A Hierarchical Path Planning and Obstacle Avoidance Framework for the Autonomous Heavy Vehicle Considering Dynamic Properties

Abstract

Autonomous obstacle avoidance (OA) has garnered increasing attention these years, which is a demanding task especially for heavy vehicles with maneuvering difficulties. A hierarchical OA framework that consists of a new virtual state planning optimizer (VSPO) and a dynamic path follower (DPF) considering dynamic properties is proposed for an autonomous heavy vehicle. In the upper layer, a new path virtual state predictor based on the nonlinear tire is proposed, with high-fidelity modeling for path dynamics (PDs) description. A rolling optimization is solved under dynamic constraints related to vehicle lateral safety and planning barrier functions. Additionally, a multidimensional safety evaluator is designed considering collision risk anisotropy, by which the optimal state sequence is obtained iteratively through the discrete planning equation. In the lower layer, a dynamic following algorithm with dynamic cost is proposed based on NMPC, in which a high-fidelity predictive model is built to depict multiple degrees of freedom (DOF) and reflect wheel slip. In order to improve the adaptation of the optimal control, a variable weight regulation strategy is formulated with a threshold sensitivity function. The constraints related to vehicle lateral safety and wheel slip are constructed, and the initial control law based on the Lyapunov function is designed, which are respectively converted to the limitation of vehicle states for stability enhancement. Finally, the simulation platform is established, and the validation is conducted in different cases, which proves the effectiveness of reliable OA and dynamic performance improvement.