Advanced Sensor Analytics and Extreme Value Modeling: Dichotomizing Day–Night Variability in Rear-End Collisions on Expressways

Abstract

Nighttime driving presents unique challenges and risks compared to daytime driving. This study analyzed rear-end conflicts on expressways and identified thresholds for various conflict indicators under both day and night conditions. Utilizing cost-effective 3D LiDAR technology, renowned for its robustness in low-light environments, this study elucidates the multifaceted influence of various factors on traffic safety dynamics across day and night conditions. Extreme value theory was applied to evaluate safety, incorporating factors like traffic environment and driver characteristics that are often overlooked in naturalistic studies. The analysis also included the effect of percentage oblique width on safety-critical events. Interestingly, drivers experienced about three times higher crash risks during the day compared to night, likely due to increased vigilance and caution at night. These findings offer valuable recommendations for setting headway requirements based on lighting conditions and can help improve advanced driver assistance systems to detect and respond more effectively to unsafe following distances. © 2000-2011 IEEE.