PERFORMANCE VALIDATION OF TRAFFIC FLOW MODELS THROUGH LONG-TERM MONITORING OF TRAFFIC STATES

Abstract

In this study, the performance of traffic flow models is verified by long-term monitoring of traffic states. In traffic control, it is important to monitor and provide information on traffic flows. As sensors for monitoring traffic states are sparse, supplementary estimation of unobserved areas by traffic flow models is necessary. As traffic flows change from moment to moment, it is difficult to calibrate the model parameters in advance. In this study, the technical issues of traffic flow models for traffic monitoring are identified and solutions are proposed, and the performance of traffic flow models is verified through case studies. The validation results show that the model has good accuracy in both normal and accident states. The results also suggest the possibility of monitoring traffic states under states where probe data is not available, by using downstream vehicle detectors as an alternative.