Multiple criteria based cluster head selection approach for energy efficient communication in WSN based resource constrained IoT environment

ML approaches have demonstrated the potential to address WSN’s issues. Furthermore, multiple criteria decision making (MCDM) is beneficial in situations such as cluster head (CH) selection in WSNs. This paper presents an approach that integrates ML for the initial phases, followed by MCDM mechanisms in subsequent phases. The proposed approach is evaluated against various well established clustering algorithms using metrics such as energy consumption, node degree, residual energy, sink node positioning, and distance, demonstrating superior performance compared to other algorithms. In this paper, MCDM based methods are compared: Rank Centroid Allocation (RCA) maintains full network death in 6344 rounds, outperforming LEACH (581 rounds), LEACH-C (170), HEED (460), DWEHC (707), DBSCAN (574), FOHCA (5410), and DDMPEA-ANUM (660); Reciprocal Rank Allocation (RRA) provides intermediate performance, with the first node failing at round 63 and the entire network death in round 6169. Summed rank allocation (SRA) performs slightly lower than RCA and RRA, with the first node death at round 58 and the entire network death in round 5639. RCA consumes 0.001371 J of energy per round, which is better than LEACH (0.016605 J), LEACH-C (0.030212 J), HEED (0.008161 J), DWEHC (0.011572 J), DBSCAN (0.006618 J), FOHCA (0.001848 J), and DDMPEA (0.011572 J).