Self-organizing network control structures: Local algorithms for forming global hierarchies

We present a distributed clustering algorithm, executed asynchronously by each node using only local information, that builds a hierarchical cluster-based control structure for a mobile wire ss network. The clustering process consists of a we propagation phase for forming and labelling an initial set of clusters based on similarity and number of members, followed by a boundary extension phase for adjusting the size and number of clusters so that tiny clusters (created as artifacts of wave propagation) are absorbed by larger ones. Cluster absorption is also used to assimilate mobile nodes into existing clusters. We assess the performance of the algorithm in terms of the number and composition of clusters produced and the transmission and computation overhead per node. Through simulation, we illustrate that this clustering algorithm is well-suited to moderately mobile networks with limited resources.