杏吧原创

WHO would win an arm-wrestling contest between a robot and a human? That鈥檚 easy, you say 鈥� after all, machines can move boulders and crush steel, so surely a robot would overpower a human?

On 7 March 2005, 300 people gathered in a ballroom in San Diego, California, to find out. Under the watchful eye of a professional arm-wrestling official, three robotic arms took turns squaring off against a human. One by one, the electrically powered challengers were soundly defeated by their unyielding opponent, a 17-year-old girl.

It was a poor showing by the bots, but the result was not so surprising to Yoseph Bar-Cohen, the organiser of the event. The catch was that the robot arms used artificial muscles, a still-nascent technology which mimics the action of the human equivalent. 鈥淭hey look like a muscle, but the mechanism inside is different,鈥� says Bar-Cohen, a physicist at NASA鈥檚 Jet Propulsion Laboratory in Pasadena, California. 鈥淲e wanted to see how it would work.鈥�

Apparently not well enough yet. Bar-Cohen concocted the arm-wrestling challenge to boost interest in artificial muscles and advance the state of the art. He plans to hold the contest annually until a robot with synthetic sinews can defeat a human world champion. That will require great improvements in the strength, speed and endurance of artificial muscles, though recent advances in materials technology make this goal seem within reach.

Artificial muscles have been around for decades. Researchers have experimented with a wide variety of materials 鈥� plastics and rubber-like polymers, gels and even metals 鈥� that can contract and expand along their length, as real muscle fibres do. They are meant to go where clunky motors and gears are impractical 鈥� in our bodies, or in microrobots, for instance. The aim is to enable machines to move more smoothly and naturally, and perform tasks such as leaping, climbing or travelling for long stretches with only enough fuel to fill a lunch box. If they can be made reliable, artificial muscles could eventually power more lifelike robots, lighter and more dexterous prosthetic limbs and even artificial organs, such as a plastic heart or diaphragm that can contract in the same way as its biological counterpart. Engineers have imagined scores of non-biological uses as well, including microvalves, flexible music speakers and tactile interfaces such as a programmable Braille display.

鈥淎rtificial muscles could eventually power robots and prosthetics in a more natural way鈥�

A big problem at the moment is how to keep the muscles going strong without large amounts of power, bulky batteries and constant repairs. Now researchers have hit upon a solution, in the process creating some of the most lifelike and self-sustaining man-made motors to date. The inspiration? Our own muscles, which run on liquid fuel.

The field traces its roots to 1880 and the German physicist Wilhelm R枚ntgen, the discoverer of X-rays. R枚ntgen observed that a rubber band altered slightly in length when he applied an electric field to it. It is only in the past few decades, though, that researchers have developed more advanced materials capable of generating significant motions and forces. Today, dozens of research labs and companies are experimenting with artificial muscles and testing them in devices (see 鈥淐ontenders or pretenders?鈥�).

Mimicking real muscle is a surprisingly hard problem. Biological muscles are light and quick to respond to signals from the brain. They can change their length by about 20 per cent and adjust their stiffness in reaction to different loads by using more or fewer of their constituent fibres. A muscle can contract a billion times over the course of its lifetime without wearing out, thanks to the body鈥檚 regenerative ability. What鈥檚 more, muscles convert chemical energy to mechanical energy with an efficiency that often exceeds that of internal combustion engines.

Remarkably, artificial muscles can match or even improve upon biological muscles in a few of these areas. Some polymers can expand to several times their original length, while others react thousands of times faster than human muscles. But so far artificial muscles have only been able to match biological muscles in one or two areas at a time, falling woefully short in others. No artificial muscles regenerate themselves when damaged or spent, for instance; most last for only a few thousand cycles before wearing out.

They also have generally relied on batteries or power cords to bring them to life. By contrast, chemical fuels such as diesel store enormous amounts of energy 鈥� 10 to 100 times that contained in the equivalent weight of today鈥檚 batteries 鈥� not to mention that being tethered to a power source is a drag. 鈥淚t would be nice to have them self-powered by the equivalent of food,鈥� says Bar-Cohen.

It was with this problem in mind that DARPA, the US Department of Defense鈥檚 research agency, approached Ray Baughman of the NanoTech Institute of the University of Texas at Dallas. The task: develop an artificial muscle that could run on a high-energy fuel such as alcohol or diesel. DARPA envisioned nimble robots that could replace humans in hazardous situations, and could carry enough fuel to operate for days without resupply.

Baughman first turned to a material he knew well: carbon nanotubes, giant cylindrical molecules of pure carbon widely studied for their unusual electrical and mechanical properties. In 1999, Baughman and his colleagues had developed an artificial muscle made of sheets of these nanotubes. The design took advantage of the material鈥檚 natural strength and ability to hold large quantities of electric charge. By applying a voltage to the sheet, Baughman was able to make it contract.

The team then incorporated the sheet into a fuel cell 鈥� a device that converts chemical energy into electricity (Science, vol 311, p 1580). They used a 3-centimetre strip of nanotube muscle as an electrode inside a closed chamber of sulphuric acid. Oxygen gas pumped into the chamber combines with hydrogen ions in the acid to form water, using up electrons in the process. These electrons get pulled out of the carbon sheet, which contracts as it becomes more and more positively charged. Connecting the sheet to the negatively charged cathode returns electrons to the sheet, making it expand again.

Researchers have hailed the work as an important advance. 鈥淲hat Ray has been able to show is that you can directly use a chemical reaction to drive artificial muscles,鈥� says John Madden, an electrical engineer at the University of British Columbia in Vancouver, Canada. Until now, he says, artificial muscles have been separate from their energy supply. An integrated fuel source takes them a step closer to biological muscles. An added bonus is that the fuel cell can generate electricity when the muscle is not being used.

The nanotube muscle has a major limitation, though. While it can generate large forces 鈥� 100 times that of a real muscle of the same size 鈥� it contracts by less than 1 per cent of its length. Such a small movement is hard to put to use in any human-scale device. Because of this, the Texas researchers鈥� second attempt involved a wholly different material.

Nitinol, a blend of nickel and titanium, belongs to a class of metals known as shape-memory alloys, which get their name from a peculiar trait they share: the metals are easily bent or stretched, but return to their original shape when heated or cooled. The shape-memory alloys used in artificial muscles are wires that can 鈥渞emember鈥� two different lengths.

The researchers heated the nitinol wires by coating them with a platinum catalyst and exposing them to methanol vapour in air. The alcohol reacts with oxygen, creating heat and causing the wire to contract. Shut off the supply of alcohol, and the wire extends to its original length. Its great advantage is that it can contract by about 5 per cent of its length, enough for engineers to dream up applications in robot arms and prosthetics.It is not without its drawbacks, however. While the range of motion is respectable, the wire only has two 鈥渟ettings鈥� 鈥� long and short. Nor is it easy to control how fast it contracts and relaxes. To incorporate these muscles into a robot, researchers would need a circulatory system that brings fuel to the wire to heat it up at the correct rate, carries away waste and efficiently cools it back down (see Diagram). At the moment, Baughman鈥檚 fuel delivery system consists of a squeeze bottle filled with alcohol.

Farther down the road, Baughman is looking at the possibility of replacing the platinum catalyst with enzymes that can process sugars for fuel instead of methanol or hydrogen gas. Though many years away, this work could conceivably lead to devices such as an artificial heart fuelled by blood sugar. The recipient of such an organ would power their artificial heart the same way you power a real one 鈥� by eating food. 鈥淚t鈥檚 a dream at the present time,鈥� says Baughman, 鈥渂ut it鈥檚 a reasonable dream.鈥�

Will fuel-powered artificial muscles be the key to besting a human in arm wrestling? Not necessarily, and not only because Bar-Cohen鈥檚 contest is currently limited to polymers. Aside from this, the nanotube muscle cannot supply enough motion yet to drive a human-sized arm, and the shape-memory wire鈥檚 circulation system has yet to be worked out. The researchers will also need to give their muscles precise control, durability and eventually biocompatibility. 鈥淚t鈥檚 more of a development problem than a fundamental flaw of the materials,鈥� says Madden. 鈥淧eople have spent a long time characterising these materials, but haven鈥檛 spent a long time applying them.鈥�

鈥淭he aim is to develop devices like artificial hearts fuelled by sugars in the bloodstream鈥�

Which is all the more reason to set up engineering contests like the arm-wrestling match. While it may take years to produce a robot champion, the event鈥檚 founder hopes to see steady improvement in each successive contest. This year, instead of going head-to-head with a human, the machines competed with each other on a calibrated rig of cables and pulleys 鈥� perhaps to avoid another obvious defeat. 鈥淲e are quite far from the physical limitations of these materials,鈥� Bar-Cohen says assuredly. 鈥淏ut it doesn鈥檛 help if we keep losing.鈥�