Search

Paddy field irrigation is crucial for high crop yields and food security, yet traditional methods lack precision and adaptability to changing environmental conditions. Existing research is hindered by biased public datasets, inadequate feature extraction, and centralized processing that obstructs real-time decision-making. To address these challenges, this work develops a comprehensive testbed for collecting diverse sensor data from paddy fields under various weather conditions and seasons. We propose a novel hybrid and ensemble feature extraction (HyEn-X) method to enhance data quality and predictive accuracy. In addition, we incorporate federated learning (FL) with hyperparameter tuning and explainable AI (XAI) to validate and optimize the proposed feature extraction approach. This methodology not only reduces noise and irrelevant features but also ensures real-time, localized decision-making for farmers. The proposed methodology improves prediction accuracy, accelerates model convergence, and reduces communication overhead. Furthermore, we have developed a hardware prototype that farmers can use to receive real-time irrigation recommendations. Experimental results demonstrate that the proposed method significantly outperforms baseline feature extraction techniques and validates its effectiveness in practical agricultural settings.

Federated learning (FL) has become increasingly popular and distributes machine learning models among a large set of resource-constraint edge devices without transferring data to the centralized server. FL seeks to learn a common global model using a set of local models, trained in distributed edge devices, coordinated by a central device. However, a set of slow processing edge devices fail to transmit the locally updated model parameters to the centralized device due to a lack of bandwidth connectivity. Additionally, the data residing across edge devices is statistically diverse, i.e., the distribution of non-IID data. Such communication overhead and the variety of data distribution are two major obstacles to the practical application of FL. Motivated by the challenges, in this research work, we develop a self-knowledge distillation-enabled Personalized Federated Learning framework, namely PerFed-SKD. By enabling edge devices to transfer knowledge from older local models to more recent personalized models, the proposed PerFed-SKD speeds up the process by recalling the historical personalized knowledge for the most recent initialized model. Extensive experiments on two publicly available datasets, i.e., MNIST and EMNIST with various data distribution settings demonstrate the outperformance of the proposed PerFed-SKD over the state-of-the-art methods.

Communicable diseases are transmitted through water, food, contaminated surfaces, bodily fluids, air. In such a situation, staying in home isolation for fewer chronic health problems and monitoring health status frequently through Medical Sensors (MSs) is recommended. The use of Artificial Intelligence (AI) in smart consumer electronics and sustainable healthcare has recently demonstrated remarkable results. However, the healthcare domain requires high levels of accountability and transparency for communicable disease prediction and sustainable life in edge networks. This paper aims to present an intelligent healthcare prototype that can identify risk factors according to monitoring parameters by analyzing the Explainable XGBoost (XXGB) model. Using edge networks for sustainable living, we explore the intersection between healthcare and consumer electronics. Initially, the prototype has been trained using the XXGB model over one publicly available dataset related to communicable diseases. Next, the prototype identifies patient risk factors by analyzing real-time monitoring parameters. Simulation results illustrate the efficiency of the proposed XXGB model up to 84.2% accuracy, which is higher than existing models.

Federated Learning (FL) collaboratively trains a model while preserving privacy and providing intelligence. This makes it ideal for Consumer Electronics (CE) applications, involving continuous streaming of data. However, the existence of a single central server for orchestrating the entire process and the participation of an abundant number of clients introduce major bottlenecks like a single-point failure of the server and communication inefficiency. Additionally, in most CE applications, local data is used by the clients as a single dataset, which fails to retrieve deep insights from data features for better decision-making. In this light, we propose a multi-view ensemble FL framework for CE applications in fog networks, namely EnsembleFed. The proposed approach derives multiple views from raw data to gain deeper insights from it and trains a model for each view. Later, the output of these models is ensembled for robust predictions. The distributed fog nodes are incorporated into the FL framework to combat single-point failure risk and improve communication efficiency. Empirical results demonstrate that the proposed approach outperforms state-of-the-art techniques on a real-time dataset in terms of increased accuracy by 4.02%, 4.32%, and 5.27%, compared to FedYogi, FedAdam, and FedAvg, respectively.