Computer game playing is an interesting subject in artificial intelligence research. Reinforcement learning is a machine learning method to study Superior strategy through trial-and-error search and delayed reward from Environment-feedback. In this article I research how to design a computer Chinese chess named RL-CHESS which has self-learning ability from play itself. The article mainly include the following works:First,I describe the development of computer game and learning algorithm, and discuss their research, lacks and trends. Then we Analyzed the characteristics of the Chinese chess ,discuss the basic characteristic of the learning algorithm, make a decision of the scope of them.Then I expatiated the main frame of the computer game and the principle and method.Then I researched how to build Chinese chess which has self-learning ability from play itself with the neural network combination the reinforcement learning algorithm ,And I researched how to build Chinese chess which has self-learning ability from play itself with the database combination the reinforcement learning algorithm .Finally ,I compared and Analyzed these two methods and compared with the other Chinese chess program which has the ability of learning. Finally, I make a brief summary to the result of experiment and future works.
|