Improvement of Wireless Sensor Network Lifetime via Intelligent Clustering Under Uncertainty

This research proposes a novel framework that integrates intelligent clustering algorithms with “multi-criteria decision-making (MCDM)” techniques to enhance the longevity of WSNs in uncertain environment. Clustering techniques are crucial in WSNs for data aggregation and energy-efficient communication. To create energy efficient network, the proposed framework incorporates intelligent clustering algorithms that perform clustering dynamically in the presence of uncertain parameter and employed MCDM techniques to select of energy efficient CHs for clustering. The intelligent clustering algorithms employ data-driven approaches, machine learning and optimization algorithms to create optimal cluster formation, cluster head selection and energy efficiency. An intelligent clustering mechanism has been made using the Silhouette Index (SI) score. Utilizing the SI score as a benchmark, we conducted optimized clustering with the “Density-Based Spatial Clustering of Applications with Noise (DBSCAN)” algorithm. We employed the elbow method to validate the SI score in conjunction with the k-Means clustering algorithm. By considering uncertainty factors in the decision-making process, the proposed algorithms can effectively adapt the network’s operation to changing conditions, thus improving the overall lifetime of the WSN. Furthermore, the framework integrates MCDM approaches to prioritize cluster formation and cluster head selection criteria. Triangular Fuzzy Numbers are compatible with fuzzy logic systems, which are designed to handle uncertainty and imprecision. The triangular shape aligns well with the concept of fuzzy sets and fuzzy reasoning. Due to this reason TFNs have been considered to represent uncertain parameters. In the end, an experiment relating to WSNs has been studied and the results have been visually presented. It has been noticed that the suggested approach outperformed the “residual energy-aware clustering with isolated nodes (REAC-IN)” model, “Low-Energy Adaptive Clustering Hierarchy Fuzzy Clustering (LEACH-FC)” and “hybrid energy efficient distributed (HEED)” by 38%, 15% and 43%, respectively. The PSO and BFAO applied optimized clustering has been outperformed by 35% and 22%, respectively. To verify the simulation results, testing of hypotheses has been conducted.