A remarkably simple system produces rather dramatic results.

Our robots manage some pretty impressive feats—including back flips—through the whirring of motors and hydraulic pumps. But all of life manages to perform far more impressive feats using muscles. Muscles allow incredibly fine control of movement, along with violent bursts of exertion. As a result, there has been a steady stream of attempts to craft artificial muscles.

But a team of Harvard and MIT researchers use part of their new paper to catalog all the ways that these efforts fall a bit short: energy efficiencies below two percent, extremely high voltage requirements, or extremely slow contractions. So they decided to focus on a different approach: pressure-driven artificial muscles. They devised a system that mixes this pressure with an origami-inspired skeleton to (by some measures) outperform muscles.

Pump it up

The basic design of their muscles is ingeniously simple. The muscles are centered on a rigid yet foldable "skeleton," which could be made of plastic or metal. This ensures that as the muscle expands and contracts, it folds (or unfolds) in a specific pattern that directs the force. The skeleton is surrounded by a sealed, flexible material, typically some sort of polymer sheet—think putting the skeleton in a form-fitting plastic bag. This can be filled using either a liquid or gas.

As the filler is pumped out, the muscle contracts, causing the skeleton to fold back up. Pump it back in, and the muscle expands again.

The system is simple, because whatever medium it's in can be used to power the muscle. Using it in the water? Fill it with water. Using it in a room? Use the air. The plastic and skeleton materials can also be really basic stuff. In fact, the researchers managed to create some muscles for less than $1 in raw materials (not counting the pump that pushes the filler in and out).

In fact, they showed you can use a variety of techniques, like molding, 3D printing, and even folding by hand to generate skeletal elements. The only real "challenge" is sealing them up inside a bag with a way to draw air in and out. But here, the researchers showed that it was possible to seal things up using, "heat-pressing, gluing, welding, zippering, and sewing." By powering these with negative pressure (meaning sucking the air out), they avoided some of the risks of high-pressure hydraulic systems, such as causing these seals to burst.

Real-world performance

The team also generated modeling software that predicted the force and contraction generated by muscle designs within about a 10 percent accuracy, allowing them to test various designs before trying to build them. Their designs ranged from a simple zig-zag structure that produced a linear contraction to a contracting sphere based on an origami water bomb design, which changed in volume by more than 90 percent when contracted.

The results were rather impressive. The sub-$1 muscle, which weighed only 2.6 grams (not counting the vacuum pump) was able to lift a 3kg weight. Other muscles were able to generate about six times the sustained stress and peak power density of our own skeletal muscle fibers.

If you use a large external vacuum source, the muscles are also relatively efficient, reaching efficiencies approaching 60 percent for simple load-lifting tests. Placing miniature pumps in the muscle itself, however, killed the efficiency, dropping it down below five percent, which is similar to existing systems.

This doesn't seem to be a major drawback, though, as lots of existing robot designs have external tethers that supply power and/or control data. So, I think it's time that this research team gets to work building a robot with these things.