ITR: Computation and Communication in Sensor Webs
Project Award Date: 08-01-2003
We will advance sensor web research by prototyping a machine model that integrates computing and communications requirements. The prototype will enable
(a) accurate distributed sensing;
(b) multimode data fusion;
(c) transformation from one domain to another;
(d) extraction of key information; and
(e) detection and circumvention of faulty sensors in ultra large arrays.
We will define a new integrated communications/computation machine model for sensor web processors in terms of a Meta instruction set architecture (ISA). We are defining the Meta ISA to support primitive system operations, such as domain selection, data reduction, and broadcast. The Meta ISA supports attribute-based temporal and query processing and enables consistent, reliable system-level operation for dynamically changing numbers and locations of sensors. It will also support multimode data fusion and provide rich semantics that support composition to form other computational models of interest in sensor webs.
We will develop a Meta ISA Virtual Machine to implement the ISA. The virtual machine approach supports tightly integrated communications and computation through dynamic reprogramming, domain partitioning, expandability, and low power operation. We propose to prototype our virtual machine using asynchronous state machines to further reduce power consumption. With this approach, sensor nodes minimize power draw by reducing the clock circuitry within the node. We plan to prototype the Meta ISA virtual machine in reconfigurable FPGA's within our laboratory.
A new link layer protocol will be defined that forms a virtual global bus independent of exact sensor locations, types, or numbers. The protocol is based on attribute-based naming, allowing multiple domains to be dynamically created and destroyed. We propose this approach because it takes advantage of the law of large numbers in increasing system connectivity, while eliminating power intensive multi-hop route explorations and maintenance of routing tables in sensor node memories.
The broader impacts of this work include applications to networking and wireless devices, computer architecture, device technologies, and higher-level abstract embedded systems design. This work will lead to future collaborations, both within the department and outside it, between the investigators and researchers in these areas. In addition to generating new concepts and approaches to sensor web design, this effort supports our educational goal of establishing a focus area in engineering of computer-based systems.
Primary Sponsor(s): NSF