This page organizes papers related to artificial intelligence published around the world. This page is summarized using Google Gemini and is operated on a non-profit basis. The copyright of the paper belongs to the author and the relevant institution. When sharing, simply cite the source.
Towards Safe Maneuvering of Double-Ackermann-Steering Robots with a Soft Actor-Critic Framework
Created by
Haebom
Author
Kohio Deflesselle, Melodie Daniel, Aly Magassouba, Miguel Aranda, Olivier Ly
Outline
This paper presents a deep reinforcement learning framework based on Soft Actor-Critic (SAC) to enable safe and accurate maneuvering of Double-Ackermann-steering mobile robots (DASMRs). Unlike holonomic or differential drive robots, DASMRs face complex kinematic constraints that make classical planners vulnerable in crowded environments. This framework leverages Hindsight Experience Replay (HER) and CrossQ overlay to improve maneuvering efficiency while avoiding obstacles. Simulation results using a four-wheel steering robot demonstrate that the learned policy achieves up to 97% of the target position while avoiding obstacles. This framework does not rely on predefined trajectories or expert demonstrations.
Takeaways, Limitations
•
Takeaways:
◦
A deep reinforcement learning-based framework for safe and accurate operation of DASMR is presented.
◦
Improved maneuverability by leveraging HER and CrossQ overlays.
◦
Learning is possible without predefined trajectories or expert demonstrations.
◦
Simulation results show a 97% success rate in reaching the target location.
•
Limitations:
◦
Performance verification in real-world environments is required.
◦
Computational complexity and training time issues.
◦
Only simulation results for a specific robot model (4-wheel steering rover) are presented.
