Distributed source coding and dispersive information routing: An integrated approach with networking and database applications

This paper considers the design of optimal joint compression-routing schemes for networks with correlated sources and multiple sinks. Such a setting is typically encountered in sensor networks. It is sometimes tempting to assume that an optimal distributed compression scheme followed by standard routing (as designed for independent sources) would be nearly optimal. We instead propose a joint design approach that integrates distributed source coding with a mechanism called dispersive information routing. Unlike network coding, dispersive information routing can be realized using conventional routers without recourse to recoding at intermediate nodes. We also point out the direct connections between dispersive information routing and a related problem in sensor network databases, namely, fusion coding for selective retrieval. We propose an efficient practical design strategy, variants of which are adopted for each of the two problems. Simulation results provide evidence for substantial gains over conventional schemes.