A Cooperation Strategy Based On A Dynamic Of The Road Partition In RoboCup Rescue

RoboCup Rescue Simulation System(RCRSS) is one of challenging tasks that has vital research significance. This kind of multi-agent coordination issue takes palce in a real-time, dynamic and unknown environment with a strong engineering background. Morever, it involves many areas of application. Studies of various algorithms can be easily extended to other robotic applications based on RCRSS simulation platform, promoting the development of robot technology. It is via RCRSS simulation platform that the task does researches on multi-agent coordination.The study aims at gaining widespread application in the fields of process controlling, task scheduling,behavior coordinating, environmental predicting, robotics designing, etc through doing analysis and comparison of the effect of various algorithms in real-time dynamic environment, and exploring effective ways to make it.Firstly, the simulation environment is the foundation of the entire agent collaboration.The robot can not get the overall information; hence there will be limitations, relativity and uncertainty in performing individual tasks. In the paper, road-partition model is built to detail the map information by geometry characteristics and the surrounding environment of the road. On the basis of the static partitions, a dynamic road model is established using the dynamic partitions, convex hull, clustering, etc, which makes the agents have better comprehension of the surrounding information to do better rescue coordination.Secondly, about the Ambulance Team Agent, The biology of the Artificial Immune System(AIS) network structure can be applied to the AmbulanceTeam in order to enhance its efficiency. Specifically, the first step is to transform the ambulance environment, agent, immune network, etc to antigens, antibodies in the AIS model. Then the task on the AIS of the entire AmbulanceTeam can be completely changed by selecting affinity, motivation level of inhibition parameters, etc. Through the introduction of artificial immune algorithm, the action of AmbulanceTeam is more purposeful and targeted.Finaly rescue collaboration strategy can be optimized, improving the efficiency of the AmbulanceTeam.In the dynamic simulation environment rescue robot agent team personal ability and teamwork in the initial condition is unknown. In order to find the optimal role for the agent task allocation strategy, we use a weighted collaboration graph for the six kinds of agents in robocup rescue role modeling, and use the graph to composite an ideal team then we test the efficiency of different role assignments. We use a learning algorithm based on muti-agents collaboration to get a ideal task allocation.and we use the date got from many examples to obtain a near-optimal task allocation strategy by the learing algorithm.We use this approach in the Robocup Rescue Simulation System, a large number of experimental simulation test results show that the apporach is capable of forming an approximate optimal agent collaborative team.