A comparative study on the dynamic estimation of network origin-destination demands

The purpose of the present research is to conduct a comparative study on the dynamic estimation of network origin-destination (OD) demands using two statistical methods, that is least squares and Kalman filtering(KF) methods, and an artificial intelligence (AI) approach, i.e., Artificial Neural Network (ANN)model. The numerical test results based on field data collection and simulation experimentsindicate that the ordinary least squares (OLS) method with nonnegative constraintprovides a satisfactory resultin solvingthe intersection turning proportionsproblem. Besides, in the freeway/expressway and general network cases, both the KFand ANNmethodsshowstatistically acceptable results, even though the ANN method provides a more stable and betterresult.In accordance with the above model evaluation results, one can design beneficial traffic control and/ormanagement strategiesto achieve some system-wide objectives.