Abstract:
The usefulness of the service provided by an Advanced Traveler Information System and Advanced Traffic Management System would depend on the quality of data that the system generates and the ability of system designers to tap into the properties of the data. Probe vehicles are a major source of data in ATIS and ATMS. We investigate the quality of data generated by probe vehicles on signalized arterial streets in terms of measurement errors and analyze the properties of travel time data. A dominant property is the presence of dependence in the data. We then explore the nature of travel time estimates. Given that travel time processes have certain properties and the realizations of travel times are affected by a large number of covariates, the estimates may be of high variance. Also unless one states the travel time estimation model carefully, some estimates may be biased. The dependence property of the data further makes statistical inference difficult. We point out invalid conclusions that one may make under an independence assumption.
tr48.pdf