Research On Multi-Agent Pursuit-Evasion Games Strategy

With the wide application of multi-agents such as multi-robots,multi-drones,and multi-UAVs in the military and industry,the multi-agent pursuit-evasion game is a cutting-edge research paper in recent years and has become important research in the field of multi-agents.It mainly studies how to divide the work of multiple pursuers and coordinate the actions of each pursuer to optimally encircle and capture the evader cooperative pursuit strategy.The pursuit-evasion method is an innovative application in important military fields such as coordinated reconnaissance,coordinated attack,intercepting rogue watercraft,and emergency rescue.Therefore,this thesis focuses on the research on the pursuit-evasion strategy of multi-agents.The main contents of this thesis are as follows.Firstly,the distributed,real-time algorithms for multiple pursuers cooperating to capture an evader are developed in an obstacle-free and an obstacle-cluttered environment,respectively.The developed algorithm is based on the idea of planning the control action within its safe,collision-free region for each robot.We initially present a greedy capturing strategy for an obstacle-free environment based on the Buffered Voronoi Cell(BVC).For an environment with obstacles,the Obstacle-Aware Buffered Voronoi Cell(OABVC)is defined as the safe region,which considers the physical radius of each robot,and dynamically weights the Voronoi boundary between robot and obstacle to make it tangent to the obstacle.Each robot continually computes its safe cells and plans its control actions in a recursion fashion.In both cases,the pursuers successfully capture the evader with only relative positions of neighboring robots.A rigorous proof is provided to ensure the collision and obstacle avoidance during the pursuit-evasion games.Simulation results are presented to demonstrate the efficiency of the developed algorithms.Furthermore,aiming at the problem that multiple pursuers cooperate to capture an evader in an unbounded environment,a multi-agent distributed formation and cooperative attack strategy based on Cooperative proportional guidance law are proposed.Firstly,the pursuers are divided into one leader and multiple followers,in which the leader pursues the evader by using the proportional guidance law;Secondly,for followers,when they are far away from the evader,distributed formation control is adopted to track the leader;When the distance is close,the cooperative proportional guidance law is adopted with the leader to attack the evader at the same time.Finally,the simulation results show that the designed distributed formation control strategy and cooperative proportional navigation law are effective.Finally,considering the hunting demand of evaders with uncertain locations,a distributed capturing strategy based on the Voronoi diagram centroid is proposed.Firstly,the probability density function about the location distribution of the evader is obtained by combining the Kalman filter prediction algorithm;Furthermore,for a single evader,the optimization strategy based on Voronoi graph centroid is used to reduce the cost function and finally realize the capture;Secondly,for multiple evaders,the optimization strategy based on the minimization of Voronoi diagram area is adopted for capturing,and a local distributed pursuing strategy based on target allocation is proposed to realize the capture of multiple evaders by multiple pursuers.Finally,the simulation results show that the designed Kalman filter prediction algorithm and the optimization strategy based on the Voronoi diagram are effective.