【作者】 刘杰； 王月海； 王鸿雁；

【Author】 Liu Jie~1 Wang Yuehai~1 Wang Hongyan~2 1.College of Information Engineering,North China University of Technology,Beijing,100041 2.College of Economics,Beijing Technology and Business University,Beijing,100037

【机构】 北方工业大学信息工程学院； 北京工商大学经济学院；

【摘要】 在多智能体领域,多智能体如何自适应地协调行为一直是热点问题。本文在基于演化博弈的基础上, 利用模仿复制选择动态模型使多智能体能够自适应地调整策略,提出了建立多智能体整体状态的统计模型, 来近似估计当前智能体所采用策略的适应性,并通过反思一复制选择算法,来选择适应性更高的策略。通过对“觅食”任务的仿真实验表明,智能体通过一段时间的学习,整体性能达到较优。系统具有很好的灵活性和鲁棒性。更多还原

【Abstract】 The central issue in the field of multi-agent systems is the coordination of the agents’ behaviors.This paper proposes a new approach for learning multi-robot coordination strategies based on Evolutionary Game Theory. We then design a Statistical Model to estimate the overall state of multi-agent system and use the algorithm of Ponder—Imitate to choice the better strategy.Our predictions are evaluated in the simulation mission of“Forage”. The simulation results show that the multiple agents adaptively learn the coordination,and the system is flexible and robust.更多还原