Goldstein, Seth, Carnegie Mellon University
Mowry, Todd, Carnegie Mellon University
Campbell Jason, Intel Pittsburgh Research
Lee, Peter, Carnegie Mellon University
Pillai, Padmanabhan, Intel Pittsburgh Research
Hoburg, James, Carnegie Mellon University
Gibbons, Phil, Intel Pittsburgh Research
Guestrin, Carlos, Carnegie Mellon University
Kufner, James, Carnegie Mellon University
Kirby, Brian, Carnegie Mellon University
Rister, Ben, Carnegie Mellon University
Derosa, Mike, Carnegie Mellon University
Funiak, Stanislav, Carnegie Mellon University
Burak Asok, Carnegie Mellon University
Sukthankar, Rahul, Intel Pittsburgh Research
This is an abstract for a presentation given at the
13th Foresight Conference on Advanced Nanotechnology
Progress in nanotechnology requires more than the ability to control and manipulate the very small. It also requires the ability to manipulate and control massive numbers of the very small. In this talk, we will describe Claytronics, a form of programmable matter. Claytronics is an ensemble of massive numbers of units each of which has the capability to compute, communicate, sense, and actuate. We are investigating the design and manufacture of the individual units, which we call claytronic atoms, or catoms, and the methods to manage and control the ensemble. Our long term goal is to program the individual catoms so that they will self-assemble into arbitrary dynamic 3D objects.
Realizing this vision requires new ways of thinking about massive numbers of cooperating units. Most importantly, it demands simplifying and redesigning the software and hardware used in each catom to reduce complexity and manufacturing cost and increase robustness and reliability. Consequently, our designs strictly adhere to the ensemble principle: A module should include only enough functionality to contribute to the ensemble's desired functionality. Early results of our research indicate that obeying the ensemble principle allows us to significantly reduce the complexity of each unit. We briefly describe some early results which highlight the effect of the ensemble principle.
Our current prototypes, see attached figure, are built at the macroscale. These units move, yet they have no moving parts. They move using magnetic forces generated by two catoms cooperatively activating electromagnets. We also exploit the ensemble principle to power the catoms. The catoms themselves have no power source. Instead, the catoms cooperate to form a power grid through unary connectors between adjacent catoms. The grid provides both power and ground to each catom.
The goal of the ensemble is to create arbitrary shapes. Traditional approaches use high-dimensional search or gradient based methods to plan the motion of the individual units towards their final assembly. We have developed a method which, inspired from semiconductor physics, uses random motion of voids in the ensemble to stochastically create the desired shape. The motion of an individual catom is not planned, instead the random movement of the voids combined with the creation of voids at a surface (which moves the surface out) and the consumption of a voids (which moves the surface in) effects a desired shape of the ensemble.
The last example we discuss concerns the creation of a hierarchical communication network. The network is used in to support different types of communication between the catoms. Existing algorithms that generate hierarchical networks (e.g. in sensor networks) use techniques which result in a tree with a depth linear in the radius of the represented object, which is simply too large when dealing with the massive numbers of units in a Claytronics object. Our approach creates variable length links which combine to form am optimized network based on a single specified global metric.