Analysis of Naturalistic Driving Study Data: Roadway Departures on Rural Two-Lane Curves (2014)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

6The second Strategic Highway Research Program (SHRP 2) conducted the largest and most comprehensive naturalistic driving study (NDS) ever undertaken. The study collected data from more than 3,000 male and female volunteer passenger-vehicle drivers, aged 16 to 98, during a 3-year period, with most drivers participating from 1 year to 2 years. The study was conducted in six sites, one each in Florida, Indiana, New York, North Carolina, Pennsylvania, and Washington. Data collected include vehicle speed, acceleration, and brak- ing; vehicle controls when available; lane position; forward radar; and video views forward, to the rear, and on the driver’s face and hands. The NDS data file contains about 50 million vehicle miles, 5 million trips, more than 3,900 vehicle years, and more than 1 million hours of video—a total of about 2 peta bytes of data. In parallel, the Roadway Information Database (RID) con- tains detailed roadway data collected on more than 12,500 centerline miles of highways in and around the study sites, about 200,000 highway miles of data from the highway inven- tories of the six study states, and additional data on crash his- tories, traffic and weather conditions, work zones, and ongoing safety campaigns in the study sites. The NDS and RID data can be linked to associate driving behavior with the roadway environment. Campbell (2012) provides an excellent overview of the study. Additional details may be found at the study’s InSight website (https://insight.shrp2nds.us/). The study’s central goal is to produce unparalleled data from which to study the role of driver performance and behavior in traffic safety and how driver behavior affects the risk of crashes. This involves understanding how the driver interacts with and adapts to the vehicle, the traffic environ- ment, roadway characteristics, traffic control devices, and other environmental features. After-the-fact crash investiga- tions can do this only indirectly. The NDS data record how drivers really drive and what they are doing just before they crash or almost crash. The NDS and RID data will be used for years to come to develop and evaluate safety countermeasures designed to prevent or reduce the severity of traffic crashes and injuries. The First SHRP 2 NDS Analysis Projects Four contracts were awarded in 2012 under SHRP 2 Project S08, Analysis of the SHRP 2 Naturalistic Driving Study Data, to study specific research questions using the early SHRP 2 NDS and RID data. An open competition solicited proposals to address topics of the contractor’s own choosing that would have direct safety applications. The request for proposals required proposals that would • Lead to real-world applications and safety benefits (theo- retical knowledge without potential applications was not a priority); • Be broadly applicable to a substantial number of drivers, roadways, and/or vehicles in the United States; and • Demonstrate the use of the unique NDS data (i.e., similar results could not be obtained from existing nonnaturalistic data sets). In addition to these goals, SHRP 2 expected these projects to serve as both pilot testers and advisers. As they conducted these first substantial NDS and RID analyses, these studies’ experienced researchers would discover valuable insights on a host of both pitfalls and opportunities that others should know about when they use the data. That experience and advice can be found on the study’s InSight website (https:// insight.shrp2nds.us/). The four projects began in February 2012 and were con- ducted in two phases. In Phase 1, which concluded in Decem- ber 2012, the four contractors each obtained an initial set of data, tested and refined their research plan, and developed a detailed plan for their full analyses. Three projects, of which C H A P T e R 1 SHRP 2 Naturalistic Driving Study Background

7 data, some data had not been quality controlled and some characteristics of the data were not yet well understood. • Tools for data users. Not all crashes and near crashes had been identified, and a separate small data set containing only crashes, near crashes, and baseline exposure segments had not been built. In addition, a small trip summary file con- taining key features of each trip had not been built. Users can conduct initial analyses on many subjects quickly and easily using a trip summary file. • Other demands on data file managers. The first priority for the NDS manager, Virginia Tech Transportation Institute (VTTI), and the RID manager, Iowa State University’s Cen- ter for Transportation Research and Education (CTRE), was to complete data processing and quality control. Field data were being ingested continually. Data delivery for users sometimes was delayed due to these demands on their resources. These issues are being resolved in 2014. The NDS and RID data are complete and are being linked. Data processing and quality control are being completed. Crash and near-crash files and trip summary files are being built. If this project and the other two were to begin in 2015, each would have more data and would obtain the data far more easily and quickly. Readers should keep these constraints in mind as they read this report. Despite working under these constraints, this project and the other two NDS projects have produced valu- able new insights on important traffic safety issues that will help reduce traffic crashes and injuries. this study is one, successfully completed this proof-of-concept phase and were selected for the full Phase 2. These three proj- ects obtained and analyzed a much richer, though still pre- liminary, data set and reported their results in July 2014. Constraints of the First SHRP 2 NDS Studies These projects were conducted while the NDS and RID data files were being built. This circumstance imposed constraints that substantially affected the researchers’ work. The con- straints included the following: • Sample size. In summer 2013, when the projects requested their full data sets, the NDS data file was only 20% to 30% complete. As a result, each project could obtain only a frac- tion of the trips of interest now available in the full NDS data. • RID not complete and not linked to the NDS. Projects based on roads of specific types or locations could not identify those roads from the RID but instead had to use Google Earth or some similar database to identify them. Researchers then obtained trips of interest using less efficient searches through the NDS than will be possible when the NDS and RID are linked. • Data processing. Some data, such as radar, had not been processed from their raw state to a form where they were fully ready for analysis. • Data quality. NDS data are field data, and field data are inher- ently somewhat messy. When these projects obtained their