Can Data Generated by Connected Vehicles Enhance Safety? A proactive approach to intersection safety management

Abstract: Traditionally, evaluation of intersection safety has been largely reactive, based on historical crash frequency data. However, the emerging data from Connected and Automated Vehicles (CAVs) can complement historical data and help in proactively identify intersections which have high levels of variability in instantaneous driving behaviors prior to the occurrence of crashes. Based on data from Safety Pilot Model Deployment in Ann Arbor, Michigan, this study developed a unique database that integrates intersection crash and inventory data with more than 65 million real-world Basic Safety Messages logged by 3,000 connected vehicles, providing a more complete picture of operations and safety performance of intersections. As a proactive safety measure and a leading indicator of safety, this study introduces location-based volatility (LBV), which quantifies variability in instantaneous driving decisions at intersections. LBV represents the driving performance of connected vehicle drivers traveling through a specific intersection. As such, by using coefficient of variation as a standardized measure of relative dispersion, LBVs are calculated for 116 intersections in Ann Arbor. To quantify relationships between intersection-specific volatilities and crash frequencies, rigorous fixed- and random-parameter Poisson regression models are estimated. While controlling for exposure related factors, the results provide evidence of statistically significant (5% level) positive association between intersection-specific volatility and crash frequencies for signalized intersections. The implications of the findings for proactive intersection safety management are discussed in the paper.