Supervisor:
Southwest Ordnance Industry Bureau
Organizer:
Chongqing Ordnance Industry Society
Chongqing University of Technology
Chongqing University of Technology
Research on target following of spherical robot based on PPO
| DOI: | 10.11809/bqzbgcxb2024.03.036 |
| Keywords: | spherical robot; target following; deep reinforcement learning; PPO; artificial potential field |
| Abstract: | Spherical robots have become a popular choice for exploring underwater environments, inspecting shores and beaches, and handling other complex scenarios due to their exceptional motion performance, remarkable terrain adaptability, and anti rollover characteristics. However, the spherical robot system model has inherent features of underdrive and nonlinearity, making motion control a challenging task, especially in complex environments where following targets reliably can be difficult. To address this issue, we propose a new method for following targets using the proximal policy optimization (PPO) algorithm. Our method utilizes deep reinforcement learning theory to design the corresponding state and action spaces based on the spherical robot dynamics model. We introduce an artificial potential field in the reward function to keep the target centered in the robot’s field of view, thereby enhancing the robustness of the target following method. Our simulation results demonstrate that the proposed method meets the requirements of the given scenario, allowing the spherical robot to follow a randomly moving target reliably. |
| Published: | 2024-03-28 |