Obstacle Straddling and Avoidance Integrated Path Planning and Tracking for Autonomous Driving on Unstructured Roads

Unstructured roads frequently feature segments with a complex distribution of obstacles of varying sizes. Conventional path planning and tracking methods uniformly consider all obstacles as non-straddleable, employing a singular strategy of avoidance. However, this inflexible approach can adversely affect the vehicle's safety and stability. Therefore, the critical challenge lies in the rational management of diverse obstacles and the determination of the most effective corridor to address these complexities. To address the challenge, an obstacle straddling and avoidance integrated path planning and tracking method is proposed for unstructured roads. Specifically, to accurately represent the safe avoidance space of straddleable obstacles, obstacle safe avoidance bounding box (SABB) is designed. Concurrently, a risk field that accounts for vehicle status and road conditions has been developed to quantify the risks of non-straddleable obstacles. In addition, to ascertain the optimal corridor, multimodal evaluation paths are generated based on the SABB and the planned path, and an evaluation function is established, leading to the proposal of an optimal corridor decision-making method. Furthermore, to facilitate efficient navigation on unstructured roads, path planning and tracking are consolidated into a single optimization problem aimed at determining the optimal control inputs. Ultimately, co-simulation and experiment have been conducted to substantiate the effectiveness of the algorithm proposed. The results indicate that in complex scenario, the mean absolute steering angle and yaw rate are reduced by 10.2% and 7.6%, respectively, thereby enhancing vehicle operational safety and stability.