Examining the Relationship between Connected Vehicle Driving Event Data and Police-Reported Traffic Crash Data at the Segment- and Event Level

Abstract

Police-reported crash data have been the de facto element used by the transportation agencies in developing and implementing traffic safety projects. This approach is reactive in nature and can lead to suboptimal investment decisions owing to inherent challenges in crash data analysis. Because of their large-scale and near real-time availability, connected vehicle (CV) driving event data have emerged as a promising means of addressing these challenges. This study utilized CV event data for three different event types, namely, acceleration, braking, and cornering at three severity levels (easy, normal, and harsh), to examine the viability of using these data in traffic safety analysis. The results showed a strong correlation between crash frequency and CV driving event frequency. CV event data also improved the goodness-of-fit of crash frequency models. The results also showed that the relationship between CV driving events and traffic volume and roadway geometric data were generally consistent with the trends that crash data usually exhibit with the same predictors. This was true at both segment level and individual event level, as well as when the data were examined across different event/crash types. Overall, the results showed a strong case for these data to be used in traffic safety analyses as a complement to, or in lieu of, crash data.