In recent years, mobile robots have been widely applied in different areas, such as medicine, rescue and home service. Obstacle avoidance path planning is one of the most important issues for mobile robots. Researchers have paid attention to the methods integrated with reinforcement learning (RL). Even though the sparse reward problem of RL is proven to be solvable by Hindsight Experience Replay (HER), obstacle avoidance is still a thorny problem. Therefore, we proposed a DQN (Deep Q-learning Network)-HER RL method with proactive obstacle avoidance skill to learn the obstacle avoidance path planning task. We built simulations in Webots and designed an ablation experiment. The results showed our method improved learning stability and efficiency. In addition, we are planning to apply our method to the heterogeneous multi-robot system for multi-robot path planning in future work.
Obstacle Avoidance Path Planning Method Based on DQN-HER
Lect.Notes Computer
International Conference on Intelligent Robotics and Applications ; 2023 ; Hangzhou, China July 05, 2023 - July 07, 2023
16.10.2023
10 pages
Aufsatz/Kapitel (Buch)
Elektronische Ressource
Englisch
Hindsight Experience Replay , Deep Q-learning Network , Path Planning , Proactive Obstacle Avoidance Skill Computer Science , Artificial Intelligence , Software Engineering/Programming and Operating Systems , Computer Applications , User Interfaces and Human Computer Interaction , Computer Communication Networks , Special Purpose and Application-Based Systems
Obstacle Avoidance Path Planning Method Based on DQN-HER
TIBKAT | 2023
|Obstacle Avoidance Path Planning of Rotor UAV
British Library Conference Proceedings | 2017
|A PATH PLANNING STRATEGY FOR OBSTACLE AVOIDANCE
British Library Conference Proceedings | 2006
|Bridge crane path planning method considering obstacle avoidance
Europäisches Patentamt | 2020
|Obstacle avoidance path robust planning method and system for unknown obstacle distribution
Europäisches Patentamt | 2023
|