Yoyok Heru Prasetyo Isnomo, M. Muladi, Hakkun Elmunsyah
Clustering techniques in Wireless Sensor Networks (WSNs) effectively reduce communication energy consumption, yet they often give rise to isolated node issues. To overcome the challenges of isolated nodes and unstable clusters, we introduce MCACOF method, an adaptive clustering framework designed to improve Cluster Head (CH) selection, enhance cluster resilience, and energy efficiency and to extend operational lifetime of WSNs across diverse node densities. This method uses spectral clustering techniques and E- Silhouette evaluation in the pre-clustering stage to optimize the determination of the number of clusters. CH selection is based on four main criteria optimized using gradient descent. MCACOF involves an E-Silhouette score-based cluster reconstruction process to evaluate and relocate unsuitable Cluster Members (CMs). Nodes identified as isolated nodes are reconnected through relay-based multi-hop sub-clusters, maintaining participation in data delivery and improving cumulative throughput. Simulations were conducted in six scenarios with uniformly deployed nodes and a centrally located Sink in each case, supplemented by two additional scenarios with non-uniformly deployed nodes where the Sink was placed at five distinct static positions. Across all these evaluated scenarios, MCACOF consistently outperforms REAC-IN, SFC, and MCDM, yielding lower node isolation, improved energy conservation, higher cumulative throughput, and extended network lifetime. © 2013 IEEE.
Universitas Negeri Malang, Electronic Systems Engineering Technology, Malang, 65145, Indonesia; Universitas Negeri Malang, Informatics Engineering Education Study Program, Malang, 65145, Indonesia
