FOR VIDEO OF THE CRUSHER ROBOT, CLICK HERE.
W. WAYT GIBBS

It was the kind of coming-out party peculiar to Pittsburgh, a city of big iron and bright engineers. On April 29, a crowd gathered on bleachers inside the darkened bay of a giant warehouse in one of the rougher parts of town. A row of one-ton concrete blocks protected the spectators from the performance about to begin. At the podium was Steve Welby, who directs the office of tactical technology at the Pentagon's Defense Advanced Research Projects Agency (DARPA). "And now please join me," Welby said, "in congratulating Carnegie Mellon University's National Robotics Engineering Center, as we present Crusher."

Up went the high bay door, and out from a cloud of dry ice fog rolled two of the most intimidating robotic vehicles yet to emerge from DARPA's decade-long quest to build autonomous vehicles for the battlefield. With its tan, 6.5-ton body slung low over six massive wheels, the first Crusher robot paraded before the crowd, its hybrid electric motors running nearly silently off lithium-ion batteries. Spinning 180 degrees in place, it leaned left and right to show off its ability to raise or lower its suspension by more than 30 inches. The robot's twin, in Army green, followed behind and then paused to raise a six-meter telescopic mast--the better to aim cameras and other sensors over foliage and hills.

A highlight reel running behind the Crusher robots showed them skidding down wooded slopes, straddling rocky creek beds, and using their wheels like paws to clamber over meter-high walls in field tests. Like a proud father, John Bares, who heads the center and has led the five-year, $24 million project since it began in 2003, pointed out the many novel features of the vehicles, from their high-tech skid plates to their aluminum-titanium space frames.

But what is most impressive about such robots, of course, is what they lack: namely, a steering wheel, seat and control pedals. The day's demonstration showed the vehicles tearing up a rubble-strewn, log-littered course behind the warehouse under the remote control of a human operator. For some military missions, such as reconnaissance and fire support operations, teleoperation will do. But the ultimate goal is to imbue the robots with enough sense, both perceptive and cognitive, that they can complete missions themselves, as automated ambulances, supply mules and sentries. By August, Bares says, his team will have outfitted the Crusher vehicles with a suite of sensors that they developed on a previous automaton, named Spinner. They plan late summer field tests at Fort Lewis in eastern Washington, followed by trials next winter at Fort Hood in Texas.

At the same time as DARPA pays the NREC group to take a steady and predictable path toward autonomy, the agency is also holding a third Grand Challenge competition to encourage more out-of-the-box approaches. Two days after the Crusher roll-out, DARPA officials invited roboticists, professional and otherwise, to submit entries for an "Urban Challenge" to be held in November 2007.

The event follows two off-road races for robots that DARPA organized in the Mojave desert in March 2004 and October 2005. No robot finished the 142-mile course in 2004, but five vehicles navigated to the finish line of the 132-mile trail last October. A Stanford University Team claimed the sole $2 million prize, but two other robots, both constructed at Carnegie Mellon, were close on its heels.

Perhaps as a result of that outcome, DARPA promises to award prizes to the top three contenders in its Urban Challenge, in which autonomous cars or trucks will have to maintain an average speed of 10 miles an hour while negotiating a mock city, crossing busy intersections and even merging into moving traffic. If the prizes, of $2 million, $500,000 and $250,000, are not incentive enough, teams can apply for up to $1 million of government development funding. DARPA has also promised purses of $50,000 for any team that makes it to the semi-finals, and $100,000 for all finalists.