Heterogeneous Multi-Agent Cooperation Strategy In RoboCup Rescue

RoboCup Rescue Simulation System(RCRSS) is a kind of Multi-Agent system(MAS) based on a real-time, unknown and dynamic environment, which provides extensive technical standards for the development of artificial intelligence, MAS and corresponding fields. Based on RCRSS, the study mainly focuses on research of heterogeneous Multi-Agent cooperation. Homogeneous agent cooperation has made a certain development in the platform, but the cooperation of three kinds of heterogeneous Multi-Agent is not mature. For improving the whole rescue efficiency, main study and contribution are as follows:Firstly, the study proposed an improved blockade clearing method based on vector weighted sum for police force. At first, K-means clustering algorithm is applied to solve map partition problem. And then target selection is key point, this paper adopt target selection method based on priority to set target priority for blockade. Finally, the key point is an improved blockade clearing method based on vector weighted sum. Based on vector weighted sum for node model midline, police force determines the final guideline by the path and clears roadblocks along this guideline. Thus clearing efficiency can get improved.Secondly, the center agent, as the upper decision maker and the caller of rescue action, should be used to fully improve rescue level. The point of center agent dispatch is task allocation with features of unknown Multi-Task, working efficiency uncertainty and dynamic real-time. The paper takes ambulance center agent as instance, adopting combination of center agent dispatch and independent decision, namely the combination of centralized control and distributed control to implement the rescue work.Finally, on the basis of improved clearing method and center agent dispatch, the paper proposed heterogeneous Multi-Agent cooperation based on information fusion. Information fusion technology divides all the agents into cooperation agent and fusion agent. Fusion agent can fully perceive fusion simulation environment information according to HMM technology, and pass the information to cooperation agent to make it more effectively cooperate with other agents. By the algorithm, the agent cooperation team can accomplish many tasks smoothly and adapt various complex environments to improve the whole rescue efficiency.