ViRe: Virtual Reconfiguration Framework for Embedded Processing in Distributed Image Sensors [Report]

NESL Technical Report #: 2007-1-1


Abstract: Emergence of new technology in sensor networks, such as low-cost CMOS cameras and others, has introduced more sophisticated sensing modalities such as imaging. Applications involving image processing introduce new challenges to the design of sensor network systems. As the embedded processing becomes more complex, in-situ reconfiguration is seen as the key enabling technology to maintain and manage such systems. Reconfiguration can be used for bug-fixes, introducing new features, and tuning system parameters to the operating environment.

This paper presents the ViRe framework that provides for in-situ reconfiguration of complex image processing applications. Applications, modeled as data-flow graphs, are composed from a library of pre-defined and reusable elements. An efficient run-time system, called the wiring engine, is installed on the nodes to manage the graph and interaction between its elements. It facilitates communication by transferring data in the form of tokens. After initial deployment, the system permits reconfiguration of the graph by allowing modification to the edges and addition/removal of elements. Hence, it is able to support complex graphs comprising of elements with multiple fan-in and fan-out and feedback, while incurring a low memory and execution overhead.

Date: 2007-01-06

Place: UCLA

Public Document?: Yes

NESL Document?: Yes

Document category: Report