Quantifying the effects of environment and population diversity in multi-agent reinforcement learning

被引:0
|
作者
Kevin R. McKee
Joel Z. Leibo
Charlie Beattie
Richard Everett
机构
[1] DeepMind Technologies Ltd,
关键词
Machine learning; Deep reinforcement learning; Multi-agent; Diversity;
D O I
暂无
中图分类号
学科分类号
摘要
Generalization is a major challenge for multi-agent reinforcement learning. How well does an agent perform when placed in novel environments and in interactions with new co-players? In this paper, we investigate and quantify the relationship between generalization and diversity in the multi-agent domain. Across the range of multi-agent environments considered here, procedurally generating training levels significantly improves agent performance on held-out levels. However, agent performance on the specific levels used in training sometimes declines as a result. To better understand the effects of co-player variation, our experiments introduce a new environment-agnostic measure of behavioral diversity. Results demonstrate that population size and intrinsic motivation are both effective methods of generating greater population diversity. In turn, training with a diverse set of co-players strengthens agent performance in some (but not all) cases.
引用
收藏
相关论文
共 50 条
  • [1] Quantifying the effects of environment and population diversity in multi-agent reinforcement learning
    McKee, Kevin R.
    Leibo, Joel Z.
    Beattie, Charlie
    Everett, Richard
    [J]. AUTONOMOUS AGENTS AND MULTI-AGENT SYSTEMS, 2022, 36 (01)
  • [2] Celebrating Diversity in Shared Multi-Agent Reinforcement Learning
    Li, Chenghao
    Wang, Tonghan
    Wu, Chengjie
    Zhao, Qianchuan
    Yang, Jun
    Zhang, Chongjie
    [J]. ADVANCES IN NEURAL INFORMATION PROCESSING SYSTEMS 34 (NEURIPS 2021), 2021, 34
  • [3] Battlefield Environment Design for Multi-agent Reinforcement Learning
    Do, Seungwon
    Baek, Jaeuk
    Jun, Sungwoo
    Lee, Changeun
    [J]. 2022 IEEE INTERNATIONAL CONFERENCE ON BIG DATA AND SMART COMPUTING (IEEE BIGCOMP 2022), 2022, : 318 - 319
  • [4] Studies on hierarchical reinforcement learning in multi-agent environment
    Yu Lasheng
    Marin, Alonso
    Hong Fei
    Lin Jian
    [J]. PROCEEDINGS OF 2008 IEEE INTERNATIONAL CONFERENCE ON NETWORKING, SENSING AND CONTROL, VOLS 1 AND 2, 2008, : 1714 - 1720
  • [5] Multi-Agent Reinforcement Learning
    Stankovic, Milos
    [J]. 2016 13TH SYMPOSIUM ON NEURAL NETWORKS AND APPLICATIONS (NEUREL), 2016, : 43 - 43
  • [6] OFCOURSE: A Multi-Agent Reinforcement Learning Environment for Order Fulfillment
    Zhu, Yiheng
    Zhan, Yang
    Huang, Xuankun
    Chen, Yuwei
    Chen, Yujie
    Wei, Jiangwen
    Feng, Wei
    Zhou, Yinzhi
    Hu, Haoyuan
    Ye, Jieping
    [J]. ADVANCES IN NEURAL INFORMATION PROCESSING SYSTEMS 36 (NEURIPS 2023), 2023,
  • [7] Cooperative Multi-Agent Reinforcement Learning in a Large Stationary Environment
    Zemzem, Wiem
    Tagina, Moncef
    [J]. 2017 16TH IEEE/ACIS INTERNATIONAL CONFERENCE ON COMPUTER AND INFORMATION SCIENCE (ICIS 2017), 2017, : 365 - 371
  • [8] Coordinated Reinforcement Learning Agents in a Multi-Agent Virtual Environment
    Sause, William
    [J]. 2013 12TH INTERNATIONAL CONFERENCE ON MACHINE LEARNING AND APPLICATIONS (ICMLA 2013), VOL 1, 2013, : 227 - 230
  • [9] Effects of Spectral Normalization in Multi-Agent Reinforcement Learning
    Mehta, Kinal
    Mahajan, Anuj
    Kumar, Pawan
    [J]. 2023 INTERNATIONAL JOINT CONFERENCE ON NEURAL NETWORKS, IJCNN, 2023,
  • [10] Deep Multi-agent Reinforcement Learning in a Homogeneous Open Population
    Radulescu, Roxana
    Legrand, Manon
    Efthymiadis, Kyriakos
    Roijers, Diederik M.
    Nowe, Ann
    [J]. ARTIFICIAL INTELLIGENCE, BNAIC 2018, 2019, 1021 : 90 - 105