A Pursuit Strategy for Multi-Agent Pursuit-Evasion Game via Multi-Agent Deep Deterministic Policy Gradient Algorithm

Abstract

This paper studies a classical pursuit-evasion problem. The pursuer attempts to capture the faster evader in a bounded area. The velocity of evader is 1.2 times as fast as the pursuers'. All of them have adaptive strategies. We use game theory to model the multi-agent pursuit-evasion game and prove that the game model has Nash equilibrium. Then, we modify the multi-agent deep deterministic policy gradient (MADDPG) algorithm for seeking the Nash equilibrium. The simulation examples are given to illustrate the effectiveness of the designed method.