A Hierarchical Approach to Trajectory Planning for Autonomous Vehicles on Curvy Roads

This paper is dedicated to enhancing the planning speed and kinematics of intelligent vehicles navigating dynamic traffic environments on curvy roads by introducing a three-layer trajectory planner. Within the coarse trajectory layer, we introduce the innovative jump node dynamic programming (DP) algorithm, designed to circumvent trajectory searches for overlapping nodes. In contrast to the conventional DP algorithm, the decision-making efficiency achieved with the jump node DP algorithm demonstrates a substantial improvement of up to 30.6% within simulated environments. Transitioning to the reference trajectory layer, we put forth an improved DP strategy. This strategy is forged through the integration of the jump node DP algorithm and the joint interpolation method, permitting the generation of a crafted reference trajectory. In the alternate trajectory layer, segments of the trajectory characterized by substantial curvature, as planned by the improved DP strategy, undergo initial replanning through the adaptive cruise control (ACC) strategy. Subsequently, these trajectory segments revert to the improved DP strategy after traversing specific waypoints within the receding horizon. Simulation cases illustrate that the minimum radius of curvature in the planned trajectory can be augmented by up to 12.7%. All proposed algorithms have undergone dynamic simulation experiments, affirming their effectiveness.