Driving on Autopilot

Imagine driving down a street at rush hour. It’s a typical commute, but this time, you’re reading a newspaper in the backseat. The driver’s seat is empty—your car is driving itself. Sounds like a fantasy, right? After all, even The Jetsons’ flying car had a driver.

Wrong. Self-driving cars are the future, says UT computer science professor Peter Stone. “We have the technology—legal and culture changes are what will take time,” Stone says. “I wouldn’t be surprised to see autonomous cars on the road within 25 years.”

In Austin, a robot car is already on the road — the closed-off roads at UT’s Pickle Research Center, that is. His name is Marvin, he’s a 1999 Isuzu with a cool quarter million in hardware, and he’s getting students hooked on computer science.

One recent afternoon, about 10 people gathered at the research center to take Marvin for a spin. Piyush Khandelwal, a computer science doctoral student, teaches a two-semester Freshman Research Initiative course that lets undergrads program the car. A full year’s work has boiled down to this day. It’s time to demonstrate their final projects.

Khandelwal and his students are joined by Don McCauley and Jack O’Quin, engineers with Austin Robot Technology. Their company owns the car, which UT leases for the course.

Sophomore Corbin Montague’s project is to get Marvin to detect an obstacle and maneuver around it. A plastic cone in the left lane symbolizes a parked car.

O’Quin—whose long white beard, Tito’s Vodka cap, and faded jeans are more cowboy than computer nerd—climbs in the driver’s seat, with Montague at his side. They use a converted PlayStation joystick to set the car in autonomous mode.

Everyone stands back as the two make final adjustments. A huge laser scanner, called a velodyne, spins rapidly, giving the car’s six computers feedback on the locations of objects.

At about 15 miles per hour, the car drives toward the cone. There’s something eerie about watching the steering wheel turn on its own as engineer and student sit unmoving.

Then, just a few feet from the obstacle, the car veers into the ditch—the fault of a GPS error.

“I tested this a lot in simulation,” Montague explains, “but more can go wrong in the real world.”

The real-world factor, Stone says, is what makes the robot car a great teaching tool. “It’s tangible and it’s fun,” he says. “And it gives students a taste for how research really works.”

But in 2007, when Stone first proposed that UT undergrads work on the car, not everyone agreed.

“Researchers at other schools thought I was crazy for letting undergrads work on the car,” Stone remembers. “They said, ‘You shouldn’t let 19-year-olds near this thing with a 10-foot pole.’ But it has worked out wonderfully.”

Montague says that working with the car solidified his interest in graduate school. “No other class allowed me to do research as a freshman,” he says. “Now I know how research works, and I’m going to continue. I definitely want to get my master’s, maybe a PhD eventually.”

The numbers back him up: The 500 UT students who join a Freshman Research Initiative course each year are 23 percent more likely than their peers to attend graduate school.

And what other project gives students a front seat to technology that could revolutionize our roads? “More than 90 percent of auto accidents are caused by human error,” Stone says. “Autonomous cars may save lives.”

Photo by Matt Wright-Steel


Tags: ,


No comments

Be the first one to leave a comment.

Post a Comment