Robot Path Planning Based on Q-learning Algorithm
Zeli Yang, Haofeng Lu, Jiaqi Wang, Yang Li, Yifan Wang · 2021 International Conference on Electronic Information Engineering and Computer Science (EIECS) · 2021
When Q-learning algorithm is applied to robot path planning, the state information needs to be discretized first, and the process of discretization needs to occupy a large amount of memory space, resulting in low learning efficiency and slow convergence speed. A path planning algorithm based on BN (Batch Normalization) and Q-learning is proposed to solve the problems of low learning efficiency and slow convergence speed. The experimental results show that the algorithm can make the algorithm converge on the premise of ensuring the robot to find the best path without collision. The convergence speed of the improved algorithm is increased by 25%, and the learning reward value is increased by 34%.