EdgeAware CHNet A Federated Deep Learning Framework for Adaptive Cluster Head Selection in Scalable IoT Enabled WSNs

The emergence of Internet of Things (IoT)-enabled Wireless Sensor Networks (WSNs) has revolutionized real-time monitoring in various domains, from environmental surveillance to industrial automation. Cluster head (CH) selection is also a complex process to perform efficiently and with low energy consumption, most particularly in a large-scale dynamic network. The paper presents EdgeAware-CHNet, a new fusion-decompose architecture, implemented by deep learning that allows adaptive and intelligent CH selection, information privacy, and minimal energy costs. The proposed system involves deployment of a MobileNetV2-Temporal Convolutional Network (TCN) hybrid model at edge devices; these devices learn the local patterns of data, using which they coordinate updates of data based on the Federated Averaging (FedAvg) algorithm without raw data sharing. Further, the strategies of CH selection are improved using a Deep Q-Network (DQN)-based reinforcement optimization module on the basis of energy efficiency, latency, and feedback on packet delivery. The soft-attention layers reinforce the spatial prioritization of CH candidates, which can make the system dynamic to the topology variances and a dynamic workload. Full simulation demonstrates the superiority of EdgeAware-CHNet in the important performance scales as compared to traditional and learning-based baselines. This model suggested a 96.45% degree of accuracy of CH selection, the network lifetime of 1820 rounds, and a PDR of 97.22, which is considerably higher than such models as LEACH, TEEN, and DQN-CH. Due to the synergistic integration of federated intelligence, reinforcement learning, and edge-aware optimization, EdgeAware-CHNet is a highly efficient and secure framework that can be used to address present-day WSN deployments.