The robots' great race: a $2 million DARPA derby in the desert

Double your money, double your fun. That rationale seems to work for the Defense Advanced Research Projects Agency (DARPA). When the $1 million Grand Challenge 2004 prize for autonomous navigation of a 150-mile course went unclaimed, the agency boosted the 2005 Challenge purse to $2 million. By the February 1 initial application deadline--only the first hurdle in a long steeplechase to the October 8 finish--195 teams had signed up for the contest, nearly twice the 106 applicants who entered in 2004.
DARPA will award the $2 million cash prize to the team that builds a completely autonomous ground vehicle that travels the fastest time in under 10 hours across approximately 150 miles of mountain desert roads and trails. The agency launched the competition in 2003 to stimulate creative thinking and lead to the development of a vehicle capable of resupplying troops in the battlefield. The Defense Department has set a goal of having a third of combat vehicles operating without drivers by 2015.
The initiative stalled with conventional large contractors, but the Grand Challenge idea has sparked innovation from smaller, less encumbered groups. 2005 entrants come from 37 states and three foreign countries. Thirty-five university-based teams and three high schools teams have toed the starting line.
The next qualifier for this year's contestants falls on March 11, when teams must provide DARPA with more detailed information about their entries, including a vehicle specification sheet and a video demonstration of their vehicle in action.
Teams that pass the vehicle and video review will receive an in-person site visit in May. DARPA will then select 40 vehicles to advance as semi-finalists to the National Qualification Event in Fontana, California, September 27-October 6. There, semi-finalist vehicles will go through a series of competitive tests. DARPA will select 20 finalists to compete in the October 8 Grand Challenge event, on a 150-mile course across a South-western desert. Race start and endpoints will be known in advance, but racing teams will not learn the exact track their vehicles must follow until the early morning hours of race day.
Last year's entrants integrated an array of technologies, in all cases centered around GPS. Ancillaries included inertial navigation systems, laser rangefinders, Lidar systems, radar, stereo video, satellite imagery, and mapping suites. In most cases, GPS technology functioned as expected. The strongest entry completed 7.2 of the 146 miles before coming to grief on a switchback turn above a steep drop. Platform rocking and jarring from the rugged terrain, disturbing sensor functions, may have been the prime factor that defeated most contestants.
DARPA expressed great satisfaction with the '04 event and mounting excitement over this year's tourney. After all, they're getting a bargain: cutting-edge technology R&D at a fraction of the cost of normal channels. Some observers estimate that the teams themselves--for pride, prestige, or developmental reasons of their own--spend more than the prize money they might take home in getting their robotic vehicles to the starting line.