Distributed Learning of Hop Count Distributions in Ad Hoc Networks.

This is a study of the feasibility of learning the hop count distribution of a mobile ad hoc network using in-network data. The nodes maintain a histogram of the hop count from the source of all packets received and share the histograms with one another. This can be used to learn the distribution between all pairs of nodes or groups of nodes. The effectiveness of this method and the effect of various factors is shown by simulation. The advantage of this method over theoretical and experimental analysis of the hop count distribution is that it is not tied to specific models of node distribution, propagation, and network protocols, and can be used to learn the distribution in real time. It is useful for the dynamic optimization of methods needed for the efficient and effective operation of an ad hoc network.