A Simple Approach to Multi-Predator Multi-Prey Pursuit Domain

Abstract

We present a different approach to a class of pursuit games: the Multi-Predator Multi-Prey domain. In the typical game, a group of predators tries to capture a group of prey, and all the agents have perfect knowledge of prey and predator positions. In our problem definition the prey-agent and the predator-agent have only local information provided by its vision range, each predator independently tries to capture a prey in a one-predator-one-prey-pair way. The predator-prey-pair capture is not known in advance and both predators and prey are moving in the environment. We show that simple greedy local predator rules are enough to capture all prey.