Quasi-Dynamic Estimation of OD Flows From Traffic Counts Without Prior OD Matrix

This paper proposes a fully specified statistical model for the quasi-dynamic estimation of origin–destination (OD) flows from traffic counts for highway stretches and networks or for urban areas where the path choice is of minor importance. Hereby, the approach (E. Cascetta et al. , Transp. Res. B, Methodol. , vol. 55, pp. 171–187, 2013) is extended by eliminating the need for supplying a historic OD matrix. This is done by a combination of least squares estimation for replicating measured link flows with maximum entropy methods to fill in the non-observable part of the distribution across paths. Additionally, it is stressed that the quasi-dynamic assumption of constant path choice proportions over time-of-day-intervals for days of the same day category can be used in order to enhance estimation by including multi-day observations. Jointly one obtains a statistical framework with an explicit estimation algorithm that can be used to test the quasi-dynamic assumption. The approach is demonstrated to provide accurate results in a small-scale simulation study as well as two real-world case studies, one dealing with a highway segment where taxi floating car data provides the true OD flows for the taxis, and the other one dealing with an urban area with a very limited number of alternative paths allowing for explicit path enumeration.