A Comparative Study of Freeway Travel Times from Four Data Sources

Abstract

This study evaluates the quality of estimated travel time data from Waze, HERE, and INRIX against Bluetooth data provided by Caltrans on multiple segments of the I-80 freeway between Davis and Sacramento, CA, U.S. We used the Bluetooth travel time data as the benchmark approximating the ground truth, and we utilize travel time error bias (TEB), average absolute travel time error (ATE), and standard error of mean band satisfaction (SEMBS) to assess the quality of the travel time data reported by Waze, HERE, and INRIX. The comparison result shows that the INRIX and HERE data closely match the Bluetooth data in both the trends and values of reported travel time. However, the quality of all three vendors’ data deteriorates when traffic congestion intensifies. Furthermore, Waze travel time data has a pattern which is distinguishable from the other three sets of data. We conducted a regression analysis to explore the potential causes of this discrepancy. The result suggests that the differences between Waze and Bluetooth are systematic instead of random. Moreover, the regression analyses also indicate that the differences between Waze and Bluetooth are positively associated with common disturbances, segment length discrepancies, segment length, and traffic count.