Industry 5.0 envisions manufacturing systems that are human-centric, sustainable, and resilient. In this context, the Internet of Everything (IoE) enables integration of devices, people, and processes into a unified digital ecosystem. This paper presents a modular, semantically enriched framework that supports this transition by managing heterogeneous data sources—such as IoT sensors, wearable devices, and smart objects—through a layered architecture. The platform enables real-time data stream processing, semantic interoperability, and secure, context-aware access. Anomaly detection is enabled through a privacy-preserving mechanism based on behavioral fingerprinting and federated learning. The platform supports immersive human-machine interaction via gesture recognition, empowering workers to control and interact with industrial systems. Use cases demonstrate the system’s ability to support gesture-based control and intelligent monitoring, highlighting its potential to enhance adaptability, security, and worker empowerment in Industry 5.0 environments.

The use of wearable devices with inertial sensors for gesture recognition is becoming increasingly common in extended reality and remote control applications. Fast and accurate gesture recognition is critical for real-time services such as industrial control and thus drives the shift of computation toward edge computing. This work aims to optimize deep learning models for direct integration into real-world edge devices. In this context, several models based on convolutional neural networks with different configurations are developed and tested through standard performance benchmarks (accuracy and Macro F1 Score). Moreover, in order to decrease models’ size with minimal accuracy loss, they were quantized using full-integer quantization. Non-quantized and quantized models have been tested on different boards through the STM32 Edge AI Developer Cloud, as an additional benchmark to verify that the proposed models meet speed and accuracy requirements also on low-cost, low-power edge nodes. The results show that the best lightweight configuration with post-training quantization achieves an inference time of 17.31 ms on the STM32L4R9I-DISCO board, while maintaining an accuracy of 95.95 % ± 0.31 %, competitive against the state of the art and making it suitable for integration into industrial control frameworks.

Bluetooth low energy (BLE) Mesh networks are a promising solution for Internet of Things applications, offering scalability, low power consumption, and resilience to node loss. Different factors in BLE Mesh networks may severely affect performance, from network topology, to the position of relay nodes, to the setting of advertising and scanning parameters that regulate message relaying. The relaying process is the backbone of BLE Mesh networks, enabling many-to-many communication where devices can communicate with any other device in the network, including those outside their coverage range. This article investigates the effects on communication performance of relaying implementations with and without preemption of the scanning process. Two possible methods of preemption are implemented, compared, and analyzed: 1) stateless preemption, which starts a new scanning after the previous event has been interrupted to initiate advertising, and 2) stateful preemption, which maintains information on the current scanning across interrupts. The implementation with stateful preemption displayed better latency than the implementation without any form of preemption, while also achieving a good packet delivery ratio and counterbalancing reliability issues of the stateless preemption approach. Based on these results, stateful implementation proved to be an excellent solution for scenarios where it is essential to ensure low latencies, both in low and high traffic topologies.

Seismic wave picking is an essential task the implementation of Earthquake Early Warning (EEW) systems. While artificial intelligence picking methods show excellent accuracy, most were designed for devices with high computational resources. At the same time, distributed approaches for Early Warning systems show promise for the implementation of viable, widespread alert systems. This paper introduces a complete AIoT system for earthquake picking on resource-constrained devices. An algorithm has been developed to enable AIoT devices to switch between a detection mode, in which inferences are run on the data measured by the device, and a transmission mode, in which the device transmits alarms or other event information. To reduce inference times and input window duration, a set of deep learning pickers, called Fast PNet, derived from the PhaseNet model, were developed, achieving 74.4% inference time reduction for the shortest input model, which also showed a significant decrease in computational and power consumption. Despite a slight reduction in picking precision compared to the baseline model (from 0.09 to 0.25 s), detection performance remains high (96% precision and 98% recall). The overall system has been tested on a real event from Central Italy, displaying a 0.16-second picking error for the selected event, besides also showing its ability to reduce the amount of data to be processed for transmission to about 26.5% of the total observed input data, a great benefit compared to fully centralized EEW systems.

Insulin pumps and other smart devices have recently made significant advancements in the treatment of diabetes, a disorder that affects people all over the world. The development of medical AI has been influenced by AI methods designed to help physicians make diagnoses, choose a course of therapy, and predict outcomes. In this article, we thoroughly analyse how AI is being used to enhance and personalize diabetes treatment. The search turned up 77 original research papers, from which we've selected the most crucial information regarding the learning models employed, the data typology, the deployment stage, and the application domains. We identified two key trends, enabled mostly by AI: patient-based therapy personalization and therapeutic algorithm optimization. In the meanwhile, we point out various shortcomings in the existing literature, like a lack of multimodal database analysis or a lack of interpretability. The rapid improvements in AI and the expansion of the amount of data already available offer the possibility to overcome these difficulties shortly and enable a wider deployment of this technology in clinical settings.

Vehicle and pedestrian detection (VaPD) is one of the most critical tasks in an advanced driver assistance system which help the driver to drive safely and save the pedestrian life. VaPD is a typical object detection problem that requires a trade-off among accuracy, speed, and memory consumption. Most existing methods focus on improving detection accuracy, while ignoring VaPD requires real-time detection speed with limited computational resources. Thus, it is of primary importance to study light-weight and real-time VaPD methods for embedded devices, that is hardware platforms with limited computation and memory resources. To deal with these issues, this paper proposes a low-rank (LR) Tiny YOLO v3 architecture that meets the requirements of real-time VaPD on embedded systems. The architecture has been developed starting from Tiny YOLO v3 adopting a convolutional neural network compression technique based on Tucker tensor decomposition, able to reduce the computational complexity of the network. A wide experimentation has been carried out on two embedded platforms, Raspberry Pi 4 and NVIDIA Jetson Nano 2 GB, and two datasets commonly used for VaPD, PASCAL VOC and KITTI dataset, showing the superiority of the LR Tiny YOLO v3 with respect to the state-of-the-art networks in obtaining the best compromise between inference time, accuracy and memory occupancy. Moreover, the proposed architecture meets the requirements of VaPD on embedded systems using only 22% of the memory required by the baseline Tiny YOLO v3 Darknet, and always providing better inference time (36.46 FPS) with only a marginal decrease in accuracy (~2%).

Bluetooth mesh is a recent technology built on the Bluetooth low energy protocol stack architecture, focusing on the Internet of Things. It represents an excellent solution for commercial and industrial lighting applications, though it is still evolving. One of the biggest challenges of the Bluetooth mesh network is the improvement of confirmation messages reception. In a Bluetooth mesh network, determining the Status of the received messages is a critical aspect that can generate unexpected issues when multiple devices respond simultaneously, as it may occur in some lighting applications. This behavior can reduce the probability of message delivery due to collisions, especially when the number of devices in the network increases. This article aims to improve the reliability in receiving confirmation messages in a Bluetooth mesh network by proposing a new technique of spreading Status overtime. To evaluate the proposed technique’s performance, we compare our technique with a Bluetooth mesh network with standard configuration (SC) using real nodes experimental setup. We evaluated our results in terms of packet-loss rate, obtaining 98.84% of the Status received for the network with our optimized configuration and 96.98% for those with the SC. Finally, an in-depth performance evaluation method for the analysis of the lost Status was also conducted.

Solutions and services for e-Health and telemedicine are constantly spreading and becoming increasingly important in the health area thanks to last innovations in electronics, informatics and telecommunications. This work proposes an innovative service for the e-Health oriented to the maximum ease of use and to the sharing of vital signs. The proposal consists in a tele-counseling service based on the WebRTC technology that allows any person residing remotely from medical staff or hospital to directly interact with them. Our solution provides all the common functions of the WebRTC, such as real-time video and audio streams, instant messaging, and file sharing with the only requirement of a traditional web browser. Beyond that, we implemented the real-time transmission and visualization of vital signs and parameters acquired by biomedical sensors connected to the patient's personal device through the RTCDataChannel. Currently our solution involves the installation of a browser extension, but this operation is very simple and may be avoided when WebRTC APIs and browsers will support media streams coming from sensors at the same way as audio and video streams. Our solution demonstrates how web technologies can be applied in the health sector providing very effective services to patients and users which for various reasons have difficulties to travel to hospitals in order to have medical support.

Wireless Sensor Networks are becoming even more a key element in networking and telecommunications especially with the advent of the Internet of Things paradigm, where each single device obtains a unique IPv6 address and is potentially reachable from everywhere through the Internet. Such technologies can be applied in many application fields with great success in terms of optimization of costs and resources, variety of implemented features, level of customization and expandability of each solution. Ambient Assisted Living is definitely one of the most interesting area for WSN and IoT application. In this scenery we propose an applicative example of the use of a IPv6 self configuring WSN with mesh topology. The network is formed by low cost and low power sensor nodes that integrate sub-GHz radio connectivity, belonging to the so-called LLN (Low Power and Lossy Network) scenery. The network stack is fully compatible with the major RFCs about Internet and wireless communications (CSMA, 6LowPAN, uIP, UDP, CoAP, etc.). The case study is a localization system using RSSI feature and does not need additional expensive hardware to be integrated into the nodes. The system proves to be fundamental in indoor environments (houses, clinics, nursing homes) for AAL systems that require localization or tracking of patients and medical equipments. Our tests are performed in extremely complex environment and shown good results localizing targets with an average error of about 2 meters that allows to properly detect the room where targets are located.

A new encryption system is presented for compressed video signals. It employs the twofish algorithm, recently proposed for standardization purposes, as the core cipher but, through simple modifications, it permits to achieve the very high security levels promised by the standard with reduced overhead and processing time. So the new method is particularly adapted for bandwidth limited applications operating in real time. For the sake of clarity, the proposal is tested on H.263+ coded signals, but it can be easily adapted to other formats, of MPEG-x or H.26x type. Performance evaluation is done off-line, by using a simulator, but also implemented on-line in a practical environment. A comparison is made with the previous systems.

Turbo codes have excellent performance at low and medium signal-to-noise ratios (SNR) very close to the Shannon limit, and are at the basis of their success. However, a turbo code performance curve can change its slope at high SNR if the code free distance is small. This "error floor" phenomenon is not acceptable for applications requiring very low values of bit error rates. A knowledge of the free distance and its multiplicity allows one to analytically estimate the error floor. An algorithm for computing the turbo code free distance, based on the notion of constrained subcodes, is described. Some considerations on the free distance distribution of turbo codes with growing interleaver length are also provided.

We present a new algorithm for computing the free distance d/sub free/ of parallel and serially concatenated codes with interleavers, the parameter that dominates the code performance at very high signal-to-noise ratios (SNRs). The knowledge of d/sub free/ allows one to analytically estimate the error floor, which may prevent the use of concatenated codes in applications requiring very low error rates. The algorithm is based on the new notion of constrained subcodes, and permits the computation of large distances for large interleavers without a constraint on the input sequence weight (e.g., up to d/sub free/=40 for a rate-1/3 turbo code with interleaver length N=3568). Applications to practical cases of relevant interest, i.e., (1) the new Consultative Committee for Space Data Systems (CCSDS) standard for deep-space telemetry and (2) the new UMTS/3GPP standard for third-generation personal communications, are presented for the first time. Other related aspects, like a study on the free distance distribution of turbo codes with small/medium interleaver length, and a comparison between parallel and serial concatenation behavior, are also discussed.

A turbo code has been included in the new Consultative Committee for Space Data Systems (CCSDS) channel coding standard for space telemetry. Many future missions with critical link budgets will benefit from its large coding gain. In this paper, the properties of this turbo code are analyzed for symbol synchronization recovery, where the transition density is essential. Key parameters like transition/symbol probability and run-length distribution are studied and compared against practical requirements. Critical working conditions are considered separately and discussed. It is shown that, thanks to the interleaver action, turbo-encoded sequences have very good properties in terms of randomness.