Multi-agent Cooperation In RoboCup Rescue Simulation

Multi agent cooperation is always a key research domain in the artificial intelligence. In thefield of urban search and rescue (USAR), rescue robot is facing many challenges，such as theunknown and dynamic environment, limited communication and resource. In this paper, relevanttheories of multi-agent system applied for RoboCup Rescue Simulation System (RCRSS).Construction of heterogeneous rescue robot cooperation model is constructed including the worldmodel, task allocation mechanism based on auction algorithm and dynamic path planning method.The main work and innovations are as follows:First, the world model is a basis of collaborative rescue robot. But robot perception informationis local, relative and inaccurate. This paper begins with the construction of world model for ourteam Apollo-Rescue. The map information is preprocessed before action. Rescue agents canmaintain a global and accurate world model by the p^icle filter algorithm which is used to processperceptual noise information of dynamic objects in rescue scenario.Secondly，in order to solve the cooperation problem in robot rescue simulation system, a taskallocation mechanism based on auction is proposed. Using the mechanism, the agents can enter andleave the auction market flexibly and switch role dynmiically. Meanwhile the anxiety degree isintroduced to adjust the allocation scheme after the basic task allocation. As a result, rescue agentscan be used to dynmiic disaster environments. Experimental results show that the rescue team usingthe proposed method is more suitable for different complex disaster environments than the rescueteam using the genetic algorithm in terms of fulfilling the rescue tasks, reducing the task responsedelay and improving the efficiency of rescue system in a limit time.Finally, considering the unknown and complex environment, an improved ant colony algorithmis introduced. A new heuristic function is defined in this paper. According to the behavior, rescueagent can change the selection probability of the next node. Through the local and globalpheromone update strategy’ sharing pheromone information mnong agents by communication toaccelerate the rate of convergence and keep the diversity of solutions. So that rescue agents canobtain the optimal path in disaster city and fulfill coordination tasks.