The day when doctors routinely made house calls may be past, but that doesn't mean that someday you won't routinely see your doctor in your home -- with emphasis on "see."

		Graphic: Morph-ology, in pdf format

That is to say, your doctor could physically work out of her office. But a three-dimensional lookalike, assembled from perhaps a billion tiny, BB-like robots, could be her stand-in in your home. She could talk with you, touch you, look at you, all under the control of the real, if distant, doc.

After the examination, she could be disassembled, leaving behind a big pile of beads. Or the beads might reassemble into a piece of moving sculpture, or turn into a chair.

"I don't think 3-D TV is out of the question, either," said Seth Goldstein, a computer scientist at Carnegie Mellon University who is trying to figure out how to use, build and control these amalgams of tiny robots.

Not a single such robot yet exists; building the one-millimeter diameter robots that Goldstein envisions is beyond current technology. And he acknowledges it could be decades before a synthetic doctor is possible, much less affordable.

But it's not too soon to start thinking about it.

"It's a little like putting a man on the moon," said Todd Mowry, a Carnegie Mellon computer scientist who directs Intel Research Pittsburgh and who conceived the project along with Goldstein. It's not just a problem of building tiny robots, but figuring out how to power them, to get them to stick together and to coordinate and control millions or billions of them.

No one's even sure what to call it. "Claytronics," "synthetic reality" and "programmable matter" have been proposed. "Dynamic physical rendering" is the label Intel uses.

"I'm still working on my 'elevator pitch,' " said Randy Bryant, dean of the Carnegie Mellon School of Computer Science, noting he struggles to find a quick way to describe it to potential sponsors. The National Science Foundation and the Defense Advanced Research Projects Agency have provided some funding thus far.

"We've very excited about the project," Bryant added, noting it already involves about a half dozen core researchers and about 20 other affiliated members within the university and Intel Pittsburgh. More funding could mean broader participation.

"There are very serious researchers around the world betting that some version of this will come to pass," said Philip Kuekes, a molecular electronics researcher at HP Labs, Hewlett-Packard's central research group in Palo Alto, Calif. "We know this is pretty far out stuff," he acknowledged, but with HP already working on experimental electronic devices that are just a molecule thick "you need to prepare now."

Goldstein and Mowry began talking about the project almost three years ago. Goldstein was interested in nanotechnology -- machines that are measured in nanometers, or billionths of a meter -- while Mowry was looking for ways of improving today's flat, two-dimensional teleconferencing.

Building a moving, sensing, color-changing replica of each person out of nanotech robots seemed a possible answer. "Every meeting would be a face-to-face meeting."

He and Goldstein have even come up with a name for this new media, a successor to audio and video that they call "pario."

Interacting with a faintly glowing replica of a person might seem a little creepy, Bryant admitted. Then again, "I'm sure people of 200 years ago would be pretty creeped out if you told them about radio and television."

Each of the individual robots comprising these people or shapes would be a "claytronic atom," or catom. Likely spherical in shape, a catom would have no moving parts. Rather, it would be covered with electromagnets to attach itself to other catoms; it would move by using the electromagnets to roll itself over other catoms.

The catoms' surfaces would have light-emitting diodes to allow them to change color and photo cells to sense light, allowing the collective robot to see. Each would contain a fairly powerful, Pentium-class computer.

For now, the researchers are trying to build a two-dimensional version, with each catom being a cylindrical device a little more than an inch in diameter with its side encircled by 24 electromagnets.

In Goldstein's lab, researcher Brian Kirby can make two of these catoms twitch against each other. Within a few months, he hopes one catom will roll all the way around the other catom.

But even attaining this low level of hardware capability has been a big step. Early on, Mowry said, "people would look at us as if we had been reading too many science fiction novels." Indeed, the project causes most people to think of the liquid-metal, shape-shifting robots of Arnold Schwarzenegger's "Terminator" movies. "But once we got those two things wiggling, our credibility went up."

"A robot without moving parts," Goldstein added, "is pretty cool."

Ping-Pong to marble size

Eventually, they hope to build enough two-dimensional catoms to begin experiments with shape changing. Next, they plan to build as many as 100 Ping-Pong-ball-size robots that could move in three dimensions.

Goldstein said building smaller marble-size catoms and, ultimately, one-millimeter diameter catoms will be hard, but "sort of inevitable." So much of their research effort is focused on how to power and control so many catoms.

A large, moving shape such as a human replica might contain a billion catoms, Mowry said. A system with a billion computer nodes, he added, "is something on the scale of the entire Internet. . . . Unlike the real Internet, our thing is moving."

This will require new schemes for quickly organizing and reorganizing such a large computer network. A moving shape will necessarily force catoms to constantly and quickly change positions, breaking connections with one set of catoms and establishing new connections with others.

A number of U.S. researchers already are studying self-reconfigurable robots, but the systems Goldstein and Mowry envision are far beyond the scale of those machines.

Mark Yim, a researcher at the University of Pennsylvania, said the idea behind self-reconfigurable robots is that a robot could change shape depending on a task -- perhaps operating as a snake-like robot to wiggle through tight spaces, while taking the form of a spider or a humanoid for other types of exploration.

But existing systems consist of tens of modules. "Going to hundreds [of modules] is tough," Yim said. "Going to thousands is hard." The Carnegie Mellon group is talking about leaping directly to a scale of millions, he noted.

As Goldstein puts it: "If it's not going to work for a million or a billion, we're not really interested."

Identifying each catom by a number, like each computer on the Internet, isn't likely to work. Rather, catoms may identify themselves based on function or position -- a catom replicating a human would need to know if it was part of a pinky finger, or a mouth, or an eye.

And with so many catoms, the system will have to compensate for the inevitable failure of individual catoms, said Jason Campbell of the Intel group. "Computers don't deal well with failure," he noted. "But we don't really have a choice when we're dealing with these ensembles."

Power also has been a concern. "As we shrink things," said HP's Kuekes, who specializes in designing molecular-scale electronics, "we find that weight and bulk is primarily in the battery."

The solution, Goldstein said, is to eliminate the battery. Instead, the catoms will automatically form themselves into electrical circuits, so delivering power to one catom effectively delivers power to all of the catoms.

As the shape moves and the catoms rearrange themselves, connections will be repeatedly made and broken, Campbell said, interupting power. So the catoms will be designed with a capacitor or small battery to hold just enough charge to compensate for the momentary disconnections.

The system also will be engineered to maintain its shape even when powered off. Metin Sitti, a Carnegie Mellon engineer and roboticist, proposes covering the sides of the catoms with manmade fibers similar to the microscopic foot hairs of the gecko, a tropical lizard.

The millions of hairs on a gecko's toes allow it to cling to almost any surface. The hairs aren't sticky, but rely on weak electrodynamic forces known as the van der Waals force.

If the synthetic hairs can be fashioned out of the microscopic fibers known as carbon nanotubes, the hairs could conduct electricity and might serve as the electrical connections between catoms, Sitti said.

Even if claytronics doesn't immediately yield 3-D motion, it might be useful for producing 3-D shapes in the computer-aided design process, Goldstein said. Claytronics antennas could change shape to improve reception of different radio frequencies. A Claytronics cell phone might grow a full-size keyboard, or expand its video display as needed.

"It could be the ultimate Swiss Army knife," Mowry said.