摘要
多人不完备信息博弈是一类存在私有信息而出现信息不完备、不对称的多人博弈.以扑克牌游戏这类典型的多人不完备信息博弈为原型提出一般动态博弈模型GDGM.在该模型框架下,提出一种基于MU算法的多人不完备信息博弈算法MMU,并将MMU算法分别与经典博弈算法Paranoid和MCTS结合,消除该算法对经验值的依赖.最后实验从胜率和得分两个角度对算法进行评价.结果表明,结合了经典博弈算法Paranoid和MCTS算法的PN-MMU和MT-MMU算法可有效处理以扑克牌游戏为代表的多人不完备信息博弈问题,并且与PN-MMU相比,MT-MMU具有更好的博弈能力.
Due to the private information in multi-player imperfect information game,the information of each game player is incomplete and asymmetric.General dynamic game model（GDGM） is proposed based on poker game,which is typical multi-player imperfect information game.Under the frame of GDGM,maxn-Monte Carlo sampling-UCT（MMU） algorithm for multi-player imperfect information game is presented based on Monte Carlo sampling-UCT（MU） algorithm,and further MMU is combined with Paranoid and Monte Carlo tree search（MCTS） respectively to eliminate its dependence on experience value.Finally,both algorithms are evaluated from the perspectives of winning rate and score by experiments.The experimental results show that the Paranoild algorithm MMU（PN-MMU） and Monte Carlo three search MMU（MT-MMU） algorithm combined with Paranoid and MCTS respectively can effectively deal with the problems of poker games.Compared with PN-MMU,MT-MMU has better performance of game.
出处
《武汉大学学报（工学版）》
CAS
CSCD
北大核心
2011年第6期792-796,805,共6页
Engineering Journal of Wuhan University
基金
国家863计划课题(编号:2006AA12A106)
民航软科学研究项目(编号:MHRD201007)
