| The Robot World Cup Soccer Games, which aims at fostering the research on Artificial Intelligence and Robotic, has become one of the hotspots in Intelligent Robot domain, Multi-Agent Systems, etc. This thesis focuses on the research of decision and cooperation in Multi-Agent System in a real-time,communication-limited and adversarial environment of RoboCup simulation soccer. Effectivness of the research work was verified through simulation experiment and analyze. MAS-Soccer is used as experiment tool. Detailed research work include: (1) Proposed a communication model based on information predict and body language.Since the communication environment of RoboCup is unreliable and low band width, a new communication model which uses to enhance the Robustness of Role Exchange Strategy was proposed. In this model, Agent can tracks teammate's condition through the information forecast, transmission specific information with the aid of body language. (2) Gave a group utility-based role exchange strategy (UBRE). Agent can be regarded as one kind of limited resources in RoboCup game. In this part of work, multi-agent cooperation model has been established based on limited resources optimization use, gives a role exchange algorithm which could maximized group utility. (3) Experiment and analyze. In MAS-Soccer test platform, Agent structure and expression form of team strategy had made the partial adjustments to adapt the request of experiment. In this foundation, the new communication model and the role exchange algorithm was realized in the platform, and verified the validity of new method through experiment and analyze. The result of experiment had shown communication model could effectively enhance the Agent role exchange success ratio; By using the UBRE strategy, Agent has been able to control the stamina effectively in the game, and achieves the goal of the optimized use limited Agent resources. Finally, summarized this article research work and has presented the mentality and the method which further studies. |
