基于强化学习的多Agent协作研究

强化学习为多Agent之间的协作提供了鲁棒的学习方法．本文首先介绍了强化学习的原理和组成要素，其次描述了多Agent马尔可夫决策过程MMDP，并给出了Agent强化学习模型．在此基础上，对多Agent协作过程中存在的两种强化学习方式：IL(独立学习)和JAL(联合动作学习)进行了比较．最后分析了在有多个最优策略存在的情况下，协作多Agent系统常用的几种协调机制．

Cooperative Multi-agent Systems Based on Reinforcement Learning

Reinforcement learning can provide a robust and natural means for agents to learn how to coordinate their action choices in fully cooperative multi agent systems (MAS). This paper first introduces the basic principles and components of reinforcement learning, then describes multi agent extension MMDP and presents reinforcement learning model of agents in cooperative MAS. After that we distinguish reinforcement learners that ignore the presence of other agents from those that explicitly attempt to learn the value of joint actions and strategies of their counterparts. In the last, some simple and commonly used coordination mechanisms are examined.