The r-Safety Reserve and Adaptive Kinematic Smoothing for Car-Following Transitions

Abstract

Longitudinal spacing models describe the speed--spacing states that vehicles maintain, while longitudinal controllers regulate vehicle motion through spacing and speed errors. Less explicit is how one spacing state should be connected to another through a finite, physically meaningful vehicle transition. This paper develops such a connection through r-Safety and Adaptive Kinematic Smoothing (AKS). The r-Safety relation represents speed-dependent spacing through a braking-normalized reserve, whose change, together with leader motion, determines the follower displacement required during a transition. AKS then converts this displacement into a low-dimensional analytical trajectory with consistent position, speed, spacing, and acceleration. Experiments using controlled platoons and NGSIM trajectories identify r-Safety values of 0.307 and 0.249--0.281, respectively, and show that AKS can both represent observed transitions and generate prospective references under prescribed endpoint states, transition horizons, and leader motion. In same-controller ablations, adding AKS reduces integrated acceleration effort by 21--47% and integrated squared jerk by 73--81% across the controlled and naturalistic datasets. FHWA ACC/CACC trajectories further provide an empirical illustration of finite-horizon execution under automated following, showing how spacing adjustment can continue beyond the prescribed speed transition. The framework therefore provides a direct path from a speed--spacing relation to an analytical finite transition and its use as a longitudinal control reference.