Hybrid AI-Based Protocol for Energy-Efficient Routing in WSNs Using GA and Q-Learning

Abstract

Wireless Sensor Networks (WSNs) play a critical role in enabling autonomous data ac quisition in diverse application domains, including environmental monitoring, industrial control, and smart agriculture. However, these networks remain severely constrained by limited energy resources, unpredictable node failures, and scalability challenges. To address these issues, this thesis proposes EAGLE (Energy-Aware Genetic and Learning based Engine), a novel hybrid protocol that integrates Genetic Algorithms (GA) and Q-learning to enhance energy efficiency, reliability, and adaptability in WSNs. The proposed architecture employs a hierarchical, layered-zone topology in which sen sor nodes are organized into concentric levels and angular zones. Within each zone, cluster heads (CHs) and backup CHs are selected using a GA-based optimization strategy that favors high residual energy and spatial centrality. Routing between CHs is managed by a distributed Q-learning model, where each CH learns optimal next-hop decisions based on distance-driven reward functions and dynamic network conditions. Extensive simulations were conducted on Google Colab to compare EAGLE with a standard K-means-based clustering method. Performance metrics such as first node death (FND), half node death (HND), cumulative packet delivery, energy distribution fairness, and per-packet energy cost were analyzed. Results show that EAGLE significantly out performs the baseline in terms of network longevity, energy balance, and transmission reliability. EAGLE demonstrates that combining evolutionary computation with reinforcement learning offers a powerful and scalable solution for intelligent WSN routing. This hy brid approach provides a pathway toward autonomous, energy-aware, and self-organizing sensor networks capable of operating efficiently in dynamic environments.