Novel, Energy-Frugal Robots Walk Like Humans

WASHINGTON, Feb. 18 -- A new breed of powered, energy-efficient, two-legged robots with a surprisingly human gait were unveiled at the annual meeting of the American Association for the Advancement of Science (AAAS) yesterday.

Researchers from Cornell University and the Massachusetts Institute of Technology (MIT), and their colleagues from Delft University of Technology in the Netherlands, displayed video footage of three bipedal robots and demonstrated a biped developed at MIT.

"These innovations are a platform upon which others will build," says Michael Foster, an expert on computer and information science and engineering and one of the National Science Foundation (NSF) managers who oversaw the research. "This is the foundation for what we may see in robotic control in the future."

By applying concepts rooted in "passive-dynamic walkers" -- devices that can walk down a gentle slope powered only by the pull of gravity -- the engineers have crafted robots like the Cornell biped that walk on level ground using half the wattage of a standard, compact fluorescent light bulb.

"The biped walking mechanism in robots is limited by on-board battery power," says Junku Yuh, an NSF expert on intelligent systems, who also oversaw the research. "The Cornell team's passive mechanism helps greatly reduce the power requirement. Their work is very innovative."

Representing fundamental developments in computer and mechanical control, the robots are helping researchers understand bipedal motion and revealing processes that underlie human locomotion and motor learning. Applications are already on the horizon, with one researcher exploring how the new robotics can aid development of increasingly energy-efficient prosthetic devices.

"This is a perfect example of a single concept yielding benefits in a variety of fields, including medicine," said Gil Devey, an NSF expert on disabilities research. The MIT walker's passive-dynamic design provides a new way to study motor learning. The robot can teach itself to walk in as little as 10 minutes, adapting to terrain as it moves.

"This project is about the fundamentals of control," says Foster. "The researchers have combined our developing knowledge of computerized control with mechanical principles that the world provides for us and shown that we can integrate the two."

All three robots verify a long-held hypothesis that suggests motors can substitute for gravity in passive-dynamic walking devices. A slope is not required, only careful engineering.

For more information, visit: www.aaas.org