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

Hierarchical Reinforcement Learning for Autonomous Motion Planning with Real Robots

Research Project

Project/Area Number 19K20370
Research Category

Grant-in-Aid for Early-Career Scientists

Allocation TypeMulti-year Fund
Review Section Basic Section 61050:Intelligent robotics-related
Research InstitutionThe University of Tokyo (2022)
Kyushu Institute of Technology (2019-2021)

Principal Investigator

Osa Takayuki  東京大学, 大学院情報理工学系研究科, 准教授 (50804663)

Project Period (FY) 2019-04-01 – 2023-03-31
Project Status Completed (Fiscal Year 2022)
Budget Amount *help
¥4,030,000 (Direct Cost: ¥3,100,000、Indirect Cost: ¥930,000)
Fiscal Year 2022: ¥520,000 (Direct Cost: ¥400,000、Indirect Cost: ¥120,000)
Fiscal Year 2021: ¥520,000 (Direct Cost: ¥400,000、Indirect Cost: ¥120,000)
Fiscal Year 2020: ¥520,000 (Direct Cost: ¥400,000、Indirect Cost: ¥120,000)
Fiscal Year 2019: ¥2,470,000 (Direct Cost: ¥1,900,000、Indirect Cost: ¥570,000)
Keywords強化学習 / 軌道計画 / ロボティクス / 深層強化学習 / 動作計画 / ロボット / 軌道最適化 / 深層学習 / ニューラルネットワーク
Outline of Research at the Start

現在の深層強化学習は,その発展が社会的な注目を集める一方,複雑な動作を計画するための方策を自律的に学習するということが,現実世界のロボットに対しては実現できていない.本研究では,学習機能を備えたロボットシステムの社会実装を目指し,実社会で現実的に利用可能なレベルで,複雑な軌道を計画する方法を学習することのできる階層型深層強化学習アルゴリズムの開発に取り組む.

Outline of Final Research Achievements

Reinforcement learning (RL) is a promising approach to realizing autonomous robots that work in the real world. However, RL faces challenges in learning efficiency and adaptation to changes in the environment. This study aimed to develop a framework for deep RL that can be adapted to changes in the environment by learning various types of movements and using them differently. First, we constructed an algorithm that finds diverse solutions to the problem of trajectory optimization, which is necessary for robot motion planning. Using this knowledge, we also developed an algorithm for deep RL that can find and model a myriad of solutions. We also showed that the algorithm can efficiently adapt to changes in the environment by using a variety of behaviors obtained by the proposed algorithm.

Academic Significance and Societal Importance of the Research Achievements

従来の研究において,ロボットの動作計画問題には無数の多様な解が存在しうることが指摘されていたが,それらを一括して導出しモデル化する手法はこれまでなかった。本研究の成果は,無数の多様な軌道を一括して導出・モデル化することを可能にした点で新規性が高い。同様に,多様な挙動を一度に学習することを深層強化学習においても実現した点にも価値がある。本研究で得られた成果は,強化学習等を活用したロボットシステムにおいて環境の変化への適応を劇的に効率化する可能性を秘めており,実社会での適用先を広げると考えられる。また,これらの成果は国際的に認知され,2022年にはロボット学習分野のトップ学会にて招待講演を行った。

Report

(5 results)
  • 2022 Annual Research Report   Final Research Report ( PDF )
  • 2021 Research-status Report
  • 2020 Research-status Report
  • 2019 Research-status Report
  • Research Products

    (19 results)

All 2023 2022 2021 2020 2019 Other

All Journal Article (7 results) (of which Int'l Joint Research: 2 results,  Peer Reviewed: 5 results,  Open Access: 1 results) Presentation (10 results) (of which Int'l Joint Research: 5 results,  Invited: 4 results) Remarks (2 results)

  • [Journal Article] Discovering diverse solutions in deep reinforcement learning by maximizing state?action-based mutual information2022

    • Author(s)
      Osa Takayuki、Tangkaratt Voot、Sugiyama Masashi
    • Journal Title

      Neural Networks

      Volume: 152 Pages: 90-104

    • DOI

      10.1016/j.neunet.2022.04.009

    • Related Report
      2022 Annual Research Report
  • [Journal Article] Motion planning by learning the solution manifold in trajectory optimization2022

    • Author(s)
      Osa Takayuki
    • Journal Title

      The International Journal of Robotics Research

      Volume: - Issue: 3 Pages: 281-311

    • DOI

      10.1177/02783649211044405

    • Related Report
      2021 Research-status Report
    • Peer Reviewed
  • [Journal Article] Deep Reinforcement Learning With Adversarial Training for Automated Excavation Using Depth Images2022

    • Author(s)
      Osa Takayuki、Aizawa Masanori
    • Journal Title

      IEEE Access

      Volume: 10 Pages: 4523-4535

    • DOI

      10.1109/access.2022.3140781

    • Related Report
      2021 Research-status Report
    • Peer Reviewed / Open Access
  • [Journal Article] Hierarchical Deep Reinforcement Learning for Robots2021

    • Author(s)
      長 隆之
    • Journal Title

      Journal of the Robotics Society of Japan

      Volume: 39 Issue: 7 Pages: 613-616

    • DOI

      10.7210/jrsj.39.613

    • NAID

      130008083627

    • ISSN
      0289-1824, 1884-7145
    • Related Report
      2021 Research-status Report
  • [Journal Article] Goal-Conditioned Variational Autoencoder Trajectory Primitives with Continuous and Discrete Latent Codes2020

    • Author(s)
      Takayuki Osa, Shuhei Ikemoto
    • Journal Title

      SN Computer Science

      Volume: 1 Issue: 5

    • DOI

      10.1007/s42979-020-00324-7

    • NAID

      120007147069

    • Related Report
      2020 Research-status Report
    • Peer Reviewed
  • [Journal Article] Multimodal Trajectory Optimization for Motion Planning2020

    • Author(s)
      Takayuki Osa
    • Journal Title

      The International Journal of Robotics Research

      Volume: 39 Issue: 8 Pages: 1-19

    • DOI

      10.1177/0278364920918296

    • Related Report
      2019 Research-status Report
    • Peer Reviewed / Int'l Joint Research
  • [Journal Article] Hierarchical Stochastic Optimization with Application to Parameter Tuning for Electronically Controlled Transmissions2020

    • Author(s)
      Hiroyuki Karasawa, Tomohiro Kanemaki, Kei Oomae, Rui Fukui, Masayuki Nakao, Takayuki Osa
    • Journal Title

      IEEE Robotics and Automation Letters

      Volume: 5 Issue: 2 Pages: 628-635

    • DOI

      10.1109/lra.2020.2965085

    • Related Report
      2019 Research-status Report
    • Peer Reviewed / Int'l Joint Research
  • [Presentation] Discovering diverse solutions in reinforcement learning2023

    • Author(s)
      Takayuki Osa
    • Organizer
      Workshop on Functional Inference and Machine Intelligence
    • Related Report
      2022 Annual Research Report
    • Int'l Joint Research / Invited
  • [Presentation] Dealing with the objective function with multiple extrema in robot learning2022

    • Author(s)
      Takayuki Osa
    • Organizer
      Conference on Robot Learning
    • Related Report
      2022 Annual Research Report
    • Int'l Joint Research / Invited
  • [Presentation] 動的障害物回避のための多峰性最適化を用い たオンライン軌道計画法2022

    • Author(s)
      是澤真由、長隆之
    • Organizer
      第40回日本ロボット学会学術講演会
    • Related Report
      2022 Annual Research Report
  • [Presentation] What should we learn in a robot-learning system?2022

    • Author(s)
      Takayuki Osa
    • Organizer
      2nd RL-CONFORM Workshop at IEEE/RSJ International Conference on Intelligent Robots and Systems
    • Related Report
      2022 Annual Research Report
    • Int'l Joint Research / Invited
  • [Presentation] 深層強化学習による掘削動作の異なる土質へのfewshot adaptation2022

    • Author(s)
      筬島 直人、逢澤 正憲、長 隆之
    • Organizer
      第40回日本ロボット学会学術講演会
    • Related Report
      2022 Annual Research Report
  • [Presentation] 多様なロボットの挙動を学習する深層強化学習2021

    • Author(s)
      長隆之
    • Organizer
      第39回日本ロボット学会学術講演会
    • Related Report
      2021 Research-status Report
  • [Presentation] 解の潜在空間を用いた軌道計画2020

    • Author(s)
      長隆之
    • Organizer
      第38回日本ロボット学会学術講演会
    • Related Report
      2020 Research-status Report
  • [Presentation] Reducing Overestimation Bias in Multi-Agent Domains Using Double Centralized Critics2019

    • Author(s)
      Johannes Ackerman, Takayuki Osa, Masashi Sugiyama
    • Organizer
      NeurIPS 2019 Deep Reinforcement Learning Workshop
    • Related Report
      2019 Research-status Report
    • Int'l Joint Research
  • [Presentation] How should we design a robot learning system?2019

    • Author(s)
      Takayuki Osa
    • Organizer
      Workshop on Robot Learning: Control and Interaction in the Real World, NeurIPS 2019
    • Related Report
      2019 Research-status Report
    • Int'l Joint Research / Invited
  • [Presentation] Trajectory optimization via density estimation2019

    • Author(s)
      Takayuki Osa
    • Organizer
      第37回日本ロボット学会学術講演会
    • Related Report
      2019 Research-status Report
  • [Remarks] 九州工業大学 長研究室ホームページ

    • URL

      http://www.brain.kyutech.ac.jp/~osa/

    • Related Report
      2019 Research-status Report
  • [Remarks] 九州工業大学研究者詳細ページ

    • URL

      https://hyokadb02.jimu.kyutech.ac.jp/html/100001202_ja.html

    • Related Report
      2019 Research-status Report

URL: 

Published: 2019-04-18   Modified: 2024-01-30  

Information User Guide FAQ News Terms of Use Attribution of KAKENHI

Powered by NII kakenhi