关键词: 自动驾驶, 轨迹跟踪控制, 近似动态规划, 双层优化

Abstract: Multi-axle wheeled articulated vehicles are prone to "tail amplification" effects and lateral instability issues during trajectory tracking. Studying the stable trajectory tracking of multi-axle articulated vehicles is of significant importance. This paper focuses on multi-axle articulated vehicles where only the tractor has steering capability. We establish a 7-degree-of-freedom vehicle dynamics model and propose a Bi-level policy trajectory tracking control method: the upper-level policy addresses the tractor’s trajectory tracking problem, while the lower-level policy optimizes the trailer’s trajectory tracking, ensuring precise tracking for both tractor and trailer. To guarantee real-time computation of the control policy, the upper-level trajectory tracking policy is solved using a finite-horizon approximate dynamic programming approach, transforming the online optimization problem into an offline pre-solution of parameters, thus reducing the time required for online solving. This approach significantly reduces online computation time. Co-simulation experiments with high fidelity simulation software demonstrate that our proposed method improves the trajectory tracking accuracy of multi-axle articulated vehicles by 12.82%, with a single-step solution time below 10 ms, enhancing computational efficiency by two orders of magnitude.

Key words: self-driving, trajectory tracking, approximate dynamic programming, bi-level optimization