Robots with real muscles

A robotic fish powered by real muscles could boost artifical limbs

When Hugh Herr put his robotic fish into its tank, it swam off looking surprisingly lifelike. But a few minutes later, it was flagging-and eventually came to a complete stop. It wasn't faulty: it just needed a break. The reason? Herr's robot is the first one to be powered by real muscles.

Researchers have known for centuries that they can make muscles contract in the lab: in 1786, Luigi Galvani discovered that electricity made a dissected frog's leg twitch. But until now, no one has ever tried to harness the phenomenon to power a machine.

So Herr and his colleagues at the Massachusetts Institute of Technology's Biomechatronics group built themselves a robotic fish. Inside it, a microprocessor sends electric signals to frog muscles on either side of the fish, making them contract. Tendons on the muscles are sewn to the nose and tail so the "fish" wiggles and swims in response to the signals. The muscles get their energy from the glucose solution the fish is swimming in (see video at www-personal.umich.edu/~bobden/biomechatronic_devices.html).

One of Herr's aims is to power prosthetic limbs with real muscles. Artificial limbs tend to be much stiffer than real ones and can't adapt to different surfaces, so they behave the same way whether you are walking on cement or sand. Even a top-of-the-range prosthetic ankle won't provide active thrust, which makes walking tiring. Prototype limbs that can thrust contain noisy joint motors. "If all our muscles were motors, we couldn't hear ourselves talking," says Herr, who lost his own legs below the knee to frostbite.

Other researchers are attempting to make better prosthetics using artificial muscles made from polymer fibres. These muscles are strong and silent, and contract when a voltage is applied. But "there are issues of robustness. They oxidise easily," admits Yoseph Bar-Cohen, who heads NASA's advanced actuators laboratory in Pasadena, California.

"Why would anyone want to build artificial muscle?" asks Bob Dennis, a colleague of Herr. "Real muscle can adapt to its environment, it can heal itself, it can self-regulate." But live muscle has serious drawbacks. "People ask: how are you going to sustain a muscle and activate it?" says Zeynep Erim of the Neuromuscular Research Center at Boston University. So although many researchers had tested muscle performance in the lab, nobody had ever put them to use in a robot.

Herr wants to improve his prototype by giving it a small stomach to supply the tiny muscles with glucose. And for bigger muscles, he needs to add a circulatory system that pumps the glucose to the muscles.

The robot has since expired, because its muscles only keep for a few hours outside the body. But Dennis is now growing muscle cultures that live for several months. He has fine-tuned a muscle-culturing technology (see "Muscles to order") and aims to create muscle on demand. MIT's project is funded by the Pentagon's Defense Advanced Research Projects Agency, which wants to know if real muscle can be used as a silent actuator in robotic systems. The military's interest could revolve around "exoskeletons"-prosthetic suits that will one day let soldiers run faster, jump higher or carry more weapons. Such systems might employ real muscle.

Author: Eugenie Samuel, Boston

New Scientist issue: 24th February 2001

Please mention New Scientist as the source of this story and, if publishing online, please carry a hyperlink to: http://www.newscientist.com