Integrated safety-critical motion planning and control strategy for emergency lane changes of autonomous vehicles

Autonomous driving requires the integration of perception, decision-making, planning, and control. Learning-based perception and decision-making methods often suffer from a lack of interpretability, making it necessary to revisit safety-critical autonomous driving technologies in the context of planning and control. To address emergency lane-change maneuvers, this paper proposes an integrated safety-critical motion planning and control strategy. First, in the motion planning layer, the vehicle’s dynamic characteristics are considered, and a quintic polynomial optimization problem is reformulated to minimize lane-change time and distance, based on steady-state steering. Second, to represent inter-vehicle safety distances more consistently, a collision-free minimum safety circle is constructed using geometric constraints. Robust high-order Control Barrier Functions (HOCBFs) are then formulated for the relative-degree-two collision constraints of the control-affine prediction model. A robust tracking-error CLF certificate is constructed over the lateral-position, heading, lateral-velocity, yaw-rate, and longitudinal-speed errors. With an explicitly bounded applied-input relaxation, the tracking error is shown to be bounded under the stated model-validity and feasibility conditions. Robust HOCBF constraints are used separately to regulate the prescribed safety margins. Comprehensive simulations and vehicle experiments demonstrate that the proposed strategy maintains the prescribed collision-avoidance margins while improving tracking and lateral-dynamic performance in the evaluated scenarios.