PHOTO: Michael Goldfarb/Vanderbilt University

14 December 2007—To let Hollywood tell it, robots are ultrapowerful machines with seemingly inexhaustible stamina. In last summer’s CGI extravaganza Transformers, shape-shifting robots swatted cars and trucks aside like gnats at a barbecue and effortlessly outpaced fleeing humans—who, by comparison, were inferior in every regard.


But the physics of the cineplex don’t hold true in the real world. Here, even the best humanoid robots don’t hold a candle to the race upon which they’re based. At 129 kilograms and just 1.6 meters in height, Honda’s P3 robot, for example, is a dumpy weakling that moves no faster than a slow walk (2 kilometers per hour), largely because it’s saddled with 30 kg of batteries that can power it for only 20 minutes on a single charge, notes Michael Goldfarb, a mechanical engineering professor at Vanderbilt University, in Nashville. For nearly a decade, Goldfarb has been researching ways to develop high-power, high-energy-density approaches to powering human-scale robots and may have hit on a solution: rocket fuel. He’s demonstrating its superiority over battery power with a lifelike prosthetic arm that he and his graduate students built from scratch. 


When a small amount of hydrogen peroxide propellant used in the space industry comes in contact with small grains of iridium-coated alumina (a catalyst also developed for rockets), the reaction product is a burst of vapor comprising water and oxygen that has 200 times the volume of the propellant. This expansion is sufficient to power sets of pneumatic actuators that can open and close the elbow joint on the prosthetic arm in Goldfarb’s lab with enough force to overcome up to 11 kg of resistance. “This system is 10 times more energetic than batteries and motors,” says Goldfarb, who admits that while it’s still not nearly a match for humans, “it’s getting a lot closer.” What’s more, when a cartridge containing the propellant is depleted, it can be replaced with another within minutes. Goldfarb envisions cartridges, each containing enough of the solution to power an arm for a day, being sold in packages of a dozen or more.


The “rocket arm” not only does more work than battery-powered prosthetics, it is far more lifelike. Robotic prosthetics currently on the market have only two joints: an elbow that bends and a hand joint that is essentially a pincher. Goldfarb’s device has 21 joints that allow nine degrees of freedom. The joints that move in concert on a human hand—like the two closest to the tips of each finger—also move in tandem on the mechanical replica. “Nature figured out a long time ago that there’s very little utility in bending one independently of the other,” says Goldfarb. 


Goldfarb and his graduate students control the laboratory prototype with an exoskeleton that a user straps onto his or her arm. The rocket arm mimics the movements of the exoskeleton; the fingers respond to pressure on a hand grip that tells them how far and how fast to contract. One of his students demonstrated the hand’s fine motor coordination by gripping a lightbulb without breaking it and pulling a single sheet from a box of tissues.


Goldfarb also reports that other researchers are working to refine cosmeses—custom-made artificial skin meant to match the appearance of an amputee’s remaining arm—so they are capable of dispersing the rocket arm’s exhaust vapor the way a real limb would sweat. An inner absorbent layer would soak up the vapor, and then its porous outer layer would allow the moisture to evaporate. “The net effect is that the artificial arm ‘sweats’ in quantities similar to what a human arm would after doing work,” says Goldfarb.


The Vanderbilt professor, who is also working on an artificial leg and a six-legged search robot for disaster sites using the same propulsion technique, developed the rocket arm at the behest of the Defense Advanced Research Projects Agency (DARPA) of the U.S. Department of Defense as part of an ongoing project aimed at developing a self-contained and self-powered humanlike arm hardwired to the human nervous system. DARPA hopes that the Revolutionizing Prosthetics project, begun in 2005, will result in an artificial arm that will restore nearly all function to soldiers who lose their limbs on the battlefield. DARPA plans to begin initial clinical testing aimed at gaining approval by the U.S. Food and Drug Administration by 2009.


Goldfarb notes that his lab has not focused on the means for integrating the artificial arm with the human nervous system. Researchers at the University of Utah, Caltech, Johns Hopkins, and the Rehabilitation Institute of Chicago are trying out several methods, including a chip that processes electrical signals from the brain and sends RF signals to a receiver on the prosthetic limb. “Creating that neural interface is the hard part,” Goldfarb says, as though his lab’s efforts were trivial.