Research on the Local Dynamic Trajectory Planning Method Based on Relative Positioning

Abstract

Advanced driving assistance system (ADAS) in autonomous driving vehicles usually obtains local relative positioning information only by identifying lane lines, and how to avoid collisions when there are jumps in lane lines and encountering obstacles is a key technology that ADAS needs to solve urgently. Therefore, this paper proposes a local dynamic trajectory planning method based on relative positioning. First, the grid sampling is performed in the vehicle curve coordinate system, and the candidate path set in the vehicle coordinate system is obtained after coordinate transformation. Then, the path set is filtered through the collision check and evaluation function to obtain the optimal path without collision. Secondly, the maximum planning speed is adaptively adjusted according to the curvature and lateral deviation of the optimal path, and the speed curve is obtained by the improved trapezoidal speed planning method. Finally, the speed curve is smoothed by the gradient descent algorithm to avoid sudden acceleration. The simulation and real vehicle test results show that the algorithm adopted in this paper can stably correct deviation, avoid obstacles and follow the car in the local relative positioning scene with curves. In the scenario without obstacles, the maximum lateral deviation of the algorithm in the simulation and real vehicle experiment is reduced by 36.7% and 28.6% respectively, the average lateral acceleration is reduced by 29.6% and 46.4% respectively, and the travelling comfort and vehicle safety are high.

Keywords

• advanced driving assistance system
• lane line
• relative positioning
• local dynamic trajectory planning
• trapezoidal speed planning
• autonomous driving