Formation Strategy Online Learning Research And Application Based On Neural Network

During the RoboCup simulation competition, effective cooperation and perfect teamwork spirit is necessary, such as formation strategy. It is difficult for agents to get more information about teammate cooperation in the limiting communication environment.If agent can get some information from the opponent agents,agents can make the relatively correct decisions. So it is urgent need to research on learning unpredictable opponent information,such as role,heterogeneous player,formation strategy,etc.At present, the research hot focus on how to improve the abilities of robots’learning and multi-agent cooperation, the opponents’modeling capability is attractive as well.With the rapid development of computer technology,distributed artificial intelligence technology and multi-agent systems,the competition level of RoboCup2D has been greatly improved.The dissertation analyzes the development procedure and significance of formation strategy, then describes the system structure and working principles of the RoboCup2D.Further researches about current situation and trendy are doing well based on several world-famous simulator teams.The dissertation analyze the formation tactics by using the Delaunay triangulation method based on the design ideas and analysis of the formation agent2d strategy procedure.Finally, the dissertation research the workflow of the online coach and online opponent modeling. What’s more, coupled with research on neural networks, the dissertation presents a method that establishes net model for point of the opponent formation by online coach. And it introduced the idea in detail, analyzed the predictive accuracy and the time complexity about this method. The games show that the method has higher prediction accuracy.