In Driverless Cars, Champion Racing Skills Will Come Standard

Stanford's robotic racecar, Shelley.
Stanford’s robotic race car, Shelley.

An Audi TTS tears around a California racetrack hitting speeds of 120 miles an hour and finessing turns like an elite driver. And it is an elite driver. Just not a human one.

Stanford Dynamic Design Lab’s robotic racer, dubbed Shelley, has been approaching the performance of human race car drivers for a few years now. In 2012, it was a few seconds off the times of expert drivers. These days, it’s neck and neck.

The car, using GPS to pinpoint its position on the track, is piloted by an onboard computer and advanced algorithms. Unlike other driverless cars, like Google’s, Shelley isn’t terribly flexible. It’s been programmed to drive a specific track (Thunderhill near Sacramento) and can’t avoid unexpected obstacles.

But that’s okay. Shelley and team are digitizing the skills of the best drivers. And here’s the cool bit: like anything else digital, once you’ve written the program it’s easily transferable. If one robot car can drive like Michael Schumacher or Dale Earnhardt Jr.—they all can. No years of practice required. Just copy the code.

Shelley has been racing David Vodden, an amateur racing champion and CEO of Thunderhill racetrack near Sacramento. In a recent time trial competition, Shelley beat Vodden by 0.4 seconds. The latest robot to beat a human? Stanford’s Chris Gerdes cautions it isn’t a perfect “Deep Blue v. Kasparov” moment yet—but still pretty impressive.

“I can start this competition at a different point on the track, and David wins by 0.4 seconds, so there is still not a clear victory here,” Gerdes says. “But the point we wanted to make is that we’ve gotten fairly comparable to an expert driver in terms of our ability to drive around the track.” (Gerdes narrates footage of Shelley at a AAAS talk in San Jose below.)

Gerdes and his team have improved Shelley’s abilities by studying the brains of top racers (using EEGs) and plotting their decisions as they lap the track. They’ve found the most experienced human drivers can make split-second, intuitive, out-of-the box decisions to shave tenths of a second off times.

Vodden calls this skill “butt sense.” And it’s an open question whether computers will ever pair brute force, by-the-book logic with flexible, creative problem solving like ours. Vodden has his doubts.

“A really motivated race car driver is willing to ‘bet the car,'” he said. “I think Shelley is precise, whereas a race car driver will do whatever it takes, including betting the car.” Though the list of things computers will never do is dwindling, he may be right. But beyond the track, those tenths of a second matter less.

Why? Because although the vast majority of human drivers will never negotiate a turn or tap the brakes like a professional driver, near approximations of these skills will come stock in future driverless cars.

High-performance maneuvers like those employed by the top drivers can translate into everyday driving, especially when evasive action is needed—dodging an unexpected car or pedestrian in the road and navigating ice or snow.

So, what matters more than driverless cars beating elite race car drivers is that they vastly surpass average drivers. And this is probably the best argument for driverless technology. Put simply, who would you prefer taking the wheel in a risky situation: You, your elderly neighbor, that teenager up the road—or Mario Andretti?

Image Credit:

Video courtesy of Stanford University, edited by AAAS/Carla Schaffer

Jason Dorrier
Jason Dorrier
Jason is editorial director of Singularity Hub. He researched and wrote about finance and economics before moving on to science and technology. He's curious about pretty much everything, but especially loves learning about and sharing big ideas and advances in artificial intelligence, computing, robotics, biotech, neuroscience, and space.
Don't miss a trend
Get Hub delivered to your inbox