CIF: Small: An Integrated Framework for Distributed Source Coding and Dispersive Information Routing

University Of California-Santa Barbara, Santa Barbara CA

Investigators

Abstract

Distributed source coding and routing are strongly motivated by high density sensor networks with promising applications in numerous scientific and engineering disciplines, where it is critical to minimize resource requirements for data communication. This research project stems from the observation that traditional separate treatment of distributed coding and routing is only optimal under strict simplifying assumptions that are often invalid. Major theoretical and practical challenges emerge once these simplifying assumptions are removed, including derivation of the fundamental performance bounds, and optimal practical system design. The degree to which distributed source coding will be practically applicable to future network environments, crucially depends on the development of such technology. This research project formalizes the tradeoffs that underlie distributed coding of sources with information routing over multiple sink networks, and develops system paradigms that allow joint optimization. Having established that optimality requires the ability to disperse fragments of a source information to various sinks, and to allow related but unrequested sources to convey information to any particular sink, this research develops the "dispersive information routing" paradigm and its integration within the distributed coding system. The main research thrusts include: foundation and performance bounds (derivation of the theoretical foundation from source coding, estimation and information theory principles); joint encoder-router-decoder design (development of efficient optimization tools for joint design of all system components); distributed storage (advances on the closely related problem of data storage and retrieval on a network); large scale distributed coding (eliminating the main complexity bottlenecks for distributed coding in very dense sensor networks).