Simulated Traffic Conflicts Do They Accurately Represent Field-Measured Conflicts?

Recently, there has been growing interest in using microsimulation models to assess the safety of road facilities by analyzing vehicle trajectories and estimating conflict indicators. Using microsimulation models in safety studies can have several advantages, although concerns have been raised about the ability of these models to represent unsafe vehicle interactions and near misses realistically as well as their need for rigorous calibration. The main objective of this study was to investigate the relationship between field-measured and simulated conflicts at an urban signalized intersection in Surrey, British Columbia, Canada. Sixty hours of recorded traffic data were collected in 2 days and used in the conflict analysis. Automated video based computer vision techniques were used to extract vehicle trajectories and identify conflicts on all four approaches to the intersection. Conflict measures (e.g., time to collision) and location were determined and compared with simulated conflicts from a microscopic simulation model (VISSIM) using the Surrogate Safety Assessment Model (SSAM). A two-step calibration procedure was proposed to enhance correlation between simulated and field-measured conflicts. The first calibration step was matching actual field conditions (desired speed and arrival type) to ensure that VISSIM gives real average delay values. The second step was the use of sensitivity analysis followed by a genetic algorithm procedure to calibrate the VISSIM parameters that had the biggest effect on the simulated conflicts. Finally, conflict heat maps were provided to compare field-measured with simulated conflict locations. The results highlighted the importance of model calibration and identified several limitations of the SSAM.