Trajectory generation method using Bézier spiral curves for high-speed on-road autonomous vehicles

This paper proposes a trajectory generation method for high-speed autonomous vehicles in a structured on-road environment, such as a highway. The proposed method generates a smooth trajectory using G² cubic Bézier spiral smoothing, based on the current trajectory and the centerline model of a desired lane, and the velocity profile of the trajectory. Using the proposed method, an on-road vehicle can follow or change lanes safely. The performance of the proposed method was confirmed by offline experiments using a real data set and online experiments in a real environment.