Reference Path Generation for Autonomous Driving in Off-road Environment

Song Zhang,Huawei Liang,Pan Zhao,Zhiyuan Li,Tao Su
DOI: https://doi.org/10.1145/3529261.3529265
2022-03-25
Abstract:This paper proposes a reference path generation method for Autonomous Vehicles(AVs) in the off-road environment. Since obtaining sufficient and reliable prior information in an off-road environment is difficult, we adopt the Astar strategy to complete global routing based on the satellite topology map. Then planning result is corrected based on vehicle localization to obtain the initial reference path. Because the road geometry changes according to the road environment, we will use the centerline of the drivable area extracted with Probabilistic Neural Networks(PNN) to generate the final reference path. While the drivable area is invalid, the reference path planner is applied based on the initial reference path to form the final reference path. In the real off-road scenarios, the proposed method is tested on our experimental vehicle. Experimental results demonstrated that the proposed algorithm could generate reliable reference paths effectively and improve the performance and robustness of local trajectory planning.
What problem does this paper attempt to address?