By Danny Freedman
Drowsy drivers cause tens of thousands of crashes each year. Now a GW professor is helping to take the fight into the driver’s seat.
Azim Eskandarian, director of GW’s Center for Intelligent Systems Research, has devised a new method for detecting driver fatigue. The technology, which was recently licensed by a manufacturer for possible use on buses, could be tailored to individual drivers and integrated into a vehicle’s on-board computer.
Dr. Eskandarian and a former doctoral student have a provisional patent and are pursuing a full patent on the technology, which also could be used in cars and commercial trucks.
In 2008, the most recent year for which statistics are available, the National Highway Traffic Safety Administration estimates that drowsy drivers caused nearly 76,000 crashes involving 96,000 vehicles. More than 800 people were killed, and 40,000 others injured.
“We believe that it’s even more than what the statistics show, because after the fact it’s very hard to determine if a person has been sleepy or not at the wheel,” says Dr. Eskandarian. “So it is still a major problem.”
Researchers have studied driver fatigue for decades, exploring it from a variety of angles including measurements of brain waves and heart rates, detection of nodding heads and drooping eyelids and monitoring vehicle steering and speed. In recent years manufacturers have begun adding detection systems to high-end cars.
Dr. Eskandarian’s approach involves a new method for digesting data about a driver’s steering habits and identifying patterns.
Work on this project began a decade ago with his team conducting drowsy-driving tests at the driving simulation lab at GW’s Virginia Science and Technology Campus in Ashburn and analyzing the results along with data from simulations by the federal government.
Of all the driving functions it analyzed—such as accelerating, braking and changing lanes—Dr. Eskandarian’s team found the most significant impact of drowsiness was on steering.
When a driver is awake and alert, he says, much of the focus is on staying in one lane, which is done by constant “micro-corrections” that the driver makes with the steering wheel to stay on course. A driver who is tired, though, has perception delays and slower reactions, which lead to wider, over-compensating tugs at the steering wheel once the driver realizes the car is drifting.
While that might not even be noticeable to the drowsy driver, the steering data “looks more like a zigzag,” says Dr. Eskandarian, rather than the more taut line of data from an alert driver.
So the team built a system that uses sensors on the steering wheel—connected to the vehicle’s computer—that monitor and interpret steering-wheel movement. But in order to differentiate normal driving from drowsy driving, this initial system needed to be fed examples of both. And while impaired driving could be safely simulated in the lab, it posed a stumbling block for use in the real world.
Dr. Eskandarian and then-doctoral student Ali Mortazavi, now an engineer at the California Center for Innovative Transportation at University of California, Berkeley, went back to the drawing board and reworked the technology so that it needs only to collect data on how an alert driver steers; when the driver gets drowsy and deviates from the norm, the system notices the change.
As the concept is further developed, Dr. Eskandarian envisions the technology could be specific not to the vehicle but to the individual driver, much like programmable seat settings. Drivers might push a button in the vehicle to have on-board sensors begin recording their normal driving behavior, and the more data the system accumulates the better able it would be to recognize a problem down the road.