Local Minima Consistent Identification Method to Accelerate Multiagent Reinforcement Learning

Abstract

This paper presents a novel method in accelerating Multiagent Reinforcement Learning called Local Minima Consisten Identification or LMCI. Incorporating LMCI in Multiagent Reinforcement Learning Algorithms will successfully accelerate the learning convergence. This method scales down the state space iteratively by distinguishing insignificant states from the significant one and then eliminating them while learning, which aggressively reduces the scale of the state space in the following learning episodes. This method is generally applicable for varying Multiagent Reinforcement Learning algorithms such as Multiagent Q() and Multiagent SARSA() in order to solve multiagent task challenges or general multiagent learning with large scale state space characteristic.