Path Planning of Mobile Robot Based on Deep Reinforcement Learning and Improved Artificial Potential Field
Yujie Xu, Chenglin Dai, Shuyi Han, Mingyu Fu · 2023
The reinforcement learning methods can adapt the state-action functions to the ever-changing environment, which helps robots find their ideal behavior when finishing tasks in complicated surroundings. In this article, we design a path planning method. Firstly, the reward function is designed by using the improved APF, then the successive state and action space are established, and finally, we train the path sizing model based on DDPG and APF algorithms in a randomly generated obstacle environment. We do simulation experiments and analyze the results, compared with traditional methods in path planning tasks, the new algorithm we proposed gives certain guidelines to the model during training and increases the success rate of training, which can greatly shorten the training time and get the optimized model faster.