摘要
麻将博弈游戏作为非完美信息博弈中的典型项目,与德州扑克、桥牌等其他非完美信息博弈游戏相比,最大的不同是玩家游戏顺序时常被打乱而产生较大不确定性,导致博弈策略设计及局面评估困难。针对大众麻将游戏规则,构造了一种快速胡牌方法。为此,首先分析麻将的大众规则;其次,基于各种牌型和持有牌点数,提出一种胡牌距离概念,以最短胡牌距离为标准,作为玩家的决策依据;最后,依据已知公开的出牌和玩家手上牌张信息,融合先验知识后,进一步计算每种出牌行为的期望胜率,从而获得出牌行为评估值,作为玩家的决策依据。实验结果表明:利用维持最短胡牌距离为核心的快速胡牌方法,在有限牌对局数下,比如在中国计算机博弈锦标赛中此局数设定为10000局,依据设计的智能体,能取得46.85%以上的牌局胜利,该值远高于仅仅使用纯粹经验方法的数值。
Mahjong game is a typical item in an imperfect information game.Compared with other imperfect information games such as Texas Hold’em and bridge,the biggest difference is that the player’s game sequence is often disrupted,resulting in greater uncertainty,leading to difficulties in game strategy design and situation evaluation.The quantitative thinking of the benefits of Hu Pai in the paper constructs a quick Hu Pai method in accordance with the rules of the popular mahjong game.To this end,the paper first analyzes the popular rules of mahjong;secondly,based on various card types and the number of cards held,a concept of Hu card distance is proposed,and the shortest Hu card distance is used as the basis for players’decision-making.Finally,based on Knowing the published cards and the card information in the player’s hand,and fusing the prior knowledge,the expected winning rate of each card behavior is further calculated,so as to obtain the evaluation value of the card behavior as the player’s decision-making basis.Experiments show that using the fast Hu card method that maintains the shortest Hu card distance as the core,in a limited number of games,such as the China Computer Game Championship,the number of games setting to 10000,the intelligent body designed according to the paper method can obtain 46.85%or more of hand wins,and this value is much higher than the value of purely empirical methods.
作者
彭丽蓉
赵海璐
甘春晏
刘洁
陈俊宇
PENG Lirong;ZHAO Hailu;GAN Chunyan;LIU Jie;CHEN Junyu(School of Artificial Intelligence,Chongqing University of Technology,Chongqing 401135,China;School of Artificial Intelligence and Big Data,Chongqing Institute of Industry and Technology,Chongqing 401120,China)
出处
《重庆理工大学学报（自然科学）》
CAS
北大核心
2021年第12期127-133,共7页
Journal of Chongqing University of Technology：Natural Science
基金
国家自然科学基金项目(60443004)。
