Microsimulation Approach for Predicting Crashes at Unsignalized Intersections Using Traffic Conflicts

A microsimulation approach is presented for assessing safety at unsignalized intersections using critical traffic conflicts computed in simulation models as a surrogate safety measure. The proposed approach was applied to a crash data set, collated over a 5-year monitoring period, corresponding to urban unsignalized intersections of the city of Salerno, Italy. Given that over 45% of accidents recorded at these intersections occurred more especially in six 1-h time periods during the day and that these hourly periods corresponded to the peak traffic volumes, it was considered worthwhile to investigate each of these 1-h concentrated time intervals. Traffic flows were computed by using video cameras placed at each intersection and expressed as peak-hour volumes in the analysis. A process of microsimulation model calibration and validation was carried out, and a good level of conformity between the traffic simulated in the peak hours and the corresponding one measured in the field was obtained. Subsequently, a form of software, which is compatible with the applied microsimulation software, was used for identifying the number of critical conflicts at the intersections investigated. For this aim, a collision was assumed to be very probable when conflicts are characterized by a time to collision (TTC) and post-encroachment time (PET) below the threshold values of 1.5 and 5 s, respectively. The relationship between critical conflicts computed in microsimulation models and actual crashes was found to be statistically significant. A crash-prediction regression model was then developed for the investigated intersections, which allows predicting crashes in the field on the basis of the critical conflicts computed in simulation. The proposed model was found to fit the accident data slightly better than the traffic volume-based regression model. DOI: 10.1061/(ASCE)TE.1943-5436.0000473. (C) 2012 American Society of Civil Engineers.