UC Berkeley researchers test-drive new AI cars
BERKELEY, Calif. - UC Berkeley researchers have teamed up with three car companies and Vanderbilt University on new artificial intelligence software for cars aimed at reducing traffic jams, energy use, and accidents.
One hundred vehicles provided by Nissan, Toyota, and GM were sent onto Interstate 24 in Nashville, Tennessee last week in the middle of morning rush hour traffic.
The cars were equipped with artificial intelligence software or adaptive cruise control, aimed at changing the pattern of traffic along the highway.
"What's so exciting for me is this is the first time one manages to get 100 self-driving vehicles in the same spot and work collaboratively while being able to watch everything from the sky," said Alex Bayen, is a Professor with UC Berkeley's Department of Electrical Engineering and Computer Science.
He says the Circles Consortium Research team chose Nashville because it launched a brand-new traffic surveillance system.
"This is the highest density video camera network in the world to monitor traffic," said Bayen, "Every 500 meters on that freeway, there is a huge ball with 6 video cameras mounted on top of it that essentially live-streams everything it sees to our network."
Researchers estimated that making small changes to 1 out of 20 vehicles could have a big impact on smoothing out traffic flow and avoiding "phantom" backups.
"The natural tendency for a human is as soon as there's a gap in front of you, is to fill it. And that's what causes the problem on the freeways. So what the software does is the opposite. When it sees the gap, it knows that if it accelerates it will create the problem so it actually drives a bit slower," said Bayen.
"It was a team effort by everyone to translate the math into something that could be understood by the vehicle itself," said Maria Laura Delle Monache, assistant professor with UC Berkeley's Department of Civil and Environmental Engineering.
"We do most of our work in front of computers when we do simulation or with piece of papers," said Delle Monache, "And seeing all of that actually translate into the technology itself, so to the vehicle itself it was an amazing feeling."
Researchers hope the technology will help slow down traffic and also reduce the number of crashes.
"An additional benefit of the system is by pacifying traffic you make it smoother and there are a lot of studies that show that if traffic is smoother, then there are less accidents," said Bayen.
The research team plans to analyze the data from the test run in Nashville and hopes to work with car companies to roll out the technology in California within the next few years.
Jana Katsuyama is a reporter for KTVU. Email Jana at jana.katsuyama@fox.com and follow her on Twitter @JanaKTVU or Facebook @NewsJana or ktvu.com.