, Alibaba’s annual Singles Day, had been rising steadily from 2011 to 2019; however, in 2020 the sales almost doubled those in 2019 [5]. COVID-19 has sped up the ongoing shift to online retail by five years [6]. Even before COVID-19, the e-commerce growth was estimated to cause severe sustainability challenges with emissions from delivery traffic in the top 100 cities globally being said to rise by 32% from 2019 to 2030 [7]. Further, the profound impact of © The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 M. Freitag et al. (Eds.): LDIC 2022, LNLO, pp. 474–485, 2022. https://doi.org/10.1007/978-3-031-05359-7_38
The Impact of the COVID-19 Pandemic on E-commerce Consumers’
475
the coronavirus health crisis on business operations led to companies reducing their priorities on environmental issues [8]. Hence, specifically e-commerce retailers are even more likely to engage in existing poor packaging practices generating a lot of waste [9], increasing the negative externalities on the environment. However, customers are increasingly concerned about the impact their online shopping behavior has on the environment [10]. Moreover, displaying the environmental impacts of last-mile deliveries influences E-commerce customers, and generally makes them more likely to choose a more sustainable last-mile delivery [11]. Further, [12] showed that to save 100 g of CO2 emissions from the delivery of their package, consumers would be willing to pay approximately 1e more or wait about 1–2 days longer (or as long as needed) for the delivery. Thus, customer awareness of the pressing need for change towards a more sustainable direction is quite strong, however, it is uncertain whether that awareness will still result in a more environmentally responsible behavior during times of a crisis. On the one hand, a survey showed that consumer engagement in sustainability has deepened during the COVID-19 crisis [13]. 57% of the respondents have made significant changes to their lifestyles to reduce their environmental impact, and more than 60% report going out of their way to recycle and purchase products in environmentally friendly packaging [13]. On the other hand, consumers also adopted some environmentally unfriendly practices such as higher consumption of plastic-packaged food and groceries and higher use of disposable tableware [14]. With shopping patterns characterized by higher reliance on e-commerce and delivery options during-COVID-19, customers’ expectations have changed forever [15]. Moreover, consumers’ habits which have been accelerated during the pandemic rather than initiated are more likely to persist in the long-term [16]. Hence, the purpose of this paper is to examine the impact of the COVD-19 pandemic on e-commerce consumers’ pro-environmental behavior. Accordingly, we conduct an online survey exploring consumers’ online shopping frequency and engagement in environmentally friendly practices before, during and after the COVD-19 pandemic. Applying the Wilcoxon test to compare these three stages, we are able to investigate the shift in e-commerce consumer pro-environmental behavior triggered by COVID-19.
2 Hypotheses Development We develop two main hypotheses focusing on (1) online shopping frequency and (2) pro-environmental behavior. To pinpoint the pandemic’s impact on both, we compare consumers’ behavior before vs. during COVID-19, and during vs. after COVID-19. The COVID-19 pandemic led to an unprecedented expansion of e-commerce worldwide [17]. Governments’ effort to stop virus transmission resulted in the implementation of stringent social distancing measures such as lockdowns [2]. In the meantime, due to the closure of brick and mortar stores, both consumers and businesses “went digital”, ordering and providing items and services online, respectively [17]. COVID-19 triggered a jump in e-commerce’s share of global retail trade from 14% in 2019 to about 17% in 2020 [17]. Based on this, hypothesis H1a was developed: H1a: Consumers shopped online less often pre-COVID-19 than during-COVID-19.
476
S. Koleva and S. Chankov
Moreover, consumers behavior in online shopping during-COVID-19 might serve as a predictor of consumers’ purchasing behavior after the pandemic is over. In the postpandemic world, safety will remain of high importance, therefore, some time is necessary before consumers feel confident to return to in-person interaction with retailers [18]. A survey released in light of the pandemic in 2020 found that 86% of the consumers will continue ordering groceries online even when the preventive measures are lifted [18]. Further, consumers, who recently started shopping online, are likely to continue using e-commerce sites even after the pandemic is over due to their overall positive experience throughout the pandemic [19]. This led to the development of H1b: H1b: Consumers will shop online post-COVID-19 as often as during-COVID-19. Since the beginning of the pandemic, various surveys show that consumers have made more environmentally friendly decisions in the direction of reducing carbon footprint [13, 20, 21]. 57% of the respondents have made significant changes to their lifestyles to reduce their environmental impact, and more than 60% report going out of their way to recycle and purchase products in environmentally friendly packaging [13]. 68% of respondents preferred locally made products [20], while 53% of consumers say they have switched to lesser known brand(s)/organization(s) whose products/services they perceive as sustainable [20]. Finally, 46% showed willingness to pay more for sustainable packaging [21]. Thus, hypothesis H2a was developed: H2a: Consumers engaged in environmentally friendly practices when shopping online pre-COVID-19 to a lesser degree than they do during-COVID-19. The desire for seeking eco-friendly options when shopping online is likely to prevail even after the pandemic [15]. 9 out of 10 respondents in a survey who reported placing environmentally friendly, ethical, or sustainable online orders during the COVID-19 health crisis stated that they are likely to be doing the same in the future [22]. Moreover, consumers’ habits which have been accelerated during the pandemic rather than initiated are more likely to persist in the long-term [16]. This clearly applies to pro-environmental habits as environmental awareness was already growing before COVID-19 emerged. Thus, H2b was developed: H2b: Consumers will continue to engage in environmentally friendly practices when shopping online post-COVID-19 as much as they did during-COVID-19.
3 Research Design To test the developed hypotheses, we needed to investigate e-commerce consumers’ behavior in three stages: pre-, during-, and post-COVID-19. Thus, to test H1a and H1b, participants were asked how often they: (1) shopped online before the start of the pandemic, (2) have shopped online since the beginning of the pandemic and (3) anticipate to order products online after the pandemic. Similarly, to test H2a and H2b, we designed questions about consumers’ involvement in eco-practices pre-, during-, and
The Impact of the COVID-19 Pandemic on E-commerce Consumers’
477
post-COVID-19. Seven different environmentally friendly practices were selected from literature in order to cover a broader spectrum [7, 9, 11, 20, 23]: (1) avoidance of retailers that use excessive packaging, (2) willingness to wait a longer time for shipping, (3) consideration of sustainability as a factor influencing purchase decisions, (4) preference for sustainable brands, (5) preference for local suppliers, (6) preference of pick-up location over home delivery, and (7) avoidance of shopping and returning clothes in multiple sizes. This resulted in a total of 24 Likert-scale questions. The main limitation of the survey lies in the fact that we gather input on consumers’ stated behavior and not actual behavior. This is especially true for post-COVID-19, where participants state anticipated future behavior. This is a common problem in studies on stated actions [24]. The survey was advertised on social media and survey sharing platforms for 2 weeks in May 2021. The participants were at least 18 years old and have shopped online at least once. 420 people took part in the survey on a voluntary basis.
4 Results 4.1 Sample Description and Frequency Distributions Table 1 shows a detailed overview of the sample. Most of the respondents in the survey are women (53.8%), followed by men (45.7%). The age range is diverse, with a greater percentage of the participants being between the age of 25 and 40 (43.3%). 43.8% of the respondents have a Bachelor’s degree, 26.2% hold an advanced education degree. A large portion is employed (37.4%) and almost as many are university students (32.1%). Most participants (24.5%) reported an income less than 500 EUR, followed by participants with an income between 501–1500 EUR (23.6%) and 1501–2500 EUR (20.5%). The majority of participants are residing in the United States (31.2%), then in Germany (18.3%), India (11.4%), Bulgaria (10.0%), and other (29.0%). The descriptive statistics of the online shopping frequency (see Table 2) show that consumers shop more frequently during- and post-COVID-19 than pre-COVID19. Comparing the during- and post-COVID-19 phases, a slight decline in shopping frequency is expected after the pandemic is over. Table 3 shows the descriptive statistics of e-consumers’ engagement in environmentally friendly practices through the stages before, during and after the pandemic. Although the median and mode remain the same across the three stages for several of the statements, we can also observe consumers plan to increase their engagement in eco-practices after the pandemic specifically when it comes to avoidance of retailers that use excessive packaging and preference for local suppliers. Moreover, the number of consumers who strongly agreed with the statement on consideration of sustainability as a factor influencing purchase decisions was highest in the post-COVID-19 phase. 4.2 Methodology The dependent t-test is considered the norm when analyzing the differences in experiments with repeated measures (pre-, during-, and post-COVID-19) [25]. However, it has been argued that Likert-type items are ordinal in nature and should be analyzed with
478
S. Koleva and S. Chankov Table 1. Sample description Gender Female Male Transgender Prefer not to say Other Age range 18-20 21-24 25-40 41-55 56-70 71 or older Occupation University student Employee Civil servant Self-employed Unemployed Training/Apprentice Other
53.8% 45.7% 0.2% 0.2% 0.0% 11.7% 27.6% 43.3% 13.3% 3.8% 0.2% 32.1% 37.4% 6.0% 15.7% 4.8% 1.2% 2.9%
Country of residence United States 31.2% Germany 18.3% India 11.4% Bulgaria 10.0% Other 29.0% Level of education High school 25.5% Bachelor’s degree 43.8% Master’s degree 23.3% Ph.D. or higher 2.9% Associates degree 3.1% Other 1.4% Income 4500 Euro 5.7% Prefer not to say 8.8%
Table 2. Descriptive statistics – online shopping frequency pre-, during-, post-COVID-19 Question Pre-COVID-19, how often did you order products online? Since the start of COVID19, how often have you been ordering products online? Post-COVID-19, how often do you anticipate to order products online?
Median Mode
Extremely Someti Occasio Not at Often Seldom often mes nally all often
4
4
4.3% 16.9% 18.3% 27.6% 24.5%
8.3%
5
5
1.9%
5.0%
9.3% 25.0% 36.2%
22.6%
5
5
2.4%
7.9% 11.0% 25.7% 35.7%
17.4%
non-parametric tests and median, mode, and frequency as descriptive statistics [26]. Hence, for the purpose of this study, the non-parametric Wilcoxon test was chosen to compare pre-COVID-19 vs. during-COVID-19, during-COVID-19 & post-COVID-19, and in some cases pre-COVID-19 vs. post-COVID-19. Furthermore, the effect size was analyzed by observing the r value [25]. The interpretation of the effect size was based on the well-established guidelines: 0.1–0.3 (small effect), 0.3–0.5 (medium effect) and ≥0.5 (large effect) [27].
Pre-COVID-19, I didn't buy from retailers that use excessive packaging Since the start of COVID-19, I haven’t bought from retailers that use exc. packaging Post-COVID-19, I won’t buy from retailer that use excessive packaging Pre-COVID-19, I was willing to wait a longer time for shipping Since the start of COVID-19, I have been willing to wait a longer time for shipping Post-COVID-19, I will be willing to wait a longer time for shipping Pre-COVID-19, sustainability was a factor that influenced my purchase decisions Since the start of COVID-19, sustainability has been a factor that influences my purchase decisions Post-COVID-19, sustainability will be a factor that influences my purchase decisions Pre-COVID-19, I only chose brands that are sustainable Since the start of COVID-19, I have only been selecting brands that are sustainable Post-COVID-19, I will only choose brands that are sustainable Pre-COVID-19, I only considered buying from local suppliers Since the start of COVID-19, I have only considered buying from local suppliers Post-COVID-19, I will only consider buying from local suppliers Pre-COVID-19, I chose to ship orders to my place of residence rather than a pick-up station or post office Since the start of COVID-19, I have been choosing to ship an order to my place of residence rather than a pick-up station or post office Post-COVID-19, I will choose to ship an order to my place of residence rather than a pick-up station or post office Pre-COVID-19, I bought multiple sizes of the same item and returned most of them Since the start of COVID-19, I have bought multiple sizes of the same item and returned most of them Post-COVID-19, I will buy multiple sizes of the same item and return most of them
Statement 2 2 4 5 5 5 4 4 4 4 4 4 3 2 4 6 6 6 1 1 1
3 3 4 4 4 4 4 4 4 3 4 4 3 3 4 5 5 5 2 2 2
Median Mode
38.1%
39.8%
41.9%
4.3%
5.7%
4.5%
6.7% 10.2% 11.2% 7.9% 15.2% 15.2% 11.0%
10.0%
12.4% 13.8% 11.4% 8.3% 9.0% 8.8% 7.1%
16.7%
16.9%
19.5%
10.0%
8.3%
8.3%
7.9% 17.9% 16.9% 11.7% 23.8% 22.6% 16.2%
12.4%
Strongly Disagree disagree
12.9%
8.8%
11.0%
13.8%
11.7%
10.2%
16.4% 23.6% 20.0% 21.2% 24.0% 22.4% 21.7%
16.9%
11.0%
14.0% 10.5%
14.3% 12.6%
12.6% 12.6%
16.7% 23.6%
18.3% 18.8%
7.9%
7.6%
2.4%
31.7%
37.1%
32.1%
24.0% 14.8% 16.0% 7.4% 17.4% 9.5% 22.6% 11.9% 12.1% 3.1% 13.3% 6.4% 17.1% 10.2% 22.4% 22.4%
30.2% 25.0% 25.0% 24.8% 21.7% 20.0% 23.8%
Strongly agree
11.2% 5.5% 13.3% 7.4% 18.3% 10.5% 24.5% 6.5% 25.7% 11.0% 25.2% 9.0% 23.6% 6.7%
Agree
30.7% 19.0%
Somewh Somewh at at agree disagree 24.3% 23.6% 23.1% 22.6% 22.4% 20.5% 18.8% 17.6% 23.3% 14.8% 21.7% 24.3% 12.1% 17.9% 24.3% 15.0% 18.6% 23.3% 14.8% 16.4% 31.4%
Table 3. Descriptive statistics – engagement in environmentally friendly practices pre-, during-, post-COVID-19
The Impact of the COVID-19 Pandemic on E-commerce Consumers’ 479
480
S. Koleva and S. Chankov
4.3 Analysis and Results Table 4 presents the test results from comparing the frequency of online shopping. Table 4. Wilcoxon test results for online shopping frequency
Effect Positive Negative size (r) Ranks† Ranks†† Pre-COVID-19 & During-COVID-19 -10.579 *** -0.37 222 51 During-COVID-19 & Post-COVID-19 -3.572 *** -0.12 67 121 Pre-COVID-19 & Post-COVID-19 -9.634 *** -0.33 186 44 * p < 0.05, ** p < 0.01, *** p < 0.001 † Pre- < During- | During- < Post- | Pre- < Post†† Pre- > During- | During- > Post- | Pre- > PostOnline shopping frequency
Z
Ties 147 232 190
To check the validity of the hypothesis H1a, a test comparing the shopping frequency pre-COVID-19 and during-COVID-19 was conducted. The comparison showed a significant difference in the frequency of online shopping pre-COVID-19 and duringCOVID-19 (p < 0.001) with a medium effect (r = −0.37). Hence, it can be concluded that the frequency of online shopping has increased (indicated by the larger number of positive ranks) significantly since the declaration of the pandemic. Consumers started ordering online more frequently during the pandemic than they used to shop before it, hence, H1a is supported. To check the validity of the hypothesis H1b, a test comparing the shopping frequency during-COVID-19 and post-COVID-19 was conducted. The comparison showed a significant difference in the frequency of online shopping during-COVID-19 and postCOVID-19 (p < 0.001) with a small effect (r = −0.12). The frequency of online shopping is expected to decrease once the pandemic is over (indicated by the larger number of negative ranks). Hence, H1b is not supported. In order to gain a better understanding of the effect of the pandemic on the frequency of shopping of e-consumers, a third test was run comparing consumers’ shopping frequency pre-COVID-19 and post-COVID-19. The findings demonstrated that even though consumers intend to reduce the amount of times they shop online after the pandemic compared to during the pandemic, they will still shop significantly more often postCOVID-19 than they used to shop before the coronavirus pandemic (p < 0.001, r = − 0.33 (medium effect), higher number of positive ranks). Table 5 displays the results from analyzing the pro-environmental behavior of consumers across the selected seven eco-practices comparing: pre-COVID-19 and duringCOVID-19, and during-COVID-19 and post-COVID-19. To accept H2a consumers would need to have adopted a more environmentally sustainable behavior duringCOVID-19 compared to pre-COVID-19. For H2b to be accepted, consumers will continue their eco-engagement acquired during the pandemic also after it is over. In this way, H2b builds on H2a, and hence it is important to note that H2b can only be accepted, if H2a holds. If the test results for H2a conclude that e-consumers have not engaged in more environmentally friendly practices during the pandemic than before it, then H2b
The Impact of the COVID-19 Pandemic on E-commerce Consumers’
481
Table 5. Wilcoxon test results for engagement in eco-practices
Engagement in Eco-Practice
Z
Effect Positive Negative size (r) Ranks† Ranks††
Excessive packaging Pre-COVID-19 & During-COVID-19 -1.126 99 94 During-COVID-19 & Post-COVID-19 -6.023*** -0.12 136 52 Longer waiting time Pre-COVID-19 & During-COVID-19 -2.772* -0.10 136 96 During-COVID-19 & Post-COVID-19 -1.657 84 113 Sustainability as a factor Pre-COVID-19 & During-COVID-19 -0.036 94 96 During-COVID-19 & Post-COVID-19 -6.622*** -0.23 138 48 Sustainable brands Pre-COVID-19 & During-COVID-19 -1.499 106 90 During-COVID-19 & Post-COVID-19 -5.859*** -0.20 136 44 Local suppliers Pre-COVID-19 & During-COVID-19 -2.119* -0.07 124 92 During-COVID-19 & Post-COVID-19 -6.307*** -0.22 156 57 Home deliveries Pre-COVID-19 & During-COVID-19 -0.388 100 82 During-COVID-19 & Post-COVID-19 -1.493 75 100 Returns Pre-COVID-19 & During-COVID-19 -4.109*** -0.14 93 55 During-COVID-19 & Post-COVID-19 -0.080 84 75 * p < 0.05, ** p < 0.01, *** p < 0.001 † Pre-COVID-19 < During-COVID-19 | During-COVID-19 < Post-COVID-19 †† Pre-COVID-19 > During-COVID-19 | During-COVID-19 > Post-COVID-19
Ties 227 232 188 223 230 234 224 240 204 207 238 245 272 261
is to be rejected. If, there is no newly acquired eco-engagement during-COVID-19, it cannot be continued post-COVID-19. Excessive Packaging. The pandemic did not change consumers’ behavior when it comes to avoidance of excessive packaging. Consumers continued avoiding buying items with excessive packaging during the pandemic to the same extent as they did before the pandemic (p ≥ 0.05). Thus, H2a is rejected for this eco-practice and H2b is also rejected. In fact, the comparison during- and post-COVID-19 shows that consumers will avoid buying products with excessive packaging after the pandemic to a larger extent than they are during the pandemic. Longer Waiting Time. The pandemic has had a positive effect on the amount of time consumers are willing to wait until their delivery arrives. During-COVID-19 consumers indicated willingness to wait longer compared to pre-COVID-19 (p < 0.05), H2a is accepted. The comparison of during-COVID-19 and post-COVID-19 is not significant,
482
S. Koleva and S. Chankov
showing no difference in consumer’s behavior, hence H2b is also accepted. Online shoppers’ eagerness to wait longer will remain the same after the pandemic. Sustainability as a Factor. The level of consumer consciousness for sustainability remained the same during-COVID-19 as it was pre-COVID-19 (H2a is rejected, and thus H2b is also rejected). In fact, the comparison of during-COVID-19 and post-COVID-19 showed that consumers are more likely to consider sustainability as an influencing factor after the pandemic than they are now. Sustainable Brands. During COVID-19, consumers did not prefer sustainable brands compared to before the pandemic (H2a is rejected, and thus H2b is also rejected). In fact, the comparison of during-COVID-19 and post-COVID-19 showed that consumers will buy from sustainable brands more often in the future than they do now. Local Suppliers. During-COVID-19 consumers had a stronger preference for local suppliers than pre-COVID-19 (p < 0.05). Hence, H2a is accepted. Moreover, after the pandemic consumers will prefer buying locally made products even more than they do during the pandemic (the comparison of during-COVID-19 and post-COVID-19 is significant, p < 0.001); therefore, H2b is not supported. Home Deliveries. Both hypotheses H2a and H2b are rejected, because the comparisons are not significant and thus show there is no preference of home deliveries or pick-up stations among consumers for online orders. Returns. In light of the pandemic, consumers are more likely to buy multiple sizes and make a return, p < 0.001, which points towards less eco-friendly behavior compared to before the pandemic. As a result, H2a is rejected, and thus H2b is also rejected. The comparison of during-COVID-19 and post-COVID-19 showed that consumers will continue to engage in this environmentally unfriendly practice after the pandemic too.
5 Discussion Table 6 shows the summary of hypotheses support. With regard to shopping frequency, H1a is supported and H1b is rejected. While consumers shop more frequently during the pandemic than they did before it, their shopping frequency will decrease once the pandemic is over. Still, the online shopping frequency is unlikely to return to its prepandemic state (post-COVID-19 frequency is higher than pre-COVID-19 frequency), showing that e-commerce is likely to continue growing in the long-term. H2a was supported for two out of the seven investigated eco-practices (willingness to wait a longer time for shipping and preference for local suppliers), while H2b was supported only for one eco-practice (willingness to wait a longer time for shipping). Overall, the COVID-19 pandemic did not lead to more pro-environmental behavior among e-commerce consumers. The before vs. during pandemic comparison revealed that for four of the studied eco-practices there was no positive change in consumers’ engagement in more environmentally friendly behavior since the start of the pandemic:
The Impact of the COVID-19 Pandemic on E-commerce Consumers’
483
(1) avoidance of retailers that use excessive packaging, (2) consideration of sustainability as a factor influencing purchase decisions, (3) preference for sustainable brands, and (4) preference of pick-up location over home delivery. In fact, consumers worsened their behavior with regard to shopping and returning clothes in multiple sizes. Looking into the future, our analysis shows that consumers have strong intentions to adopt more eco-friendly practices after the pandemic ends. For four of the studied eco-practices, consumers indicated a positive change in their anticipated engagement in more environmentally friendly behavior post-COVID-19: (1) avoidance of retailers that use excessive packaging, (2) consideration of sustainability as a factor influencing purchase decisions, (3) preference for sustainable brands, and (4) preference for local suppliers. For the other three eco-practices, the level of engagement was indicated to remain the same as during-COVID-19. Thus, it is highly likely that consumers will adopt a more environmentally friendly behavior in the future than they do now. Therefore, we can conclude the COVID-19 pandemic had a negative impact on the pro-environmental behavior of e-commerce customers. During the pandemic, there was more online shopping without consideration of eco-practices, while after the pandemic, consumers will shop slightly less often online and will be more likely to engage in eco-practices. This indicates that during the times of a global health crisis consumers’ prioritization of environmental sustainability was reduced, but it will be strengthened after the world recovers from the pandemic. Table 6. Hypotheses support overview Hypothesis
Support Hypothesis
H1a: Consumers shopped online less often pre-COVID-19 than duringYes COVID-19. H2a: Consumers engaged in environmentally friendly practices when shopping online preCOVID-19 to a lesser degree than they do during-COVID-19. Excessive packaging No Wait longer Yes Sustainability as a factor No Sustainable brands No Local suppliers Yes Home delivery No Returns No
Support
H1b: Consumers will shop online post-COVID-19 as often as duringNo COVID-19. H2b: Consumers will continue to engage in environmentally friendly practices when shopping online post-COVID-19 as much as they do during-COVID-19. Excessive packaging No Wait longer Yes Sustainability as a factor No Sustainable brands No Local suppliers No Home delivery No Returns No
484
S. Koleva and S. Chankov
6 Conclusion The purpose of this paper was to examine the impact of the COVD-19 pandemic on ecommerce consumers’ pro-environmental behavior. The results indicated that the shopping frequency has increased substantially since the start of the pandemic, but will drop down after the end of the pandemic. Moreover, the COVID-19 pandemic was detrimental to consumers’ pro-environmental behavior: during the pandemic consumers showed a tendency towards less environmentally friendly behavior but they have strong intentions to adopt more eco-friendly practices after the pandemic ends. The main limitation of the paper lies in the fact that we analyzed consumers’ stated behavior and not actual behavior. This is especially true in the survey section dedicated to post-COVID-19 behavior, as it asks participants to speculate on their anticipated future behavior. Hence, future research should study consumers’ behavior after the pandemic is over to verify our findings.
References 1. UNCTAD. COVID-19 has changed online shopping forever, survey shows. https://unctad.org/ news/covid-19-has-changed-online-shopping-forever-survey-shows. Accessed 11 Mar 2021 2. Koch, J., Frommeyer, B., Schewe, G.: Online shopping motives during the COVID-19 pandemic — lessons from the crisis. Sustainability 12(24), 10247 (2020) 3. Unnikrishnan, A., Figliozzi, M.A.: A Study of the Impact of COVID-19 on Home Delivery Purchases and Expenditures (2020) 4. Kapser, S., Simon, Z.: Covid-19 as a catalyst for autonomous last-mile delivery - CAMELOT Blog 5. Statista. Alibaba’s gross merchandise volume on Singles’ Day from 2011 to 2020. https:// www.statista.com/statistics/364543/alibaba-singles-day-1111-gmv/. Accessed 07 May 2021 6. Perez, S.: COVID-19 pandemic accelerated shift to e-commerce by 5 years, new report says. https://techcrunch.com/2020/08/24/covid-19-pandemic-accelerated-shift-to-ecommerce-by-5-years-new-report-says/. Accessed 30 Mar 2021 7. World Economic Forum. The Future of the Last-Mile Ecosystem (2020) 8. Barreiro-Gen, M., Lozano, R., Zafar, A.: Changes in sustainability priorities in organisations due to the COVID-19 outbreak: averting environmental rebound effects on society. Sustain. 12(12), 5031 (2020). https://doi.org/10.3390/su12125031 9. Malley, B.: The Inconvenient Truth About Packaging Waste and Last Mile Delivery - PARCEL Industry. https://parcelindustry.com/article-5067-The-Inconvenient-Truth-About-PackagingWaste-and-Last-Mile-Delivery.html. Accessed 15 Mar 2021 10. MetaPack. 2018 State of eCommerce Delivery Consumer Research Report (2018) 11. Ignat, B., Chankov, S.: Do e-commerce customers change their preferred last-mile delivery based on its sustainability impact? Int. J. Logist. Manag. 31, 521–548 (2020). https://doi.org/ 10.1108/IJLM-11-2019-0305 12. Salihu, E., Chankov, S.: How much value do consumers put on environmental sustainability when choosing last-mile delivery ? In: 2021 BVL International Scientific Symposium on Logistics, Dresden, pp. 113–123 (2021) 13. McKinsey & Company. Survey: consumer sentiment on sustainability in fashion. https:// www.mckinsey.com/industries/retail/our-insights/survey-consumer-sentiment-on-sustainab ility-in-fashion. Accessed 09 Apr 2021
The Impact of the COVID-19 Pandemic on E-commerce Consumers’
485
14. Raja, K., Bhakta, H., Prakash, V., Samal, B., Bhattacharya, J.: Science of the total environment challenges and strategies for effective plastic waste management during and post COVID-19 pandemic. Sci. Total Environ. 750, 141514 (2021). https://doi.org/10.1016/j.scitotenv.2020. 141514 15. Deloitte. Understanding the Impact of COVID-19. Grocery & Food Retail (2020) 16. McKinsey & Company. Meet the next-normal consumer. https://www.mckinsey.com/ business-functions/marketing-and-sales/our-insights/meet-the-next-normal-consumer. Accessed 09 Apr 2021 17. UNCTAD. How COVID-19 triggered the digital and e-commerce turning point. https://unctad. org/news/how-covid-19-triggered-digital-and-e-commerce-turning-point. Accessed 21 May 2021 18. PwC. Global Consumer Insights Survey 2020 (2020) 19. Kim, R.Y.: The impact of COVID-19 on consumers: preparing for digital sales. IEEE Eng. Manag. Rev. 48, 212–218 (2020). https://doi.org/10.1109/EMR.2020.2990115 20. Capgemini. How sustainability is fundamentally changing consumer preferences. https:// www.capgemini.com/wp-content/uploads/2020/07/Final-Infographic.pdf 21. PWC. PwC’s March 2021 Global Consumer Insights Pulse Survey (2021) 22. BBC News. Has coronavirus made us more ethical consumers? https://www.bbc.com/news/ business-55630144 23. Velazquez, R., Chankov, S.M.: Environmental impact of last mile deliveries and returns in fashion E-commerce: a cross-case analysis of six retailers. In: 2019 IEEE International Conference on Industrial Engineering and Engineering Management (IEEM), pp. 1099–1103. IEEE (2019). https://doi.org/10.1109/IEEM44572.2019.8978705 24. Beck, L., Ajzen, I.: Predicting dishonest actions using the theory of planned behavior. J. Res. Pers. 25, 285–301 (1991). https://doi.org/10.1016/0092-6566(91)90021-H 25. Field, A., Miles, J., Field, Z.: Discovering Statistics Using R. Sage, Los Angeles, London, New Delhi, Singapore, Washington DC (2003) 26. Carifio, J., Perla, R.: Resolving the 50-year debate around using and misusing Likert scales. Med. Educ. 42, 1150–1152 (2008). https://doi.org/10.1111/j.1365-2923.2008.03172.x 27. Cohen, J.: Statistical Power Analysis for the Behavioral Sciences. L. Erlbaum Associates, New York (1988)
Developing a Serious Game for Intelligent Transportation Systems Ilja Bäumler(B)
, Moritz Elfers , Okan Dogtas , Fynn Gresens , and Sercan Eyigün University of Bremen, Bremen, Germany [email protected]
Abstract. Considering the increasing freight traffic and the associated negative effects, (e.g.: congestion, unpunctuality, waste of energy, harmful environmental influences etc.), systems for improving the status quo are being researched and implemented. The understanding and the acceptance of possible future applications in road freight transport must be brought to the users by appropriate means. This paper presents a prototype game on the topic of “Intelligent Transport Systems for Road Freight Transport”. Based on the preliminary work, a prototype game is designed. Questionnaire results show that more than half of the participants (n = 32, average age = 43 years) could not relate to intelligent transport systems. By using the serious game presented here, the understanding of and need for intelligent transportation systems in society can be achieved. The decision to use intelligent transport systems in the logistics industry can be simplified for decision makers by playful interaction and beneficial consequences between the game elements. This paper describes the development stages of a serious game from problem definition to prototyping in the field of intelligent transport systems. Keywords: Intelligent transportation systems · Road freight transport · Gamification · Serious game
1 Introduction 1.1 Problem Background/Motivation Growing e-commerce, market liberalization, just-in-time deliveries and globalization lead to higher demand in transportation services and thus to slow-moving traffic and congestions (Savelsbergh and Van Woensel 2016). The damage is predicted to be in the three-digit billion range in 2030 (Centre for Economics and Business Research 2014). In many areas, an expansion of the road infrastructure could ease the situation and mitigate slow-moving traffic. However, this is not always possible due to geographical, building or policy regulations. Intelligent transportation systems (ITS) promise a relief in this matter by optimizing utilization of transport infrastructure, resources and capacity. Assuming new IT-based services in the transportation sector will rise in the coming years, the acceptance of these new services on the user side is a prerequisite for this © The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 M. Freitag et al. (Eds.): LDIC 2022, LNLO, pp. 486–497, 2022. https://doi.org/10.1007/978-3-031-05359-7_39
Developing a Serious Game for Intelligent Transportation Systems
487
development (Bäumler and Kotzab 2020). Survey results (n = 31, average age = 44) at a games convention show that more than 69% of the participants are not familiar with ITS although more than 84% at the same time answered that they are familiar with logistics (games). A game-based approach on teaching and a positive correlation between gamebased learning and educational as well as motivational outcome is discussed in several books and papers (Abt 1987; Cowley et al. 2011; Gee 2003; Jacob and Teuteberg 2017). Gee (2003) states that games allow the player to make decisions as a producer and thus better reflect on the interactions between the game elements. The trial-anderror approach in a risk-free serious game environment allows the player to explore the challenges and problems and to make mistakes without fearing harsh consequences (Abt 1987). In this respect, this paper proposes a logistics game that takes up ITS and explains the underlying concepts and interactions of the game elements. This game may be used to teach environmental and operational problems that arise in the transport sector and train skills to solve these problems with the means of ITS. 1.2 Research Question and Methodological Approach To our best knowledge there does not exist any serious game which deals with application of ITS from a shipping company’s perspective. The purpose of this paper is to develop a serious game which can teach ITS and their main interrelations from the perspective of a shipping company. This will be achieved by a developmental process divided into three phases (see Sect. 3.1). This paper intends to help recognize serious games as a practical fit for comprehensive teaching in complex systems. Consequently, the research question is the following: Is it possible to integrate intelligent transportation systems into a serious (transport) game? To answer the research question a methodological triangulation was conducted. Firstly, we carried out theoretical research on the topics of game-based learning and ITS for road freight transport. Secondly, we created a questionnaire and gathered survey results from a games convention. Thirdly, we developed a serious game on a three-phased approach. The remainder of the paper is as follows: After having introduced the motivation and the research question, we present a short clarification on the two most prominent terms of game-based learning “gamification” and “serious game” and assign the developed game accordingly. Subsequently, we outline the developmental phases of the game, the necessary theoretical background of ITS for road freight transport and present the gameplay phases. Afterwards, the game design is critically assessed and discussed from the research question’s perspective. The paper closes with a critical reflection of the results and an outlook for future research.
2 Gamification or Serious Game? 2.1 Gamification Deterding et al. (2011, p. 10) defines Gamification as “[…] the use of game design elements in non-game context”. Developers use gamification as a tool to apply a system
488
I. Bäumler et al.
of rewards, points, challenges or similar systems to specific topics (Deterding et al. 2011). The basic idea of gamification is to use game functions as a means of solving real tasks and thus to increase motivation to achieve learning goals. The learning effect occurs unintentionally and indirectly with the achievement of the goals. However, it should also be noted that the behavior of players in the gamified environment differs from that in reality, since, for example, higher risks are taken in such environments that do not bring any real consequences. Experiments have shown that a sense of achievement increases the motivation of the players and thus has a positive influence on learning efficiency (Frank 2014). However, such games do not directly lead to concrete competences. The focus is rather on teaching facts, details and processes (Landers and Callan 2011). Hence, gamification is a strategic decision to enhance learning outcome in a working environment (Wanick and Bui 2019). 2.2 Serious Games As a commonly accepted definition for gamification exists, it is not the case with a definition for serious games (Eckardt et al. 2017). Sawyer and Smith (2008) try to pinpoint and analyze typical characteristics as well as design and engineering patterns of serious games. They conclude that serious games are game based simulation which are made for government foresight and public policy. They do not narrow serious games to learning or training, but instead broaden its understanding as an idea what games can be in different market segments such as healthcare, education, industry and many more. As this might be true for Sawyer and Smith (2008), other authors state that learning interrelations between game elements and skill acquisition are primary characteristics of serious games (Connolly et al. 2012; Jacob and Teuteberg 2017; Ritterfeld et al. 2009). Wouters et al. (2013) derive their definition of serious (computer) games from different authors. For them serious games are “[…] interactive, based on a set of agreed rules and constraints, and directed towards a clear goal […] [and] constantly provide feedback, either as a score or as changes in the game world, to enable players to monitor their progress toward the goal.” (Wouters et al. 2013, p. 250). Although, this definition combines all the aspects of a game, this definition lacks the connection with learning and behavioral outcomes. Most of the definitions have the same idea in mind when it comes to basic characteristics of serious games such as educational and/or behavioral outcomes (see Michael and Chen (2006) or Zyda (2005)). To have a common denominator in this matter we follow Susi et al.’s (2007, p. 1) notion, which states that “serious games are (digital) games used for purposes other than mere entertainment.” 2.3 What is Logistics Empire? The subject area of logistics is becoming increasingly complex and is mainly taught in the form of classical teaching methods such as frontal teaching or lectures (Wood and Reiners 2012). In the area of logistics, elements of gamification are advantageous in order to illustrate the complex interrelationships at a game level and to create a learning motivation with challenges.
Developing a Serious Game for Intelligent Transportation Systems
489
Well-known examples from the logistics sector include the Beer Distribution Game. It is a simulation of a supply chain, in which the players try to balance demand and supply in a beer supply chain (Steinegger and Barbey 2014). The game tries to demonstrate and teach the bullwhip effect which occurs through order batching, demand signaling, fluctuating prices and product shortages across the supply chain. These upstream information in a supply chain may lead to the phenomenon, that the demand at the manufacturers side does not match the actual sales demand at the consumer side (Lee et al. 2004). The environment in which road freight transport takes place is subject to many influences. Thus, for a serious game, the reduced complexity still represents the reality to a degree which allows the player to recreate reality. The game interface should offer an abstract representation of the necessary environment, in which important and only necessary elements can be represented (Hamayon 2012). “Games, as metaphors, distort the actual world that they are supposed to model. As is the case for any system of representation, certain parts of “what’s real” are present within its framework” (Savignac 2017). For the game developed here, called Logistics Empire, the player takes on the role of a forwarding company. He must plan the routes of his trucks by means of acquired delivery orders and must observe the environmental and traffic regulations. Using Intelligent Transportation Systems (ITS), the player can bypass some of these regulations, earn bonuses and manage his company more efficient. It is therefore an attempt to turn a reallife example into a board game and to explain the importance of ITS and the interplay between modular upgrades and transport and environmental issues. The real day-to-day life of a freight forwarding company is simplified with necessary abstractions to be able to show the thought processes of a freight forwarding company during the execution of an order. Hence, the focus of Logistics Empire lies on mastering the challenge in the game and not directly pursuing the learning objectives. In addition, the player is made aware of the necessity of ITS, which leads this game into the direction of a serious game and does not pursue the goals of gamification.
3 Logistics Empire – The Game 3.1 Phases of Game Development The development of the game is divided into three phases (see Fig. 1). The first phase comprises the clarification of the game environment and game goal. This phase is characterized by the following core questions: • In which thematic context does the game take place? • On which game platform should the game be realized? • How can the game be presented realistically without losing the main intensity in its complexity? The second phase deals with the elaboration and refinement of the game concept. Therefore, a questionnaire was distributed among participants of a game convention in order to ensure the necessity and requirements of a serious game in terms of playability and reasonable game elements. Afterwards, participants of a workshop commented and
490
I. Bäumler et al.
discussed the existing game idea which was development in phase 1 and on basis of the questionnaire results. The game was further improved through the participants’ feedback. The third phase develops a prototype game and is characterized by ongoing tests and continuous adjustments of game dimensions and the interdependencies of the game elements.
Fig. 1. Schematic illustration of the game’s development process
3.2 ITS for Road Freight Transport General Information. In Germany, approximately 3,16 billion ton were driven on the roads in 2018, which is 77% of all tonnage distributed in Germany in 2018 (Statistisches Bundesamt 2020). In the case of trucks, the type of transported goods determines the required equipment of the truck in order to guarantee a goods-based delivery. For example, fresh products such as fruits and vegetables or milk-based products necessitate cooling in the loading area to ensure quality. When transporting hazardous goods, international regulations must be observed. These are constantly being checked and improved, which is why transport companies may have to upgrade or expand their fleets. The technology surrounding vehicle construction has also evolved. Modern hybrid engines are based on an internal combustion engine, which is supported by an electric motor for short distances and especially when starting up. The battery is charged while driving, e.g. when braking, excess energy is stored (Hofmann 2014). Pure electric motors, on the other hand, run solely on electrical energy. These are mainly charged by a power cable, although charging can also be integrated during braking. In comparison to a hybrid motor, there are no CO2 emissions with an electric motor (Bundesministerium für Umwelt, Naturschutz und nukleare Sicherheit - BMU 2019). Since the transport of goods on the road is a complex process, full integration of all factors into a game would be a major challenge. Thus, this serious game represents only the major aspects of road freight transport amid ITS.
Developing a Serious Game for Intelligent Transportation Systems
491
ITS is defined by the European Commission as “[…] advanced applications which without embodying intelligence as such aim to provide innovative services relating to different modes of transport and traffic management and enable various users to be better informed and make safer, more coordinated and ‘smarter’ use of transport networks” (Smith 2015, p. 1). For road freight transport ITS means improvement of the transport process in the business and operations area such as management, maintenance, monitoring, control, and security. Therefore, it provides real-time information for the observation, evaluation and control of processes, enabling interventions in ongoing operations (García-Zuazola et al. 2015). This is done by wireless communication between infrastructure, machines and different actors in the transport process (Kala 2016). Bäumler (2019) evaluated research areas of interest for ITS in road freight transport and concluded with four main areas: Fleet management, city logistics, advanced driver assistance systems (ADAS) and toll systems. Fleet Management. Fleet management increases the efficiency of a company’s fleet by continuously improving the activities of planning, management, control and communication of the fleet with the aim to achieve the best performance to cost ratio (Jung and van Laak 2001; Vahrenkamp and Kotzab 2012). An important factor of “intelligent” fleet management is the exchange of information and communication between drivers and logistics service providers (Evers and Kasties 1999). Sensors and data computing technologies alongside with wireless communication systems provide the customer, vehicle fleet and traffic control centers automatically with necessary information. Information include data about the route, status and position of the vehicle (Evers and Kasties 1999). The constant exchange of information between the various actors in the transport process optimizes capacity utilization, and improves order fulfilment and control of transport processes (Evers and Kasties 1999). City Logistics. City Logistics refers to transport, handling and storage processes in inner-city areas (Strauß 1997). According to Bekta¸s et al. (2017) literature seems to agree on three points regarding the definition of city logistics: 1. City logistics refers to urban freight transport. 2. City logistics is a system that pursues the consolidation and coordination of freight transport within a city. 3. City logistics tries to make the logistic processes within a city more efficient and to reduce the negative environmental impact. Especially the third point is achieved by data processing technologies which include the collection of information on transport routes, times, loads and positions, and enables optimization in route planning. Hence a reduction in negative emissions is achieved by improved capacity utilization (Organisation for Economic Co-operation and Development 2003). Advanced Driver Assistance Systems (ADAS). ADAS support the driver in information processing throughout the driving process. By using different sensors and cameras, ADAS analyze the incoming data and support the driver in stabilizing, guiding and
492
I. Bäumler et al.
navigating the vehicle (Mörbe 2012). Through continuing evaluation of incoming data, ADAS decide when to influence the driver by information or intervention (Heinrichs 2018). Toll Systems. Toll Systems provide a source of revenue to finance road repairs and function to manage traffic. The development of modern technologies such as wireless communication, wireless data transmission and global navigation satellite systems has made toll collection faster and cheaper (Iseki and Demisch 2012).
3.3 Game Design Gameplay. Basically, the game is designed for four players. In the game, each player represents his own shipping company which fulfills transport orders. These transport orders provide the player with money and victory points. While the latter lead to victory of the game (when reaching ten victory points), the money can be used to improve the company and pay toll costs. In this way, the player can gain advantages over the opposing companies. The goal of the game is to build a flourishing shipping company, while managing his fleet, buying upgrades to improve the management of the order processing and executing transport orders. Each player starts his shipping company with a truck in his fleet and a fixed starting amount of in-game money. In addition, each player randomly receives one of the predefined starting roles and a starting transport order. Each truck is placed on the starting order city. A game move is structured in four phases and is timed according to a regular working day. Phase 1 to 3 is played consecutively by all players. Phase 4 is played simultaneously by all players. The starting player is drawn randomly and rotates clockwise each turn. In the first phase order picking takes place. Each player may pick an order from the order distribution platform in a clockwise order beginning from the starting player. After a player picks an order a new order card is placed on the vacant field. In case of an auction, which occurs after the game has progressed to a certain point, an auction order is auctioned among the players on the auction order platform. All accepted orders will be placed in the designated fields in the players’ head office. The second phase is characterized by the operational activities of the company. Here, the cargo for the transport order is collected or delivered to the destination. A player can move his truck by 8 fields per move, which represents 400 km. Load acceptance and load delivery takes an hour each, which corresponds to a range of 50 km. If an order is now completed in this phase, the corresponding payment is received, and the player drives his remaining steps or chooses to stay. Further components to be considered during transport are event fields, toll and environmental measures. Event measures are triggered when a truck drives on or over a field with an exclamation mark. Toll must be paid when driving on motorways with toll measures. When loading or unloading a truck in a state with environmental measures, penalties or bonuses must be considered. In the third phase the player can acquire upgrades in different ITS categories, trucks and equipment or new staff regarding the strategic orientation of the company. This will generate various advantages in the game, which will take effect from the next round.
Developing a Serious Game for Intelligent Transportation Systems
493
Fig. 2. Logistics Empire’s game board
Game Board Design. In the final phase, the timers for the individual transport orders are moved forward by one day. If an order has now exceeded the delivery time, a certain amount of money must be paid as a penalty. In case an order exceeds the delivery time on the order distribution platform, the order is discarded, and a new order is drawn. The playing field on which the player executes transport orders represents Germany (see Fig. 2). Here, the top 20 cities in Germany in terms of gross domestic products represent the loading and unloading locations of the logistics orders. Another key element is the transport network on which road freight transport mainly takes place. This is represented in the form of motorways. The selection of motorways has been limited to the 5 most frequented motorways (BASt 2018). In addition, some roads were added to connect all cities with the main motorways. The average working time of a truck driver is eight hours per day which is regulated by the Working Time Act (Arbeitszeitgesetz – ArbZG). This includes the truck driving time, loading and unloading time, time spent repairing and cleaning the truck as well as other waiting times during
494
I. Bäumler et al.
the journey (§ 21a ArbZG). Taking all regular factors into account, the average distance travelled by a truck driver is 50 km per hour, which can result in a daily driven distance of 400 km (Rieckenberg 2004). Taking this information into account, the motorway routes were divided into fields where each field reflects a length of 50 km. Head Office. The control center is used to monitor and control the company’s own orders and trucks. Each truck has a maximum load of 30 tons. The accepted orders are positioned in the designated fields in the head office. The open orders can be subsequently assigned to the trucks. To do this, the player must place the order on an empty truck field. Below each order field is a number bar from 1 to 10. Here the continuing delivery time is tracked after the order has been accepted. An order is considered accepted when a player places it in his head office. After each round a game chip is moved on the number bar to represent the expiring delivery time. This allows the head office to keep track of all trucks and the respective orders and their delivery dates.
4 Discussion and Conclusions The game’s core element are transport orders. Transport orders determine the company’s strategic orientation. If they want to focus on refrigerated or dangerous goods, they need to buy certain upgrades and or hire suitable employers with specific training. If they want to focus on large quantity transport orders, they need to consider buying additional trucks. The player can also trade his acquired orders when its desirable. Despite the fact, that the order trade mechanism is only suitable in an open information environment, which does not exist in reality (yet), we implemented this mechanism to induce further problem thinking regarding the optimal transport order allocation. This means that every player knows his opponents’ transport orders in stock at any time and can offer to fulfill the opponent’s orders instead for a negotiated compensation on the order trade platform. The game neglects some of the daily work tasks for truck drivers, such as refueling, maintenance and repair. Because these tasks can be time-consuming and from the game design perspective tedious to implement, we decided to include some of these tasks through game events. These game events can trigger along the road, as most of the mentioned task origin from driving. There is also a manageable number of ITS upgrades. As Logistics Empire will undergo further balancing of interdependencies more ITS will be added. All four relevant areas for ITS for road freight transport (fleet management, advanced driver assistance systems, toll systems and city logistics) offer further upgrade potential, which improves safety, reliability and punctuality of transport orders (e.g. driver fatigue detection, lane keeping assistance, autonomous driving, on-board units for faster payment or priority passing, anticipated transport orders due to big data management). Logistics Empire is a serious (transport) game which tries to involve the player into typical decision-making situations of a daily shipping company. Although the game design offers the player to make sophisticated decisions, there is luck involved in terms of occurring events and offered transport orders. However, in former case the player can avoid critical routes by driving alternative and less optimal routes or buying upgrade
Developing a Serious Game for Intelligent Transportation Systems
495
to prevent negative events. In latter case of unfavorable orders on the order distribution platform we implemented the order exchange platform which a player can use to buy orders from another player. The first player to reach 10 victory points wins the game. The problem with such a winning condition is, that it lacks reference to reality. Since it is possible to sell some upgrades in the game, a winning condition regarding highest amount of money earned after set number of rounds, lacks reference to reality as well. In that case players will sell necessary upgrades in order to temporarily increase their financial assets. Nevertheless, replacing the victory points with a value that relates to real-life aspects can improve the game experience. A possible approach to a more realistic winning condition could be the company value. Components such as upgrades, the amount of successfully executed transport orders, business popularity and environmental awareness could be used to determine a company’s value. The game shows the road freight traffic on a national level and gives no insight into city logistics besides an upgrade which improves loading and unloading times in all cities on the map. This means that transport processes and possible challenges that take place in inner city areas are not communicated. This could be realized in the future by extending the playing field with certain city maps and maps of urban road networks. Real-life influences, such as weather condition or seasonal demand, could be subject to continuous improvement processes. Also, from trust building perspective, executed transport orders could build up the relationship between a client and a shipping company. Thus, implementation of a trust point system could show how satisfied or dissatisfied clients are with the shipping company. Additional game expansions could be implemented through modules which represent real-life examples. Regarding the research question Logistics Empire incorporated several ITS into a serious (transport) game and still has potential for additional ITS extensions. It offers a sound basis for further empirical work which will start when restrictions on personal meetings due to Covid-19 will be eased. The empirical evaluation of behavioral, motivational, physiological and social outcomes including perceptual and cognitive knowledge acquisition as suggested by Connolly et al. (2012) will be conducted by means of a questionnaire. Future research could support game design elements with mathematical and or conceptual models.
References Abt, C.C.: Serious Games. University Press of America (1987) BASt, 2018. Automatische Straßenverkehrszählung. https://www.bast.de/BASt_2017/DE/Verkeh rstechnik/Fachthemen/v2-verkehrszaehlung/Aktuell/zaehl_aktuell_node.html. Accessed 22 Apr 2020 Bäumler, I.: Entwicklungspfade Intelligenter Transportsysteme für den Straßengüterverkehr (Monographie). Universität Bremen, Bremen (2019) Bäumler, I., Kotzab, H.: Scenario-based development of intelligent transportation systems for road freight transport in Germany. In: Urban Freight Transportation Systems, pp. 183–202. Elsevier (2020). https://doi.org/10.1016/B978-0-12-817362-6.00010-0 Bekta¸s, T., Crainic, T.G., Van Woensel, T.: From managing urban freight to smart city logistics networks. In: Network Design and Optimization for Smart Cities, pp. 143–188. World Scientific (2017)
496
I. Bäumler et al.
Centre for Economics and Business Research: The Future Economic and Environmental Costs of Gridlock in 2030. Centre for Economic and Business Research, London (2014) Connolly, T.M., Boyle, E.A., MacArthur, E., Hainey, T., Boyle, J.M.: A systematic literature review of empirical evidence on computer games and serious games. Comput. Educ. 59, 661–686 (2012). https://doi.org/10.1016/j.compedu.2012.03.004 Cowley, B., Moutinho, J.L., Bateman, C., Oliveira, A.: Learning principles and interaction design for ‘Green My Place’: a massively multiplayer serious game. Entertain. Comput. Ser. Games Develop. Appl. 2, 103–113 (2011). https://doi.org/10.1016/j.entcom.2011.01.001 Deterding, S., Dixon, D., Khaled, R., Nacke, L.: From game design elements to gamefulness: defining “gamification.” In: Proceedings of the 15th International Academic MindTrek Conference: Envisioning Future Media Environments, MindTrek 2011, pp. 9–15. Association for Computing Machinery, Tampere (2011). https://doi.org/10.1145/2181037.2181040 Eckardt, L., Körber, S., Becht, E.J., Plath, A., Al Falah, S., Robra-Bissantz, S.: Führen serious games zu Lernerfolg? – Ein Vergleich zum Frontalunterricht. In: Strahringer, S., Leyh, C. (eds.) Gamification und Serious Games. EH, pp. 139–150. Springer, Wiesbaden (2017). https://doi. org/10.1007/978-3-658-16742-4_11 Evers, H., Kasties, G.: Kompendium der Verkehrstelematik - Technische Hochschule Wildau. TÜV-Verlag, Köln (1999) Frank, A.: Achieving game goals at all costs? In: Meijer, S.A., Smeds, R. (eds.) ISAGA 2013. LNCS, vol. 8264, pp. 13–20. Springer, Cham (2014). https://doi.org/10.1007/978-3-319-049 54-0_2 García-Zuazola, I.J., Hernandez-Jayo, U., Perallos, A., Onieva, E.: Intelligent Transport Systems, 1st edn. Wiley (2015). https://doi.org/10.1002/9781118894774 Gee, J.P.: What video games have to teach us about learning and literacy. Comput. Entertain. 1, 20 (2003). https://doi.org/10.1145/950566.950595 Hamayon, R.:. Jouer. Une Etude Anthropologique. La Découverte, Paris (2012) Heinrichs, R.: Lernen von Sensormesswerten zur Verbesserung von Fahrerassistenzsystemen am Beispiel eines adaptiven Kurvenwarners. Universität Berlin, Berlin (2018) Hofmann, P.: Hybridfahrzeuge: Ein Alternatives Antriebssystem für die Zukunft, 2nd edn. Springer, Wine (2014). https://doi.org/10.1007/978-3-7091-1780-4 Iseki, H., Demisch, A.: Examining the linkages between electronic roadway tolling technologies and road pricing policy objectives. Res. Transp. Econ. Selected Papers from the 12th WCTR Topic Area Transport Economics and Finance 36, 121–132 (2012). https://doi.org/10.1016/j. retrec.2012.03.008 Jacob, A., Teuteberg, F.: Game-Based Learning, Serious Games, Business Games und Gamification –Lernförderliche Anwendungsszenarien, gewonnene Erkenntnisse und Handlungsempfehlungen. In: Strahringer, S., Leyh, C. (eds.) Gamification und Serious Games. EH, pp. 97–112. Springer, Wiesbaden (2017). https://doi.org/10.1007/978-3-658-16742-4_8 Jung, J., van Laak, B.L.: Flottenmanagementsysteme - Grundlegende Technologien. Funktionen und Marktüberblick (Arbeitsbericht No. Nr 28) (2001) Landers, R.N., Callan, R.C.: Casual social games as serious games: the psychology of gamification in undergraduate education and employee training. In: Ma, M., Oikonomou, A., Jain, L.C. (eds.) Serious Games and Edutainment Applications, pp. 399–423. Springer, London (2011). https:// doi.org/10.1007/978-1-4471-2161-9_20 Lee, H.L., Padmanabhan, V., Whang, S.: Information distortion in a supply chain: the bullwhip effect. Manag. Sci. 50, 1875–1886 (2004). https://doi.org/10.1287/mnsc.1040.0266 Michael, D., Chen, S.: Serious Games: Games That Educate, Train and Inform. Thomson Course Technology, Boston (2006) Mörbe, M.: Fahrdynamik-Sensoren für FAS, in: Handbuch Fahrerassistenzsysteme, pp. 94–109. Vieweg+Teubner Verlag, Wiesbaden (2012). https://doi.org/10.1007/978-3-8348-8619-4_11
Developing a Serious Game for Intelligent Transportation Systems
497
Organisation for Economic Co-operation and Development: Delivering the Goods: 21st Century Challenges to Urban Goods Transport. Transport Organisation for Economic Co-operation and Development, Paris (2003) Rieckenberg, T.: Telematik im Schienengüterverkehr (PhD Thesis). TU Berlin (2004) Ritterfeld, U., Cody, M., Vorderer, P.: Serious Games: Mechanisms and Effects, 1st edn. Routledge, London (2009). https://doi.org/10.4324/9780203891650 Savelsbergh, M., Van Woensel, T.: 50th anniversary invited article—city logistics: challenges and opportunities. Transp. Sci. 50(2), 579–590 (2016) Savignac, E.: The Gamification of Work: The Use of Games in the Workplace. Wiley, Hoboken (2017) Sawyer, B., Smith, P.: Serious Games Taxonomy. Serious Games Initiative (2008). Retrieved 28 June 2019 Smith, R.: Directive 2010/41/EU of the European Parliament and of the Council of 7 July 2010. In: Core EU Legislation, pp. 352–355. Macmillan Education UK, London (2015). https://doi. org/10.1007/978-1-137-54482-7_33 Statistisches Bundesamt. Verkehr (No. Fachserie 8), Verkehr aktuell. Destatis (2020) Steinegger, R., Barbey, H.-P.: Beer distribution game: erfolgreich dank Regelungstechnik. Presented at the 22. Symposium Simulationstechnik (ASIM 2014), Berlin, Deutschland, 3–5 September 2014. ZHAW Zürcher Hochschule für Angewandte Wissenschaften (2014). https:// doi.org/10.21256/zhaw-3477 Strauß, S.: City-Logistik - Ein Instrument zur Verringerung des Städtischen Güterverkehrs, 7th edn. Kassel (1997) Susi, T., Johannesson, M., Backlund, P.: Serious Games: An Overview. Institutionen för Kommunikation och information (2007) Vahrenkamp, R., Kotzab, H.: Logistik Management und Strategien, 7th edn. Oldenburg Verlag (2012) Wanick, V., Bui, H.: Gamification in management: a systematic review and research directions. Int. J. Ser. Games 6, 57–74 (2019). https://doi.org/10.17083/ijsg.v6i2.282 Wie umweltfreundlich sind Elektroautos? Eine ganzheitliche Bilanz: Bundesministerium für Umwelt. Naturschutz und nukleare Sicherheit - BMU, Berlin (2019) Wood, L., Reiners, T.: Gamification in logistics and supply chain education: extending active learning. Internet Technol. Soc. 2012, 101–108 (2012) Wouters, P., van Nimwegen, C., van Oostendorp, H., van Der Spek, E.D.: A meta-analysis of the cognitive and motivational effects of serious games. J. Educ. Psychol. 105, 249–265 (2013). https://doi.org/10.1037/a0031311 Zyda, M.: From visual simulation to virtual reality to games. Computer 38, 25–32 (2005). https:// doi.org/10.1109/MC.2005.297
Author Index
B Bäumler, Ilja, 38, 486 Berger, Bernhard J., 349 Bhardwaj, Debarshee, 3 Bode, Christopher, 376 Börold, Axel, 462 Brandtner, Patrick, 435 Broda, Eike, 290, 462 C Capayova, Silvia, 449 Castillo-Villagra, Raúl, 92 Chankov, Stanislav, 474 D Dogtas, Okan, 486 Drechsler, Rolf, 349 E Elfers, Moritz, 486 Engelhardt, Maximilian, 275 Ernits, Rafael Mortensen, 52 Eyigün, Sercan, 486 F Fatahi Valilai, Omid, 26, 301 Fischer, Julia, 15 Flores da Silva, Mauricio Randolfo, 386 Forcellini, Fernando Antônio, 409 Frazzon, Enzo Morosini, 144, 409 Freitag, Michael, 52, 290, 423, 462 Fu, Shan, 326
G Galkin, Andrii, 449 Geier, Ben, 275 Geyer, Martin, 217 Grasse, Ole, 133, 156 Gresens, Fynn, 486 Gyulyev, Nizami, 449 H Haasis, Hans-Dietrich, 117 Hämäläinen, Esa, 204 Heger, Jens, 376 Herman, Katharina, 435 Hinckeldeyn, Johannes, 179 Hüseyino˘glu, I¸sık Özge Yumurtacı, 38 I Inkinen, Tommi, 204 J Jahn, Carlos, 133, 156, 179 Jathe, Nicolas, 462 Jedermann, Reiner, 217 K Karimi, Hamid Reza, 326 Kastner, Marvin, 133 Keiser, Dennis, 52 Keuschen, Thomas, 314 Khaturia, Roshaali, 301 Kießner, Phillip, 363 Kinra, Aseem, 3, 65, 104 Klumpp, Matthias, 314 Koleva, Simona, 474
© The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 M. Freitag et al. (Eds.): LDIC 2022, LNLO, pp. 499–500, 2022. https://doi.org/10.1007/978-3-031-05359-7
500 Kolmykova, Anna, 82 Kotzab, Herbert, 15, 38, 104 Kreutzfeldt, Jochen, 179 Kühn, Mathias, 396
Author Index Rose, Hendrik, 179 Roß, Antje, 168
M Mahajan, Pramod, 217 Massimiani, Andrea, 435 Mendes, Lúcio Galvão, 409 Morosini Frazzon, Enzo, 386 Müller, Alina, 314
S Schauer, Oliver, 435 Schlosser, Tibor, 449 Schmidt, Thorsten, 396 Schulte, Frederik, 244, 261 Schulz, Arne, 231 Schütze, Felix, 156 Schweers, Dirk, 462 Schwemmer, Julia, 396 Schwientek, Anne Kathrina, 156 Seeck, Stephan, 275 Serkowsky, Janik, 15 Siekmann, Fabian, 104 Sousa Agostino, Ícaro Romolo, 386 Spaan, Matthijs T. J., 244, 261 Sprodowski, Tobias, 462 Steinbacher, Lennart, 423
N Navendan, Karthikeyan, 26 Negenborn, Rudy R., 244, 261 Neto, Luciana Amaral Stradioto, 386
T Teucke, Michael, 290 Thoben, Klaus-Dieter, 92 Triska, Yuri, 144
O Olaniyi, Eunice O., 204
V Veigt, Marius, 423 Völker, Michael, 396 Voss, Thomas, 376
L Lang, Walter, 217 Lange, Ann-Kathrin, 179 Lange, Eva Ricarda, 191 Lange, Kerstin, 117, 168 Lobashov, Oleksii, 449 Los, Johan, 244, 261 Lütjen, Michael, 52
P Perera, H. Niles, 337, 363 Perera, H. Y. Ranjit, 337 Pfeiffer, Christian, 231 Pfoser, Sarah, 435 Plump, Christina, 349 Pupkes, Birte, 52 R Ren, Bingxuan, 326 Ribeiro, Danilo Ribamar Sá, 409 Rohde, Ann-Kathrin, 52
W Warmbier, Piotr, 65 Wicaksono, Hendro, 26, 301 Y Yin, Tangwen, 326 Z Zander, Bennet, 117 Zeitler, Waldemar, 462