Study On Integrated Approach For OD Estimation And Link Travel Time Function Calibration Based On Bi-level Model

The models which used to estimate OD matrix based on traffic counts are paid more attention. The traditional method divides this course into two steps : first to estimate the OD matrix under the conditions of determinated road network parameters and assignment model parameters; then modify road network parameters and assignment model parameters under the conditions of determinated OD matrix. However, the problem in this method is that the parameters which got from step2 are not consistent with the parameters got from step1. Repeat iteration is necessary to get steady and convergent result.However, the steady and convergent parameter is usually can not be got and sometimes not consistent, which will influence the fitting precision.The influence factors of OD estimation precision are analyzed in this paper. On the basis of OD estimation theory, current OD estimation methods and Bi-level programming model, the Bi-level programming model used to estimate OD which integrates OD estimation and BPR function is put forward in this paper. The improved method is: up level subject function is a least squares model which makes the quadratic sum error of estimation matrix, count volume and impedance minimum, and low level subject function is a stochastic user equilibrium assignment Pr obit model with constraint conditions in which user route choice behavior is well considered. Then, the solution procedure is put forward which based on genetic algorithm, and programmed with Matlab . Finally, the model and algorithm are applied in the example network, the rationality and validity of the model and algorithm are also verified, which indicate that the method this paper puts forward can improve the precision of fitting than traditional method.