UI Researchers Use Car Sensors to Detect Drunk Driving
University of Iowa News Release
Researchers at the University of Iowa have moved one step closer to the day when built-in, drunk driving devices in cars and other vehicles will be as common as today's anti-lock braking systems.
A two-and-one-half-year-long study conducted at the National Advanced Driving Simulator (NADS), a research unit of the University of Iowa College of Engineering, successfully demonstrated the feasibility of a non-breathalyzer, vehicle-based system to detect alcohol impairment based on driver behavior. The U.S. Department of Transportation's National Highway Traffic Safety Administration (NHTSA), which funded the study, has since awarded the researchers a new $1 million contract to explore the possibility of using similar systems to detect driver drowsiness
The alcohol study was successful in several ways, according to Timothy L. Brown, co-principal investigator and NADS driver impairment program manager.
"The goal of the alcohol study was to demonstrate that existing sensors on a typical vehicle can be used to detect impaired driving from alcohol," said Brown. "These are common sensors that look at how drivers steer and use the accelerator and brakes -- and they do not include breathalyzer technology."
"The project was a good example of NADS leading the research activity and working with experts around the country at various institutions, including the University of Wisconsin, Montana State University, Southern California Research Institute, and Pacific Institute of Research and Evaluation," said Omar Ahmad, project manager and NADS assistant director. "This research also provides a concrete example of the ground-breaking research that the NADS was designed to conduct -- research that could not have been safely and ethically conducted with an on-road naturalistic study."
Here's how the study worked:
--Participants were required to take simulated, 24-minute drives home from a bar using three types of roadways: rural, urban and interstate.
--Some 108 drivers in three age groupings (21-34, 38-51 and 55-68) took part at three blood alcohol concentrations (BAC) -- 0.00, 0.05 and 0.10 percent.
--The data collected from the scenarios supported the development of three algorithms, or mathematical formulas. One algorithm used speed and lane-keeping measures, another was based on cues that police officers often use to identify impaired drivers, and a third used support vector machines (a data mining technique).
The results showed that it is possible to design a vehicle-based system that will detect alcohol impairment based only on driver behavior. In addition, the algorithms differentiated between drivers with BAC levels at and above 0.08 percent and below 0.08 percent with an accuracy of 73 to 86 percent, a result comparable to a standardized field sobriety test. This accuracy can be achieved with approximately eight minutes of driving performance data under ideal road conditions. Both algorithms that are general in their application and those that are specific to individual drivers show great promise; however, individualization rather than the "one size fits all" approach will allow algorithms to be tailored to individual driving styles and varying roadway situations.
Although it could be years before such technology will begin showing up in showroom vehicles, Brown noted that the research findings are an important early step. Statistics show that about 31 percent of all U.S. traffic fatalities involve alcohol-impaired driving.
"The ultimate aim of the impairment detection algorithms is to support interventions that guide the driver to safer behavior," he said.
He said that a new, $1 million NHTSA-funded follow-on study on designing algorithms to detect driver drowsiness likely won't be the last of its kind. "We anticipate that this research will be the foundation of several future studies looking at impairment from drugs and other medical conditions," he said.
Brown's colleagues on the study, in addition to Ahmad, include John D. Lee of the University of Wisconsin-Madison, Dary Fiorentino of the Southern California Research Institute, Michelle L. Reyes of the University of Iowa's Public Policy Center, James Fell of the Pacific Institute of Research and Evaluation, and Nic Ward of Montana State University.
The full report can be viewed online at:
STORY SOURCE: University of Iowa News Services, 300 Plaza Centre One, Suite 371, Iowa City, Iowa 52242-2500
MEDIA CONTACT: Gary Galluzzo, writer, 319-384-0009, firstname.lastname@example.org