In large-scale resource-constrained systems, such as sensor networks, global objectives should be achieved through in expensive local interactions. A well-known technique satisfying these requirements is information potentials. Information potentials map local information to values that are more meaningful within the global context of the network. For example, the local information may be the node's surrounding density, and the otential could be a relative value that states how high (or low) this density is with respect to other areas in the network – to facilitate, for instance, load balanced routing.
In this talk, I will present some of our recent work aimed at uncovering the deep connection between the aggregation scope of information potentials (which determines how far, or close, information is disseminated) and the network's Laplacian (which captures the connectivity graph of the network). Our results, based on linear algebra and spectral graph theory, have an important practical implication: sensor nodes would be able to shape the landscape of information potentials based on the unique topological properties of their network. This ability would allow, for instance, the use of fully-distributed low-cost communication to prevent greedy routing techniques from getting stuck on undesirable local maxima.
Marco Zúñiga is an Assistant Professor of Computer Science at Delft University of Technology, The Netherlands. He obtained his B.S. in Electrical Engineering from the Pontificia Universidad Catolica del Peru, and his M.S. and Ph.D. in Electrical Engineering from the University of Southern California, Los Angeles – in May 2002 and December 2006, respectively. After his PhD, he joined Xerox Research Labs (2006-2008) as a member of the research staff, and then, joined the National University of Ireland, Galway as an IRCSET Fellow (2008-2010) and the University of Duisburg-Essen, Germany as a Senior Researcher (2010-2012). His research interests are in the areas of cyber-physical systems, embedded networks and mobile computing.
PDF-Announcement: click here