Research On Multi-Robot Pursuit-evasion Problem Based On Game Theory

The multi-robot pursuit-evasion problem is a problem in which a team of mobile robots act as pursuers to pursue other mobile robots act as evaders in a pursuit-evasion system composed of multiple mobile robots. It is a game between two teams that through mutual cooperation within team and competition in teams to beat each other. Pursuit-evasion problem is a complicated system of real-time dynamic coordination multi-robot confrontation and cooperation game problem. Many key technology of pursuit-evasion problem has been widely used in industry, military, etc.Before widely applied to other fields, game theory was first used in the analysis of economic behavior. In computer science, especially multi-agent systems can find game theory everywhere,such as multi-agent negotiation,auction, etc. This paper studies multi-pursuer vs one-evader games with the evader has resisting arrest behavior under the framework of game theory, main content includes:Firstly, we introduce the background of multi-robot pursuit-evasion problem, research status at home and abroad, and the related knowledge of game theory in this paper.Secondly, the paper gives the evader resisting arrest behavior, adds some action to robot and gives the corresponding definition. The payoff function is redefined by improving the weights allocation of each influence coefficients and introducing the resistance coefficient of evader, based on the original game model of multi-robot pursuit problem.Thirdly, this paper studies the multi-robot pursuit problem under incomplete information,with the type of robot ("brave "or" cowardly ") is taken as unknown information.Fourthly, this paper explores the resisting arrest behavior and the weight distribution of influence coefficients can do what a result on capture time and result with MATLAB. The effectiveness of the algorithm also be verified after the model is modified. The simulation results show that resisting arrest behavior can bring difficulties to pursuers,and it can the shorten the time the pursuers cost to capture the evader if weight allocation of influence coefficients is reasonable. The algorithm in this paper performs better compared to pursue directly even the evader has resisting arrest behavior.