This weekend, 23 robots will line up in a Nevada desert and attempt to race along a 150-mile course rife with sharp turns and steep cliffs.

If there's a winner -- a big if considering that none reached the finish line last year -- the team behind the vehicle will walk away with a $2 million prize and hefty bragging rights.

The competition, called the Grand Challenge, is sponsored by the Defense Advanced Research Projects Agency, an arm of the Defense Department that seeks out new technologies for the military. It attracted 193 teams, most of them from universities and technology companies such as Mitre Corp., a McLean-based engineering firm, and Ensco Inc., a Falls Church government contractor.

The event was created last year to encourage thinking that might eventually lead to road-worthy vehicles that can perform dangerous military tasks -- such as making war-zone deliveries -- and save lives. Most of the vehicles use a combination of global positioning systems and advanced radar to detect obstacles and figure out a route.

Last year, the farthest a vehicle went was 7.4 miles, and this year's course is much harder, DARPA officials say. But, they add, the field of competitors has improved dramatically.

"Last year, we would declare success if we got somebody to start their vehicle," said Anthony J. Tether, DARPA's director. "This year, we have seven teams that have a high probability of making it all the way through."

DARPA officials examined the original 193 and narrowed the field to 43 teams, which went on to a qualifying race this week at the California Speedway. Virginia sent five teams, the most of any state except California, which had 17. Along with Mitre and Ensco, two from Virginia Tech qualified, as did a team comprising engineers from the Charlottesville area.

Gary Carr, Ensco's team leader, said the group has spent 6,000 hours trying to construct a machine that can drive like a human. It took so many nights and weekends that he says -- not quite jokingly -- his family had to come to the shop if they wanted to see him.

Charlottesville's Team Jefferson became a sentimental favorite this week after its vehicle, "Tommy," crashed into a concrete barrier at more than 60 mph, nearly hitting an onlooker and destroying many of its systems. The nine teammates worked 48 hours straight to rebuild, getting Tommy back on the track for a final try yesterday.

"Everyone's got their heart in their throat," Michael L. Woosley, of Team Jefferson, said before the team's last set of runs.

Carr, of Ensco, believes his team's robot, named Dexter, "has the capability to go the full route," though it may be challenged by the 10-hour time limit. He'll have the chance to find out Saturday: Dexter advanced to the finals.

Yesterday, in fact, turned out to be a good day for Virginia's robotics community. Four of the five teams from the commonwealth were named among the 23 that will compete in the big race. (Only Team Jefferson did not.)

Now victory is just a treacherous 150 miles or so away.

From Tech to Finance

Nelson A. Carbonell Jr. is reluctant to admit this, but he wasn't all that up to speed on the latest technologies by the time he got through selling off the assets of his once-high-flying technology company, Cysive Inc.

The decade he spent running Cysive -- a Web development firm that was founded in 1993 in Reston, followed the dot-com boom to a public offering in 1999 and was dismantled in 2003 after demand for its services evaporated -- included more hours looking at financial plans than computer codes.

Starting a new tech firm on that standing wouldn't exactly be prudent, Carbonell concluded. So he did the next best thing: started an investment firm.

Early this year, Carbonell and several former Cysive colleagues quietly launched Snowbird Capital, a $50 million fund made up solely of the partners' money. Snowbird will lend up to $5 million to companies that have some revenue and are growing. The firm takes a 2 or 3 percent equity stake in the company but intends to make most of its money on interest from the loans.

Unlike venture funds, Snowbird isn't looking just for start-ups that have potential for initial public offerings or high-priced acquisitions -- if they appear stable enough to grow and pay the money back, they are candidates for investment. "We're not necessarily looking for home runs," he said. "We're looking to hit a lot of singles."

Ellen McCarthy writes about the local tech scene. Her e-mail is

Ensco's robot, Dexter, will be among 23 finalists vying in the Grand Challenge on Saturday.