Reinforcement Learning Methods Based On The Evolution Of Multi-robot Map Building

Recently, the study about Agent and MAS has been a growing interest in the area of DAI. Knowledge and ability of one agent is not enough, so study about MAS is developing rapidly. MAS is a aggregation consists of multiple Agent. It can harmonize the action of a group of Agents, so as to cooperate with each other acting and solving problems. Multi-Robotic System based on MAS is a concrete application of Multi-Agent conception in robotics field.Map building process is the basis for an autonomous mobile robot to execute various works. This paper will study Multi-Agent Reinforcement learning method and Co-Evolutionary Algorithm on the basis of MAS. Then propose an Evolutionary Reinforcement learning algorithm for map building using multi-agent mobile robots. The algorithm is applied in a decentralized homogeneous multiagent system. All of these are proved by theory study and simulation experiment. As seen from computer simulations, the proposed method is useful for building a map. And through compare with non-Evolutionary Map building method, the Evolutionary Reinforcement learning algorithm can increase search map efficiency and expedite convergence speed of global map.