Research Of Multi-Agent Cooperation In RoboCup Rescue Simulation System

The harm brought about by multiple natural disasters and man-made armed conflict in recent years, seriously threats to the safety of the people and has been aroused widespread concerns. RoboCup rescue simulation system simulates compute a real urban disaster situation by computer. The purpose is to perform the effective relief work in disaster scenarios offered by the simulation system, and to rescue the injured people as soon as possible, to rescue people’s lives and property, reducing disaster losses to a minimum. RoboCup rescue simulation has varieties of possible developments in different ranges. It can provide decision support for the people in the actual rescue operations, thus the study has a great significance.RoboCup rescue simulation system provides six agents:Police Force, Police Office, Ambulance Team, Ambulance Center, Fire Brigade and Fire Station. Each agent has a number of samples,for this reason, RoboCup rescue simulation system exists the mutual cooperation between the isomorphism agents. Because of the cooperation, the RoboCup rescue simulation system also exists mutual cooperation among heterogeneous agents. The research of this paper is based on RoboCup rescue simulation of robotic systems, studying the multi-agent collaboration in the rescue simulation in the robot system. The decision-making process of the agent facing multiple tasks, which are viewed as a multiple target. But agent task selection is evolved to a multi-objective optimization problem. The main content is as follows:(1)Multi-agent collaborative research based on binary particle swarm optimization Rescue evaluation of effects of a variety of tasks faced in the decision-making process of the rescue simulation robot agent is selected to perform the task as the optimization objective. Build a particle dimension of the particle swarm of the total number of tasks, through particle swarm optimization method obtains the best mission options. Whether the task is executed, the states only exists "yes or no". Therefore, this paper uses binary particle swarm optimization process.(2)Multi-agent collaborative research based on multiple evaluation function and the constraints hybrid particle swarm optimization methodIn order to improve particle swarm optimization to solve the problem of multi-agent collaboration, which is that the number of tasks are so much large as that the particle dimension is large and couldn’t convergence timely. And also, for evaluation of the effect of task rescue predominantly choosing the human experience, the single cause of the fitness function of the particle swarm makes the lacking of evaluation, in order to particle swarm optimization method using in multi-agent collaboration preferably. At the same time, the particle dimension is no longer subjected to the increase of a number of tasks, the particle dimensions remain unchanged and the dimension is not high. It avoids the problem which is that the after-algorithm can’t timely convergence.