A novel approach for safety assessment at unsignalized intersections: Integrating spatial metrics into postencroachment time analysis under heterogeneous traffic conditions.

Abstract

Objective: Evaluating the safety risks at unsignalized intersections becomes increasingly complex amid conditions of dense traffic flow, a heterogeneous mix of vehicles, nonadherence to lane demarcations, and reactive driving techniques. Understanding driver behavior under such varying circumstances is crucial for accurately assessing the potential hazards present at these intersections. The study aims to assess the safety of unsignalized intersections by incorporating both spatial and temporal variables under heterogeneous conditions.

Methods: The study presents a new safety indicator, dynamic postencroachment time (DPET), formulated to encapsulate both the spatial and temporal variables of heterogeneous traffic. Six unsignalized intersections were selected as the study areas in Assam, India, to assess the new indicator for merging and crossing conflicts. A videographic survey of the intersections was done to obtain vehicles' trajectory data and capture their conflict behaviors based on their position, speed, and steering angle. The peak over threshold (POT) approach of extreme value theory (EVT) was used to examine the feasibility of the indicator, and the methodology was validated using 4 years of crash data.

Results: The result showed that a common threshold of 0.7 s from the POT approach is sufficient to identify severe conflicts. Furthermore, the threshold level yielded a shape parameter greater than -0.5, affirming that the maximum likelihood estimates retain the standard asymptotic attributes associated with EVT. The DPET approach estimated more crashes than observed fatal crashes, reflecting its ability to capture extreme events at a lower threshold. In comparison, the traditional PET approach estimated fewer crashes due to higher values influenced by evasive actions. Graphical analysis demonstrated a strong correlation between the observed crash data over 4 years and the estimated crashes derived from the EVT models.

Conclusions: Integrating spatial variables in PET analysis provides a more robust measure for conflict analysis and assessment of potential traffic conflicts at unsignalized intersections. The subsequent validation of the model with actual crash data highlights its practical applicability in enhancing road safety. The findings from the study provide a promising direction for future research and the potential for widespread implementation in traffic management systems.