Research on Robot Path Planning Based on Reinforcement Learning

• DOI: 10.48550/arxiv.2404.14077
• Publication Date: 2024-04-22
• Authors: Ruiqi Wang
• Keywords: FOS: Computer and information sciences; Computer Science - Robotics; Computer Vision and Pattern Recognition (cs.CV); Computer Science - Computer Vision and Pattern Recognition; Robotics (cs.RO)

This project has conducted research on robot path planning based Visual SLAM. The main work of this is as follows: (1) Construction SLAM system. Research been the basic architecture A system developed ORB-SLAM3 system, which can conduct dense point cloud mapping. (2) map suitable for two-dimensional obtained through conversion. part converts by into an octomap and then performs projection transformation to grid map. conversion containing a large amount redundant information extremely lightweight planning. (3) algorithm reinforcement learning. experimental comparisons between Q-learning algorithm, DQN SARSA found that with fastest convergence best performance in high-dimensional complex environments. verification simulation environment. results open-source dataset self-made prove feasibility effectiveness designed At same time, also comparative experiments three learning algorithms under condition obtain optimal condition.