Enhancing Energy Efficiency in Wireless Sensor Grids Through Skeleton-Guided Convolutional Neural Networks for Multi-Stage Delay-Aware Routing and Optimized Network Performance

The Internet of Things (IoT) enables autonomous sensors to collect and transmit data in remote environments via cluster heads to a base station. However, congestion at cluster head nodes often degrades network performance and Quality of Service (QoS). To overcome these challenges, enhancing energy efficiency in wireless sensor grids through skeleton-guided convolutional neural networks for multi-stage delay-aware routing and optimized network performance (EE-WSG-SGCNN) is proposed.

The residual energy and the separation between the sink and nodes in each cell are then used to select a cluster head node (CHN). Then, Raindrop Optimization Algorithm (ROA) is used to optimize energy efficiency by selecting Super-CHNs near the sink with high residual energy. Then, the Supercell Thunderstorm Algorithm (STA) is used to efficiently connect nodes within the cell and route data to the sink. For, data transmission, the Skeleton Guided Convolutional Neural Network (SGCNN) is used to predicting energy consumption and identifying the most efficient transmission routes.

The proposed EE-WSG-SGCNN protocol achieves the highest throughput, reaching 998 kbps at 80 nodes, outperforming other models in large-scale networks. It enhances energy efficiency and reduces delay through optimized routing, ensuring reliable performance in energy-constrained IoT environments.

This work is novel in that it combines skeleton-guided deep learning with a two-stage metaheuristic optimization (ROA and STA) to energy-conscious clustering and routing. This hybrid strategy allows routing decisions in dynamically changing IoT sensor networks that are adaptive, intelligent, and energy efficient.