This paper concentrates on the research of cooperation strategy in dynamic games of multi-robot systems based on MAS.Firstly,under the study of multi-agent system theory,a multi-robot architecture based on MAS is discussed.Secondly,based on the research of the multi-agent collaborative strategies and multi-robot systems' collaborative strategies,a common multi-agent collaboration model is presented.Again, through the Combination of human social organization and division of labor,the multi-robot cooperation strategy is implemented by dynamically assigning the role and reliability of the robots.Finally,after the study of multi-agent learning problems in dynamic game environment,a multi-robot cooperation strategy based on multi-agent reinforcement learning is presented.All cooperation methods are used in robot soccer SimuroSot 5vs5 platform.By actual experiment competitions,the results verify the proposed collaboration strategies are effective and feasible. |