Driving trajectory planning based on spatio-temporal navigation map in dynamic traffic scenes

Aiming at the path planning problem of driverless vehicles in high-speed autonomous driving scenes on highly dynamic structured roads, a driving path planning method based on space-time navigation map is proposed. The time dimension as a reference is introduced, combining with multi-targets behavior prediction, the perception results are projected onto a three-dimensional spatio-temporal navigation map. Thus, by increasing the time dimension, static and dynamic obstacles are unified into the same parameter space. In this space, the control points of uniform B-spline curves are initialized by the front-end A* path searching, the trajectory cost function is designed, and nonlinear optimization is performed to generate a collision-free and kinematically feasible (limited by speed, acceleration, etc.) spatio-temporal trajectory. As a result, the decision-making and planning problem in the two-dimensional Frenet dynamic physical space is transformed into a static scene decision-making and planning problem in the three-dimensional spatio-temporal coordinate system. Through simulation verification, the whole process of the proposed trajectory planning method takes an average of 51.27 ms, which meets the requirements of high-speed autonomous driving. Moreover, the proposed method adjusts the search conditions of the A* algorithm, thus increasing the searching speed by 27.86% compared with original algorithms, and improving the overall planning efficiency. © 2021, Editorial Department of Journal of Chinese Inertial Technology. All right reserved.