Multi-Agent Game Abstraction via Graph Attention Neural Network

被引:0
|
作者
Liu, Yong [1 ]
Wang, Weixun [2 ]
Hu, Yujing [3 ]
Hao, Jianye [2 ,4 ]
Chen, Xingguo [5 ]
Gao, Yang [1 ]
机构
[1] Nanjing Univ, Natl Key Lab Novel Software Technol, Nanjing, Peoples R China
[2] Tianjin Univ, Tianjin, Peoples R China
[3] NetEase Fuxi AI Lab, Beijing, Peoples R China
[4] Huawei, Noahs Ark Lab, Shenzhen, Peoples R China
[5] Nanjing Univ Posts & Telecommun, Jiangsu Key Lab Big Data Secur & Intelligent Proc, Nanjing, Peoples R China
基金
中国国家自然科学基金;
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 large-scale multi-agent systems, the large number of agents and complex game relationship cause great difficulty for policy learning. Therefore, simplifying the learning process is an important research issue. In many multi-agent systems, the interactions between agents often happen locally, which means that agents neither need to coordinate with all other agents nor need to coordinate with others all the time. Traditional methods attempt to use pre-defined rules to capture the interaction relationship between agents. However, the methods cannot be directly used in a large-scale environment due to the difficulty of transforming the complex interactions between agents into rules. In this paper, we model the relationship between agents by a complete graph and propose a novel game abstraction mechanism based on two-stage attention network (G2ANet), which can indicate whether there is an interaction between two agents and the importance of the interaction. We integrate this detection mechanism into graph neural network-based multi-agent reinforcement learning for conducting game abstraction and propose two novel learning algorithms GA-Comm and GA-AC. We conduct experiments in Traffic Junction and Predator-Prey. The results indicate that the proposed methods can simplify the learning process and meanwhile get better asymptotic performance compared with state-of-the-art algorithms.
引用
收藏
页码:7211 / 7218
页数:8
相关论文
共 50 条
  • [41] A multi-agent approach to process the distributed computing of neural network
    Shi, WR
    Zhang, L
    Qin, LX
    [J]. 7TH WORLD MULTICONFERENCE ON SYSTEMICS, CYBERNETICS AND INFORMATICS, VOL II, PROCEEDINGS: COMPUTER SCIENCE AND ENGINEERING, 2003, : 300 - 303
  • [42] Multi-Agent Stock Trading Algorithm - A Neural Network Approach
    Tirea, Monica
    Tandau, Ioan
    Negru, Viorel
    [J]. ADVANCES IN KNOWLEDGE-BASED AND INTELLIGENT INFORMATION AND ENGINEERING SYSTEMS, 2012, 243 : 706 - 715
  • [43] Routing strategy based on multi-agent systems and neural network
    Dong, Jun
    Pan, Yun-He
    [J]. Zidonghua Xuebao/Acta Automatica Sinica, 2002, 28 (04): : 505 - 512
  • [44] Coalition Formation for Multi-agent Pursuit Based on Neural Network
    Pei, Zhaoyi
    Piao, Songhao
    Souidi, Mohammed El Habib
    Qadir, Muhammad Zuhair
    Li, Guo
    [J]. JOURNAL OF INTELLIGENT & ROBOTIC SYSTEMS, 2019, 95 (3-4) : 887 - 899
  • [45] Coalition Formation for Multi-agent Pursuit Based on Neural Network
    Zhaoyi Pei
    Songhao Piao
    Mohammed El Habib Souidi
    Muhammad Zuhair Qadir
    Guo Li
    [J]. Journal of Intelligent & Robotic Systems, 2019, 95 : 887 - 899
  • [46] A multi-agent architecture for game playing
    Kobti, Ziad
    Sharma, Shiven
    [J]. 2007 IEEE SYMPOSIUM ON COMPUTATIONAL INTELLIGENCE AND GAMES, 2007, : 276 - 281
  • [47] Multi-Agent Word Guessing Game
    Luis, Gabino
    de la Pena, Jaime
    Suarez, David
    Mateos, Alfonso J.
    [J]. ADCAIJ-ADVANCES IN DISTRIBUTED COMPUTING AND ARTIFICIAL INTELLIGENCE JOURNAL, 2018, 7 (04): : 17 - 26
  • [48] Multi-agent Deployment on a Ring Graph
    Elor, Yotam
    Bruckstein, Alfred M.
    [J]. SWARM INTELLIGENCE, 2010, 6234 : 215 - 226
  • [49] A Multi-agent Approach for Graph Classification
    Baldini, Luca
    Rizzi, Antonello
    [J]. PROCEEDINGS OF THE 13TH INTERNATIONAL JOINT CONFERENCE ON COMPUTATIONAL INTELLIGENCE (IJCCI), 2021, : 334 - 343
  • [50] A graph neural network based deep reinforcement learning algorithm for multi-agent leader-follower flocking
    Xiao, Jian
    Wang, Zhuoran
    He, Jinhui
    Yuan, Guohui
    [J]. INFORMATION SCIENCES, 2023, 641