Transportation institute will share massive trove of driver research data
The Virginia Tech Transportation Institute has won a $16.4 million contract from the Transportation Research Board that will make available thousands of hours’ worth of naturalistic driving data available to researchers across the auto and highway industry.
The project is a follow up to the second Strategic Highway Research Program (SHRP2) naturalistic driving collection effort, started in 2006 and completed with data enhancement projects in 2015. These projects were also funded by the Transportation Research Board, part of the National Academies.
The effort captured volunteer motorists driving their own vehicles in everyday traffic as they normally would, an effort dubbed by institute researchers as “naturalistic driving.” Here, driver actions are captured using a sophisticated data acquisition system developed by Virginia Tech engineers that continuously collects and stores information, including acceleration, braking, GPS location, and other instrumentation data, with video views of the driver’s face, interactions within the vehicle cab, and views outside the vehicle. These data were captured from October 2010 to December 2013 in six U.S. cities.
The huge collection of video and accompanying data – totaling over 2 petabytes and representing the equivalent of four millennia of driving time – will be shared with researchers that wish to use the data to improve driving safety for all motorists, said Jon Hankey, senior associate director for research and development at the institute.
“The data can really be used by anyone who wants to evaluate and improve transportation, such as road designers, operations researchers, vehicle manufacturers, tier one suppliers, transportation researchers, students, or safety advocates,” said Hankey.
As part of the continuing SHRP2 project, a web portal called the InSight Data Access Website is being further developed by institute engineers. It allows qualified researchers to explore some of the most meaningful parts of the captured naturalistic data without having to visit the institute itself, said Hankey.
From InSight, users can obtain information on driver assessment, vehicle type, trip summary, and crash and near crash information. The data available can be used by itself to accomplish research, or it can be used to define larger projects that sample additional data from the larger and more detailed dataset.
Hankey is leading the four-year data sharing effort with co-investigators Miguel Perez, director of the institute’s Center for Data Reduction and Analysis Support, and Clark Gaylord, chief information officer for the institute and architect of the computing infrastructure that holds and processes the data. Suzie Lee, director of research compliance at the institute, is ensuring participant protection by developing and tracking all data-use licenses.
The collected, reviewed, and annotated data can be used to study a wide range of topics – from drowsy or distracted driving to near collisions and impacts – just as similar footage taken by the institute previously has been used in dozens of other studies. For instance, recorded footage of a car crash can be annotated as to whether the driver appeared fatigued before a crash, was gazing at their cell phone, or was texting.
“We believe the next big improvements in safety will come from a more thorough understanding of how the driver, vehicle, and infrastructure all interact in the driving environment,” added Hankey. “What is it about a specific intersection design that leads some drivers to crash and what countermeasures can be used to mitigate this?”
More than 3,500 volunteers from the following areas were selected and paid for their efforts: Tampa, Florida; the Raleigh-Durham metro region in North Carolina; Buffalo, New York; Seattle, Washington; State College, Pennsylvania; and Bloomington, Indiana. Over 5 million trips were captured during the study’s time frame; participants were generally involved in the study for either one or two years.
“Data were collected any time participants drove their vehicles, shortly after ignition on to ignition off,” said Perez. “We’ll be able to quantify the interplay between driver behavior, vehicle performance, and roadway environment in ways that have not been possible before.”
Only cars, pickup trucks, and SUVs were used in the study. Despite the initial thought that motorists would drive with extra caution while being filmed, participants quickly ignored the unobtrusive camera rigs and fell into their regular driving routines. This type of naturalistic driving is considered far more valuable than filming drivers in computer simulators or at closed-off roads designed for research.
“We faced great challenges in convincing more than 3,500 volunteers to let us record their driving for months or years at a time,” said Lee. “This next phase will also be challenging, but it will allow us to leverage the data collected from participants to improve road safety and roadway design.”
Virginia Tech researchers will be using this new data for the first time alongside outside parties. Data can be used to back up previous institute research findings or for future highway studies yet to be conceived.
“This collection is a key data asset in the research portfolio and a cornerstone of VTTI's data repository,” said Gaylord. “By integrating petascale storage and data analysis systems with high-performance computing clusters, the infrastructure fuels the naturalistic driving research program.”