Center for Traffic Simulation Studies

Studying the variability of travel time is very important in ensuring that intelligent transportation systems provide drivers with useful and accurate route guidance. Much of the research on travel time variability has concentrated on long-term variability, e.g. peak hours, non-peak hours and daily and seasonal variability. This research concluded that travel time would follow a normal or log-normal distribution under the conditions that: (a) travel times on all separate route sections are independent and that (b)trip times per unit distance on all sections are identically distributed. This paper addresses the issue of short-term temporal and spatial variations and correlations of travel times. Because of the inadequacy of GPS-based probe vehicle data, vehicle tracking data from a Paramics simulation of the Orange County, California, network is used in this study. The results provided strong evidence of a significant correlation between link travel times and between link travel time and link arrival time. Conditional link travel time distributions can be minimally represented by normal or lognormal distributions and vary by links and arrival times. This has important implications for calibrating travel time estimation and prediction.