(FORTUNE Small Business) – Brian Gregory lives on a quiet, family-friendly street in Newbury Park, a leafy, lazy California suburb. He sports a beard, geeky glasses, and a small paunch. He does not, by any stretch, look like a guy who'd possess, say, a Special Forces Prowler assault vehicle with LIDAR sensors, waterproof laptops, and actuator servo-motors. And yet there it sits in his garage, a dune buggy on steroids, right where you'd expect to see the minivan. Gregory--working with a group of teammates assembled under the name Sciautonics--hopes that on March 13 the Prowler, guided by onboard computers, will drive itself from somewhere near Barstow in Southern California through the desert to Las Vegas. If it completes the circuit faster than other, enemy robot cars, Sciautonics will win a million bucks from the Department of Defense. But to do so, it will have to beat out about 19 other teams--from small companies and university departments, and one made up of precocious students from Palos Verdes High School. "It is not easy," says Sciautonics team leader Reinhold Behringer. "The logic of this is easy, but to implement it ... There are so many variables."

The million-dollar prize is almost beside the point. The Sciautonics team hopes that a good showing in March will result in lucrative contracts from the Department of Defense and the automotive industry, which would help turn the ragtag group of volunteer engineers into a legitimate small company.

And Sciautonics is not alone. Down the coast in Irvine, Team Cyberrider has been working on high-tech water pumps that could move robotic parts. "Robotics is the future," says leader Ivar Schoenmeyr, who sees massive economic potential in the field and a possible new business stemming from his work on the race. "The robotic revolution will be like the Industrial Revolution."

And in Westlake Village, Calif., just a few towns over from Gregory's garage, Bill Kehaley has assembled a group of programmers, cartographers, and even buxom blond models to form Axion Racing. Kehaley hopes his team will create some salable technology, and he also would like to use his vehicle as a marketing platform for Melo Water, a spring water shipped from Micronesia. (Kehaley owns the U.S. distribution rights.) Before race day he plans on releasing what he refers to as the "world's first autonomous robot vehicle calendar." Did we mention that the blonds are twins?

If it sounds strange that a robot race might launch so many companies, it's even stranger that the prize money is coming from the U.S. government. But it all makes sense--sort of--once you realize that DARPA, a small bureau within the Department of Defense, is involved. DARPA, which stands for the Defense Advanced Research Projects Agency, was established in 1958 with the sole mission of funding bizarre experiments. DARPA-backed research has led to the Internet, stealth aircraft, and--a less celebrated moment--the terrorism futures market (in which traders were to calmly place bets on the specifics of the next horrific attack). In the past, DARPA-funded studies have helped create such companies as Cisco, Sun Microsystems, and Silicon Graphics.

DARPA announced the Grand Challenge for Autonomous Ground Vehicles in July 2002 to spark new technology that could be used in the military. On the day of the race, DARPA will distribute a map of the route, which the competing cars will then have about ten hours to navigate, with no human assistance. (That means no remote control--except for the kill switch DARPA mans in case one of the vehicles endangers someone.)

The military uses for this kind of technology are far-flung. "These cars could be used for surveillance and reconnaissance, going into environments that are too hazardous for people," says DARPA spokesperson Jan Walker. "They could carry things on the battlefield. Or they could be used for search-and-rescue missions."

But the Department of Defense isn't the only audience for the technology. Detroit may be watching too. Behringer previously worked with DaimlerChrysler on road-recognition sensors, which the company thought could be used in collision-avoidance systems. General Motors also has a group working on collision avoidance, so the market for those ideas is already established. "We've been testing vehicles equipped with collision-warning technology and adaptive cruise control," says GM's Jim Schell, "which could alert you to slowed or stopped vehicles ahead and even let the car automatically slow down itself." The potential is staggering, going beyond collision avoidance to the Holy Grail: fully autonomous driving. Imagine leaving a rental car at the airport gate and letting it return itself. Or climbing into your car at home, taking a nap, and waking up in the parking lot at work. "We dream of a day when kids get a car on their first day of kindergarten," says Behringer, "and it drives them to school, ballet, and soccer practice, so Mom and Dad don't have to."

In an empty parking lot a few miles from Gregory's garage, that goal seems far away. Bob Addison and Wayne Gothmiller, two Sciautonics team members, are slowly guiding an old golf cart around the pavement. A LIDAR sensor--a sort of laser-mapping device--is mounted on the front of the cart, hooked up to a laptop computer in the passenger seat, connected in turn to a small generator. They drive a few feet forward, crawling along, then stop and check the computer screen to figure out what the sensor saw. On race day the technology on this puttering, halting golf cart will need to guide the Prowler as it bounces through riverbeds at 60 mph. Nobody said building a robot car would be easy.

Or cheap. Sciautonics expects that it will cost about $250,000 for the team to put together its vehicle. And unlike the teams from Caltech, Virginia Tech, and Carnegie Mellon, members can't rely on large amounts of university funding (though they have done some work with a group at Auburn). Instead, they've devised a plan whereby anyone who donates cash or equipment will get membership certificates (essentially stock) in Sciautonics, commensurate with the size of the donation. One current major equity holder is ATV Corp., which makes the Prowler and donated it to Sciautonics. (If Prowlers were for sale to the nonmilitary public--they're not--they might cost around $25,000.) But by far the biggest investor is Rockwell Scientific, where many of the Sciautonics crew still work. Rockwell has donated money, facilities, and equipment, and of course much of the intellectual capital comes from moonlighting Rockwell employees. In return, Rockwell holds more than 50% of Sciautonics' membership certificates, though both Rockwell and Sciautonics would like to see that stake cut by at least half when new investors come onboard.

Sciautonics, naturally, feels that it is a very attractive investment. "I think we already have enough here to form a company if we want to," says Behringer. If it creates some new methodology or line of code, Sciautonics will have instant value, and the team will have to make major decisions. First will be to grant membership certificates to team members in chunks that reflect what each person contributed to the effort. "There will be a lot of heartache over that," predicts one team leader. Then those on the team might decide to quit their Rockwell jobs and go full speed ahead with Sciautonics as a small business. Or they could sell Sciautonics and its intellectual property. In that case, one principal suitor might be Rockwell itself, which could fold in the team under a new department heading. Rockwell already understands that if that occurs it will be paying market price to buy the company (less, of course, the shares it already owns).

All that entrepreneurialism is heartening. But it's also a bit befuddling. There was nothing to prevent Behringer from building a robot car any time he wanted. Why wait for DARPA and its piddling $1 million to get started?

As it turns out, the contest format has helped Sciautonics combat a number of hurdles common to young businesses. Most traditional startups, for instance, have trouble attracting press. But DARPA teams have already been mentioned in feature stories by Wired News and the New York Times. Japanese TV network NHK has begun filming a documentary. The narrative power of a robot race has brought journalists out in droves. (Ahem.)

Then there's the matter of motivating staff. That, of course, wasn't a huge problem a few years ago, when every small business was counting down to its IPO. But in the post-boom years, most startups don't have such a definable endpoint. When there's no goal in view and no clear target date for success, and the work stretches on with big rewards far out of sight, it's tough to maintain employee loyalty. But race day provides a handy cutoff point: Sciautonics workers can come in the day after the race, reassess the prospects of the company, and instantly exit stage left if they decide there's no future. They'll have devoted less than a year to it, mostly on weekends, and could chalk it up to the fun of joining a team and solving a puzzle. As Kishor Bapat, Sciautonics' business manager, puts it, "It's a lot less stressful than a regular startup. It's not like if I fail, my kids won't go to college."

In part, that's because he kept his day job. Which addresses perhaps the No. 1 pitfall for startups: payroll costs. Those have been no problem for Sciautonics. That is, there are none. Every employee is a volunteer, often contributing 20 to 30 hours of work each week, just for a chance at the thrill of victory. "With no salaries to pay, it's a whole different ballgame," says Bapat. As a tech veteran, Bapat has worked on several startups, but none like this one. "We expect about $800,000 worth of free, highly skilled labor, right up until the race in March," he says.

From DARPA's angle, the Challenge format would seem to be a no-brainer. DARPA has spent millions and millions over the years funding autonomous-vehicle projects with only a few research groups. The Grand Challenge puts up just $1 million (plus race administration costs) yet has already brought to life scores of new research teams, each working on the problem with its own approach--and its own money. "We wanted to cast a wider net," says DARPA's Jan Walker. "With the Challenge format, we can get around some of the accounting standards the DOD would require for a grant proposal, and teams can get involved without acquiring a mastery of all the paperwork and regulations."

So if you're going to bet on this race, which team should you put your money on? Probably none. Most of the teams don't expect anyone to maintain the 25 mph pace necessary to complete the challenge. But in the end, whether Sciautonics' Prowler growls to victory or no one even completes the course, it won't really matter. The real prize here lies beyond the finish line.