Design News for Mechanical and Design Engineers

Faced with such a challenge, rescue workers are turning to technology for help in finding the dead and wounded, but they've met with limited success. Firemen may have had trouble finding people in the seven-story stack of bricks and steel, but they could easily scramble over any obstacles. In contrast, robots can locate survivors through body heat or knocking sounds, but they can't negotiate the rubble. Still, companies from around the country sent their best robots to New York, including MIT spinoffs iRobot and IS Robotics (both of Somerville MA), Foster-Miller Inc. (Waltham, MA), and the University of South Florida's Perceptual Robotics Lab (Tampa, FL).

There are four major tasks robots can perform at a disaster site, according to the FAQ page at the Perceptual Robotics Lab:

• reconnaissance of the hazards
• telling when it's safe for human rescuers
• detecting victims in the rubble
• mapping debris for removal

But robot developers must be careful not to try to perform all these tasks at once. "People have tried to make Swiss Army Knife robots, but that doesn't work," says Hagen Schempf, a scientist in the Field Robotics Center at Carnegie Mellon University, and CEO of robotics company Automatika, Inc. (Pittsburgh, PA). "The best thing for search and rescue would be a robotic gerbil or a robotic cat."

Today, live dogs are often used to find bodies, because of their fantastic sense of smell. "Sensors will eventually get there, but the problem is delivery," says Schempf. He admires the gerbil for its size ("how small a hole can you crawl into?") and its agility ("how nimble are you? If it's too big, you've got to climb over or leap on top").

The top three challenges in creating robots for USAR (Urban Search and Rescue) are locomotion, power (source), and size, Scheupf says. An engineer who solves these challenges must also keep the cost down, and then achieve acceptance—convincing proud firefighters to try the new tool. Because of these staggering challenges, many search and rescue robots are relatively simple, tethered to a power source and steered by a human driver instead of navigating themselves.

By comparison, robots used for pedestrian tasks like scrubbing office floors or patrolling warehouses can be nearly autonomous. A floor scrubbing robot like the ScrubberVac from INtelliBOT (Richmond, VA) can clean 60,000 square feet of floor space with no supervision. But merely to navigate through an office with perfectly smooth floors and square geometry, it needs a broad array of sensors, including ultrasonic, infrared, touch-sensitive strips, a gyroscope, and current sensing (of the electricity in its own motors), says VP Allen Bancroft.

More advanced versions can store maps, so they can "learn" new routes when a person walks them through the path. But even with a map, such a machine still needs the full sensor array to avoid being sidetracked by obstacles like chairs left in a hallway, or wheel slippage on soapy floors—simple challenges compared to the chaos of a disaster site.

Another approach is the CyberGuard from CyberMotion (Roanoke, VA), a human-sized, 400-lb robot designed to patrol warehouses. The U.S. Army currently uses this machine to inventory the weapons in its armories. "We're far more technically advanced than the search and rescue machines used in New York," says John Holland, company president and founder. "Those are not autonomous, they're remote vehicles which just use a camera. They don't even have a collision avoidance system."

Instead, the security guard navigates by sonar or LIDAR, then uses a microwave motion detector, as well as ultraviolet, poisonous gas sensors, and pyrometers (to look at a door and tell if it's hot on the other side).

"Anything that people make a portable detector for, we can put it onboard," Holland says. "These robots normally run autonomously; the operator's just there to monitor the results."