Spatiotemporal Variation of Traffic Conflict at Signalized Intersections under Non-Lane-Based Traffic Condition

Abstract

The majority of accidents observed at signalized intersections comprise rear-end collisions resulting from a diversity of actions happening during signal change. This study explores the spatial and temporal variation of rear-end conflict in non-lane-based multi-class traffic conditions. Precise positions of the vehicles extracted from traffic video data collected from two major cities in India are utilized to identify critical vehicle interactions. Multiple surrogate safety measures (time to collision and deceleration rate to avoid crash) are used to identify the rear-end conflicts. More than 9,000 vehicle trajectories extracted from four signalized intersections were used to develop heat maps showing space-time variation of rear-end conflict at varying threshold values. Results showed that the majority of rear-end conflicts happened at the beginning of red time and green time at the upstream side of signalized intersections. Separate heat maps were developed to explore the variation in rear-end conflict caused by different vehicle types and lateral movement of vehicles. In addition, this study develops conflict-based safety models to investigate how changes in traffic parameters including traffic volume, vehicle speed, platoon ratio, queue length, and signal control affect the conflict hotspot (most intense conflict location) at the signalized intersections. The results showed that the higher the vehicle speed and queue length the greater the shift of the conflict hotspot away from the stop line in the upstream direction. The developed models can give insight into how changes in the traffic parameters and the signal cycle time affect the safety of signalized intersections.