摘要
德州扑克是一种计算机博弈牌类项目,作为不完美信息博弈,其牌力评估较为困难。为了提高针对不同对手的评估效益,提出一种基于对手模型的牌力评估方法。首先用树存储对手的动作频率信息,然后根据树构建一个与对手行动类似的虚拟对手智能体,最后在对局中计算虚拟对手的策略,更新对手的手牌权重,缩小对手可能手牌范围,完成牌力评估。实验结果显示:使用基于对手模型牌力评估方法的智能体击败了不同风格的对手,且总体赢得的筹码比使用静态评估方法的智能体高。与传统方法相比能有效针对不同的对手类型,提高评估效益。
Texas Hold’em poker is a kind of card game in computer games.As an imperfect information game,it is difficult to evaluate the card strength.In order to improve the evaluation efficiency when versus different opponents,a new method based on opponent model is proposed.Firstly,the tree is used to store the information of the opponent’s action frequency,and then construct a virtual agent whose action is similar to the opponent’s.Finally,after calculating the virtual opponent’s strategy in the game and updating the weight of opponent’s hand cards,the opponent’s possible hand range is reduced so that we can complete the card strength evaluation.The experiments show that agent using the evaluation method based opponent model beat opponents with different styles,and the overall winning chips are higher than the agent using static evaluation method.Compared with traditional methods,it can effectively target different opponents and improve the evaluation efficiency.
作者
张小川
杜松
赵海璐
刘贺
伍帆
ZHANG Xiaochuan;DU Song;ZHAO Hailu;LIU He;WU Fan(School of Artificial Intelligence, Chongqing University of Technology, Chongqing 401135, China;School of Computer Science and Engineering, Chongqing University of Technology, Chongqing 400054, China)
出处
《重庆理工大学学报（自然科学）》
北大核心
2021年第9期130-135,共6页
Journal of Chongqing University of Technology：Natural Science
基金
NSFC青基项目(61702063)
重庆市自然科学基金项目(cstc2019jcyj-msxmX0544)。
关键词
牌力评估
德州扑克
对手模型
计算机博弈
hand evaluation
Texas Hold’em poker
opponent model
computer game
