| With the rapid development of computer science,artificial intelligence technology has become the focus of scholars’ research.As an important platform to test the level of development of artificial intelligence,computer game has also been developed by leaps and bounds.At present,many perfect information game problem has been solved,and not perfect information game is also to solve the perfect information game problem in the process of slowly get a certain amount of development space.Unlike the perfect information game problem,imperfect information game problems usually hide some of the key game information.Because of these hidden information,the method for the perfect information game problem can’t be used to solve the problem of imperfect information game.To solve the problem of imperfect information game,we have to deal with the problem of randomness,opponent modeling,risk management and information unreliability caused by imperfect information game.However,the problem of imperfect information game is closer to the game problem encountered in real society than the perfect information game problem.To solve the problem of imperfect information game can provide a decision support system for the game problem in real society.It is more practical to solve the problem of perfect information game than to solve the perfect information game.In imperfect information game,the imperfect information determines the imperfect information game problem formed by the scale of the game tree is huge.This leads to the general game tree search algorithm is not suitable for the application of imperfect information game tree search.Monte Carlo tree search method(MCTS: Monte-Carlo Tree Search)provides a way to solve the problem of large-scale game tree search.Based on the problem of Texas poker game,this paper uses the Monte Carlo game tree search method to study the opponent modeling and risk management problem in the imperfect information game problem,and puts forward the method of calculating the risk loss based on the opponent’s modeling.Based on this improved approach,a strategy based on risk loss is calculated for the Texas Poker game problem.In order to avoid the single mode strategy is used by opponents players,an adaptive strategy selection method based on voting mechanism is put forward,which makes use of the excellent algorithm strategy of Texas Hold’em game research.The method provides a reliable and stable solution for imperfect information Game strategy selection problem.In order to test the performance of the game strategy,this paper designs and develops a visualized Texas poker game platform,and designs the performance of the experimental test strategy on the game platform.Experiments show that the strategy selection method based on the random voting mechanism in multi-strategy mode can effectively combat all kinds of players,and provide a reliable and stable strategy choice for the players. |
