Effects of calibration process on the simulation of rear-end conflicts at roundabouts

A methodology for calibrating and validating VISSIM simulation model is presented that allows to replicate the observed vehicles conflicts. A roundabout case study has been selected to test the usefulness of a combined approach of VISSIM simulation package and the surrogate safety assessment model (SSAM) for providing reliable estimates of traffic conflicts. Safety performance has been assessed from the field by video-recording vehicle interactions at the roundabout, and then expressed in terms of time to collision (TTC) values.The proposed calibration procedure has been performed by a multistage methodology involving microscopic drivers' car following behavior parameters to enhance the correlation between observed and simulated queue lengths at the roundabout's entries. The calibration procedure is based on a statistical screening of inputs leading to a linear expression relating significant parameters to the queue length. The best estimates of the model's parameters have been determined using a genetic algorithm technique.The spatial distribution of the rear-end conflicts and the TTC values determined by SSAM have been finally compared with the observed ones to analyze the capability of the model of replicating rear-end conflicts.The results suggest to this calibration procedure impacts positively on the estimate of the safety performance measures obtained through the simulation processes.Notwithstanding the good results in the evaluation of the model's accuracy, the simulation seems to fail in reproducing the traffic phenomena linked to unusual driving behavior, and therefore it is not able to replicate forced drivers' maneuvers that can lead to a conflict situation.