Simulation-based cluster head selection using multi-objective particle swarm optimization for IoT: Trade-off between intra-cluster distance and energy efficiency

Abstract

Internet of Things (IoT) has transformed modern life by enabling interconnected systems through distributed sensor nodes that collect data from remote locations such as agriculture, wildlife monitoring, and forestry. However, challenges arise due to the limited battery capacity of sensor nodes, affecting network lifetime and energy efficiency. To improve energy conservation and extend network longevity, clustering techniques play a vital role. Although various clustering protocols have been proposed, many still face the "energy-hole" problem caused by inefficient Cluster Head (CH) selection methods. CHs are responsible for managing intra-cluster communication and tend to exhaust energy rapidly, especially those near the base station due to excessive relay traffic. To address this, a Multi-Objective Particle Swarm Optimization (MOPSO) technique is proposed for CH selection to ensure better energy efficiency and intra-cluster distance in IoT-based wireless sensor networks (WSNs). This technique operates in two phases: cluster formation and CH selection. Euclidean distance is used for clustering member nodes, and high-energy nodes are adaptively chosen as CHs using the MOPSO method. Simulations conducted in MATLAB R2019a validate that the proposed MOPSO outperforms existing algorithms such as LEACH, LEACH-FL, LEACH-FC, KM-PSO, EECHS-ARO, HSWO, and EECHIGWO by mitigating premature convergence and enhancing CH selection accuracy. The proposed technique achieves improvements of 10.02% in packet delivery rate and 9.68% in network lifetime. Results indicate that 105 nodes remain active after the final simulation round, with lower average energy consumption of 0.0570 Joules.