Affinity Propagation-driven Distributed clustering approach to tackle greedy heuristics in Mobile Ad-hoc Networks.

Graphical abstract Abstract This paper presents a Gauss Markov Mobility-based Affinity Propagation-driven Distributed (GMM-APD) clustering approach which primarily tackles the greedy clustering heuristics in MANETs. GMM-APD addresses two fundamental short-comings of greedy approach namely: (a) Use of local optima concept and (b) frequent event-driven broadcasts by mapping evolution of the network topology into clusters. Node mobility is considered as the proximity criteria while the node movement pattern is represented as a Gauss Markov (GM) distribution. The temporal dependency of GM distribution is used to compute the proximity criteria which is further optimised by allowing the network to evolve using a modified Affinity Propagation technique; followed by the clustering decisions. An analytical model to explain the dynamics of GMM-APD is also presented describing the occurrences of various events taking place on the timeline of cluster and node. Finally, GMM-APD clustering technique is represented as an Integer Linear Problem formulation focussing on minimising the dominating set. The simulations carried out using NS-2 reveal that GMM-APD generates a quasi-optimal set of cluster heads and exhibits enhanced performance as compared to some of the existing clustering techniques in terms of cluster stability, quality and cost. [ABSTRACT FROM AUTHOR]