Arterial Performance and Evaluation using Bluetooth and GPS Data



Journal Title

Journal ISSN

Volume Title



Accurate travel time data are necessary to monitor and evaluate traffic conditions effectively. In the past 20 years, the hours per year lost by the average driver have increased by 300% in the 85 largest U.S. cities, which translates into lost productivity and increased costs. State department of transportation (DOT) agencies and other government organizations need accurate travel time and speed information to better combat this congestion faced by motorists. In the past, ground truth travel time information was typically collected with probe vehicles using the ?floating car? method. However, new methods using data collected from global positioning systems by private companies such as INRIX?, Navteq?, and TomTom? have emerged that allow travel time data to be obtained more cheaply and quickly. The Urban Mobility Report (UMR) has turned to these companies, specifically INRIX?, for calculating congestion indices across the United States. This is done by analyzing average speeds and reference speeds supplied by INRIX.

The UMR analysis relies on INRIX-supplied reference speeds to calculate delay, which produces artificially high delay on many suburban arterials. Currently, these reference speeds are determined by taking the 85th percentile of weekly speeds (typically overnight hours [10PM to 6AM]). There is a need to refine the reference speeds on arterials in order to account for signal operations, particularly during the daytime hours, so that the UMR more accurately reflects arterial congestion across the nation. Using Bluetooth and INRIX speed data, this thesis develops a new reference speed methodology that accurately reflects arterial delay during daytime hours. This study found that a 60% daytime free-flow reference speed best represents arterial congestion.

Using Highway Capacity Manual (HCM) guidelines, this thesis also explores the use of Bluetooth data for arterial and intersection level of service (LOS) analysis under both HCM 2000 and HCM 2010 methodologies. Through analysis, it was found that Bluetooth data capture more of the high and low LOS values compared to the HCM methodology based on segment speed calculations. These high and low LOS values, as well as the rapidly changing LOS between 15-minute intervals, could be attributed to an insufficient sample size.