ESTIMATION OF AN ORIGIN-DESTINATION TRIP MATRIX FROM LINK TRAFFIC COUNTS FOR LARGE NETWORKS

Abstract

Existing approaches to estimating origin-destination trip matrices (O-D) from traffic counts are often limited by the size of a network. These approaches concentrate on achieving a target O-D matrix obtained elsewhere (eg from previous data or a sample survey). With the inconsistency in flow sampling and bias target matrix estimation, incorrect results may be obtained. This paper recognises that specifying a target matrix in the objective function of some existing models may not resolve the problem of uniqueness for congested networks. As an alternative approach to overcoming the bias estimation often associated with the specification of a target trip matrix, the paper formulates a non-linear programming model which incorporates production and attraction information and inconsistency in the traffic counts, and a network model for extending sampled traffic counts to network population flows. Heuristics for solving the formulated models are presented, together with a case study.