Stein Variational Goal Generation for adaptive Exploration in Multi-Goal Reinforcement Learning

被引:0
|
作者
Castanet, Nicolas [1 ]
Sigaud, Olivier [1 ]
Lamprier, Sylvain [2 ]
机构
[1] Sorbonne Univ, ISIR, Paris, France
[2] Univ Angers, LERIA, SFR MATHSTIC, F-49000 Angers, France
关键词
D O I
暂无
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
In multi-goal Reinforcement Learning, an agent can share experience between related training tasks, resulting in better generalization for new tasks at test time. However, when the goal space has discontinuities and the reward is sparse, a majority of goals are difficult to reach. In this context, a curriculum over goals helps agents learn by adapting training tasks to their current capabilities. In this work we propose Stein Variational Goal Generation (SVGG), which samples goals of intermediate difficulty for the agent, by leveraging a learned predictive model of its goal reaching capabilities. The distribution of goals is modeled with particles that are attracted in areas of appropriate difficulty using Stein Variational Gradient Descent. We show that SVGG outperforms state-of-the-art multi-goal Reinforcement Learning methods in terms of success coverage in hard exploration problems, and demonstrate that it is endowed with a useful recovery property when the environment changes.
引用
收藏
页数:18
相关论文
共 50 条
  • [21] Reinforcement Learning Control Based on Multi-Goal Representation Using Hierarchical Heuristic Dynamic Programming
    Ni, Zhen
    He, Haibo
    Zhao, Dongbin
    Prokhorov, Danil V.
    2012 INTERNATIONAL JOINT CONFERENCE ON NEURAL NETWORKS (IJCNN), 2012,
  • [22] Two-stage visual navigation by deep neural networks and multi-goal reinforcement learning
    Shantia, Amirhossein
    Timmers, Rik
    Chong, Yiebo
    Kuiper, Cornel
    Bidoia, Francesco
    Schomaker, Lambert
    Wiering, Marco
    ROBOTICS AND AUTONOMOUS SYSTEMS, 2021, 138
  • [23] MULTI-GOAL OPTIMIZATION IN MANAGERIAL SCIENCE
    BAUM, S
    CARLSON, RC
    OMEGA-INTERNATIONAL JOURNAL OF MANAGEMENT SCIENCE, 1974, 2 (05): : 607 - 623
  • [24] Pragmatically Learning from Pedagogical Demonstrations in Multi-Goal Environments
    Caselles-Dupre, Hugo
    Sigaud, Olivier
    Chetouani, Mohamed
    ADVANCES IN NEURAL INFORMATION PROCESSING SYSTEMS 35 (NEURIPS 2022), 2022,
  • [25] MULTI-GOAL ANALYZER OF THE SOCIOLOGICAL INFORMATION
    STANILEVICH, VI
    FEDOTOV, YV
    FEDOROV, SV
    SOTSIOLOGICHESKIE ISSLEDOVANIYA, 1985, (04): : 114 - 118
  • [26] Automatic Goal Generation for Reinforcement Learning Agents
    Florensa, Carlos
    Held, David
    Geng, Xinyang
    Abbeel, Pieter
    INTERNATIONAL CONFERENCE ON MACHINE LEARNING, VOL 80, 2018, 80
  • [27] Reward-weighted DHER Mechanism For Multi-goal Reinforcement Learning With Application To Robotic Manipulation Control
    Wei, Xueyu
    Duan, Lilong
    Xue, Wei
    JOURNAL OF APPLIED SCIENCE AND ENGINEERING, 2023, 26 (12): : 1829 - 1841
  • [28] Graph-Structured Policy Learning for Multi-Goal Manipulation Tasks
    Klee, David
    Biza, Ondrej
    Platt, Robert
    2022 IEEE/RSJ INTERNATIONAL CONFERENCE ON INTELLIGENT ROBOTS AND SYSTEMS (IROS), 2022, : 4765 - 4772
  • [29] Variational Empowerment as Representation Learning for Goal-Based Reinforcement Learning
    Choi, Jongwook
    Sharma, Archit
    Lee, Honglak
    Levine, Sergey
    Gu, Shixiang Shane
    INTERNATIONAL CONFERENCE ON MACHINE LEARNING, VOL 139, 2021, 139
  • [30] Planning multi-goal tours for robot arms
    Saha, M
    Sánchez-Ante, G
    Latombe, JC
    2003 IEEE INTERNATIONAL CONFERENCE ON ROBOTICS AND AUTOMATION, VOLS 1-3, PROCEEDINGS, 2003, : 3797 - 3803