CSR-DMSS,TM: Distributed Computing With an Ad-Hoc Network

Cornell University, Ithaca NY

Investigators

Abstract

Abstract for Dr Alyssa Apsel An integrated low power radio transceiver combined with a backend processor in a single networkable node enables applications that are not possible with traditional sensor node platforms. If such a node can be mass produced in a CMOS process without a costly external crystal such that the network formed by a set of homogenous nodes is robust to network changes or any node failing, then reliable ad-hoc networks may be formed from sets of these nodes with group intelligence akin to a distributed machine. This project focuses on the development and exploitation of a distributed computing machine formed out of a network of integrated single chip microsystems that combine both communication and computation functions with a total power budget that is under 50ìW. Low power performance, critical to long lifetimes is achieved through duty cycling of the transceiver and establishment of a global clock via natural phenomenon of pulse coupled oscillators (PCOs) as well as the use of simple event driven computation. Although each node has limited computing capability, the power of such a network will arise from aggregate capabilities to perform distributed computing tasks. This project explores three issues; the construction of the network from simple nodes, communication within the network, and the computing potential for such a constrained platform. The final goal of this work is to develop a complete network with computing models and primitives capable of exploiting the network?s unique features.