The Study Of Multi Robots Coordinated Pursuit-Evasion Task Based On NEAT

Multi robots coordinated control is a very important research field in Robotics. While themulti robots pursuit-evasion task is a classic problem in that field. This task requires multiautonomous mobile robots to form a pursuit team and cooperate with each other to captureanother team of evader robots. The most difficult part lays on that single robot isunavailable to capture any evader; instead they have to cooperate to finish the task.This paper is based on the2-Demension environment multi robots pursuit task, and a newsolution according to the swarm intelligence theory is proposed to solve the multi robotscoordinated control problem. At first, the paper introduced the background and studysituation in and out of China of that field, and also the purpose and meaning of the research.And then, a method based on swarm intelligence on multi robots coordinated control wasproposed. The definition and foundation of the swarm intelligence theory was introduce,and the tools for implementing the theory was analyzed, including the Artificial NeuralIntelligence, Optimization Algorithm and simulation environment, etc. After that, a2-dimension simulation environment for multi robots pursuit was created, and a set ofpursuit robots was added. By implementing sensors, brains and actuators, a simplemanmade life was created. Via the iteration in the simulation environment, and guided bythe fitness function, the parameters of the ANN brain of the individual was optimized,which brings on the meaningful actions of the robot, like avoid the obstacle and pursuit theevader. Finally the team can capture the evaders according to the coordination of theindividuals. The experiment proves the feasibility of the control method based on swarmintelligence. At last, the PSO and NEAT methods were compared on optimizing the ANNbrain of the pursuit robots, and the simulation experiment proves the feasibility of thealgorithm.