基于值分解的2对2三国杀多智能体强化学习算法OA

There are many problems in the multi-agent reinforcement learning environment,and the card-based multiplayer strategy game is a classic multi-agent system,which is always a problem that needs to be solved by game AI because of its unique characteristics,but there are few existing card game AI research,and most of them come from Dou Di Zhu,Texas Hold'em,Mah-jong,etc.In order to expand the research and application of reinforcement learning in card game AI,this paper proposes a three-kingdom killing multi-agent reinforcement learning method based on value decomposition,and self-builds a battle game scene with the three-kingdom killing game as the background as the environment of multiple agents,models the cooperative multi-ple agents,and solves the problem of instability in the environment of multiple agents.