In urban and suburban areas alike, it’s an all-too-familiar accident scenario: Cars zipping along a road; suddenly, a pedestrian or bicyclist veers into traffic.
In the future, a warning system could alert drivers before a collision occurs between vehicles and walkers or cyclists.
An ambitious new project led by Kassem Fawaz, an assistant professor of electrical and computer engineering at the University of Wisconsin-Madison, aims to create such an alert system. With a $924,000 grant from the U.S. Department of Transportation’s Federal Highway Administration, he and his collaborators are leveraging the power of mobile computing to make the streets safer for everyone.
“The core idea is relying on devices that people already use, like smartphones or wearables, to monitor their mobility and to predict when there could be a potential conflict with other road users,” says Fawaz.
The researchers plan to create an app that gathers GPS location and motion data from the accelerometers and sensors already present in most people’s cell phones or smartwatches, and then leverage the power of machine learning algorithms to identify subtle movement signatures that predict unexpected pedestrian or cyclist actions such as jaywalking or swerving across traffic.
When the app anticipates a potential conflict, it will send out messages to nearby drivers, warning them to be extra-cautious—especially on stretches of road between traffic lights and crosswalks. A gentle notification could remind drivers to keep their eyes open, and might help prevent some of the more than 5,000 pedestrian deaths that occur every year in the United States.
“Most of these accidents don’t happen at intersections,” says Fawaz. “People are less attentive when they’re just driving down the street.”
Crucially, the system won’t require any new hardware or infrastructure; all the analysis and messaging will occur on people’s own devices.
And in case people might feel squeamish about an app that tracks their movements, the scientists are building strict privacy protections into the system.
“All of the processing will stay on the device,” says Fawaz, “There’s no server that will collect your data.”
Security also is a primary concern, and the scientists are incorporating strategies to ensure that bad actors won’t be able to jam up the system by sending fake messages, which could snarl traffic.
Fawaz and colleagues plan to demonstrate a test of the platform on the roads of Madison within three years.
His UW-Madison collaborators include Parmesh Ramanathan, a professor in electrical and computer engineering; and Associate Professor Soyoung Ahn, Assistant Scientist Madhav Chitturi and Arthur F. Hawnn Professor David Noyce, all of the Department of Civil and Environmental Engineering.