A scalable approach for serial data fusion in Wireless Sensor Networks

In Wireless Sensor Networks (WSNs), serial data fusion approaches, in which a parameter of interest is estimated through the serial communication of nodes, have shown their effectiveness over centralized and distributed ones. Nevertheless, they still suffer two major drawbacks: (i) they require the construction of a path passing through every node in the network exactly one time, which is known to be a NP-Complete problem and (ii) they experience poor scalability, which is an important concern in large scale WSNs. In this paper, we tackle these issues by proposing a novel localized serial algorithm, called Peeling Algorithm (PA). In the proposed algorithm, a packet travels serially from node to node, carrying with it the parameter estimate. Each visited node determines locally the next hop for the packet and does not need to store any information about the network topology. This unique feature allows a very good scalability of our approach. We also present a second algorithm, called Enhanced PA (EPA). We discuss their implementations, provide proof of correctness and report on their performance evaluation through an extensive set of simulation experiments using OMNET++ simulator. Our results indicate clearly that our proposed algorithms outperform previously known and existing ones.