基于随机博弈的Agent协同强化学习方法  

A Cooperative Reinforcement Learning Method Based on the Stochastic Game in Multi-Agent Systems

作　　者：王长缨[1] 尹晓虎[1] 鲍翊平[1] 姚莉[1] 

机构地区：[1]国防科技大学管理科学与工程系,湖南长沙410073

出　　处：《计算机工程与科学》2006年第2期107-110,共4页Computer Engineering & Science

基　　金：国家自然科学基金资助项目(70371008)

摘　　要：本文针对一类追求系统得益最大化的协作团队的学习问题,基于随机博弈的思想,提出了一种新的多Agent协同强化学习方法。协作团队中的每个Agent通过观察协作相识者的历史行为,依照随机博弈模型预测其行为策略,进而得出最优的联合行为策略。This paper aims at the learning process of a kind of cooperative teams, which pursue the maximum benefit of a whole system. We propose a new cooperative reinforcement learning method based on the stochastic game in multi-agent systems. Each agent of the team decides its behaviors after forecasting the behavior strategy of acquaintances according to the stochastic game structure and their historical behaviors, and then a jointly optimal behavior strategy is obtained.

关 键 词：强化学习 多AGENT系统 随机博弈 协作 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程] TP391[自动化与计算机技术—控制科学与工程]