Informatisches Kolloquium Wintersemester 2008/2009

Montag, 3. November 2008
um 17 Uhr c.t.
Vogt-Kölln-Straße 30
Konrad-Zuse-Hörsaal
Gebäude B

Computational Sensor Networks

Prof. Thomas C. Henderson,
School of Computing, University of Utah

Computational sensor networks (CSN) provide a conceptual framework which offers insight into the design, analysis, development and execution of distributed sensing and actuation systems. The method depends on a set of models describing the constituent components: sensors, actuators, computation, communication, and physical phenomena.

The standpoint from which this work proceeds is that CSNs are measurement systems which are embedded in a continuous phenomenon for which they build or exploit models, and which can perform experiments to validate those models. There should be well-defined measurement goals, as well as error measures, and mechanisms (algorithms) to reduce the error to within a desired tolerance. Furthermore, nodes are generally viewed as equivalent; that is, all have the same computational, sensing, energy, and communication power, run the same algorithms, and are otherwise interchangeable; of course, the roles played by individual nodes in a specific computation may differ.

CSN offers a unique vantage point as well with respect to the physical phenomena in which the system is embedded. Given a valid forward solution for the phenomenon of interest (e.g., the heat equation), it may be possible to formulate questions about the structure of the sensor network as inverse problems. For example, the heat equation gives rise to a set of nonlinear equations whose solution solves the sensor node localization problem.

This talk will describe our work which has mainly addressed the creation of an information layer on top of the sensor nodes. This includes distributed algorithms for leadership protocols, coordinate frame and gradient calculation, reaction-diffusion pattern formation, level set methods to compute shortest paths through the net, and sensor node localization using inverse models.

Kontakt

Prof. Dr. Jianwei Zhang
Telefon 2431