Robots Meet Reality
Andy Greenberg , 11.08.07, 6:00AM ET
Drive all by themselves. But TerraMax’s autonomous driving technology could save lives by doing more mundane chores, including automatically following another car in a convoy or providing a warning system aimed at preventing a human driver from making dangerous mistakes.
TerraMax is the largest–and easily the most terrifying–of the 11 robotic vehicles that participated in the final race of DARPA’s Urban Challenge in early November, a milestone event that showcased the robotic cars’ ability to follow complex routes and negotiate traffic completely under their own control through 60 city miles. (See: “Viva La Robot Revolution!”) The race, sponsored by the Defense Advanced Research Projects Agency, the Pentagon’s research wing, offered $3.5 million in prizes designed to springboard the robotics industry and help fulfill Congress’s ambitious mandate that one-third of all military vehicles be unmanned by the year 2015.
But the race also underscored how far away that goal still is: At one point, two robotic SUVs collided. Another mistook a driveway for a road. TerraMax itself came within inches of plowing into a concrete pillar and had to be taken off the course.
Taken together, all of these imperfections prove to many roboticists that the dream of a totally driverless fleet of military vehicles is still too complex–both technically and politically–to be more than science fiction. But that doesn’t mean it’s a waste of time. What DARPA’s race really demonstrated, they argue, is that robotic driving technology is ready to work together with human drivers–not to replace them.
“This was a fun event, but it clearly shows that the world is not ready for autonomous driving,” says Sebastian Thrun, the head of the Stanford team whose robotic Passat, “Junior,” took the competition’s second-place prize. In the near term, Thrun says, these autonomous driving technologies should be put to work in warning systems and automatic stopping controls, devices that he says could reduce the 95% of vehicular deaths that are caused by human error. Thrun points out that more than 42,000 automobile casualties occur in the United States every year. “It’s a number that keeps me up at night,” Thrun says. “If we could cut that in half, it would be an incredible achievement.”
The key to applying imperfect robotic technology to present problems, says autonomous-driving researcher Jay Gowdy, is to combine humans’ ability to understand their surroundings with a robot’s ability to measure and react consistently.
“We’re not building autonomous chauffeurs,” says Gowdy. “We’re building robotic horses.” Like a horse, Gowdy says, a robotic car of the near future might control much of the moment-by-moment decision-making that goes into getting from point A to point B. But if the robotic car were “spooked,” he says, a human driver could take control.
That kind of robotic integration is well on its way. Gowdy works for Natick, Mass.-based Cognex (nasdaq: CGNX – news – people ), a company that has developed lane-departure warnings systems that “watch” the lane lines on the road. Installed in trucks, those sensors can alert a sleepy driver who is weaving out of his or her lane.
Adaptive cruise control, pioneered by companies like Mercedes-Benz and Lexus, uses the same laser and radar scanners installed on DARPA’s robotic cars to maintain a set distance from other vehicles on a highway. Sensor developers like IBEO and its parent company, SICK, in Walkirch, Germany, are working on electronic eyes that could one day help cars spot–and so avoid–pedestrians, animals or other obstacles.
Off-road, where traffic doesn’t complicate matters, robotic driving is even more practical. Caterpillar Construction (nyse: CAT – news – people ), which sponsored the three top teams in this year’s DARPA’s challenge, now equips some of its bulldozers with a combination of GPS and laser scanners to allow for semi-autonomous earth-moving. The driver has merely to guide the vehicle back and forth, and the blade robotically positions itself to create a perfectly flat surface.
Red Whittaker, the head of Carnegie Mellon’s Tartan Racing Team, whose robotic Chevy Tahoe called “Boss” took the top prize of $2 million in the most recent DARPA race, cites another off-road application: farming. Whittaker, who farms about 300 acres of land in his spare time, points out that Trimble, the company that created global-positioning satellite systems for many of the robots in the race, also sells a system called “EZ Steer,” a small steering-wheel attachment that robotically guides tractors. “Farmland goes for miles–you want straight, even, careful rows. You don’t want to compact the land you’re driving on, so you drive in the same tracks year after year after year,” he says. “A good guidance system creates much higher quality and higher performance.”
If all of these developing technologies mean that DARPA’s dollars are funding commercial applications more than military advances, it wouldn’t be the first time, says Stanford’s Sebastian Thrun. The Internet, Thrun points out, was also originally sponsored by DARPA, with technology built by university and industry teams. “Did the military intend to foster porn-surfing on the Web?” he asks. “I doubt it.”
Whether DARPA’s autonomous driving initiative spurs more military or civilian spin-offs isn’t as important as simply making driving safer, Thrun says.
“A life saved is a life saved,” he says. “In these moments of disruptive technology, everyone benefits.”
Sourced and published by Henry Sapiecha 1st May 2009
Share and Enjoy
