A collaboration method of MAS based on information fusion and its application in RoboCupRescue simulation system

Collaboration is one of the key issues in multi-agent system. This paper presents a multi-agent collaboration method based on the concept of information fusion in which the tasks can be divided quickly without task overlapping and resource conflicting. In this model we classify the agents into two categories: sensor agents and decision agents. Sensor agents with the abilities of sensing and action convert the raw sensor information into a definite protocol and transmit them to decision agents. Decision agents adopt a method based on dynamic Bayesian networks to infer the situation assessment which is reliable for decision making from large quantities of sensor information, and then retrieve a best action set which will be sent to sensor agents to execute. This model has applied effectively in RoboCupRescue simulation competition.