Dynamic estimation of origin-destination matrices for an urban network using traffic counts

The purpose of this research is to design a model for estimating a dynamic or time dependent origin-destination trip table for a large urban area and to test the model using traffic count data obtained from inductive loops on a sample of arterial streets. So far the only reliable way to obtain an origin destination trip table has been through surveys at origins and destinations. These surveys are expensive and they do not reflect changes in traffic flow that occur over time. The hope of this research is to provide a reliable substitute for the traditional method of obtaining an O-D trip table. The network chosen to test the model is downtown Boston.; The method of O-D flow estimation used in this study is sequential. The algorithm used is the Generalized Least Square Method. The input data used to test the model is the on-line traffic information obtained from inductive loops on a sample of local streets. The study period is from 6:00 a.m. to 9:00 a.m. The length of this study is 5 weekdays. The 1987 O-D table provided by the Boston Area Regional Planning Authority is used as the apriori O-D table for the first interval of day one.; Local street count data were provided by the City of Boston Traffic Control Center. The validity of O-D estimates were checked by examining how closely the estimated link counts derived from the O-D estimates match the actual link counts. Statistics used to measure the degree of closeness are Root Mean Square Error, and Percent Root Mean Square Error.; The model was validated based on testing using synthetic data where the true O-D volumes were known. The results of the validation proved the mathematical integrity of the model and its robustness. The model, when applied to the urban traffic data, did not produce estimates of link counts which would be suitable for real time traffic control due to the following causes: (1) partial traffic count data; (2) insufficient detector coverage; (3) lack of speed and delay information on all the links in the network; (4) lack of sufficient centroid connectors.