Fault tolerance and energy efficiency of data aggregation schemes for sensor networks

A sensor network is modelled as a platform to gather correlated data from sensor nodes and transfer the data to a sink node over a multi-hop network. Two coding strategies are analyzed: Slepian-Wolf coding and explicit communication. We show that while Slepian-Wolf coding is more energy efficient than explicit communication, it is less tolerant to link failures since decoding the encoded data from one node depends on data from other nodes. For the case of explicit communication, we propose multipath routing schemes based on finding k spanning trees such that all nodes remain connected to the sink after the failure of fewer than k links. We show that clustering schemes with large and small cluster size perform well for high and low correlation respectively even for the multipath case.