Dynamic origin-destination demand estimation using separable programming approach

This paper presents a novel methodology for the dynamic origin-destination demand estimation problem using separable programming. The methodology takes advantage of the separable structure of the ordinary least-squared error minimization formulation of the problem. The problem is formulated as a linear mathematical program through approximating each nonlinear term in the objective function by a piecewise-linear curve. An iterative procedure that integrates the demand estimation model and a dynamic traffic assignment model is used to ensure consistency between the demand and network congestion dynamics. A set of experiments are performed to test the sensitivity of the demand estimation results against several endogenous and exogenous model parameters. A real-world application is presented where demand pattern for the US 75 Corridor in Dallas, Texas is estimated. The results illustrate that the new methodology is adequate in terms of its estimation accuracy as well as execution time.