Research on The Safety Boundary of Intersection Car-Following Scenarios Based on Naturalistic Driving Behavior Study

The current autonomous car-following function poses a potential safety risk at intersection areas because of the road construction and non-physical driving lanes. We propose to extract the naturalistic driving behavior to study the features of intersection car-following scenarios. We extracted the car-following scenarios from Next-Generation Simulation (NGSIM) database and presented the driving behaviors of different intersection car-following scenarios. The naturalistic driving characteristics of intersection car-following scenarios are obtained through the Multivariate Kernel Density Estimation (MKDE) method. The bivariate probability density distribution and normal distribution probability density formula of velocity, acceleration, and reciprocal Time Headway (THW) are established to represent the safety boundary of intersection car-following scenarios. Our work is the preliminary study of regional group safety research