Advancing Crash Prediction Models Based on Simulated Conflicts and Exploring Their Predictive Capabilities and Transferability

Abstract

<p>Evaluating the impacts of planned or implemented road safety treatments could be challenging as limited information on crash effects may be accessible. Thus, surrogate measures for safety assessments could be considered as an alternative approach in evaluating the effects of various road safety treatments. The main objective of this study was to investigate various approaches for developing crash prediction models for four-legged signalized intersections in the City of Toronto based on simulated traffic conflicts, including the speed of conflicting vehicles, a variable that has received little emphasis in previous research. A safety evaluation of these intersections with automated vehicles (AVs) was conducted and the transferability of the models to two Canadian jurisdictions was investigated. Results indicate that the safety of intersections may improve with the presence of AVs in cautious operation mode and that these types of models may be transferred for use with caution in other jurisdictions. The primary outcome of this study was the establishment of improved relationships between surrogate safety measures and crashes to swiftly evaluate planned or implemented road safety treatments with and without the presence of AVs.</p>