\\ Home Page : Articolo : Stampa
Programmable matter could self-assemble into anything from a Mars robot to a smart surgical pill.
By Admin (from 18/07/2011 @ 14:00:21, in en - Science and Society, read 2175 times)

It's code red on Mars. A storm has knocked out our Martian colony's main power supply and the base has been running on backup power from a nearby array of solar panels.

But as the storm continues to rage, the winds tear the solar panels from their anchors, dashing them across the Martian rocks. The colonists can't risk venturing outside in these treacherous conditions. There's nothing else for it: time to deploy the programmable matter.

That scenario is a long way in the future, but the first programmable materials are already on their way. Essentially, programmable matter is a clay-like, electronic material that can shape itself into different functional configurations at your command. The simulated Martian scene, which recently played out at a conference on robotics and automation technology, illustrates its potential. One day some innocuous-looking, amorphous material could, at the touch of a button, turn itself into a tool to repair a broken solar panel, an entirely new solar panel, or even a robot to install it for you. And it is not just future Martian pioneers who will benefit. The stuff could make the Swiss army knife look as inadequate as a "spork", put an end to electronic waste, and even revolutionise medicine.

The idea of programmable matter dates back to 1952, though not under that name. John Von Neumann, the father of the architecture used almost ubiquitously in today's computers, was on the hunt for a mechanism that could ape, in silicon, biology's trick of self-reproduction. But his so-called cellular automata stayed in the realm of mathematical fantasy until 1991, when two engineers, Tommaso Toffoli and Norman Margolus, penned a manifesto calling for experiments on machines based on the principles of cellular automata, which they rechristened programmable matter (Physica D, vol 47, p 263). But even by that point, computers were neither small nor powerful enough to bring the idea of microscopic, self-replicating machines much closer to reality.

That's about to change. The astonishing computing advances of the past 10 years have convinced many engineers that programmable matter's time is at hand. To that end, the engineers working on the problem have defined the criteria programmable matter must meet in order to fulfil Von Neumann's original dream. Most crucially, it needs building blocks that can act individually but also as part of a network. Each of these must have an energy source, some way of receiving and processing information and some physical means of acting on that information. And to get to the stage where it can form itself into something useful, these building blocks need to be very small indeed, and they must be mobilised in enormous numbers.

Natural inspiration

These criteria might seem familiar to biologists: they are the same for every multicellular organism on Earth. Consider the basic piece of biological hardware, the cell. It can grow, obtain fuel and carry out complex operations at the behest of its DNA software, while also operating as part of a much larger whole. If part of the organism is damaged, the cells recognise and process the signals, and effect a repair. And when the organism dies, it disintegrates, breaking down into a molecular buffet that can be reused by other organisms. Electronics don't do such things. A smashed-up cellphone stays broken, and when it becomes obsolete, it cannot recycle itself.

The obvious temptation, then, is to hack actual biological cells to our own advantage. At Harvard University, David Liu and his group started down this avenue. Liu's group used complementary DNA base pairs as a "Velcro" to bind together millimetre-scale synthetic building blocks (Nature, vol 394, p 539).

But there is a problem with turning biological systems into programmable matter - the strength of the bond. "It was not nearly as strong as conventional adhesives such as epoxy," Liu says. "The bond's strong enough to hold a glass of water, maybe, but not much more." And that is why other teams have looked to electronics instead.

But while plastic and metal are strong, electronics has its own limitations. Chief among them is the way its intelligence is distributed: whereas a cell has all its sensing and processing mechanisms contained within, electronic devices have just a few "thinking" parts - microchips - in charge of a whole lot of dumb metal. The key to turning an electronic device into programmable matter is to give it more thinking parts. In fact, the intelligence of the whole should be imbued throughout the parts. "We want a robot's body to also be its brain," says Daniela Rus, a roboticist at the Massachusetts Institute of Technology, who is working on several approaches to programmable matter.

And indeed, some of the first working prototypes of programmable matter consist of groups of scaled-down, reconfigurable robots. Mark Yim of the University of Pennsylvania, Philadelphia, has built a set of programmable, 6-centimetre-wide "CKBots"(pictured, right) that can sense and communicate with each other and work as a team. Some of the cubes have cameras, LED lights and accelerometers, while others have embedded computers, positioning sensors and servo motors that allow them to rotate and manoeuvre.

In 2008, Yim's lab demonstrated the potential of the CKBots by assembling 15 cubes into a crude humanoid shape with two legs and a torso. With a swift kick the modules broke apart and scattered, lying on their sides with their LED lights blinking like the eyes of stunned animals. In reality, though, they were communicating. Each module pulsed a pattern of flashing light that identified its position in the original configuration, and the cameras scanned the scene to catch the signal from their former neighbours. Soon, the clusters began to twitch, flipped up and snaked towards each other. Within a few moments, they had reassumed their humanoid shape and the reconfigured robot took a couple of tentative steps. When Yim and his team uploaded a video of the reassembling robot to the web, it quickly went viral.

Yim's CKBots were largely the tools of choice for the Mars simulation, part of the Planetary Contingency Challenge at last year's International Conference on Robotics and Automation in Anchorage, Alaska. Teams were tasked with designing and assembling a remotely operated robot that could scramble across the uneven surface of Mars, pick up the strewn solar panels and reattach them to their original structure. Unless they brought systems of their own devising, the contestants were supplied with CKBots that they could assemble into modular robots with various appendages. The CKBot-based team that covered the ground and reattached the most solar panels in the time allowed was from Harvard University.

Compared with nature's efforts, though, the CKBot is still a decidedly clunky affair. To create real programmable matter, consisting of thousands or millions of bits that can be moulded and remoulded like clay, those units will need to be far smaller than Yim's.

The solution might just be to start smaller. That's what Seth Goldstein is doing at Carnegie Mellon University in Pittsburgh, Pennsylvania, using building blocks he calls claytronic atoms, or catoms (IEEE Computer, vol 38, issue 6, p 99). Rather than designing a fully endowed robotic cube, he has coaxed flat silicon to curl into tiny spheres and cylinders 1 millimetre across or less. The obvious problem with this approach, though, is finding space for the robot's innards.

Fluid movement

Goldstein solved the problem by reducing the robot's insides to circuitry and electrodes printed onto the silicon. The fundamental idea is that when two catoms get near each other, their electrodes become either positively or negatively charged, and that is how they stick to each other. Movement and power comes from an external track that is akin to the third rail of a railroad track, and the electrostatic force it creates attracts the catoms to it. Changing the current running in separate sections along the track drags the catoms along.

Now the team is trying to work out whether they can find a way to get the catoms to move without the track, using only the electrostatic forces between catoms.

Goldstein is dreaming big: hundreds of thousands, if not millions, of these catoms could be deployed at a moment's notice. That could have applications in medicine. "If you could take a pill of this stuff," he says, "it would change its shape inside you, do its job, and then flush itself out." Such applications depend on making the basic unit of programmable matter a lot smaller than a millimetre, and it is as yet not clear how that can be achieved while still allowing them to be smart enough to communicate with each other.

But is that component-to-component coordination strictly necessary? Hod Lipson, a researcher at Cornell University in Ithaca, New York, thinks not. Instead of imbuing individual components with the intelligence to adhere to each other, he and his team are manipulating the surrounding environment to coax the components into place. After all, that's how nature does it: biological structures assemble themselves in liquid environments, using random fluid motion as a transit system. Amino acids and proteins might eddy about for several minutes before they find another molecule to which they can attach.

Lipson simply translated that principle. His centimetre-sized cubes are perforated, waterproof computers with a valve on each face. Submerged in water, each cube can autonomously open and close any of its valves. The flow of water through one free-floating cube will subtly pull a second cube toward it. When the second cube latches onto the first, both automatically open the correct valves that will form a new path for the water to flow through. This attracts more and more cubes. Using this method, the team was able to assemble a structure of 10 cubes (IEEE Transactions on Robotics, vol 26, p 518). They have similarly experimented with assembling structures out of 500-micrometre silicon tiles (Applied Physics Letters, vol 93, p 254105). "Ultimately, we want to find ways to assemble a million millimetre-scale cubes," Lipson says. "So we still have some way to go."

But there's another option for programmable matter, and it's a bit of a radical departure. Perhaps programmable matter doesn't need to consist of separate bits, says Rus. She and her colleague Robert Wood, at Harvard University, have taken their inspiration from origami to reconceive programmable matter as a single, rigid sheet with inlaid circuitry and predefined creases (Proceedings of the National Academy of Sciences, vol 107, p 12441). This approach bypasses the tricky problem of getting large numbers of robots to talk to each other.

Each crease is created from a shape-memory alloy - a metal that "remembers" its original shape - which bends when electricity flows through the sheet. The sheets are identical, consisting of interconnected triangles, but different programming induces them to adopt one of a number of predetermined shapes. So far, the shapes are small and fairly rudimentary, but it's a good start. Eventually, the researchers hope the shapes could be determined by stickers that would contain programming instructions that direct electricity through the desired circuitry. So, for example, applying a "tripod" sticker would cause the sheet to fold itself into the shape of a tripod. "If you had a stack of smart sheets in your backpack, you could grab one and make it turn into anything on demand," says Rus.

Some of these approaches might not see applications for decades. That's why some researchers, like Neil Gershenfeld of MIT, are focusing instead on a parallel research effort called functional digital materials, which may deliver benefits in the shorter term. Like programmable matter, these building blocks can be assembled into relatively complex structures. Unlike programmable matter, however, they can't do it themselves, instead requiring a machine to assemble them. Gershenfeld's mechanism is a cross between a 3D printer and a rapid-prototyping machine. The machine can assemble complex structures out of different building blocks, printing, for example, a cellphone's circuit board out of lego-like conductors and insulators. But here's the kicker. After the machine has created a structure, it can disassemble it back into the original constituent bits, and then reuse those parts to make something new. "It's just like when a plant dies in the forest, and its components turn to mulch," he says. "It's going to mean the end of trash."

Gershenfeld predicts that his prototypes will be mass-produced within about five years. Commercial applications of programmable matter, by contrast, are at least 20 years away. Nonetheless, he says, all the research now under way is solving different parts of what were once thought fundamental roadblocks. Catoms are small; shape-memory sheets can create specific designs; fluid cubes can self-assemble. One thing is for sure, however. The precursor fields to programmable matter - including functional digital materials and reconfigurable robotics - will soon bear fruit. For reconfigurable robots in particular, real-life applications aren't far off: Yim is already using CKBots to build search-and-rescue robots for operations that are too dangerous for humans.

Meanwhile, competitors are flexing their muscles for a rerun of last year's modular robotics competition in Alaska, which this year will be in Shanghai. And here's another sign that the field is advancing: unlike in previous years, when most contenders had to be furnished with CKBots, all the contestaVnts are now expected to bring their own robotic modules.

Source: NewScientist