Improving sustainability, flexibility, and robustness of artifactitious systems using emergence of divisional cooperation

2021 Fiscal Year Final Research Report

• Project/Area Number: 17KT0044
• Research Category: Grant-in-Aid for Scientific Research (B)
• Allocation Type: Multi-year Fund
• Section: 特設分野
• Research Field: Intensification of Artifact Systems
• Research Institution: Waseda University
• Principal Investigator: Sugawara Toshiharu 早稲田大学, 理工学術院, 教授 (70396133)
• Project Period (FY): 2017-07-18 – 2022-03-31
• Keywords: マルチエージェントシステム / 持続可能性 / 分業 / 強化学習 / 交渉 / 自律分散システム

Outline of Final Research Achievements

Recent developments in computer/AI and machine technology have led to promising applications of multi-agent systems consisting of multiple intelligent agents (e.g., self-driving robots) that make decisions autonomously and cooperate/coordinate with each other. Because agents are often software programs running on computers and/or controlling machines, their replacement, renewal, and periodic inspections are mandatory to maintain the sustainability and robustness of the system. However, there is a temporary but significant loss of performance that occurs when they are stopped for these purposes. To mitigate this, we proposed a negotiation method in which agents delegate tasks, especially important ones, to others. We also pursued a learning method that builds organization and division of labor among agents in a bottom-up manner to increase overall efficiency. We believe that our results have received academic recognition, including presentations at top-level conferences in this 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

少子化問題や危険箇所での作業など、人間の代理として作業する機械（ロボット等）が期待されている。特に、広大な範囲や複雑な作業が必要な時には、複数のロボットなどの協力が必要である。エージェントは、これらの機械を制御するソフトウェアであり、中心的な存在である。本研究では、これらの知的なエージェントが、学習を通じて自ら担当する作業を決定する分業化により全体の効率を上げると共に、更新や定期点検などが予定されている場合には、やはり自律的な交渉を通して協力的に仕事を補完・委託し合い、その効率低下を最小限に押さえる手法を提案している。