This study characterised real-world intersection traversals using naturalistic driving datasets. The real-world intersection traversals were selected from the Second Strategic Highway Research Program (SHRP-2) and the Virginia Traffic Cameras for Advanced Safety Technologies (VT-CAST) 2020 datasets. The VT-CAST dataset contains real-world trajectories from 2,800 hours of vehicle traversals across 126 intersections. A step-by-step approach was taken to create an algorithm that can identify three different intersection traversal trajectories: straight crossing path (SCP); left turn across path opposite direction (LTAP/OD); and left turn across path lateral direction (LTAP/LD). Crashes and near-crashes in SHRP-2 were manually reviewed and characterised into the same three scenarios. For every encounter, the velocity, acceleration, and estimated time to collision (eTTC) were calculated. The average velocities of the traversing vehicles were found to be about 7 m/s for all three intersection traversal scenarios. The average maximum deceleration was at minimum 17 times greater for crash and nearcrash scenarios compared to the everyday driving. The VT-CAST dataset allows for a very large quantity of intersection traversals to be recorded and identified. This data on standard driving traversals from VT-CAST and crashes/near-crashes from SHRP-2 may be useful for developing detailed intersection driver behaviour models for I-ADAS development.
Keywords:
Advanced Driver Assist Systems; crashes; driver behaviour; intersection; real-world data