Innovative Application of Computer Game Algorithms of Surakarta Based on Reinforcement Learning

The traditional methods to solve the computer games problem are to use various search algorithms on the game search tree and to combine the situation evaluation to generate the corresponding methods. The paper discusses the principle of MCTS based reinforcement learning. Based on this principle, BP algorithm is combined, it is designed a self-learning game playing program of Surakarta. The experiment shows that the methods can improve the performance of the agent, avoiding design static evaluating function by manual.