A coordination model using fuzzy reinforcement learning for multi-agent system

It is important for multi-agent system that the functionally independent agents apply negotiation, coordination and cooperation to perform some set of tasks or to satisfy some set of goals. In this paper, we propose a two-layer architecture coordination model based on fuzzy reinforcement learning for multi-agent system. Agents make use of fuzzy inference system to choose the optimal behavior locally and confer the intentions and actions of others according to their state information. Then coordination layer harmonizes sub-goals among agents and assigns rational task to each agent while learning the strategies of agents using fuzzy reinforcement learning. As a result, agents choose and execute proper action to accomplish the desired task together in action layer. The simulation results showed that the performance of attacking is obviously improved in the RoboCup soccer simulation game.