Application of High-Resolution Vehicle Data for Free-Flow Speed Estimation. Issue 1 (2017)

• Record Type: Journal Article
• Authors: Rouphail, Nagui M.; Kim, SangKey; Aghdashi, Seyedbehzad
• Is Part Of: Transportation research record. Volume 2615:Issue 1(2017)
• Page Start: 105
• Page End: 112
• Publication Date: 2017
• Subjects: Transportation -- Periodicals; Roads; Transport -- Périodiques; Routes -- Périodiques; Routes -- Conception et construction -- Périodiques; Roads; Transportation; 388.05
• DOI: 10.3141/2615-12
• Languages: English
• ISSNs: 0361-1981

Abstract

The use of probe vehicle data for highway performance monitoring is increasingly being adopted in many countries. In the United States, third-party data provider entities such as Google, INRIX, HERE, and TomTom are delivering products to state and local transportation agencies that are enabling them to identify bottlenecks, incidents, and other key operational events on the basis of probe vehicle speed and travel time. However, the capacity analysis methods in the U.S. Highway Capacity Manual continue, for the most part, to rely on the analyst's ability to gather data at fixed points, whether manually or from fixed point sensors. This paper explores the use of intelligence to drive (i2D) high-resolution vehicle data to assess several research questions related to free-flow speed (FFS) estimation, a key parameter in freeway segment analyses. On the basis of 1 year of high-resolution data collected from a local fleet of about 20 vehicles driven by volunteer drivers, researchers accumulated more than 20 million s of driving, which when filtered were used to evaluate research questions and develop enhanced predictive models for FFS. Speed limit and section ramp density (i.e., those ramps within the segment proper only) were found to have had a strong effect on the value of FFS. Driver familiarity was found to have an effect also, although this effect was not conclusive across 10 study sites. Finally, an FFS predictive model that incorporates speed limit and section ramp density was found to fit the high-resolution data quite well, generating an absolute error of only 1.3% across all sites. That finding compares with an error of 6.6% with the current Highway Capacity Manual 2010 model predictions.