Estimating Time-Varying Origin-Destination Flows from Link Traffic Counts

Abstract

A dynamic model for estimating real-time origin-destination flows from time-series of traffic counts is presented. The time variation of flows is explicitly treated as a dynamic process. The model is formulated based on minimizing the integrated squared error between predicted and observed output traffic counts over the period of observation. An efficient solution method is developed by using Fourier transformation and illustrated with numerical examples. The numerical simulation experiment shows that the system dynamic approach may be particularly suitable for on-line traffic management and control in urban transportation systems.