计算机应用 ›› 2012, Vol. 32 ›› Issue (10): 2875-2878.DOI: 10.3724/SP.J.1087.2012.02875

• 人工智能 • 上一篇    下一篇

多智能体系统分散式通信决策研究

郑延斌1,郭凌云2,刘晶晶2   

  1. 1. 河南师范大学 计算机与信息技术学院,河南 新乡453007
    2. 河南师范大学 计算机与信息技术学院,河南 新乡 453007
  • 收稿日期:2012-04-27 修回日期:2012-05-28 发布日期:2012-10-23 出版日期:2012-10-01
  • 通讯作者: 郭凌云
  • 作者简介:郑延斌(1964-),男,河南内乡人,教授,博士,主要研究方向:虚拟现实、多智能体系统、对策论;郭凌云(1987-),女,河南林州人,硕士研究生,主要研究方向:虚拟现实;刘晶晶(1986-),女,河南兰考人,硕士研究生,主要研究方向:虚拟现实。
  • 基金资助:
    河南省重点科技攻关项目(102102210176)

Research on decentralized communication decision in multi-Agent system

ZHENG Yan-bin,GUO Ling-yun,LIU Jing-jing   

  1. College of Computer and Information Technology, Henan Normal University, Xinxiang Henan 453007, China
  • Received:2012-04-27 Revised:2012-05-28 Online:2012-10-23 Published:2012-10-01
  • Contact: GUO Ling-yun

摘要: 通信是多智能体系统(MAS)之间协调与协作的最有效和最直接的方法,然而通信的代价却限制了该方法的使用。为了减少MAS协调过程中的通信量,提出一种启发式算法,使Agent仅选择能够改善团队期望回报的观察信息进行通信。实验结果证明,对通信信息的选择能够高效的利用通信带宽,有助于提高系统的性能。

关键词: 多智能体系统, 协作, 分散式通信, 马尔可夫决策过程, 部分可观察马尔可夫决策过程

Abstract: Communication is the most effective and direct method of coordinating and cooperating among multi-Agents, but the cost of communication restricts the use of this method. In order to reduce traffic subject in the coordination of Multi-Agent System (MAS), this paper put forward a heuristic algorithm, which would make Agents choose the observation that is beneficial to team performance to communicate. The experimental results show that choosing beneficial observation to communicate could ensure the efficiency of limited communication bandwidth and improve system performance.

Key words: Multi-Agent System (MAS), cooperation, decentralized communication, Markov Decision Process (MDP), Partially Observable Markov Decision Process (POMDP)