Performance evaluation of gossip algorithms in WSNs using outage probability

Abstract

This paper studies the performance of asynchronous consensus algorithms in wireless sensors networks (WSNs). The application of such distributed consensus algorithms for the estimation of parameters or detection of events in WSNs has attracted increased attention recently. In this work the reliability of the wireless links between nodes is modeled by the outage probability. The outage probability formulation embeds the effect of small and large scale channel fading and allows a more precise performance evaluation of consensus algorithms for wireless applications. We investigate the performance of the following algorithms: pairwise, neighborhood, geographic and broadcasting. Simulation results show which algorithm minimizes the number of exchanged messages and/or obtains a faster convergence for different scenarios. The best strategy depends both on the wireless channel characteristics and the network topology.