Trajectory tracking for substation inspection robots using deep reinforcement learning and tube model predictive control

This study addresses the limited adaptability of the traditional Tube Model Predictive Control (Tube-MPC) path-tracking algorithm, which fails to meet the operational requirements of substation inspection robots. Therefore, an improved Tube-MPC method based on Deep Deterministic Policy Gradient (DDPG) is proposed for robot trajectory tracking control. First, the structure and kinematic model of the inspection robot were analyzed and discretized to accommodate practical operating conditions. Second, to overcome the limitations of Tube-MPC in handling nonlinear systems and enhancing robustness, the DDPG algorithm is integrated with Tube-MPC to design a DDPG Tube-MPC trajectory tracking controller. Finally, comparative simulations verified the superiority of the DDPG Tube-MPC in terms of trajectory tracking accuracy and robustness. The simulation results show that DDPG Tube-MPC is better than traditional MPC and Tube-MPC when it comes to tracking accuracy, response speed and disturbance rejection capability. This makes the control performance of the inspection robot better and provides a new way to track a trajectory in complex environments.