Comparison of Averaging Algorithms for Wireless Sensor Networks

We consider a particular class of ad-hoc sensor networks, in which information must be exchanged among nodes in order the value at each node to converge to the average of the sensed quantity. Several algorithms have been proposed to accomplish this task, based on the gossip principle. Gossip is an epidemic protocol, in which nodes iteratively contact one or more neighbors in order to spread their own information throughout the network. Such algorithms have been already investigated from the theoretical viewpoint, but scarcely assessed in practical scenarios. We consider the model of a real network, with the position of each node chosen at random on a two-dimensional map, and apply different kinds of averaging algorithms in such practical case. We assess their performance through numerical simulations, and compare them in terms of convergence speed and complexity.