Claytronics: An Instance of (mimicking a human form) would then be specified ab-

Programmable Matter stractly, perhaps as a series of “snapshots” or as a collec-

Seth Copen Goldstein and Todd Mowry tion of virtual deforming “forces”, and then broadcast to
{seth,tcm}@cs.cmu.edu the catoms. Compilation of the specification would then
Carnegie Mellon University provide each catom with a local plan for achieving the de-
sired global shape. At this point, the catoms would start to
Programmable matter refers to a technology that will move around each other using forces generated on-board,
allow one to control and manipulate three-dimensional either magnetically or electrostatically, and adhere to each
physical artifacts (similar to how we already control other using, for example, a nanofiber-adhesive mecha-
and manipulate two-dimensional images with computer nism [4]. Finally, the catoms on the surface would display
graphics). In other words, programmable matter will al- an image; rendering the color and texture characteristics
low us to take a (big) step beyond virtual reality, to syn- of the source object. If the source object begins to move, a
thetic reality, an environment in which all the objects in concise description of the movements would be broadcast
a user’s environment (including the ones inserted by the allowing the catoms to update their positions by moving
computer) are physically realized. Note that the idea is not around each other. The end result is that the system ap-
to transport objects nor is it to recreate an objects chem- pears to be a single coordinated system.
ical composition, but rather to create a physical artifact One key motivation for our work is that technology has
that will mimic the shape, movement, visual appearance, reached a point where we can realistically build a pro-
sound, and tactile qualities of the original object. grammable matter system which is guided by design prin-
The enabling hardware technology behind synthetic re- ciples which will allow it to ultimately scale to millions of
ality is Claytronics, a form of programmable matter that sub-millimeter catoms. In fact, we expect our prototype
can organize itself into the shape of an object and render for 2D Claytronics to be operational before ASPLOS’04.
its outer surface to match the visual appearance of that Our goal is that the system be usable now and scal-
object. Claytronics is made up of individual components, able for the future. Thus, the guiding design principle,
called catoms—for Claytronic atoms—that can move in behind both the hardware and the software, is scalabil-
three dimensions (in relation to other catoms), adhere to ity. Hardware mechanisms need to scale towards micron-
other catoms to maintain a 3D shape, and compute state sized catoms and million-catom ensembles. For example,
information (with possible assistance from other catoms the catom hardware minimizes static power consumption
in the ensemble). In our preliminary designs, each catom (e.g., no static power is used for adhesion) and avoids
is a self-contained unit with a CPU, an energy store, a net- moving parts (e.g., the locomotion mechanism currently
work device, a video output device, one or more sensors, uses magnetic forces). Software mechanisms need to be
a means of locomotion, and a mechanism for adhering to scale invariant. For example, our localization and orienta-
other catoms. tion algorithms are completely distributed, parallel, and,
Creating a physical replica of an arbitrary moving 3D are indifferent to catom size.
object that can be updated in real time involves many chal- Claytronics will be a test-bed for solving some of the
lenges. The research involved in addressing these scien- most challenging problems we face today: how to build
tific challenges is likely to have broad impact beyond syn- complex, massively distributed dynamic systems. It is
thetic reality. Particularly relevant to the ASPLOS com- also a step towards truly integrating computers into our
munity, for example, is how to build and program a robust lives—by having them integrated into the very artifacts
distributed system containing millions of computers that around us and allowing them to interact with the world.
must cooperate extensively in a harsh environment where
their configuration and goals are constantly changing? It References
is already possible to reproduce static 3D objects [5, 2].
[1] H. Abelson, et al. Amorphous computing. Communications of the
To create a dynamic 3D object, however, we will build ACM, 43(4):74–82, May 2000.
upon ideas from such diverse areas as modular and recon- [2] Z. Corporation. http://www.zcorp.com.
figurable robotics [6, 3] and amorphous computing [1]. [3] M. Fromherz, et al. Distributed adaptive constrained optimization
A Claytronics system forms a shape through the inter- for smart matter systems. In AAAI Spring Symp. for Intelligent Dist.
action of the individual catoms. For example, suppose we and Embedded Systems, 2002.
wish to synthesize a physical “copy” of a person. The [4] M. Sitti, et al. Synthetic gecko foot-hair micro/nano-structures
catoms would first determine their relative location and as dry adhesives. Journal of Adhesion Science and Technology,
17(8):1055–1074, 2003.
orientation. Using that information they would then form
[5] D. Systems. http://www.3dsystems.com.
a network in a distributed fashion and organize themselves [6] M. Yim, et al. Polybot: a modular reconfigurable robot. In Proc.
into a hierarchical structure, both to improve locality and of the IEEE Int. Conf. on Robotics and Automation, pp. 514–520,
to facilitate the planning and coordination tasks. The goal April 2000.