The odds-on favorite to win the today's s high-tech robotic vehicle race in Southern California blew an engine after it traveled just seven miles from the starting line. Most of the other vehicles competing for a $1 million prize in the Pentagon's "Grand Challenge" race failed to travel more than a few hundred yards from the starting point near Barstow, Calif.
Sandstorm, the Humvee modified by a team from Carnegie Mellon University in Pittsburgh to compete for the $1 million prize, had impressed race observers during qualification rounds earlier this week at California Speedway in Fontana. But its exit from the competition reflects the enormously difficult challenge of building a vehicle smart enough to navigate across hundreds of miles of desert landscape.
Both Virginia-based teams are also out of the running today. The modified Honda all-terrain vehicle assembled by Team ENSCO from Falls Church only made it a few hundred yards out of the launching area before it flipped over. The golf cart entered by a team from Virginia Tech University made it to the edge of the launching area before its breaks locked up. Several other vehicles hit retaining walls or fences.
Fifteen vehicles qualified for today's race. The vehicles set out early this morning from an area near Barstow, Calif. Teams got course information about two hours before the race. Waypoints -- a series of global positioning system coordinates -- were programmed into onboard navigation systems and the vehicles set off independently.
As of 8:30 a.m. local time, eight vehicles were already out of the race due to technical or other problems. Two vehicles had yet to start out on the course, while five others were still attempting to navigate the course, which begins in Barstow and ends just across the California state line in Primm, Nev.
The race sponsor, the Defense Advanced Research Projects Agency, received more than 100 submissions from teams wanting to participate. That list was winnowed down to 25 teams, comprised of engineering and R&D firms, colleges and volunteers. Of the 21 teams that attempted to qualify over four days of trials, just seven completed a flat, 1.36-mile obstacle course at the California Speedway.
More than two decades ago, military efforts to research autonomous technology produced large, slow vehicles that could only traverse flat terrain. California officials also successfully tested an automated highway system in San Diego in 1997. But a vehicle that can move quickly over a variety of landscapes and around or over natural and man-made obstructions has remained elusive.
In sponsoring today's race, DARPA is responding to a congressional mandate that one-third of U.S. military operational ground combat vehicles by unmanned by 2015. For example, robotic vehicles one day could run supply routes, eliminating the threat to drivers and security personnel assigned to vehicle convoys.
The agency spent $13 million on the race. It estimates competitors laid out four to five times that amount developing their entries, which rely on global positioning satellites as well as a variety of sensors, lasers, radar and cameras to orient themselves and detect and avoid obstacles.
Commercial considerations also sparked some teams to participate in today's race. Scott Gray, a spokesman for the Carnegie Mellon University team, said he envisions vehicles one day that could be programmed to let blind people travel independently.
Autonomous vehicles make decisions based on their knowledge of the terrain. If a vehicle's cameras or radar detect an obstacle, onboard computers make decisions to go around, or back up, or change gears before moving toward the next waypoint.
The DARPA Grand Challenge did not escape controversy in the planning stages. Axion Racing, based near Los Angeles, earlier this month objected to a rule change that allowed humans to refuel the vehicles if they wound up spending the night in the desert. Axion team leader Bill Kehaly said entrants with larger vehicles would benefit from the revision.
And one of the 25 teams selected to participate in the challenge, Northern California-based Team Overbot, dropped out in February. Overbot's John Nagle said he ran out of time to complete his vehicle, noting that an improving economy in Silicon Valley late last year took away many of his volunteers.
Nagle also questioned DARPA's decision to increase the number of course waypoints. He says a heavily preplanned approach "doesn't lead anywhere," saying the technique was proven in the California highway test in 1997. The current level of waypoints favored Carnegie Mellon's Red Team, Nagle said.
But Gary Carr, team leader from ENSCO, a Northern Virginia engineering firm, said the route is not simply a matter of "connecting the dots." He said the vehicles will have to do a lot of their own thinking on the course, noting a lot of turns can happen in the quarter mile average distance between waypoints.
No comments:
Post a Comment