This paper presents the main findings of a British research project, which combines complementary theoretical and practical approaches to estimating origin-destination matrices from traffic counts and other sources of low-cost data. The project's approach is to unify existing ideas and methods, and extend them by further theoretical and numerical analysis where appropriate. The main issues addressed here relate to ways of collecting and using data from various sources, especially in view of variations in their quality and the influence of this on the quality of the estimated trip matrix. A robust and flexible analytical formulation is described, that uses an objective function uniformly treating various sources of information. Analytical results are presented, that (1) indicate the quality of estimated trip matrix elements; (2) support survey design; and (3) suggest appropriate weights for each of the data to be used. The results of some calculations for a specific numerical example are presented, and contrasted with corresponding results from Willumsen's ME2 method. The new method given here is computationally practical, and can estimate trip matrices reliably from prior estimates and various kinds of traffic flow observations.
Abstract