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
  • [41] TOWARD A MULTI-LEVEL, MULTI-GOAL INFORMATION SYSTEM
    FIELD, JE
    ACCOUNTING REVIEW, 1969, 44 (03): : 593 - 599
  • [42] Learning behavior-selection by emotions and cognition in a multi-goal robot task
    Gadanho, SC
    JOURNAL OF MACHINE LEARNING RESEARCH, 2004, 4 (03) : 385 - 412
  • [43] Generalizing Goal-Conditioned Reinforcement Learning with Variational Causal Reasoning
    Ding, Wenhao
    Lin, Haohong
    Li, Bo
    Zhao, Ding
    ADVANCES IN NEURAL INFORMATION PROCESSING SYSTEMS 35 (NEURIPS 2022), 2022,
  • [44] Multi-goal programming model for quality function deployment
    Xi'an Jiaotong Univ, Shanxi, China
    Jisuanji Jicheng Zhizao Xitong, 6 (26-30):
  • [45] ON THE FORMULATION AND USE OF A MULTI-GOAL FUNCTION FOR AN INDUSTRIAL FIRM
    JOHNSEN, E
    OPERATIONS RESEARCH, 1961, 9 : B31 - B31
  • [46] CPA/Tiger-MGP: test-goal set partitioning for efficient multi-goal test-suite generation
    Sebastian Ruland
    Malte Lochau
    Oliver Fehse
    Andy Schürr
    International Journal on Software Tools for Technology Transfer, 2021, 23 : 853 - 856
  • [47] Multi-Goal Multi-Agent Path Finding via Decoupled and Integrated Goal Vertex Ordering
    Surynek, Pavel
    THIRTY-FIFTH AAAI CONFERENCE ON ARTIFICIAL INTELLIGENCE, THIRTY-THIRD CONFERENCE ON INNOVATIVE APPLICATIONS OF ARTIFICIAL INTELLIGENCE AND THE ELEVENTH SYMPOSIUM ON EDUCATIONAL ADVANCES IN ARTIFICIAL INTELLIGENCE, 2021, 35 : 12409 - 12417
  • [48] Evaluation of a Multi-Goal Solver for Use in a Blackboard Architecture
    Straub, Jeremy
    INTERNATIONAL JOURNAL OF DECISION SUPPORT SYSTEM TECHNOLOGY, 2014, 6 (01) : 1 - 13
  • [49] MULTI-GOAL TRAJECTORY PLANNING FOR REDUNDANT SPACE ROBOT
    Zhao, Suping
    Zhu, Zhanxia
    Siciliano, Bruno
    Gutierrez-Giles, Alejandro
    Feng, Jing Lang
    Luo, Jianjun
    SPACEFLIGHT MECHANICS 2017, PTS I - IV, 2017, 160 : 1503 - 1514
  • [50] Neuroevolution of sequential behavior in multi-goal navigation task
    Muratov, Sergey
    Lakhman, Konstantin
    Burtsev, Mikhail
    ALIFE 2014: THE FOURTEENTH INTERNATIONAL CONFERENCE ON THE SYNTHESIS AND SIMULATION OF LIVING SYSTEMS, 2014, : 772 - 778