Machine-learning-enhanced glowworm swarm optimization for energy-efficient multi-hop routing in wireless sensor networks

Energy-efficient wireless sensor networks (WSNs) are critical for modern intelligent applications, yet maintaining balanced battery power across the network remains a persistent challenge. Numerous algorithms have been proposed to address this issue, but notable research gaps still exist. This study introduces a novel multi-hop routing framework designed to enhance energy efficiency and prolong network lifetime. The framework integrates six key components: a network model, a radio and energy model, cluster deployment using a low-energy adaptive clustering hierarchy (LEACH), cluster-head selection via a glowworm swarm optimization algorithm, multi-hop routing through a Taylor-based cat-salp swarm method, and machine-learning driven analysis employing both k-means and hierarchical clustering. The simulation results demonstrate superior performance over existing state-of-the-art approaches, achieving higher total residual energy, reduced end-to-end delay, lower packet drop rates, improved packet delivery ratio, and greater overall throughput. Hierarchical clustering with Ward linkage achieved an accuracy of 96.76% with an R 2 value of 0.9852 for packet drop and an accuracy of 97.37% with an R 2 value of 0.9831 for network throughput, while K-means clustering attained an accuracy of 96.25% with an R 2 value of 0.9725 for packet delivery ratio. The novelty of this work is the synergistic integration of optimization-assisted machine learning clustering with adaptive multi-hop routing, enabling dynamic and energy-aware cluster head selection while explicitly minimizing both intra- and inter-cluster communication overhead, an aspect largely unaddressed in conventional WSN protocols.