MIT Department of Electrical Engineering & Computer Science

Computational Matter: A New Frontier for Engineering Design

Andrew A. Berlin
Xerox Palo Alto Research Center

Thursday, April 29, 1999
4:30 PM (refreshments 4:15)
Room NE43-518
EECS Special Seminar

Abstract

A key research challenge for the next century will be finding ways to artfully blur the boundaries between the physical world and the world of computation. As batch fabrication technologies such as MEMS make it possible to distribute dense arrays of sensors and actuators throughout the physical world, and computation becomes inexpensive and ubiquitous, a new setting for information technology emerges. In this "computational matter" regime, products and processes rely on massively parallel, fine-grained interactions between computation and matter to achieve their goals.

Making the vision of "computational matter" a reality requires developing ways to coordinate the actions of a large number of micro-scale sensors and actuators in a way that substantially influences the macro-scale world. In this talk, I'll focus on two novel approaches to achieving this micro/macro coupling. One approach involves making small modifications to the dynamic behavior of individual structural members so as to stabilize them against Euler buckling. A prototype system employing this approach, that I constructed while at MIT, uses an array of piezo-electric actuators to achieve a factor of 5.6x increase in the load-bearing capability of compressively-loaded columns. Another approach to micro/macro coupling is to create an active surface that uses tiny directed jets of air, each with an associated MEMS valve, to levitate and transport a planar object (such as a sheet of paper) at high speeds without contacting it. To prototype this approach, we have developed a hybrid fabrication process that combines aspects of MEMS with aspects of PC-board processing, and are using that process to construct an array of 576 valves and 32,000 grey-level sensors, integrated with parallel computation, as a testbed for parallel algorithm development.

As we enter the 21st century, the central challenge of the "computation/matter" frontier is shifting from creating basic fabrication technologies to developing system-level engineering paradigms: developing artful ways to architect, design, and program massively-parallel, real-time systems that span the computation/matter boundary.

HOSTS: Professor G. Sussman, Professor M. Schmidt, & Professor R. Brooks