On the Robustness of Cooperative Multi-Agent Reinforcement Learning

被引:32
|
作者
Lin, Jieyu [1 ]
Dzeparoska, Kristina [1 ]
Zhang, Sai Qian [2 ]
Leon-Garcia, Alberto [1 ]
Papernot, Nicolas [1 ,3 ]
机构
[1] Univ Toronto, Toronto, ON, Canada
[2] Harvard Univ, Cambridge, MA 02138 USA
[3] Vector Inst, Toronto, ON, Canada
关键词
D O I
10.1109/SPW50608.2020.00027
中图分类号
TP [自动化技术、计算机技术];
学科分类号
0812 ;
摘要
In cooperative multi-agent reinforcement learning (c-MARL), agents learn to cooperatively take actions as a team to maximize a total team reward. We analyze the robustness of c-MARL to adversaries capable of attacking one of the agents on a team. Through the ability to manipulate this agent's observations, the adversary seeks to decrease the total team reward. Attacking c-MARL is challenging for three reasons: first, it is difficult to estimate team rewards or how they are impacted by an agent mispredicting; second, models are non-differentiable; and third, the feature space is low-dimensional. Thus, we introduce a novel attack. The attacker first trains a policy network with reinforcement learning to find a wrong action it should encourage the victim agent to take. Then, the adversary uses targeted adversarial examples to force the victim to take this action. Our results on the StartCraft II multi-agent benchmark demonstrate that c-MARL teams are highly vulnerable to perturbations applied to one of their agent's observations. By attacking a single agent, our attack method has highly negative impact on the overall team reward, reducing it from 20 to 9.4. This results in the team's winning rate to go down from 98.9% to 0%.
引用
收藏
页码:62 / 68
页数:7
相关论文
共 50 条
  • [21] Reinforcement learning of coordination in cooperative multi-agent systems
    Kapetanakis, S
    Kudenko, D
    [J]. EIGHTEENTH NATIONAL CONFERENCE ON ARTIFICIAL INTELLIGENCE (AAAI-02)/FOURTEENTH INNOVATIVE APPLICATIONS OF ARTIFICIAL INTELLIGENCE CONFERENCE (IAAI-02), PROCEEDINGS, 2002, : 326 - 331
  • [22] Learning Distinct Strategies for Heterogeneous Cooperative Multi-agent Reinforcement Learning
    Wan, Kejia
    Xu, Xinhai
    Li, Yuan
    [J]. ARTIFICIAL NEURAL NETWORKS AND MACHINE LEARNING - ICANN 2021, PT IV, 2021, 12894 : 544 - 555
  • [23] Pacesetter Learning for Large Scale Cooperative Multi-Agent Reinforcement Learning
    Zhou, Pingqi
    Li, Chao
    Qiu, Mengwei
    Liu, Jun
    Ma, Chennan
    Yan, Ming
    [J]. ARTIFICIAL NEURAL NETWORKS AND MACHINE LEARNING, ICANN 2023, PT VI, 2023, 14259 : 115 - 126
  • [24] Learning Fair Policies in Decentralized Cooperative Multi-Agent Reinforcement Learning
    Zimmer, Matthieu
    Glanois, Claire
    Siddique, Umer
    Weng, Paul
    [J]. INTERNATIONAL CONFERENCE ON MACHINE LEARNING, VOL 139, 2021, 139
  • [25] QTRAN: Learning to Factorize with Transformation for Cooperative Multi-Agent Reinforcement learning
    Son, Kyunghwan
    Kim, Daewoo
    Kang, Wan Ju
    Hostallero, David
    Yi, Yung
    [J]. INTERNATIONAL CONFERENCE ON MACHINE LEARNING, VOL 97, 2019, 97
  • [26] Learning Implicit Credit Assignment for Cooperative Multi-Agent Reinforcement Learning
    Zhou, Meng
    Liu, Ziyu
    Sui, Pengwei
    Li, Yixuan
    Chung, Yuk Ying
    [J]. ADVANCES IN NEURAL INFORMATION PROCESSING SYSTEMS 33, NEURIPS 2020, 2020, 33
  • [27] Certified Policy Smoothing for Cooperative Multi-Agent Reinforcement Learning
    Mu, Ronghui
    Ruan, Wenjie
    Marcolino, Leandro Soriano
    Jin, Gaojie
    Ni, Qiang
    [J]. THIRTY-SEVENTH AAAI CONFERENCE ON ARTIFICIAL INTELLIGENCE, VOL 37 NO 12, 2023, : 15046 - 15054
  • [28] Optimistic Value Instructors for Cooperative Multi-Agent Reinforcement Learning
    Li, Chao
    Zhang, Yupeng
    Wang, Jianqi
    Hu, Yujing
    Dong, Shaokang
    Li, Wenbin
    Lv, Tangjie
    Fan, Changjie
    Gao, Yang
    [J]. THIRTY-EIGHTH AAAI CONFERENCE ON ARTIFICIAL INTELLIGENCE, VOL 38 NO 16, 2024, : 17453 - 17460
  • [29] Reinforcement Learning Approach for Cooperative Control of Multi-Agent Systems
    Javalera-Rincon, Valeria
    Puig Cayuela, Vicenc
    Morcego Seix, Bernardo
    Orduna-Cabrera, Fernando
    [J]. PROCEEDINGS OF THE 11TH INTERNATIONAL CONFERENCE ON AGENTS AND ARTIFICIAL INTELLIGENCE (ICAART), VOL 2, 2019, : 80 - 91
  • [30] Cooperative targets assignment based on multi-agent reinforcement learning
    Ma, Yue
    Wu, Lin
    Xu, Xiao
    [J]. Xi Tong Gong Cheng Yu Dian Zi Ji Shu/Systems Engineering and Electronics, 2023, 45 (09): : 2793 - 2801