A decentralized approach to pursuer-evader games with multiple superior evaders

Abstract

Few approaches dealing with pursuer-evader (PE) game allow real-time applications for multi-pursuer-multi-evader situations, especially for cases of evaders with superior control resources such as higher speeds. In addition, existing implementations are often centralized and require prohibitive computation even for very small-scale case. This paper introduces a real-time decentralized approach for such games. A sufficient condition for capturing a superior evader is proposed. Evader value, time-discounting payoff factor, and different priorities, such as "to catch as many evaders as possible" and "to achieve highest team payoff", is considered. Simulations confirm the efficiency of this approach.

Scalable and practical pursuit-evasion with networked robots

“…For superior evaders, Seymour [34] showed that finding the necessary number of pursuers to capture a single evader with infinite velocity in a graph G when pursuers and evader move simultaneously is equivalent to finding the treewidth of G. Other works [40,41] focus on an Euclidean open plane and provide sub-optimal real-time approaches.…”