Adaptive Clustering in D2D Networks: A Density-Aware Approach for Dynamic Cluster Construction

In device-to-device (D2D) networks, clustering is of critical importance in terms of improving the performance of the network, extending the coverage area, and reducing communication delays. While traditional clustering algorithms require the number of clusters to be determined in advance, this number may be far from optimum in dynamic and variable-density networks. In this study, a method is proposed that dynamically determines the number of clusters depending on the edge density of the network. The proposed method reduces the number of clusters as the network gets denser by using the inversely proportional relationship between the edge density and the number of clusters. Experiments with common clustering algorithms such as K-Means and Gaussian Mixture Model (GMM) are conducted and the results are analyzed. Our results indicated that the proposed method increases the clustering performance. In addition, comparisons are made with the clustering results of Louvain and Girvan-Newman algorithms and the advantages of the proposed approach are highlighted. This study can provide significant contributions especially for applications such as positioning UAVs (Unmanned Aerial Vehicles) as mobile base stations in disaster scenarios.