Optimizing Low Carbon Sustainable Environmental Monitoring With Consumer Technology: An IoT-Driven Federated Learning Approach for Edge Computing Optimization

Transmitting data from low-carbon sustainable environmental monitoring systems to a central server for processing is inefficient, especially when utilizing consumer devices inside IoT networks. The large volume of data generated by numerous sensors and consumer technology devices significantly increases the bandwidth required for data transmission. To address these challenges, this study proposes an adaptive edge computing-based federated learning (FL) technique that leverages IoT features and consumer applications to enhance the effectiveness of low-carbon sustainable environmental monitoring. The proposed model processes data directly on peripheral consumer devices, adjusting the global model’s aggregation frequency based on each device’s available resources. This reduces latency and computational load, optimizing bandwidth usage. The model’s performance is compared to state-of-the-art techniques such as Federated Averaging (FedAvg), Hierarchical Federated Edge Learning (HFEL), and IoT Convolutional Neural Network Federated Learning (IoT-CNN-FL). The proposed methodology offers substantial accuracy, efficiency, and cost-effectiveness improvements, particularly for consumer applications that rely on IoT devices. The approach minimizes bandwidth usage and processing costs by transmitting only updated model parameters to the central server, accelerating convergence and making it highly suitable for low-carbon environmental monitoring systems.