This paper introduces the concept of the Point of No Return as a physically grounded safety indicator for rear-end vehicle conflicts, addressing fundamental limitations of the widely used time-to-collision metric. Unlike purely kinematic approaches, the proposed formulation incorporates braking capability and reaction constraints, enabling a direct assessment of whether a collision can still be avoided. To illustrate the applicability of the concept, a vision-based framework using a single camera is developed based on dashcam data, combining YOLO-based object detection, Kalman-filter tracking, and geometric distance estimation derived from bounding-box features and camera projection models. The estimated distance is further processed to obtain relative motion, allowing a unified analysis of time to collision and the Point of No Return within the same evaluation pipeline. Experimental results on real-world driving sequences show that the Point of No Return consistently precedes critical conditions identified by time to collision and provides a more stable and physically interpretable characterization of the transition toward collision inevitability. The results also highlight the sensitivity of the proposed indicator to braking capability, while showing lower sensitivity to variations in relative speed. Overall, this study demonstrates the relevance of the Point of No Return as a complementary indicator for collision risk assessment, offering a physically meaningful basis for decision-making in driver assistance systems and improving the interpretation of critical traffic situations. The proposed approach supports sustainable urban mobility by enabling earlier and more reliable intervention strategies, contributing to improved traffic safety, smoother traffic flow, and reduced environmental impact.
Adrian Soica (Thu,) studied this question.