Research on lane change trajectory planning model for autonomous vehicles considering U-turns

To address the safety challenges of lane changes in autonomous vehicles, this paper proposes a trajectory planning method that incorporates lane change fallback. First, a virtual lane change trajectory is generated using fifth-order polynomial curves and numerical optimization techniques. By integrating lateral and longitudinal motion relationships, a constraint analysis method centered on deceleration time intersecting with the motion segment is introduced to establish a kinematics-based lane change safety domain. A collision severity calculation model is further established to define thresholds, enabling unified decision-making for lane change, no-lane-change, and fallback behaviors. Subsequently, a two-stage trajectory planning method is proposed: Stage 1 simultaneously plans the overlapping segment of both successful lane change and fallback trajectories, targeting overall performance optimization. Stage 2 independently optimizes the terminal segment trajectory based on actual decision outcomes, using the termination state of the overlapping segment as the initial condition. In the simulation analysis, our method is compared with critical comfortable reversal and critical fast reversal methods. The observed results demonstrate that the determination of collision severity thresholds not only influences the likelihood of autonomous vehicles successfully completing lane changes, but also affects the suitability of different trajectory planning methods. These findings support the development of a tailored lane-change trajectory planning model, which can enhance both the safety and comfort of lane-changing maneuvers in autonomous vehicles.