Inferring Network Origin-Destination Matrices Using Partial Link Traffic Flow Information

Abstract

Inferring trip origin-destination (O-D) matrix in a vehicular network using partial link flow information is a critical issue for transportation planning and traffic management. However, due to a budgetary constraint, the minimum subset of links to be equipped with vehicle detectors (VDs) and their installation locations need to be determined to provide a desirable O-D matrix estimate. In the present research, a linear algebra-based method was developed to deal with the network location problem. Given a network structure represented by its link-path incidence matrix, some crucial links were strategically identified by a basis link model (Hu et al., 2009a, b) and the collected link flow information were further incorporated into an O-D matrix estimation model based on link flow conservation rule without any unreasonable assumptions, such as known prior O-D information and/or users' route choice probability. Numerical results based on a simplified real highway network indicated that the O-D matrix estimates given by the basis link model combining with the proposed O-D estimation model are generally satisfactory.