• Search Research Projects
  • Search Researchers
  • How to Use
  1. Back to previous page

Knowledge Transfer in Multi-Agent Reinforcement Learning for Unknown Cooperative Environments

Research Project

Project/Area Number 21K17807
Research Category

Grant-in-Aid for Early-Career Scientists

Allocation TypeMulti-year Fund
Review Section Basic Section 61030:Intelligent informatics-related
Research InstitutionOkayama University

Principal Investigator

Uwano Fumito  岡山大学, 環境生命自然科学学域, 助教 (30880687)

Project Period (FY) 2021-04-01 – 2024-03-31
Project Status Completed (Fiscal Year 2023)
Budget Amount *help
¥4,680,000 (Direct Cost: ¥3,600,000、Indirect Cost: ¥1,080,000)
Fiscal Year 2023: ¥1,040,000 (Direct Cost: ¥800,000、Indirect Cost: ¥240,000)
Fiscal Year 2022: ¥1,820,000 (Direct Cost: ¥1,400,000、Indirect Cost: ¥420,000)
Fiscal Year 2021: ¥1,820,000 (Direct Cost: ¥1,400,000、Indirect Cost: ¥420,000)
Keywordsマルチエージェントシステム / 強化学習 / ニューラルネットワーク / 未知環境 / 知識 / 知識転移
Outline of Research at the Start

本研究では,ロボットなどの活動主体(エージェント)が複数存在するときの協調制御ルールを,周囲環境から得た情報から各々が学習するマルチエージェント強化学習において,学習すべき協調や環境が未知であるときに適応した協調行動学習法を提案する.具体的には,他の環境などで今まで学習したエージェントの学習結果を各要素に分割し,階層的に抽象化することで生成した知識を組み合わせて学習することで未知の協調・環境に適応する.

Outline of Final Research Achievements

In this study, we clarified an efficient method for utilizing knowledge in unknown cooperative and environmental contexts and developed a methodology for multi-agent reinforcement learning based on this approach. We demonstrated its effectiveness through experiments. Specifically, we proposed a knowledge module that extracts environmental information using neural networks and employs a tree-structured function. By reconfiguring the branches and leaves of the tree-structured function and optimizing the parameters through reinforcement learning, we adapted to unknown cooperation and environmental contexts. The results of this research have been reported in one English journal, two international conferences, one invited lecture, and four domestic academic conferences.

Academic Significance and Societal Importance of the Research Achievements

本研究は、構造型メタ知識に基づくマルチエージェント学習という従来の研究領域に対して、要素型メタ知識を扱えるように拡張することから、学術的観点から見ると新たな研究領域を開く位置づけとなり、学術的意義が大きい。更に、この基盤技術が確立すれば、ロボットの学習結果を未知問題に転移可能となることから、複数の災害救助ロボットや宇宙探査機による問題解決が可能となる。また、災害救助ロボットで得た学習結果が宇宙探査機に活用可能となるなど、知識の相互利用が可能となるため、産業的意義そして社会的意義も極めて大きい。

Report

(4 results)
  • 2023 Annual Research Report   Final Research Report ( PDF )
  • 2022 Research-status Report
  • 2021 Research-status Report
  • Research Products

    (15 results)

All 2024 2023 2022 2021 Other

All Int'l Joint Research (2 results) Journal Article (2 results) (of which Peer Reviewed: 1 results,  Open Access: 1 results) Presentation (10 results) (of which Int'l Joint Research: 5 results,  Invited: 1 results) Book (1 results)

  • [Int'l Joint Research] Queensland University of Technology(オーストラリア)

    • Related Report
      2023 Annual Research Report
  • [Int'l Joint Research] Queensland University of Technology(オーストラリア)

    • Related Report
      2022 Research-status Report
  • [Journal Article] Inverse Reinforcement Learning with Agents’ Biased Exploration Based on Sub-Optimal Sequential Action Data2024

    • Author(s)
      Uwano Fumito、Hasegawa Satoshi、Takadama Keiki
    • Journal Title

      Journal of Advanced Computational Intelligence and Intelligent Informatics

      Volume: 28 Issue: 2 Pages: 380-392

    • DOI

      10.20965/jaciii.2024.p0380

    • ISSN
      1343-0130, 1883-8014
    • Year and Date
      2024-03-20
    • Related Report
      2023 Annual Research Report
    • Peer Reviewed / Open Access
  • [Journal Article] Multi-Agent Reinforcement Learning in Different Granularity of Observations2023

    • Author(s)
      上野 史
    • Journal Title

      Journal of The Society of Instrument and Control Engineers

      Volume: 62 Issue: 2 Pages: 104-104

    • DOI

      10.11499/sicejl.62.104

    • ISSN
      0453-4662, 1883-8170
    • Year and Date
      2023-02-10
    • Related Report
      2022 Research-status Report
  • [Presentation] Reinforcement Learning in Cyclic Environmental Change for Non-Communicative Agents: A Theoretical Approach2023

    • Author(s)
      Fumito Uwano
    • Organizer
      The 5th International Workshop on Explainable and Transparent AI and Multi-Agent Systems
    • Related Report
      2023 Annual Research Report
    • Int'l Joint Research
  • [Presentation] Hierarchical Frames-of-References in Learning Classifier Systems2023

    • Author(s)
      Fumito Uwano
    • Organizer
      The Genetic and Evolutionary Computation Conference 2023
    • Related Report
      2023 Annual Research Report
    • Int'l Joint Research
  • [Presentation] Implicit Cooperative Learning on Distribution of Received Reward in Multi-agent System2023

    • Author(s)
      Fumito Uwano
    • Organizer
      15th International Conference on Agents and Artificial Intelligence (ICAART 2023)
    • Related Report
      2022 Research-status Report
    • Int'l Joint Research
  • [Presentation] Reinforcement Learning in Cyclic Environmental Change for Non-Communicative Agents: A Theoretical Approach2023

    • Author(s)
      Fumito Uwano
    • Organizer
      5th International Workshop on Explainable and Transparent AI and Multi-Agent Systems (EXTRAAMAS 2023)
    • Related Report
      2022 Research-status Report
    • Int'l Joint Research
  • [Presentation] Design of Human-Agent-Group Interaction for Correct Opinion Sharing on Social Media2022

    • Author(s)
      Fumito Uwano
    • Organizer
      25th International Conference on Human-Computer Interaction
    • Related Report
      2022 Research-status Report
    • Int'l Joint Research
  • [Presentation] マルチエージェント強化学習における知識とその境界2022

    • Author(s)
      上野 史
    • Organizer
      第69回自律分散システム部会研究会「若手を中心とした模倣学習・強化学習」
    • Related Report
      2022 Research-status Report
    • Invited
  • [Presentation] マルチエージェント強化学習の報酬設計による知識の蒸留と転移に関する一考察2022

    • Author(s)
      上野 史
    • Organizer
      第36回人工知能学会全国大会
    • Related Report
      2022 Research-status Report
  • [Presentation] 獲得報酬の分布に基づくエージェント間の暗黙的協調行動学習とその効果の検証2022

    • Author(s)
      上野 史
    • Organizer
      SMASH22 Winter Symposium
    • Related Report
      2021 Research-status Report
  • [Presentation] 未知の協調・環境を想定したマルチエージェント強化学習の知識転移2021

    • Author(s)
      上野 史
    • Organizer
      境界と関係性を視座とするシステムズアプローチ調査研究会
    • Related Report
      2021 Research-status Report
  • [Presentation] 未知の環境に適応する学習エージェント群の知識利用法の検討2021

    • Author(s)
      上野 史
    • Organizer
      計測自動制御学会システム・情報部門学術講演会2021
    • Related Report
      2021 Research-status Report
  • [Book] Explainable and Transparent AI and Multi-Agent Systems2023

    • Author(s)
      Fumito Uwano and Keiki Takadama
    • Total Pages
      281
    • Publisher
      Springer
    • Related Report
      2023 Annual Research Report

URL: 

Published: 2021-04-28   Modified: 2025-01-30  

Information User Guide FAQ News Terms of Use Attribution of KAKENHI

Powered by NII kakenhi