Robocar race winner: Aging drivers may get self-guided autos

In the not-too-distant future, aging baby boomers may have self-guided vehicles to drive them around when they get too old to get behind the wheel themselves, according to one of the leaders of the team that won DARPA's Urban Challenge race over the weekend.

Chris Urmson, director of technology for Carnegie Mellon University's Tartan Racing team, said he wouldn't let the team's so-called autonomous vehicle drive his wife and children around city streets just yet. But he added that he thinks we're only 10 to 20 years away from having driverless cars motoring around the roadways. And that could be perfect timing for all of the baby boomers who may be losing their ability to drive safely around that time.

"There's a tremendous wealth of wisdom and knowledge in the elderly," said Urmson, just a day after his team won the $2 million first prize at the Urban Challenge finals. "If they can't drive to see their family and friends, we lose that [knowledge] as a society. If they could come out of their house, get in an autonomous car and say, 'Take me to my grandson's house,' that would really be something. We're working on it. It will be a little while, but it's coming."

At this point, even Tartan Racing's winning car, called Boss, is making on-road mistakes that a production-quality car would have to be able to avoid, Urmson said. "When I get in my car, I don't want it to make a mistake," he noted. "For us to get to that level of reliability, we're maybe a decade or a little further out."

This past weekend, Urmson and the rest of the 60-member Tartan Racing team topped rivals from Stanford University and Virginia Tech during the Urban Challenge, which was held by the Defense Advanced Research Projects Agency (DARPA) in an effort to help spur the development of autonomous vehicles for battlefield applications.

In all, 11 robot-driven vehicles, which had been winnowed down from 35 semifinalists, raced at the abandoned George Air Force Base in Victorville, Calif. The cars that competed in the 60-mile race were judged by DARPA officials on both time and how well they performed on the course.

The autonomous vehicles and their self-guidance systems had to find their way through an urban street layout that included multiple lanes, traffic circles and four-way stops, while obeying traffic laws and completing maneuvers such as merging into traffic, responding to blocked roads and passing by oncoming cars on narrow roads. To complicate matters, DARPA's event planners added about 40 human-driven cars to the mix.

"To have 11 self-driving cars out on the road was just amazing," Urmson said. "I, obviously, wanted our team to win. But at one point, I stood near the course, and I saw six robots driving around and interacting, and it was magical."

The stunt drivers whose cars mingled with the autonomous vehicles would "stop on the road and cause the robots to go around them," he said. "They added extra traffic at intersections. [DARPA officials] put an awful lot of faith in the vehicles making smart decisions, and for a large part, they did."

Page Break

In fact, some of the most interesting technologies developed for the race were the tools that enabled the self-guided cars to make those kinds of "smart" decisions, said Urmson, whose team worked on the Urban Challenge for the past 18 months.

"One of the neatest things was the behavioral [technology], getting the robot to act reasonably," he said. "When a person comes up to an intersection and you're not sure who should go first, at some point, one of us would go. You need to get a robot to think, 'OK, it's not officially my turn, but one of us should go, so I'm going to just go.' It needs to act reasonably, and then [be able to] recover and do something reasonable when it gets confused."

Programming that level of artificial intelligence into Boss gave Tartan Racing a leg up on the competition, Urmson added, noting that it helped the vehicle avoid another robotic car that was coming toward it -- partly in the lane that Boss was rightfully occupying.

"Apparently, the other robot was oblivious that it was in the wrong lane," Urmson said. Boss "dodged out of the way" and then stopped short after it swerved toward a concrete barrier -- seeming to think at first that it was stuck, according to Urmson.

"From a human point of view, you would have said, 'Whoa, that was close' and continued driving," he said. "Boss sat there for a minute or two and then backed up and continued driving down the road. It got a little confused and then figured out something to try, and away it went."

The car's ability to figure out that something had gone wrong and what it should do to get out of trouble is "very exciting," Urmson continued. "Obviously, we'd like to build a system that's perfect. But that's not going to happen, so you need to build a system that will recover when it makes mistakes. That was great to see."

Although it may be another 10 years or more before autonomous vehicles are actually driving people to and from the grocery store, Urmson said that radar systems and vehicle tracking technology could start to be incorporated into production vehicles right away. That, he added, could help to curb the average annual count of 43,000 deadly auto accidents that occur in the U.S. alone.

Urmson said that Tartan Racing teamed up with General Motors as part of its preparations for the Urban Challenge, and that automaker has expressed interested in the technology that the team has been developing.

"I know that [GM] is excited about the race and that they see autonomous vehicles as an important part of the future," he added. "It's such a ground-breaking demonstration for robotics. We showed the world what could be done and hopefully opened the door to taking this technology to make cars safer and save lives."