Multi-agent Reinforcement Learning: an Approach Based on AgentsCooperation for a Common Goal

摘要：

This paper is devoted to the problem of reinforcementlearning in multi-agent systems. Multi-agent systems forma particular type of distributed artificial intelligencesystem. This paper presents an approach based onagents cooperation for a common goal. By using otheragents experiences and knowledge, an agent may learnfaster, make fewer mistakes, and create some rules forunseen situations. But the information communion amongagents is deficient and limited. In this paper, we assumethat every agent can only observe its neighbors currentpositions and can see whether or not they reach the goalafter the actions have been taken. Experimental resultsshow the effectiveness of the proposed approach.

作者: Guo-quan Wang Hai-bin Yu

作者单位: Shenyang Institute of Automation Chinese Academy of Sciences 114 Nanta Street,Shenyang,P.R.China 110016

会议类型: 国际会议

会议名称: The 8th International Conference on Computer Supported Cooperative Work in Design(第八届计算机支持的协同工作设计国际会议)(CSCWD2004)

会议地点: 厦门

会议语种:英文

页码: 336-339

在线出版日期: 2004-05-26（万方平台首次上网日期，不代表论文的发表时间）