Validation of traffic conflicts extracted from a simulation model (VISSIM)

Traffic safety considers an essential aspect of transportation engineering, yet the results of the various methods used in their evaluation often differ significantly. Lately, there has been increasing interest in utilising micro-simulation software to evaluate traffic safety through analysing the file trajectories of vehicles and counting various indicators of conflicts. The major objective of this study is thus to validate such simulation through examining the relationship between traffic conflicts extracted from coupling the VISSIM simulation model with both the Surrogate Safety Assessment Model (SSAM) and field measures of conflicts. Thirty-two hours of traffic data were collected from four cameras at the Arbein Street signalised intersection over two days; these were installed at the four intersection approaches, and each camera recorded for four hours per day. The famous safety indicator Time-To-Collision (TTC) was then used to estimate simulation traffic conflicts. Such conflicts were then extracted from the field data and compared with the simulated conflicts to assess the consistency between simulated and observed conflicts and to calibrate the VISSIM simulation model. The results showed that calibration of the model reduced mean absolute preventable error (MAPE) from 55% to 33% for rear-end conflicts, from 71.5% to 38.5% for lane-change conflicts, and from 67% to 47% for crossing conflicts; in general, total conflicts decreased from 62% to 37%. This calibration thus led to an improvement in the consistency of the two approaches used. A linear regression model and Spearman Rank Correlation Coefficient (SRCC) were then developed to examine the relationship between observed and simulated conflicts. The R² values of the linear regression models varied, ranging from over 60% to 76% for different types of conflicts. The Spearman Rank Correlation Coefficient (SRCC) was 86%, however, indicating reasonable consistency between the two approaches.