Analyzing rear-end collision risks in foggy weather through a generalized extreme value model

Area of Science:

• Road safety engineering
• Traffic management systems
• Data-driven risk assessment

Background:

• Fog severely degrades visibility and driver perception, escalating crash risks, particularly rear-end collisions.
• Traditional crash analysis methods are limited by underreporting and low event frequency, hindering proactive risk assessment in adverse weather.

Purpose of the Study:

• To develop a proactive, data-driven framework for estimating fog-related crash risks on Indian multilane highways.
• To integrate naturalistic vehicle trajectory data, anticipated collision time (ACT), and extreme value theory (EVT) for enhanced safety evaluation.

Main Methods:

• Collected vehicle trajectory and fog data from 8 Indian highway sites (4- and 6-lane divided segments).
• Estimated visibility using contrast-based image processing and extracted vehicle trajectories via object detection and tracking.
• Classified 21,461 vehicle conflicts using ACT profiles and applied EVT models (GEV distribution) to estimate crash probabilities across varying visibility and road types.

Main Results:

• Rear-end conflicts present a higher safety risk than sideswipe conflicts due to car-following dynamics in fog.
• Safety margins decrease significantly in dense fog, particularly for cars and two-wheelers on 4-lane highways, increasing crash potential in mixed traffic.
• Two-wheeler crash risk on 6-lane highways increases sevenfold in dense fog; heavy commercial vehicles show improved safety on 7-lane highways.

Conclusions:

• Demonstrated the feasibility of ACT-based EVT modeling with field trajectory data for proactive safety assessment in low-visibility conditions.
• The developed framework enables real-time risk estimation to inform adaptive countermeasures like variable speed limits and enhanced warning systems.
• Proposed solutions include fog-responsive warning systems and improved lane delineation for overall roadway safety enhancement.