Mathematical Optimization of Energy-Efficient Wireless Sensor Networks: A Genetic Algorithm-Based Framework for Cluster-Head Selection and Reliable Data Transmission

Amer O. Abu Salem
Department of Computer Science, Faculty of Information Technology, Zarqa University, Zarqa, Jordan

Hebatullah Khattab
Department of Computer Science, Faculty of Information Technology, Zarqa University, Zarqa, Jordan

Louai A. Maghrabi
Department of Software Engineering, College of Engineering, University of Business and Technology, Jeddah 21448, Saudi Arabia

Abstract

Clustering protocols derived from Low-Energy Adaptive Clustering Hierarchy (LEACH) prolong the lifetime of wire- less sensor networks, but the cluster head remains a single point of failure: when it runs out of energy in the middle of a round, the readings it has buffered are lost and its members are orphaned.  In earlier work we proposed LEACH-GV, which elects cluster heads with genetically tuned weights and assigns each one a vice cluster head that takes over after the head dies. That scheme improved the half-dead-node lifetime but brought the first node death forward. This paper traces that weakness to the reactive trigger of the vice and proposes LEACH-AGV, an adaptive genetic vice-cluster-head protocol. LEACH-AGV combines a distributed election weighted by residual energy, distance to the sink and node density; a vice chosen for its ability to finish the round; and a proactive handover that transfers the cluster-head role at a frame boundary once the head’s energy falls below a safety multiple of its frame budget. A genetic algorithm tunes the weights and the safety factor offline on training deployments.  LEACH-AGV was evaluated on 30 unseen deployments in each of three scenarios with a packet-level energy model, and compared with LEACH, LEACH-G, LEACH-V and LEACH-GV. It reduced readings lost to cluster-head failure by at least 97.9% and delayed the first node death by 2.5–5.2% relative to LEACH and by 1.9–5.9% relative to LEACH-GV. All of these improvements were significant under Holm-corrected Wilcoxon tests. Energy per delivered reading stayed practically the same, and half-network lifetime showed no significant loss. An ablation study shows that the proactive handover is the essential component, and that genetic tuning provides one half to two thirds of the stability gain.

Keywords: Wireless Sensor Networks, LEACH, Genetic Algorithm, Cluster Head Selection, Vice Cluster Head, Fault Tolerance, Energy Efficiency, Network Lifetime