Safety-Critical Control for Multi-Agent Collaborative Pursuit

With the development of multi-agent systems, the pursuit-evasion of multiple agents has emerged as a cutting-edge research area in recent years. Focusing on the problem of multiple pursuers collaboratively capturing multiple evaders in bounded convex environments while ensuring collision avoidance, this paper designs an effective pursuit algorithm by integrating Voronoi partitioning and utilizing quadratic programming (QP) with Control Barrier Functions (CBF) to outline safety constraints. In this paper, Voronoi partitioning is used to allocate pursuit targets to each pursuer, planning the optimal pursuit actions. In conjunction, a minimally invasive controller designed with QP-CBF ensures that robots achieve effective pursuit control while guaranteeing safety. Additionally, the developed algorithm is distributed and can be implemented online, allowing pursuers to safely and effectively capture evaders by merely communicating positional information with their Voronoi neighbors. Simulation results have validated the effectiveness of the developed algorithm.