2023 Fiscal Year Final Research Report
Studies on autonomous learning of agents' organizational formation for system efficiency
| Project/Area Number |
20H04245
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (B)
|
| Allocation Type | Single-year Grants |
|---|
| Section | 一般 |
|---|
| Review Section |
Basic Section 61030:Intelligent informatics-related
|
|---|
| Research Institution | Waseda University |
|---|
Principal Investigator |
|
|---|
| Project Period (FY) |
2020-04-01 – 2024-03-31
|
| Keywords | マルチエージェントシステム / 組織行動 / 社会学習 / 深層強化学習 / 組織化 / マルチエージェントプランニング |
|---|
| Outline of Final Research Achievements |
In this study, we have been developed methods for learning appropriate cooperative and coordinated behaviors in a multi-agent system consisting of multiple agents that make decisions autonomously, either organizationally or from the perspective of group behaviors. Specifically, we proposed (1) methods for learning to act while dynamically observing the movements of its neighboring agents in tasks that requires multi-step cooperative actions, and (2) learning methods for determining its own cooperative and coordinated behavior based on its internal estimation of the neighbors' behaviors. Furthermore, we also pursued (3) techniques for identifying the objects that the agent is focusing on in their observations to confirm the validity of the selected cooperative and coordinated behaviors for improving the explainability. We believe that our results have been well received academically, with presentations at top-level international conferences in the field.
|
| Free Research Field |
マルチエージェントシステム
|
| Academic Significance and Societal Importance of the Research Achievements |
人工知能により人間の代理としての役割をもつ知的ソフトウェア（エージェント）が社会に広く普及したときに、これらを活用した協力・協働行動や、それらの間での干渉を避けるための調整行動が必要となる。しかしこれらの行動は個々の利得だけではなく、相互のあるいは社会の観点からの利得を考慮した行動が必要である。ここでは、近年の深層（強化）学習が発展し、ある程度の知的な行動が可能となったとき、単純な最適化、つまり個々の利得を越えた行動の学習の実現が必要となる。ここでは、グループ作業を対象に、組織的な行動の発現とその判断根拠を提示した説明性の可能性について貢献した成果と考える。
|
