The Research And Application On The Behavior Eelection Of Connect6 Based On Intelligent Algorithms

Behavior is the manifestation of life in vitro, and it is the external appearance of life's intelligent. Tyrell said:"Behavior selection is that selecting the most appropriate behavior from a group of potential candidates."Therefore, behavior selection is an advanced form of intelligent life. Behavior selection is the core issue of artificial life research. Artificial life is the development of artificial intelligence. Artificial life, as an interdisciplinary subject of information science, life science and system science, doesn't understand life by analyzing and anatomizing, but uses an integrated approach to understand life. It emphasizes on the systematic and holistic. In addition, the process of computer game is essentially a process of highly antagonism and intelligent behavior. Therefore, taking the game robot which based on computer game system as artificial life, and using intelligent methods to research the intelligent systems are feasible and have important research significance.By mimicking the process of human's behavior of the game selection, this article divided the game robot into four parts - "brain", "visual", "memory", "control". The ultimate goal of the research project was constructing a game robot which could play game on physical board with human. The main work of this paper is designing the "brains". This article addressed the following four aspects:First, a game system has been designed in this article. It includes the expression of the board and pieces in the computer move generation, search technology and evaluation functions.Second, for the problem that it's difficult to judge and count the chess pieces in the Connect6 game system, the thinking of"path"has been proposed and standardized. It makes the preparatory work before evaluating the situation greatly simplified.Third, for the problem that the static evaluation function relying on human's chess knowledge and it is not accurate enough, based on previous studies, the application of genetic algorithms has been improved. It overcomes the original problem of premature convergence of genetic algorithm, and it increases the diversity of population. The experimental results show that the method is effective.Fourth, for the problem that the static evaluation function relying on human's chess knowledge and it is not accurate enough, a new adaptive Particle Swarm Optimization algorithm which dynamically changes inertia weight based on the current rate of cluster focus distance was proposed to improve the PSO algorithm and optimize parameters. Experiments show that the algorithm is faster than the genetic algorithm and the result is better than genetic algorithms. It is an effective way to solve the optimization of the parameters of evaluation function.