Full Bayesian conflict-based models for real time safety evaluation of signalized intersections

Area of Science:

• Traffic Engineering
• Transportation Safety
• Intelligent Transportation Systems

Background:

• Advanced traffic management and connected vehicles (CVs) generate data crucial for real-time intersection safety.
• Understanding signal control's impact on safety is essential for optimization.
• Existing methods require enhancement to leverage dynamic traffic data effectively.

Purpose of the Study:

• To develop conflict-based safety performance functions (SPFs) for signalized intersections at the cycle level.
• To relate dynamic traffic parameters to rear-end conflicts using multiple traffic conflict indicators.
• To provide a foundation for real-time safety optimization of signalized intersections.

Main Methods:

• Utilized traffic video data from six signalized intersections.
• Developed cycle-level SPFs incorporating traffic parameters like queue length, shock wave dynamics, and platoon ratio.
• Employed multiple traffic conflict indicators: Time-to-Collision, Modified-Time-to-Collision, and Deceleration Rate to Avoid the Crash.
• Applied the Full Bayesian approach to model heterogeneity and site variations.

Main Results:

• Developed SPFs demonstrated a good fit, with statistically significant explanatory variables.
• Identified the highest conflict frequency at the onset of green signal time.
• Observed the highest conflict severity at the beginning of red signal time.

Conclusions:

• The developed conflict-based SPFs effectively link dynamic traffic parameters to intersection safety.
• Findings provide valuable insights for real-time safety adjustments in signalized intersections.
• The study supports the integration of CV data for proactive traffic safety management.