In multi-cell wireless networks, providing intelligent services via federated learning (FL) becomes more challenging due to multi-level distribution shifts across clients and regions, as well as additional communication delays among edge and cloud servers. To address these issues, we propose SplitOMC, a split learning framework that integrates overlapping-area clients and a multi-exit neural architecture to jointly handle (i) client-preferred, (ii) out-of-preference, and (iii) out-of-region tasks. By strategically leveraging clients in overlapping regions, SplitOMC accelerates training without excessive backhaul communication, while maintaining both personalization and generalization. We theoretically analyze the convergence behavior of the proposed algorithm, ensuring performance stability under heterogeneous data and communication conditions. Extensive experiments on MNIST, CIFAR-10/100, and a real-world Jetson Nano testbed demonstrate that SplitOMC consistently achieves faster training and inference with improved accuracy compared to state-of-the-art methods. In particular, the framework shows robustness in resource-constrained and unstable network environments, highlighting its practical value for next-generation wireless intelligent services.

The construction industry remains one of the most hazardous sectors, requiring innovative solutions to safeguard workers, especially in complex, dynamic outdoor environments. Digital Twin (DT) technology offers promising capabilities for real-time safety monitoring through virtual replicas of physical systems. However, existing DT frameworks rarely integrate comprehensive safety monitoring via a centralized model consolidation mechanism, such as Federated Learning (FL), which is explicitly tailored for multi-worker scenarios. Addressing this challenge, this paper proposes a FL-based Safety Monitoring Digital Twin (SMDT) framework designed to enhance multi-worker safety in resource-constrained settings. This enables real-time safety monitoring and control by representing on-site workers as virtual objects within a synchronized DT environment. A dynamic node selection mechanism based on client performance is employed to optimize global model convergence in FL. To validate the proposed approach, an edge computing-based experimental testbed using actual Raspberry Pi devices was implemented, using real-world construction safety data including worker status, weather conditions, and building structural parameters. Experimental results demonstrate the effectiveness of the proposed framework in significantly improving safety predictions and real-time monitoring efficiency. This research establishes a foundational work towards safer construction sites through intelligent, synchronized safety monitoring systems.

The rapid growth of the Federated Internet of Things ecosystem has introduced new challenges in achieving seamless connectivity and interoperability across heterogeneous IoT networks. The presence of heterogeneous platforms and protocols creates significant obstacles for effective communication among cross-silos federated IoT nodes. To tackle this challenge, we have developed Heterogeneous Federated Internet of Things (Hetero-FedIoT), an innovative rule-based interworking architecture enabling interoperability and seamless connectivity among heterogeneous federated IoT networks (oneM2M, OCF and EdgeX). Hetero-FedIoT offers a two-faceted solution to address these challenges. Firstly, it incorporates a rule-based interworking mechanism that fosters effective collaboration among Hetero-FedIoT networks. Additionally, it introduces a novel aggregation function capable of achieving accelerated convergence, effectively handling both system and statistical heterogeneity. By leveraging device proxies, Hetero-FedIoT enables interoperability among heterogeneous FedIoT networks by translating protocols from platform-native formats to a common format and vice versa. As a result, collaborative model training can be seamlessly conducted without the need to consider underlying frameworks. Additionally, the novel aggregation algorithm employed by Hetero-FedIoT empowers nodes to customize the complexity of local models according to their communication and computation capabilities. This is accomplished through the dynamic adjustment of hidden channel widths, ensuring that the overall performance of the global model remains unaffected. This groundbreaking Hetero-FedIoT architecture establishes a foundation for enhanced interoperability and optimal performance. Extensive evaluation of Hetero-FedIoT has demonstrated superior computational and communication efficiency over baseline schemes. The Hetero-FedIoT system revolutionizes decentralized training under heterogeneous conditions, fostering widespread adoption.

Abstract Cognitive disorders affect various cognitive functions that can have a substantial impact on individual’s daily life. Alzheimer’s disease (AD) is one of such well-known cognitive disorders. Early detection and treatment of cognitive diseases using artificial intelligence can help contain them. However, the complex spatial relationships and long-range dependencies found in medical imaging data present challenges in achieving the objective. Moreover, for a few years, the application of transformers in imaging has emerged as a promising area of research. A reason can be transformer’s impressive capabilities of tackling spatial relationships and long-range dependency challenges in two ways, i.e., (1) using their self-attention mechanism to generate comprehensive features, and (2) capture complex patterns by incorporating global context and long-range dependencies. In this work, a Bi-Vision Transformer (BiViT) architecture is proposed for classifying different stages of AD, and multiple types of cognitive disorders from 2-dimensional MRI imaging data. More specifically, the transformer is composed of two novel modules, namely Mutual Latent Fusion (MLF) and Parallel Coupled Encoding Strategy (PCES), for effective feature learning. Two different datasets have been used to evaluate the performance of proposed BiViT-based architecture. The first dataset contain several classes such as mild or moderate demented stages of the AD. The other dataset is composed of samples from patients with AD and different cognitive disorders such as mild, early, or moderate impairments. For comprehensive comparison, a multiple transfer learning algorithm and a deep autoencoder have been each trained on both datasets. The results show that the proposed BiViT-based model achieves an accuracy of 96.38% on the AD dataset. However, when applied to cognitive disease data, the accuracy slightly decreases below 96% which can be resulted due to smaller amount of data and imbalance in data distribution. Nevertheless, given the results, it can be hypothesized that the proposed algorithm can perform better if the imbalanced distribution and limited availability problems in data can be addressed. Graphical abstract

Greenhouses are a productive system that allows us to respond to the growing global demand for fresh and healthy food throughout the year, but the greenhouse environment is not easily controlled because its climate parameters are interrelated. However, the numbers of the actuator are operated parallelly to maintain the greenhouse environment; as a result, the energy consumption of greenhouses is high. In this study, we presented the optimization module by considering the outdoor environment with the aim of minimum energy consumption. Metaheuristic-based differential evaluation (DE) is used to optimize the climate parameters by considering indoor and outdoor environmental constraints. Furthermore, the long short-term memory (LSTM)-based inference model is offloaded on the Internet of Things (IoT) device to predict the next environmental situation. The objective function selects the optimal parameters within user preferences with minimum energy consumption based on the inferred parameter value. The open-source software framework IoTivity, implementing open connectivity foundation (OCF) technical standards, is used for the real-time connection between IoT devices and the IoT platform. Greenhouse owners can set the preferences based on the requirements of plants in the greenhouse by using a smart and remotely accessible Android-based interface. A fuzzy logic-based control module operates on an IoT device that maps the optimized parameters with the actuator and operates accordingly. The proposed model is analyzed, and the performance is evaluated in terms of energy consumption for each climate parameter and actuator in the greenhouse. The results show that the proposed mechanism saves 36% of energy.