This paper presents a study on the control of unmanned bicycle motion based on inertial wheels in the pybullet physics simulation environment. To address the nonlinear and strongly coupled nature of unmanned bicycles, the Proximal Policy Optimization deep reinforcement learning algorithm is employed in this study to achieve stable motion control of the bicycle. Two sets of independent reward functions are designed for balancing and path tracking of the unmanned bicycle to improve control effectiveness. Simulation results demonstrate that the designed control strategy can effectively maintain the balance of the unmanned bicycle and accurately track the desired path. This research provides a novel approach and insight into the intelligent control of unmanned bicycles, showcasing the application potential of reinforcement learning-based motion control in highly complex systems.
Research on Bicycle Motion Control Based on Reinforcement Learning
2024-07-26
448081 byte
Conference paper
Electronic Resource
English
Natural Residual Reinforcement Learning for Bicycle Robot Control
British Library Conference Proceedings | 2021
|