Estimating travel time mean and variance using intelligent transportation systems data for real-time and off-line transportation applications

The accurate estimation of travel time data is valuable for a variety of real-time and off-line transportation applications. This dissertation includes methodologies for estimating link and corridor travel time mean and variance. The test sites are two limited-access freeway corridors—one instrumented with AVI antennas and one instrumented with dual inductance loop detectors at 0.5-mile spacings. The estimates using the ITS data were compared to simultaneous instrumented test vehicle and commercial vehicle travel time data.; A procedure was outlined for using the loess non-parametric statistical technique to obtain corridor travel time mean and variance estimates from each ITS data source, commercial vehicles, and instrumented test vehicles. The estimates from each data source were then aggregated to five minutes and the ITS data source estimates were compared to the commercial vehicle and instrumented test vehicle corridor travel time estimates. A method is also presented for estimating corridor mean and variance from the continuous and disaggregate ITS sub-link mean and variance data using loess and fourth-order polynomials. In addition, a methodology for testing the accuracy of instrumented test vehicle drivers along a corridor was developed.; The research demonstrates that commercial vehicles have statistically different travel time mean and standard deviation than AVI-equipped vehicles which suggests it may be beneficial to provide traveler information in real-time for commercial vehicles. It was also found that AVI-equipped vehicles were not statistically different than the instrumented test vehicles and that an AVI system with an adequate number of tag reads could replace traditional data collection methods. By comparing inductance loop travel time estimates to the commercial vehicle and test vehicle data sources, the research quantifies how aggregated inductance loop detector travel time estimates do not capture the travel time variance characteristics of individual vehicles.; Finally, the research shows that the covariance between links along the corridor is positive using the AVI ITS data source, though the additional programming of a first- and second-order Taylor's Series approximation of corridor travel time mean and variance is not likely justified from a practical perspective for the approximately two-mile corridor studied here. Further work is recommended on longer corridors as the practical impact of the link covariance identified in this dissertation could be higher for longer corridors.