有限理性模型在游戏智能决策中的研究与应用

Research and application of bounded rationality model used in game intelligent decision-making

为了让游戏NPC能够学习和模拟玩家在游戏中的策略和行为方式,在基于模型的智能决策方法基础上,结合行为决策理论中的有限理性模型提出了一种新的游戏智能方法。该方法分别从有限理性模型的两个核心原则———有限理性和满意准则来改进过去的方法在感知和决策过程中所面对的问题,从而使得游戏NPC行为决策方式更加人性化。最后,通过在Starcraft平台上与其他方法的对抗性实验来进一步验证该方法的优势。

This article presented a novel game intelligent decision-making method to make game NPC（ non-player character） learn and simulate strategies and behavioral pattern which the players used in the video game. This method improved the model-based method, and introduced the bounded rationality model of the behavioural decision theory as the basic decision-making model. So that it would drawback the problems confronted with the past methods in the perception and decision-making process from two aspects-bounded rationality and satisfactory criterion, which was the core of the bounded rationality model. Finally, the advantages of this method were verified by the experiments on the platform of Starcraft.