Multi-Agent Group Interaction Research Based On Game Learning

Group interaction behavior is human behavior which happens all the time in one’s life, and it is the most common activity in human society. As an important part of MAS (multi-agent systems) research, group interaction behavior plays an important role in everyday life, commerce and trade, social relations, and teaching research as well as political activity. With the development of artificial intelligence, computer graphics, and coordinating science, the study on behavior of group interaction which is simulated with the help of virtual reality technology gradually turns to be mature. Research on group interaction behavior has been widely used in many areas of our daily life, such as commerce transactions, political activities, teaching and military simulation. And it has made great contributions to human progress and social development. The Study and simulating on Interactive behavior in daily life can not only better reflect the real group interactions, but also increase people’s awareness of group interaction behavior and help people to deal with complex groups interactions in reality.Based on multi-Agent or game theory, early studies of group interaction behaviors still have some limitations, the role of intelligence some cannot be guaranteed, the others analog simulation of group behavior are untrue. Participants’characteristics and performance in the interactive group interactions are major components of group interaction behavior research. Moreover, they are also key elements to reflect the group interaction behavior.Thus, it has become the focus of most scholars to study the method of group interaction behaviors research. In this paper, according to the game learning thought to study the group interaction behaviors,the author mainly do the following work:(1)In this paper, the author makes an overview about the status of population interactions study, analyzes the advantages and disadvantages of this approach. In addition, the author proposes some improved methods.(2)The author describes game theory, Nash equilibrium, game learning concepts and related theoretical knowledge and ideas. Besides, the author also analyzes the model of the common groups interaction, improves the learning in the game based on the idea, and builds a multi-Agent group interaction model based on game learning. (3)For group participants in games, assuming according to bounded rationality, the author builds an multi-agent game learning coordination algorithm on the basis of games learning thought. The smooth flow of traffic in the city and optimization of the global transportation is achieved after analyzing and correcting travelers’ coordinated behavior by using learning algorithm. Finally, the author utilizes example simulation to verify its feasibility.(4)By studying the game thinking basing on multi-agent interaction model of the game, the author builds a multi-agent, which is used to amend teamwork members’conduct, achieve partial equilibrium by updating learning factors in game learning methods and ultimately attain the global optimization of benefits. At last, utilizing an example simulation to verify its feasibility.