Gossip Algorithms for Average Energy Estimation in Wireless Sensor Networks

Abstract

Gossip algorithms which belong to a kind of distributed algorithms can be used to compute the (possibly weighted) average of the initial measurements of the nodes at every node in the network. This paper proposes an improved broadcast gossip algorithm to estimate average energy with better accuracy. In this proposed algorithm, we use some companion variables to save the sum, and convergence error is decreasing with the number of companion variables increasing. Convergence of the proposed algorithm is studied theoretically and verified by simulations. Although the estimated value is random, we show that the novel algorithm can converge closer to average than broadcast gossip algorithm in probability.