A Non-Uniform Sampling and Dynamic Constrained Optimization-Based Trajectory Planning Method for Autonomous Vehicles

Abstract

The quintic polynomial algorithm is not suitable for the sudden change of road curvature in the roundabout overtaking scenario, resulting in large curvature fluctuation and velocity oscillation. A trajectory planning method is proposed based on non-uniform sampling (NUS) and dynamic constraint optimization (DCO). Based on the NUS mechanism, the curvature fluctuation can be effectively reduced by using different center angles and road radii for sampling. Additionally, arc-length parameterization and time mapping techniques are adopted to extract velocity data, thereby achieving real-time adjustment of interpolation points. Centripetal acceleration constraints are employed to suppress velocity oscillations under dynamic conditions, significantly improving the smoothness of trajectories and comfort of vehicle driving. The simulation results show that the path length and curvature are optimized, and the centripetal acceleration is reduced, which significantly improves trajectory smoothness and passenger comfort.