Citation

Huang F, Liu P, Yu H, Wang W. Accid. Anal. Prev. 2013; 50: 1014-1024.

Affiliation

School of Transportation, Southeast University, Si Pai Lou #2, Nanjing 210096, China. Electronic address: flyh630@gmail.com.

Copyright

(Copyright © 2013, Elsevier Publishing)

DOI

10.1016/j.aap.2012.08.018

PMID

23000074

Abstract

The primary objective of this study was to identify if the VISSIM simulation model and the Surrogate Safety Assessment Model (SSAM) approach provided reasonable estimates for the traffic conflicts measured at signalized intersections. A total of 80h of traffic data and traffic conflicts data were collected at ten signalized intersections. Simulated conflicts generated by the VISSIM simulation model and identified by SSAM were compared to the traffic conflicts measured in the field. Of particular interest was to identify if the consistency between the simulated and the observed conflicts could be improved by calibrating VISSIM simulation models and adjusting threshold values used for defining simulated conflicts in SSAM. A two-stage procedure was proposed in this study to calibrate and validate the VISSIM simulation models. It was found that the two-stage calibration procedure improved the goodness-of-fit between the simulated conflicts and the real-world conflicts. Linear regression models were developed to study the relationship between the simulated conflicts and the observed conflicts. Results of data analysis showed that there was a reasonable goodness-of-fit between the simulated and the observed rear-end and total conflicts. However, it was also found that the simulated conflicts were not good indicators for the traffic conflicts generated by unexpected driving maneuvers such as illegal lane-changes in the real world. The research team further tested the prediction performance of the conflict prediction models using the simulated conflicts as independent variables. It was found that the conflict prediction models provided acceptable prediction performance for the total and the rear-end conflicts with a MAPE value of 18% and 20%, respectively. However, the prediction performance of the conflict prediction models for the crossing and the lane change conflicts was only moderate with a MAPE value of 31% and 38%, respectively.

Language: en