Prediction of rear-end conflict frequency using multiple-location traffic parameters

Area of Science:

• Traffic Engineering
• Transportation Safety
• Autonomous Systems

Background:

• Traffic collisions are often preceded by traffic conflicts, offering insights into crash mechanisms.
• Existing research on traffic conflicts often uses aggregated data and focuses on specific environments, limiting broader applicability.
• Autonomous vehicles (AVs) are emerging, necessitating advanced traffic management and safety analyses.

Purpose of the Study:

• To investigate rear-end traffic conflict frequency at an urban network level using vehicle-to-vehicle interactions.
• To correlate conflict frequency with network traffic states.
• To analyze conflict frequency under scenarios including AV characteristics.

Main Methods:

• Utilized Time-To-Collision (TTC) and Deceleration Rate to Avoid Crash (DRAC) metrics to estimate conflict frequency.
• Defined critical conflict points based on TTC and DRAC thresholds.
• Employed Zero-inflated, Negative Binomial, and quasi-Poisson models, controlling for endogeneity, to analyze contributory factors.

Main Results:

• Conflict frequency is significantly higher during congested traffic conditions.
• Increased variations in vehicle speed correlate with higher conflict counts.
• Analysis provides a foundation for understanding traffic conflicts in relation to traffic states.

Conclusions:

• Urban network-level analysis of rear-end conflicts reveals significant correlations with traffic congestion and speed variability.
• Findings highlight the importance of considering traffic state in proactive traffic management and AV safety.
• Future research should incorporate simulated AV traffic and additional surrogate safety indicators for deeper insights.