The $1 Million Challenge
Published by Florida Times-Union on March 13, 2004.
The qualifying run was going really well.
Then the concrete wall got in the way.
Not the type of thing you want happening to a handcrafted robotic car that, you hope, will soon be driving itself across the desert.
But for the University of Florida robotics experts who have turned a collection of automotive parts, computers, sensors and software into an example of bleeding-edge technology, concrete walls were just one more challenge to overcome.
"We know what went wrong, and the team's in a frenzy right now fixing what went wrong," said Bob Touchton, a Jacksonville resident and UF doctoral student working on the robot.
Fifteen robots -- all that's remaining from the 25 teams selected for the qualification tests -- will be dashing through the sands of the Mojave Desert beginning at 9:30 this morning, hoping to make it from Los Angeles to Las Vegas in less than 10 hours. At stake: $1 million.
The race, known as the Grand Challenge, is sponsored by the Defense Advanced Research Projects Agency, a Department of Defense organization that deals with cutting-edge technology. Through the race, DARPA hopes to spur innovations in the development of autonomous vehicle technologies for military use.
Among the Humvees, motorcycles and other vehicles tackling the course will be the NaviGator, an autonomous car built by a team of scientists at the University of Florida.
Their vehicle might be missing a few parts. There might be some stubborn technical glitches that were never fully worked out. But for the roboticists, being on the course will mean more than getting a shot at a bunch of money; it will give them a chance to see their knowledge put to a real-world test.
"It's the technological challenge of it," said Mary Ahmed, a graduate student from Jacksonville who worked on teaching the robot what the world looks like. "This is exciting to people like me, people who are geeky like me."
If any vehicle is able to complete the race, added doctoral student Sanjay Solanki, it will make the impossible seem possible. "It's a big challenge," he said. "It almost looks impossible."
NaviGator began its life ignobly, as a smashed 1993 Isuzu Trooper, rescued from a junkyard for a couple of bucks. The Trooper is still the vehicle's heart; despite the computer brains, "ladar" eyes and armored shell, the UF robot runs on a gasoline engine and rolls on standard tires.
But that's just what the vehicle looks like. The important parts, the groundbreaking technology, is mostly hidden away inside.
Much of NaviGator, which all told cost less than $100,000, came together in a tiny suite of rooms tucked away at the university's Gainesville campus, amidst stacks on books on computer-aided design and fluid mechanics, worked on by students squeezing in a few minutes between classes and studying for tests.
To bolster their chances, the group from the school's Center for Intelligent Machines and Robotics partnered with Autonomous Solutions Inc. The robotics department has worked with the Utah-based company in the past in an effort to create standardized robot parts, some of which ended up in NaviGator.
In fact, much of what's in the car isn't new, so much as re-utilized. Various members of the team have been creating autonomous vehicles for decades -- machines that do things like discover land mines and help farmers. Participating in the race is way to showcase that experience in front of a wider audience.
"We felt like we had to be there, on the national stage," said Carl Crane, a professor who heads up the UF robot center. "If you're a player in this universe, you have to show that it's not all talk."
'Milestone in robotics'
Despite their experience, though, creating NaviGator hasn't been easy. "We've been working on this a long time. We're building on successes we've had in the past," said Project Manager David Armstrong.
Even the trip from Gainesville to Los Angeles was fraught with difficulty: While transporting NaviGator to the shop in Utah, the team's trailer broke down, stranding them by the side of the road until rescuers arrived -- "The guys all brought computers and worked on the trip," Armstrong said -- and when they finally got to their destination, snow and cold weather wracked havoc with the computer systems.
Those systems can be finicky things: NaviGator boasts a dizzying array of technology, mechanically reproducing in many ways the biological systems humans use when they're driving down the highway.
The navigational system starts out with a general map of the area -- a translation of the terrain into a format that the computer can understand -- which the vehicle can roughly locate itself on through the signals from Global Positioning Satellites.
On a smooth, straight track, that might be enough, but the Grand Challenge will wind its way over cliffs, across gullies and next to brush, requiring the vehicles to be able to pick the best place to drive, and will feature obstacles the cars must navigate around. To handle these tasks, the team outfitted NaviGator with sensors such as ladars, which use light beams to scan the area in a way similar to radar and cameras that the vehicle can use to differentiate between roads and dirt.
These data are then fed through software programs -- the brains of the vehicle -- that decide where the vehicle should go and at what speed. Once the start button is pushed, the human teams have no control over the machine.
Seeing that combination of technology in action, Armstrong said, will be like watching Charles Lindbergh cross the Atlantic Ocean. "I compare this to the Spirit of St. Louis," he said. "That event marked a milestone in aviation. This race will be a milestone in robotics."
Two hours before the race today, the participants will be given details on the 210-mile route their vehicles have to follow. Preliminary reports Friday pointed to a relatively easy 5-mile start leading into a mountaintop route featuring hairpin turns and sheer drops from the side of the 10-foot-wide road.
But even before they got out West, the roboticians knew the race would be tough. "Our strongest competitor is the desert," Armstrong said. "Rocks, hills, sand, going through water. We're going to give it our best shot. A lot of fingers will be crossed."
But the chance to see their creation in action has been enough to keep them going.
"It's almost too good to be true," said Touchton, who's doing his doctoral research on real-world robotic applications. "My personal learning has been incredible. It's hard to create something like this in a university environment, but everyone knows that this is the real deal."
Touchton, who handled logistics for the team, said he doesn't think anyone will claim the million dollars, an outcome he has no problem with.
"I'm counting on no one winning this year," he said. "I'm looking forward to tackling these problems next year."
If no one wins the race, DARPA has said, they'll hold it again, probably annually, until they give the money away. That might be the best shot for NaviGator.
"We have a solid plan, but could use some more time," Armstrong said. "If we go again, the question will not be if we win, it will be by how much."
ROBOTS: FACT VS. FICTION
Robots, or the popular conception of them, have changed a lot since playwright Karel Capek coined the term, based on the Czech word for "serf," in his 1921 play R.U.R.
(The idea of artificial men has, of course, been around much longer, dating back to Jewish golems and ancient Greek statues that came to life.)
Like with so many fantastic ideas, robots have become much more prosaic during the translation from literature to reality: Rather than the friendly, beeping R2-D2, robot companionship can be found with Sony's mechanical dog Aibo. The fake chess-playing robots of the 1800s have given way to machines like Big Blue that play the game at a stratospheric level but have no arms. And rather than the sleek mechanical sexiness of Fritz Lang's Hel, real robots look functional, industrial.
But is the change in vision a bad thing? Sure, George Jetson's Rosie isn't running a vacuum around the house, but the Roomba -- a pizza-sized robot that can roam the house picking up dirt -- does just as good a job and isn't as likely to end up sulking in the kitchen. And though creating functioning legs has proved a difficult challenge for robot makers, that's offset by the fact that wheels and treads often provide better manoeuvrability.
Our robots don't come equipped with the Three Laws of Robotics (which, in Isaac Asimov's books, make the robots obey humans and stopped them from harming themselves or others) but that, quite simply, is because there's no need for them.
"I don't think any machine in the world," said David Armstrong, the project leader for the University of Florida's entry in the Grand Challenge, "even comes close to being human."