Energy-Efficient Federated Learning for Internet of Things: Leveraging In-Network Processing and Hierarchical Clustering

Federated learning (FL) has emerged as a promising solution for the Internet of Things (IoT), facilitating distributed artificial intelligence while ensuring communication efficiency and data privacy. Traditional methods involve transmitting raw sensory data from IoT devices to servers or base-stations for processing, resulting in significant communication overhead. This overhead not only increases energy consumption but also diminishes device longevity within IoT networks. By focusing on model updates rather than raw data transmission, FL reduces the volume of data communicated to the base-station; however, FL still faces challenges due to the multiple communication rounds required for convergence. This research introduces an innovative approach that leverages the in-network processing capabilities of IoT devices by integrating a hierarchical clustering routing protocol with FL. This approach enhances energy efficiency through single-round pattern recognition, minimizing the need for multiple communication rounds to achieve convergence. It is envisaged that the proposed approach will prolong the lifespan of IoT devices and maintain high accuracy in event detection, all while ensuring robust data privacy.