Cooperative Pursuit in a Non-closed Bounded Domain

This paper considers a multi-agent pursuit-evasion game in a bounded domain with multiple exits and obstacles. First, the problem modeling is performed, and the control strategy of the evaders is assumed. Then, based on Voronoi partition and potential field gradient, a pursuit strategy for N-pursuer/M-evader in the non-closed bounded domain is proposed, where the pursuers are classified into three autonomous switchable categories: the goalkeeper, the striker and the defender. The proposed strategy is a low-dimensional algorithm, which can greatly reduce computational costs. Finally, the feasibility and effectiveness are demonstrated by numerical case studies.