Attack of the killer prototype robots

Combine the right voltage and software program with a flat piece of fabric and you get dynamic physical rendering. Photos: Robot prototypes

SAN FRANCISCO--Intel is trying to see if millions of tiny robots can work together to create a coffee cup, or a model of a truck.

Intel's lab in Pittsburgh, affiliated with Carnegie Mellon University, is showing off a technology concept at the Intel Developer Forum here this week called Dynamic Physical Rendering, which could ultimately lead to a shape-shifting fabric.

Photos: Prototype robots

Apply the right voltage and software program and the flat piece of fabric turns into a 3D model of a car. Change those parameters and it transforms into a cube. Dynamic Physical Rendering has grown out of the ongoing Claytronics project headed up by CMU professor Seth Goldstein.

"Rather than look at a 3D model on a CAD (computer-aided design) program, a physical model would be manifested on your desk," said Babu Pillai, who, along with Jason Campbell, is heading up the project. "The material would change shape under software control."

The trick is that the fabric would not be a continuous piece of material. Instead, it would be composed of millions of independent silicon spheres covered in electronic actuators--half-capacitors or electromagnets. By applying charges to different actuators, different points on the sphere would be repelled or attracted to similar points on other spheres. The coordinated movement of the spheres would then cause the fabric to assume a shape.

The intelligent fabric doesn't exist yet, but the Air Force Research Laboratory has created prototypes of the components that make up intelligent fabric. The spheres measure about a millimeter in diameter. First, a piece of silicon cut into the shape of a star with many arms is produced. The stress of the material causes it to be rolled up into a ball. Intel uses these spheres in its prototypes.

The group has also demonstrated how small groups of actuators could move objects. In one demonstration, two cylinders covered in rows of electromagnets are propelled across a surface through repulsion and attraction. The cylinders do not contain moving parts.

The spheres and the actuators right now are made separately, but in the future they could be made together on standard silicon processes, said Pillai. A layer containing the actuators would be laid down in a silicon wafer and then the material that would form the skeleton of the sphere would be applied on top of it. Star shapes would then be cut from the wafer. When stress forces the silicon to curl up, the layer of actuators would be there on the outside, sort of like the shell of a potato bug.

Hardware prototypes of the fabric could be ready in about five years, Pillai predicted.

That, however, is the easy part. Coming up with software to control the movement of the spheres looks to be far more difficult.

"How do you program 10 million nodes to work together," he asked rhetorically. "Essentially it is a robotic system with thousands of robots moving against each other in a constrained way."

Pillai and Campbell think that the solution may lie in developing programs where every step does not have to be planned. A simulation program demonstrates this. In the demo, 40,000 robots, represented by light dots, move about as each tries to accomplish a single task: fill in any gaps between itself and other robots. The computer, meanwhile, sets external boundaries to the field over which the robots can jostle.

In the simulation, the software robots are grouped into five squares. After a few minutes of bumping and grinding, the five squares begin to spell out the I-n-t-e-l of the company logo.

Even if they can solve this problem, they would then have to conquer the problem of getting the spheres to coordinate their action in three dimensions.

Around 20 people are working on the project. The Pittsburgh lab is one of a collection that Intel has built next to prestigious universities. The company also has labs in Berkeley and Cambridge, for instance. Carnegie Mellon is one of the chief centers for robotics research.