A constrained generalized least squares technique for origin-destination trip table estimation: An application in paramics

An origin-destination (O-D) trip table is one of primary inputs required in transportation studies including traffic simulation models. In this study, we apply the concept of constrained generalized least squares (CGLS) within the bilevel programming framework to estimate O-D trip tables for PARAMICS, a microscopic traffic simulator. Zonal capacity constraints are added to account for unreleased vehicles during the simulation such that a reasonable estimate of O-D demand can be determined. In addition, the gradient projection (GP) technique is proposed for solving the CGLS sub-problem due to its efficiency in handling large-scale problems. Numerical results are provided using a freeway corridor network to demonstrate the capability of the proposed procedure. Preliminary results are encouraging.