Path Planning for Mobile Robot Based on Deep Reinforcement Learning and Fuzzy Control
Chunling Liu, Jun Xu, Kaiwen Guo · 2022 International Conference on Image Processing, Computer Vision and Machine Learning (ICICML) · 2022
Aiming at the problems of slow training speed and poor generalization ability of deep reinforcement learning model in path planning, this paper proposes a multi-controller model of Dueling DQN combined with fuzzy control(DDFC). At the beginning of training, fuzzy control is used to provide a large number of positive samples for the Dueling DQN model, so as to improve the training efficiency of the model while ensuring that the mobile robot has certain obstacle avoidance ability at the beginning of training. The negative feedback shaping reward function and state space are designed to alleviate the problem of sparse reward. Aiming at the fact that the membership function of traditional fuzzy control can’t deal with difference situations in the process of moving, an improved membership function is designed, which can change according to the change of the situation. The simulation results show that the improved model can make the mobile robot avoid obstacles effectively and improve the rate of convergence. It also has good performance in different scenes and improves the generalization ability of the model.