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
  • [11] Goal Density-based Hindsight Experience Prioritization for Multi-Goal Robot Manipulation Reinforcement Learning
    Kuang, Yingyi
    Weinberg, Abraham Itzhak
    Vogiatzis, George
    Faria, Diego R.
    2020 29TH IEEE INTERNATIONAL CONFERENCE ON ROBOT AND HUMAN INTERACTIVE COMMUNICATION (RO-MAN), 2020, : 432 - 437
  • [12] Scalable Multi-Robot Cooperation for Multi-Goal Tasks Using Reinforcement Learning
    An, Tianxu
    Lee, Joonho
    Bjelonic, Marko
    De Vincenti, Flavio
    Hutter, Marco
    IEEE ROBOTICS AND AUTOMATION LETTERS, 2025, 10 (02): : 1585 - 1592
  • [13] GOAL AND RESOURCE TRANSFERS IN MULTI-GOAL ORGANIZATION
    BAILEY, AD
    BOE, WJ
    ACCOUNTING REVIEW, 1976, 51 (03): : 559 - 573
  • [14] Improving multi-goal and target-driven reinforcement learning with supervised auxiliary task
    Horita, Luiz R. T.
    Nakamura, Angelica T. M.
    Wolf, Denis F.
    Grassi Junior, Valdir
    2021 20TH INTERNATIONAL CONFERENCE ON ADVANCED ROBOTICS (ICAR), 2021, : 290 - 295
  • [15] A Multi-goal Oriented Approach for Adaptation Rules Generation
    Liu, Yang
    Chen, Zhengyin
    Jiao, Wenpin
    2018 25TH ASIA-PACIFIC SOFTWARE ENGINEERING CONFERENCE (APSEC 2018), 2018, : 249 - 257
  • [16] Biologically inspired reinforcement learning: Reward-based decomposition for multi-goal environments
    Zhou, WD
    Coggins, R
    BIOLOGICALLY INSPIRED APPROACHES TO ADVANCED INFORMATION TECHNOLOGY, 2004, 3141 : 80 - 94
  • [17] Multi-goal Q-learning of cooperative teams
    Li, Jing
    Sheng, Zhaohan
    Ng, KwanChew
    EXPERT SYSTEMS WITH APPLICATIONS, 2011, 38 (03) : 1565 - 1574
  • [18] ADAPTABLE MULTI-GOAL CONTROL
    CARLSSON, C
    EKONOMISKA SAMFUNDETS TIDSKRIFT, 1978, 31 (01): : 75 - 83
  • [19] Learning Multi-Goal Dialogue Strategies Using Reinforcement Learning With Reduced State-Action Spaces
    Cuayahuitl, Heriberto
    Renals, Steve
    Lemon, Oliver
    Shimodaira, Hiroshi
    INTERSPEECH 2006 AND 9TH INTERNATIONAL CONFERENCE ON SPOKEN LANGUAGE PROCESSING, VOLS 1-5, 2006, : 469 - +
  • [20] Hindsight Curriculum Generation Based Multi-Goal Experience Replay
    Feng, Xiaoyun
    IMAGE AND GRAPHICS (ICIG 2021), PT III, 2021, 12890 : 182 - 194