Group Sharing Risk Avoidance Method Based on Fuzzy Reinforcement Learning Strategy

Abstract

Aiming at the problem of autonomous risk avoidance and efficiency optimization of swarm agents in unknown environment, a swarm sharing avoidance method combining distributed fuzzy control and artificial potential field is proposed. Fuzzy division, parameter learning, group sharing and motion control are carried out by taking the force resultant of artificial potential field of agent as the input of the fuzzy control system. Through reinforcement learning to learn and adjust the control parameters, the fuzzy set division and control output in the control system are more effective. In order to improve the efficiency of group learning, agents with better performance of avoiding risk are screened and learning parameters are shared within the group. Simulation results show that the proposed risk avoidance method has good performance and learning efficiency.