Hybrid Multiagent Reinforcement Learning Approach: The Pursuit Problem

How to coordinate the behaviors of the agents through learning is a challenging problem within cooperative multi-agent domains. As a generic machine learning method, reinforcement learning can suit the needs of multi-agent learning in many aspects. This study presents a novel approach, which using reinforcement learning to learn the coordinated actions of a group of cooperative agents, without requiring explicit communication among them. The proposed approach combines advantages of the modular architecture, profit-sharing learning and opponent modeling technique in a single multi-agent framework. The effectiveness of the technique is demonstrated using the pursuit problem.