Agents and Artificial Intelligence: 13th International Conference, ICAART 2021, Virtual Event, February 4–6, 2021, Revised Selected Papers (Lecture Notes in Computer Science, 13251)

This book constitutes selected papers from the refereed proceedings of the 13th International Conference on Agents and Artificial Intelligence, ICAART 2021, which was held online during February 4–6, 2021. A total of 72 full and 99 short papers were carefully reviewed and selected for the conference from a total of 298 submissions; 17 selected full papers are included in this book. They were organized in topical sections named agents and artificial intelligence. Preface Organization Contents Agents Specification Aware Multi-Agent Reinforcement Learning 1 Introduction 2 Foundations 2.1 Problem Formulation 2.2 Reinforcement Learning 2.3 Multi-Agent Reinforcement Learning 2.4 Reward Shaping in RL 3 Related Work 4 Smart Factory Domain 5 Specification Transfer 6 Evaluation 6.1 Experimental Setup 6.2 Analysis of Separate Reward Components 6.3 Scaling Challenges and Solutions 6.4 Safety Constraints as Reward Components 6.5 Case Study: Specification Compliant Run-Time Behavior 7 Discussion 8 Conclusion and Future Work References Task Bundle Delegation for Reducing the Flowtime 1 Introduction 2 Related Work 3 Multi-agent Situated Task Allocation 4 Consumption and Reallocation 5 Negotiation Strategy 5.1 Peer Modelling 5.2 Acceptability Rule 5.3 Offer Strategy 5.4 Acceptation Strategy 5.5 Agent Behaviour 6 Results and Discussion 6.1 Context of Experiments 6.2 Classical Heuristic and Acceptability Criterion 6.3 N-Ary Delegation 6.4 Distributed Constraint Optimization Problem (DCOP) 7 Conclusion References A Detailed Analysis of a Systematic Review About Requirements Engineering Processes for Multi-agent Systems 1 Introduction 2 Background 2.1 Requirements Engineering 2.2 Belief-Desire-Intention Model 3 Related Works 4 Research Method 4.1 The Research Questions 4.2 Identifying and Selecting Primary Studies 4.3 Inclusion and Exclusion Criteria 4.4 Studies Quality Assessment 4.5 Data Extraction Strategy 4.6 Conducting the Review 4.7 Data Extraction 5 Results 5.1 Methodologies Analysed in the Systematic Review 5.2 Coverage in Relation to the Requirements Engineering Subareas Defined in SWEBOK 5.3 Methodologies Supporting the BDI Model 5.4 Gaps Found in This Review 6 Threats to Validity 7 Insights for Future Works 8 Conclusions and Future Works References Automatically-Generated Agent Organizations for Flexible Workflow Enactment 1 Introduction 2 A Running Example 3 From Business Process Models to Agent Organizations 3.1 Ontology and Goals from a Business Process 3.2 Mapping Goals to Agents via Organization 3.3 The BPMN2MOISE Tool 4 The Automatic Definition of Organizations 4.1 Heuristic for the Structural Specification: Roles and Groups 4.2 Heuristic for the Functional Specification: Goals and Plans 4.3 Heuristic for the Normative Specification 5 Conclusions References Negotiation Considering Privacy Loss on Asymmetric Multi-objective Decentralized Constraint Optimization Problem 1 Introduction 2 Background 2.1 Asymmetric Constraint Optimization Problem with Cost of Private Information to Be Published 2.2 Criteria and Measurement of Social Welfare 2.3 Decentralized Complete Solution Method for Asymmetric Multi-objective Constraint Optimization Problems Based on Pseudo-trees and Dynamic Programming 3 Decentralized Solution Framework for Selection of Utility Values to Be Published and Solution of Published Problems 3.1 Basic Design of Proposed Framework 3.2 Selection of Newly Published Utility Values 3.3 Evaluation Criteria for Utility Values to Be Published 3.4 Hierarchically Structured Cost Vector Integrating Criteria for Publication of Utility Values 3.5 Solving Problems with Published Utility Values 4 Additional Preprocessing Methods 4.1 Trading Utility Between Neighborhood Agents 4.2 Approximation of Binary Functions 5 Evaluation 5.1 Settings of Experiment 5.2 Experimental Results 6 Discussion 7 Conclusion References Artificial Intelligence Utilizing Out-Domain Datasets to Enhance Multi-task Citation Analysis 1 Introduction 2 Related Work 2.1 Sentiment Classification 2.2 Intent Classification 2.3 Out-Domain Data Utilization 3 Datasets 3.1 Sentiment Datasets 3.2 Intent Dataset 4 Contributions 4.1 ImpactCite 4.2 Overcoming Data Scarcity and Data Feeding Techniques 4.3 Fusion Approach 5 Experiments and Analysis 5.1 Intent Classification 5.2 Sentiment Classification 5.3 Out-Domain: Evaluating Impact of Additional Data 5.4 Multi-task Model: Fusing Scientific Sentiment and Intent 6 Discussion 7 Conclusion References Using Possibilistic Networks to Compute Learning Course Indicators 1 Introduction 2 Possibility Theory 3 Message Passing Inference 4 Compiling Possibilistic Networks 5 Experimentation 5.1 Presentation 5.2 Results 6 Conclusion References Assured Deep Multi-Agent Reinforcement Learning for Safe Robotic Systems 1 Introduction 2 Background 2.1 Markov Decision Process 2.2 Multi-Agent Markov Decision Process 2.3 Abstract Markov Decision Process 2.4 Single-Agent Reinforcement Learning 2.5 Multi-Agent Reinforcement Learning 2.6 Deep Reinforcement Learning 2.7 Quantitative Verification 3 Domain Example 4 Approach 5 Evaluation 5.1 Experimental Set-Up 5.2 Radiation Avoidance Patrolling Domain 5.3 Multi-Agent Guarded Flag Collection 6 Related Work 7 Conclusion References How to Segment Handwritten Historical Chronicles Using Fully Convolutional Networks? 1 Introduction 2 Related Work 2.1 Methods 2.2 Datasets 3 FCN Architecture 4 Experimental Setup 5 Experimental Results 5.1 Input Resolution 5.2 Loss Function Weighting 5.3 Training Data Extension 5.4 Combined Setup 5.5 Post-processing 5.6 Transfer Learning 6 Porta Fontium Integration and Method Tuning 7 Conclusions and Future Work References On the Relationship with Toulmin Method to Logic-Based Argumentation 1 Introduction 2 Preliminaries 2.1 Toulmin Model of Argumentation 2.2 NDSA: Natural Deduction for Structured Argumentation 3 The Reception and Refinement of Toulmin's Model in Logic-Based Argumentation 3.1 Reasoning on NDSA and Admissible Sets 3.2 2-Tier AF: Two-Tier Argumentation Framework 4 Related Work 5 Conclusion and Future Direction References Informer: An Efficient Transformer Architecture Using Convolutional Layers 1 Introduction 2 Previous Work 3 Information Organization Layer 4 Experiments 4.1 Training Details 4.2 Results Analysis 5 Conclusions References Improving the Generalization of Deep Learning Classification Models in Medical Imaging Using Transfer Learning and Generative Adversarial Networks 1 Introduction 2 Related Work 2.1 Transfer Learning in Medical Imaging 2.2 Generative Adversarial Networks in Medical Imaging 3 Materials and Methods 3.1 Dataset Description and Pre-processing 3.2 Transfer Learning Models 3.3 Generative Adversarial Networks (GAN) 4 Results 4.1 Comparison of Results with Other Recent Similar Works 5 Discussion 6 Conclusions and Future Work References An Interpretable Word Sense Classifier for Human Explainable Chatbot 1 Introduction 2 Related Work 3 Proposed Tsetlin Machine Based Word Sense Disambiguation 3.1 Tsetlin Machine 3.2 Training of the Proposed WSD Model 4 Performance and Interpretation of WSD 4.1 Results 4.2 Explainable WSD 4.3 Application of Interpretation in Chatbot 5 Conclusion References A Tsetlin Machine Framework for Universal Outlier and Novelty Detection 1 Introduction 2 Related Work 3 Tsetlin Machine Framework 3.1 TM Architecture for Generator 3.2 Classifier 4 Experiments and Results 4.1 Outlier Detection 4.2 Novelty Detection in Text 5 Conclusions References Adding Supply/Demand Imbalance-Sensitivity to Simple Automated Trader-Agents 1 Introduction 2 Prior Work 2.1 Automated Traders 2.2 Critique of Church & Cliff 2.3 Measuring Imbalance 3 Adding MLOFI-Impact to Robot Traders 3.1 Simple Robot Traders with Impact: AA, ZIP, and ISHV 3.2 MLOFI Opinionated PRZI Traders for Narrative Economics 4 Discussion and Conclusion References Advances in Measuring Inflation Within Virtual Economies Using Deep Reinforcement Learning 1 Introduction 1.1 Proposed System 2 Related Works 2.1 Parameter Tuning 2.2 Game Balance 2.3 Automated Game Testing 3 History of Economies in Games 3.1 Eve Online 4 Testing Economies 5 Parameter Tuning 6 Methods 6.1 Economy Design 6.2 Adventure Agents 6.3 Co-operative Behaviours 6.4 Crafting Agents 6.5 Training 6.6 Economy Sinks 6.7 Training 6.8 Parameter Tuning 6.9 Data Collection 7 Inflation Results 7.1 Results - Parameter Tuning 8 Discussion and Future Work References Practical City Scale Stochastic Path Planning with Pre-computation 1 Introduction 2 Previous Work 3 Framework 3.1 City and Edge Weights 3.2 Traffic Data 3.3 Open Street Map 3.4 Agents 3.5 City Graph Partitioning 3.6 Exemplar Assignment 3.7 Base Path Planning Framework 3.8 Pre-processing: Building Distance Oracles 3.9 Scalable Algorithm 4 Experiments and Results 4.1 How Many Partitions Are Needed to Represent the City Graph? 4.2 Which Partitioning Method We Picked? 4.3 Which Exemplar Assignment Approach Is the Best? 4.4 How Is the Quality of Approximate Paths? 4.5 What Is the Time and Space Complexity of Scalable Algorithm? 5 Conclusion and Future Work References Author Index