Estimating Travel Time Summary Statistics of Larger Intervals from Smaller Intervals Without Storing Individual Data

Abstract

Historically, real-time intelligent transportation systems data are aggregated into discrete periods, typically of 5 to 10 min duration, and are subsequently used for travel time estimation and forecasting. In a previous study of link and corridor travel time estimation and forecasting by using probe vehicles, it was shown that the optimal aggregation interval size is a function of the traffic condition and the application. It is expected that traffic management centers will continue to collect travel time statistics (e.g., mean and variance) from probe vehicles and archive this data at a minimum time interval. Statistical models are developed for estimating the mean and variance of the link and route or corridor travel time for a larger interval by using only the observed mean travel time and variance for each smaller or basic interval. The proposed models are demonstrated by using travel time data obtained from Houston, Texas, which were collected as part of the automatic vehicle identification system of the Houston TranStar system. It was found that the proposed models for estimating link travel time mean and variance for a larger interval were easy to implement and provided results that had minimal error. The route or corridor travel time mean and variance model had considerable error compared with the link travel time mean and variance models.

Traffic Probe Data Processing for Full-Scale Deployment of Vehicle-Infrastructure Integration

“…The primary challenge could quickly shift from determining how to extract as much information as possible from limited available data, to determining how to efficiently identify traffic detectors (9). A similar experiment showed how to use toll tags to estimate freeway travel times, with minimum aggregation intervals of 5 min because of the limited density of equipped vehicles (10). This approach should be scalable for higher-density applications with shorter aggregation intervals.…”