Accurate Data-Driven Sliding Mode Parking Control for Autonomous Ground Vehicles with Efficient Trajectory Planning in Dynamic Industrial Scenarios

Autonomous Ground Vehicles (AGVs) can transfer or load and unload material in industrial scenarios due to their flexibility and operability, freeing people from tedious and repetitive labour. However, the dynamic scene and the system model uncertainty reduce the parking accuracy of AGVs in industrial scenes, which seriously affects the autonomous operation robustness and accuracy. This paper proposes a data-driven parking control-planning integration solution for AGVs in complex industrial scenes, which combines sliding mode parking control with trajectory planning based on iterative error compensation, allowing AGVs to park accurately and converge to the target parking site quickly. First, a data-driven discrete sliding mode controller is designed to iteratively reduce parking error, which is insensitive to disturbance during erratic iterative error compensation, ensuring the rapid and asymptotic convergence of the parking error in industrial scenarios. Then, to achieve efficient planning with the target parking site constantly being corrected, an improved Bi-RRT based trajectory planning scheme considering operational constraints and node expanding region division is proposed, which provides the trajectory that contributes to parking convergence for the proposed controller promptly. Finally, the efficiency of the proposed method is verified by real-world experiments with self-developed AGV in industrial scenes, and experimental results show that the proposed method achieves accurate parking control with efficient trajectory planning and rapid parking error convergence.