A Novel Many-Objective Clustering Algorithm in Mobile Ad Hoc Networks

In mobile ad hoc networks, clustering refers to the process of identifying the set of clusterheads that optimize one or more network objectives. To optimize each objective, the nodes of the network should be evaluated and compared in terms of one or more corresponding attributes. In many-objective problems, as the number of favorable network objectives increases, the number of assessment attributes can increase significantly. Based on such attributes, two clusterhead selection approaches have been proposed: weight-based and dominance-based methods. In the weight-based methods, the large number of attributes in the weight equation reduces the accuracy of the weight factors. In dominance-based methods, the large number of attributes in the comparison process enlarges the Pareto set and reduces the convergence speed. In this paper, we propose an approach that decomposes the main objectives into intermediate sub-objectives in a hierarchical manner. Common sub-objectives can then be estimated based on the measurable node attributes. We combine these sub-objectives, rather than the raw attributes, in the weight equation. By exploiting this approach, we reduce five different objectives to just two sub-objectives for use in our proposed clustering algorithm. The results indicate that the proposed clustering algorithm is considerably more efficient than the well-known weighted clustering algorithm and its fast version, in terms of network objectives.