Browse Theses

A paintable computer is defined as an agglomerate of numerous, finely dispersed, ultra-miniaturized computing particles; each positioned randomly, running asynchronously and communicating locally. Individual particles are tightly resource bound, and processing is necessarily distributed. Yet computing elements are vanishingly cheap and are regarded as freely expendable. In this regime, a limiting problem is the distribution of processing over a particle ensemble whose topology can vary unexpectedly.

The principles of material self-assembly are employed to guide the positioning of “process fragments”—autonomous, mobile pieces of a larger process. These fragments spatially position themselves and re-aggregate into a running process. We present the results of simulations to show that “process self-assembly” is viable, robust and supports a variety of useful applications on a paintable computer.

We describe a hardware reference platform as an initial guide to the application domain. We describe a programming model which normatively defines the term process fragment and which provides environmental support for the fragment's mobility, scheduling and data exchange. The programming model is embodied in a simulator that supports development, test and visualization on a 2D particle ensemble.

Experiments on simple combinations of fragments demonstrate robustness and explore the limits of scale invariance. Process fragments are shown interacting to approximate conservative fields, and using these fields to implement scaffolded and thermodynamic self-assembly. Four applications demonstrate practical relevance, delineate the application domain and collectively illustrate the paintable's capacity for storage, communication and signal processing. These four applications are Audio Streaming, Holistic Data Storage, Surface Bus and Image Segmentation. (Copies available exclusively from MIT Libraries, Rm. 14-0551, Cambridge, MA 02139-4307. Ph. 617-253-5668; Fax 617-253-1690.)