Multi-agent Collaborative Research In RoboCup2D

Research of multi-Agent collaborative is popular recently. In the RoboCup 2D match, the result of match decides on how to resolve collaboration. According to the fact in RoboCup 2D match, the dissertation firstly analyses background and actuality in multi-Agent collaborative research; meanwhile, the dissertation introduces study status of multi agent system.Secondly, studies teammate model theory, pops up Change-Role model and gives its application in collaborative attack,and forecasts Agent model according to the Agent's environment and gets more accurate collaborative decision. Tradition coding methods only consider decide-Agent's environment.Thirdly, analyzes role, active area and heterogeneous types of opponents, and gives application in collaborative defensive according to Q learning method.Chang-Role idea in collaborative attack application describe as as follows: forecast teammates'model in case of environment of teammates, the model may be pass ball, dribble ball, or shoot ball. Then, the Agent makes collaborative decision according to the model of teammates. Chang-Role idea in collaborative defensive application describe as follows: analyze role, active area and heterogeneous types of opponents, forecast model of opponents, then make collaborative defensive decision.