基于博弈学习多Agent交互研究  

Research on interaction of multi-Agents based on game theory learning

作　　者：陶雪丽[1] 郑延斌[1] 王宁[1] 

机构地区：[1]河南师范大学计算机与信息工程学院,河南新乡453007

出　　处：《计算机工程与设计》2014年第12期4320-4323,4328,共5页Computer Engineering and Design

基　　金：河南省科技攻关基金项目(122102210086;132102210537;132102210538);河南省教育厅自然科学技术研究重点基金项目(13A520530)

摘　　要：由于人们之间的博弈行为受多种因素的制约和影响,而传统的博弈方法很难处理这种影响因素多变、交互关系复杂的博弈问题,给出一个基于博弈学习的多智能体(multi-Agent)交互模型,并以此为基础构建多Agent交互的博弈学习方法。对合作小组中成员的行为进行修正,通过博弈学习中学习因子的更新得到局部均衡,达到全局利益优化。实例仿真验证了该方法的可行性。Taking the increasingly complex sociability into consideration,people’s behavior is restricted and influenced by many factors.However,it is very difficult for the traditional game method to deal with the game problems with polytrophic influence factors and complex interactive relations.A multi-Agent interaction model based on game theory learning was given to solve this problem.On this basis of the model,a multi-Agent’s interactive game learning methods was constructed,which was used to correct the behaviors of team members,and partial equilibrium was achieved by updating the learning factors of game learning,so that the global optimization was ultimately achieved.The simulation results show that the method is feasible.

关 键 词：交互 博弈学习 多智能体 均衡 学习因子 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]