Research On The Bi-level O-D Matrix Estimation Model Based On The Confidence Coefficient

As the basic material for research, planning, controlling and management of urban transportation, OD (origin-destination) matrix data is commonly used for revealing traffic problems and analyzing the relationship between traffic demand and urban economy and land-use. In light of the high cost of manpower, finance and time, the large-scale sampling investigation, the traditional way to get OD matrix data has been nowadays replaced by a more efficient method which estimates OD matrix data through link traffic volume and given OD matrix.Based on the review of the theories and applications on OD matrix estimation at home and abroad, this article first of all classified different estimation methods with different criterions and briefly introduced the main trend of current research. Then a bi-level OD matrix estimation model with a confidence coefficient was put forward. This model, which was based on reference of prior researches, contains two parts. The upper level improved the common generalized least squares model by introducing a confidence coefficient. The lower level adopted the widely-used Wardrop user equilibrium model.The above mentioned confidence coefficient was used to describe the reliability of the two reciprocal error parameters in the upper level, i.e. the one between the given OD matrix and the estimated OD matrix, and the one between link flow and distributed flow. The value of confidence coefficient increases with the growth of the reliability of the first error parameter. In order to determine the value of the confidence coefficient, some qualitative methods were explained such as Delphi and trial method. The algorithm of this new bi-level model was then put forward through some in-depth discussions on models of this kind.In the end, a case was investigated to show the advantages of this model. In the end, this research was summarized into several conclusions and expectation was made for future work.