On the distributed binary consensus algorithm in wireless sensor networks

We consider the binary consensus problem over the wireless sensor networks (WSN), where each node of the network initially observes one of two states and the aim of each node is to decide which one of the two states was held by the majority of nodes. In this paper we consider the averaging of a measurement in a WSN composed of N nodes.We evaluate the distributed binary consensus algorithm by deriving the average convergence time of the algorithm. Since this time has been evaluated theoretically for mathematical aspects [1], we complete this work by considering real life environment including packet loss phenomenon and automatic repeat request (ARQ) protocol used in real message transmissions. In addition we compare the new analytical results with those obtained by a hardware emulation of the same binary consensus algorithm under TinyOS and TOSSIM. In performance evaluation, we consider the analysis of the average convergence time for node states. In the simulations, we apply the distributed binary consensus algorithm for fully connected, ring, path, Erdos Reny random, and star-shaped topologies.