ASPIRE: Antipodal Staged Processing in Role-adaptive  Embedded Systems

This project investigates how to leverage the complimentary features of a large variety of processors, radios and sensors (low-power vs. power-efficient) to design sensor nodes that can operate over a high dynamic range. With such functionality, nodes can dynamically assume roles inside a sensor network, switching from cluster head to leaf nodes in a power efficient manner.

As part of this project, ENALAB has been designing models for sensor node architectures that analyze the asymptotic lifetime properties with respect to the different components and overheads. The ultimate goal is to design a new interconnect architecture that would allow the node to dynamically switch between low-end and high-end processors,  low-bandwidth (low -power) and high-bandwidth (power-efficient) radios and a wide variety of sensors. With such an architecture a node is expected to leverage a highly dynamic range of operation from simple sampling tasks to higher performance data processing.

The main contributions of this research include:

• A detailed asymptotic lifetime model that analyses event-driven and trigger driven operation with respect to application properties, sensor event arrival rates and power mode transition overheads. See [1] and [3] for the details.
• MATSNL, a MATLAB tool for performing lifetime analysis based on the developed models. MATSNL computes the energy expenditure breakdowns for existing and hypothetical node architectures. It can be used to evaluate components and conceptual designs, compare the performance of existing sensor nodes against certain application properties or better understand how energy gets expended in various power-saving or communication schemes.
• A conceptual node architecture composed of high-end/low-end processors, radios and sensors. This demonstrates how transitioning between modes can conserve energy resources and allows you to estimate the energy budgets and overheads of the required interconnect for such an architecture. For more details see [2] below. 

The project has conducted/collected detailed power measurements from various node architectures including, Intel's iMote2 and its VGA camera module developed at Yale, Yale's XYZ sensor node, the mPlatform from Microsoft, Telos and Mica2.

Node Lifetime Analysis Tool (MATSNL)

MATSNL is a MATLAB tool for performing node lifetime analysis based the node lifetime models we have developed. For the most recent documents and code downloads, please visit our MATSNL site and the MATSNL Manual.

People

Deokwoo Jung, Ph.D Candidate

Andreas Savvides, Principal Investigator

Publications:

[1] D. Jung, T. Teixeira and A. Savvides, Sensor Node Lifetime Analysis: Models and Tools, Proceedings of ACM Transactions on Sensor Networks, February 2009 (MATSNL slides)

[2] D. Jung, T. Teixeira and A. Savvides, An Energy Efficiency Evaluation for Sensor Nodes with Multiple Processors, Radios and Sensors, proceedings of IEEE Infocom 2008  (slides)

[3] D. Jung, T. Teixeira, A. Barton-Sweeney and A. Savvides, Model- Based Design Exploration of Wireless Sensor Node Lifetimes,  Proceedings of the Fourth European Conference on Wireless Sensor  Networks, EWSN 2007, January 29-31, Delft, Netherlands (slides)

This project is sponsored in part  by the National Science Foundation CISE under contracts 0615226 and 0448082. ASPIRE is a collaborative project between UMASS Amherst (Deepak Ganesan, Mark Corner and Prashant Shenoy), UCLA (Mani Srivastava) and Yale (Andreas Savvides).