Uncertainty Prediction-Based Left-Turning Trajectory Planning for Automated Vehicles at Signal-Free Intersections

This paper presents a trajectory planning framework aimed at facilitating safe and efficient left-turn maneuvers at signal-free intersections. A Sequential Quadratic Programming (SQP) method is proposed to determine the optimal turning point and timing, balancing collision avoidance, driving efficiency, and driving comfort. To account for the uncertain future motions of surrounding vehicles, an S-T-R (Space-Time-Risk) space is introduced. It extends the traditional S-T space by explicitly introducing a third dimension to quantify the probabilistic collision risk arising from prediction uncertainties. These risk-quantified uncertainties are further incorporated into the optimization process as inequality constraints the by Dynamic Programming (DP) and Quadratic Programming (QP) algorithms. Comprehensive simulation results indicate that the proposed approach can not only enhance the safety of autonomous left turning at signal-free intersections, but also improve overall driving comfort.