In DARPA race, no driver, no problem

The countdown clock on Team Caltech’s website drives home the fact that precious few days remain to get all of Bob’s systems on line and perfectly synchronized before the day of the big race.

Who’s Bob, and what race is he entered in? None other than the DARPA Grand Challenge, in which 20 self-driving and self-navigating vehicles will sail for 200 miles across the Mojave desert for a $1 million purse. Caltech’s entrant has been nicknamed Bob.

Conceived by the U.S. government’s Defense Advanced Research Projects Agency, the DARPA Grand Challenge pits modified vehicles against each other and the clock. The gist of the contest is that the vehicles are completely autonomous: each vehicle must successfully drive along a predetermined route, navigate itself past obstacles, and reach the goal before all others to win. They have 10 hours to do it.

“It’s a great engineering course, very much in the style of engineering that takes place in industry and in research,” says Richard Murray, professor of mechanical engineering and the team’s sponsor. “It requires not only understanding the big picture but also solving the individual details and problems.”

One of the prerequisites to making it into the contest is passing DARPA’s Qualification, Inspection, and Demonstration phase, taking place the second week of March at the California Speedway in Fontana. That’s where Bob will be put through its paces during its time slots on Monday the 8th and Wednesday the 10th.

“The QID is an obstacle course that contains dirt hills, a tower, and other cars,” says David van Gogh, the team’s project manager and a Caltech staff member. “It has to go down a hill, across a sand trap, a ditch, a cow guard, a straightaway, over a washboard stretch of road, go around boulders, and dodge a moving car.”

Van Gogh, who began working on this project in February of last year, says that teams of students, as well as many volunteers from Caltech, JPL, and the Northrop Grumman Corporation, do all the work themselves.

“The students are learning how to integrate a really complex system—not just work on one part of it—and getting it to all work together,” van Gogh says. They balance their time in the garage with related electrical and mechanical engineering courses and computer science classes.

Several hundred thousands of dollars have already gone into Bob. This doesn’t take into account the brainpower, provided courtesy of Caltech professors and staffers, JPL scientists, and other interested researchers, or the innumerable hours of labor, provided mainly by some 80 Caltech undergrads.

The automobile appraisers at Kelley Blue Book would probably be flummoxed by Bob, which was lifted some three inches for greater clearance and sports Kevlar-belted tires. The interior of the white ’96 Chevrolet Tahoe SUV has been ripped out—who needs a driver’s seat if there’s no driver? The extra space made way for a computer brain, eight IBM computers, a tangle of wires, a generator, and various motors to control the gas, brakes, and steering.

The SUV’s exterior is clad in a roll-cage structure that supports an array of sensors that allow Bob to “see.” Two emit laser beams up to 40 meters in front of the SUV, creating terrain maps that detect obstacles. The images that two pairs of digital cameras capture are output to software that tells the computer brain the proximity of objects around the vehicle. Bob will also use satellite maps of the desert that reveal the relative position of hills, streams, and other geographic features.

“The ability to interpret what’s in the environment: that’s a very humanlike quality,” Murray says. “The DARPA Grand Challenge is pushing us to get machines to do things that humans can do.”

Bob is also equipped with an inertial measurement unit, three accelerometers, three gyroscopes, a magnetometer, and a Global Positioning System antenna. The factory 30-gallon fuel tank will be augmented by a 15-gallon reservoir, which will give Bob a 450-mile range, van Gogh says.

The sensors act like advisors, making suggestions and constantly feeding threads of information to the decision-making software, called the arbiter. The arbiter, in turn, uses the information from each of these threads and “decides” on the best course of action.

To win, or come close to victory, Bob must haul its three-ton body an average of 21 miles per hour across difficult terrain. In the world of autonomous vehicles, this velocity is lightning fast.

An advantage that Team Caltech has over the other teams is its proximity to the desolate desert region where the contest will be held. Scanning a map on his computer, van Gogh traces lines that represent dusty, one-lane roads. A volunteer surveyed some of these trails, he says, to get a feel for what challenges Bob may face, but more work is needed.

“We’re looking for volunteers from the Caltech community to drive a truck outfitted with GPS along these roads. This information will be fed to Bob,” van Gogh says.

If Team Caltech wins, the prize money will be deposited in a Caltech undergraduate student fund. If no team wins, Murray predicts that the team will enter a future grand challenge with a superior vehicle. More details about this venture are available on Team Caltech’s website, located at http://team.caltech.edu/.
•