Estimating departure times from traffic counts using dynamic assignment

Abstract

Dynamic traffic assignment and observed traffic counts are used to estimate the distribution of departure times in an urban area trip matrix. The objective is to find the maximum entropy distribution of departure times by origin zone subject to observed traffic counts on a subset of network links. The procedure results in the estimated number of trip departures from each origin in 10–15 minute time intervals of the full analysis period. Such an analysis period will typically range from one to three hours long. We first review the dynamic user-equilibrium assignment problem (DUE) and the dynamic traffic assignment algorithm (DTA) from a previous paper, and then describe its use with traffic counts to estimate departure times in a trip matrix. We present an application to a Pittsburgh network in which trip departures are estimated for each 10 minute interval of a peak-hour survey trip matrix. Dynamic assignment and departure time estimation based on traffic counts can be used to examine the impacts of alternative transportation management strategies such as work hour changes and ramp metering. Computational advances such as parallel computing will enable the procedure to be run on large networks while counts are being monitored. Such an application may provide near real-time detection of temporal trip departure profiles by origin zone in emergency evacuations or other special events.