The image is grainy and poorly lit, but the video camera mounted atop a small, tracked robot shows the robot's progress through a service corridor of the collapsed World Trade Center's second tower.
As the robot comes up to a rubble pile, the video shows the top of a corpse's head and a wristwatch.
 |
 |
|
It looks like a toy, but don't expect to find the "snakebot" being designed by Howie Choset, a Carnegie Mellon University professor of mechanical engineering, and a group of students on store shelves any time soon. The robot's projected uses aren't child's play. It would wiggle into destroyed buildings, for example, to locate disaster survivors. (Andy Starnes, Post-Gazette) |
The chilling image, recorded a week after the Sept. 11 attack, demonstrates both a strength and a weakness of the robots now available for urban search and rescue operations, said Howie Choset, a mechanical engineer at Carnegie Mellon University. The tethered robots, similar to miniature tanks, can search areas inaccessible to or unsafe for humans, but the inability to crane their necks or probe into rubble piles limits what they can see.
Imagine, Choset continued, if the robot's camera was mounted on something akin to an elephant's trunk. The camera could peer over the edge of the rubble, or perhaps probe through spaces between the rubble to search for other, larger voids where survivors might be.
Or, separated from the tracked robot, the long, segmented device might slither like a snake deep into the rubble in its search for survivors.
The concept of snake-like robots isn't new, but Choset is leading a research effort to overcome problems with strength, navigation and locomotion that have restricted use of these robots.
His lab recently was awarded an $800,000 grant from the U.S. Department of Energy to extend the technology, perhaps making these "hyper-redundant" robots available for use in waste tank or bridge inspections, surveying of bioterrorism sites or robotic surgery, as well as for search and rescue operations.
The beauty is both the ability of these robots to do what humans can't do, such as squirming through tiny gaps, and to simply do a lot of different things, from dipping its head into a coffee cup thought to contain anthrax to serving passively as a makeshift support amid shifting debris.
"Sometimes, it's more important to have something that's versatile, rather than to have something that does one task well," Choset said.
The potential value of snake robots became obvious in the aftermath of the Twin Towers' collapse.
Within 24 hours of the attack, researchers from several research centers and robot manufacturers were on the scene with about a dozen robots, varying in size from a shoebox to a suitcase. The remote-controlled robots ventured into areas too small or too unstable for human emergency workers.
By 8 a.m. the day after the attack, a robot had found its first human victim in the rubble, noted Robin Murphy of the nonprofit Center for Robot-Assisted Search and Rescue.
At Ground Zero, time was of the essence, as it is at any such urban disaster, such as building collapses caused by earthquakes. "You've basically got 48 hours to find survivors," explained Murphy, a roboticist at the University of South Florida. But it's hard to know where large voids that contain survivors might be in the rubble.
"You just don't have enough cranes and people to move everything," she added. "You gotta know where to dig."
Within 10 years -- and maybe in as few as five -- robots will be as routine at search and rescue sites as dogs are today, Murphy predicted. By using a variety of types and sizes of robots, emergency crews might quickly identify pockets of survivors and perhaps even establish communication with them, or deliver food and medical supplies to them.
"But our robots [at the World Trade Center] were basically little tanks on a tether," Murphy said. As rugged as they were, they simply couldn't negotiate many of the voids in the rubble, voids that often were twisty and jumbled.
Cameras mounted on long poles were available to probe small cracks and it was possible to bend the poles, but a pre-bent pole can be extended only so far into these twisty voids, Murphy said. A snake, however, might have been able to follow some of these voids and more quickly penetrate the rubble piles.
"That's critically important for urban search and rescue," said Choset, who joined Carnegie Mellon six years ago after working on snake robots at the California Institute of Technology.
In Choset's lab, researchers use a 3-foot-long snake robot, built 10 years ago for NASA's Jet Propulsion Laboratory, to probe through simulated rubble. The head of the robot, outfitted with a tiny video camera, can wend its way through cracks and twist inside the pile to avoid obstacles or scan voids. The segmented robot is about 1 1/2 inches in diameter, about right for negotiating voids in building rubble.
"If it's bigger than the size of my wrist, then it's useless," Choset said. But small size has meant a tradeoff in strength. The robot is usually operated in a vertical alignment, he noted, because its joints are so weak the robot can't support its own seven-pound weight.
The design of the joints is compact. Each joint between the body segments is a sphere cut in half transversely. As the sphere halves are rotated, the attached body segments angle away from each other; a series of these joints and body segments allow the robot's body to twist in serpentine fashion.
But the motors that turn the joint segments engage at the center of the joint, Choset said. That keeps the joints compact, but also leaves them mechanically weak and subject to wobbling.
Choset's group has developed a stronger, yet similarly compact joint by using offset gears that apply energy to gears on the periphery of the joint. That provides mechanical advantage that preserves the joint's suppleness while increasing its strength.
The improved joint can make itself so stiff that a snake robot might be able to temporarily prop up unstable rubble, allowing access for human rescuers, Choset said. Students Elie Shammas and Ben Turk are among those developing the mechanical design. The university is seeking a patent on the joint.
Other work in the lab, principally by student Ji Yeong Lee, focuses on the mathematics necessary to control a robot that changes its three-dimensional shape as it navigates through a rugged landscape. Students Eleanor Moore, Renata Melamud and others are are studying methods of locomotion, using hints from Carnegie Museum of Natural History herpetologist John Wiens on how real snakes move.
Roboticists can pick up some tricks from nature -- not by copying the 200 vertebral bones found in some snakes, but by using the biology to inspire their designs and methods, Choset said.
"There's no way we're going to be able to design snake muscles," he added.
