An integrated model for estimating time-varying network origin–destination distributions

This paper presents an integrated method for estimating time-varying O–Ds in urban networks. The proposed method starts with our previously developed two-stage, non-assignment-based model that can yield a time-varying O–Ds without a reliable prior O–D set and a dynamic traffic assignment model (DTA). With the initial estimated O–D set, one can compute the link-incident matrix with any acceptable DTA model and generate a revised distribution of network O–Ds. To further improve the estimation accuracy and also account for the impact of urban signals, we have developed an intersection O–D estimation model that can produce an additional set of system observation constraints based on either existing or estimated intersection turning fractions. Although the statistical properties and variances of such a system under a large-scale network remain to be investigated, the results of simulation experiments have clearly indicated that our proposed method for integrated estimation of time-varying network O–D distributions is quite promising.