Development of a Surrogate Safety Measure for Evaluating Rear-End Collision Risk Perception

Surrogate safety measures are widely used in advanced driver assistance systems (ADASs) to detect traffic conflict and evaluate risk levels in car-following scenarios. However, the integration of driving style into surrogate safety measures has not been thoroughly investigated. To enhance the reliability and user satisfaction of ADASs, this study introduces a surrogate safety measure, risk deceleration (RD), to describe drivers’ risk perception and decision-making process in the car-following process. The parameters of RD are calibrated using a genetic algorithm to better reflect individual driving styles. The effectiveness of RD in identifying perceived risk in severe deceleration scenarios is tested and evaluated using vehicle trajectories representing three different risk levels. The evaluation framework includes classification performance metrics, timeliness, and alignment with driver behavior. Results show that RD outperforms the safety margin and deceleration rate to avoid crash across the evaluation framework and for all four vehicle type combinations. Furthermore, the calibrated reaction time value significantly enhances prediction performance. In addition, RD effectively characterizes drivers’ behavior and perceived risk levels during severe deceleration processes. Moreover, RD shows considerable potential for integration into collision warning systems, which can enhance driver satisfaction and compliance.