The University Record, March 12, 2001

Researchers working with motorists to fine-tune crash warning systems

By Nancy Ross-Flanigan
News and Information Services

With more than 1.8 million rear-end crashes occurring every year in the United States—accounting for some 2,000 deaths and 800,000 injuries—automobiles equipped with forward collision warning systems could be a boon to highway safety. But when should such a system signal a driver that a crash is likely?

If the system waits too long, the driver may not react in time to prevent an accident. But if an alarm goes off every time there’s even a remote chance of a collision, the driver may get annoyed and ignore the warnings or disconnect the device. Figuring out the timing is tricky, but experiments by U-M researcher David LeBlanc and colleagues are helping to get it just right.

LeBlanc, an assistant research scientist with the U-M Transportation Research Institute (UMTRI), presented results at the Society of Automotive Engineers’ 2001 World Congress March 6.

In the experiments, which were done when LeBlanc was with the Ford-General Motors (GM) Crash Avoidance Metrics Partnership, researchers tested 108 “off-the-street” drivers at the GM Milford Proving Ground. For each test, the subject drove an ordinary vehicle, following behind a “surrogate target”—a molded composite mock-up of the rear half of a passenger car, mounted on an impact-absorbing trailer that was towed via a collapsible beam. This setup allowed for the possibility of realistic rear-end crashes without endangering the drivers.

In the first part of the test, drivers were told to wait until the last possible moment, then brake to avoid crashing into the surrogate target when it slowed down or stopped. They were instructed to use either “normal,” “comfortable hard” or “hard” braking pressure. From this baseline information, the researchers developed a model of last-moment braking decisions that helped them calculate what seemed like the best time for a crash alert system to give its signal—after most attentive drivers would have hit the brakes, but in time to still avoid a collision.

Subjects then tested the timing in cars equipped with forward crash warning systems. In some tests, drivers were told about the system and instructed to brake when they saw or heard the warning signal—an icon on a display accompanied by a sound. In “surprise braking event” tests, drivers were unaware that the car had a warning system until the first time it alerted them. Because most rear-end collisions happen when drivers are distracted or simply not paying enough attention to driving, researchers rode along with the subjects and diverted them by chatting, peppering them with questions or asking them to search for a nonexistent indicator light on the instrument panel.

On average, the drivers rated the timing of the warning as “just right.” And it did come in time to help them avoid accidents. In the surprise braking event tests, 104 of 108 drivers responded to the alarm in time to avoid crashing into the surrogate target.

“We were pleasantly surprised that there does seem to be this zone where you can wait until most drivers would have normally hit the brakes, but still have time to help people avoid accidents,” says LeBlanc, who collaborated on the work with Raymond Kiefer and Richard Deering of GM, Michael Shulman and Melvin Palmer of Ford, and Jeremy Salinger of Veridian/ERIM International.

The next step is to test crash-avoidance systems in the real world. For that, LeBlanc and other UMTRI researchers are working with GM, Delphi Delco Electronics Systems and the U.S. Department of Transportation on a five-year, $35 million project. GM will supply a fleet of Buick LeSabres outfitted with sophisticated multi-sensor crash-avoidance technology, and UMTRI will begin testing the cars with a small group of drivers in fall 2001. In 2002, UMTRI will oversee a larger test of 120 drivers, who will use the cars as their own for two–four weeks each.