Study On Game Behavior Selection Of Artificial Life Based On Intention

Artificial life is the development of artificial intelligence. Behavior selection is the core issue of artificial life. Game to artificial life is as the same as"fly"to genetics. Chinese Chess Game is very popular, because it is moderate complexity and it is easy to judge victory or defeat. Intention is an important factor in intelligence, which study behavior trend of two sides in against environment. So intention is very important to decision-making and behavior selection.Every chess player has its own style and rule. Engine program also has a regular tendency when it chooses next jugada according to evaluation function and search algorithm. In this paper, Chinese Chess is looked as an artificial life agent. The agent can find the tendency of opponent's by several times game. So, it can judge opponent's next behavior, and take measures in advance. The competitive level of agent is improves to some extent.The HMM is widely used to study the law of development of things. The paper established an intention model based on HMM, which is formalized into Chinese Chess Agent. The agent search next behavior according to its intention, so it can reduce search space and improve search efficiency and speed. The paper has mainly studied five questions:First, the concepts of game and the basic idea of game tree were analyzed. The commonly algorithm of game search and the key technologies were also analyzed.Second, the important effect of intention to intelligence and the significance of introducing intention to Chinese Chess were analyzed.Third, the HMM, one of intention recognized methods, was introduced, including HMM's theory and application. An intention model based on HMM was established, and formalized into Chinese Chess Agent.Fourth, the method of machine-machine game was adopted. The agent studied opponent's tendency in the process of game, and established intention model for opponent. Then, gaming with opponent who was had been trained, the agent can identify opponent, and use the intention model of opponent's forecast opponent's behavior tendency. So, agent chooses game behavior according to intention.Fifth, through a lot of training, agent optimize the model parameters for certain opponents, and constantly learning. With a lot of training, agent can get better and better when game with the same opponent.