Research on Local Path Planning Algorithm Based on Frenet Coordinate System

Aiming at the problem of real-time trajectory planning of unmanned vehicles in highway environment, this paper proposes a local trajectory planning algorithm based on rule sampling. Taking the globally planned path as the reference line, the local trajectories are decoupled horizontally and vertically based on the Frenet coordinate system. According to the designed end sampling rules, polynomial fitting is used to generate horizontal and vertical trajectory clusters. A cost function based on horizontal and vertical evaluation is designed to rank the trajectory clusters respectively. A collision detection method based on the three-circle method is proposed. According to the sorting results, the trajectories are merged and collision detection is performed, and the optimal collision-free trajectory is output. Finally, for the obstacle avoidance scene in driving, Python is used to simulate the simplified algorithm. The experimental results show that the proposed local path planning algorithm meets the requirements of security and real-time performance.