Search

Electromagnetic (EM) body models predict the impact of human presence and motions on the Radio-Frequency (RF) field originated from wireless devices nearby. Despite their accuracy, EM models are time-consuming methods which prevent their adoption in strict real-time computational imaging and estimation problems, such as passive localization, RF tomography, and holography. Physicsinformed Generative Neural Network (GNN) models have recently attracted a lot of attention thanks to their potential to reproduce a process by incorporating relevant physical laws and constraints. They can be used to simulate or reconstruct missing data or samples, reproduce EM propagation effects, approximated EM fields, and learn a physics-informed data distribution, i.e., the Bayesian prior. Generative machine learning represents a multidisciplinary research area weaving together physical/EM modelling, signal processing, and Artificial Intelligence (AI). The paper discusses two popular techniques, namely Variational Auto-Encoders (VAEs) and Generative Adversarial Networks (GANs), and their adaptations to incorporate relevant EM body diffraction methods. The proposed EM-informed GNN models are verified against classical EM tools driven by diffraction theory, and validated on real data. The paper explores emerging opportunities of GNN tools targeting real-time passive RF sensing in communication systems with dense antenna arrays. Proposed tools are also designed, implemented, and verified on resource constrained wireless devices. Simulated and experimental analysis reveal that GNNs can limit the use of time-consuming and privacy-sensitive training stages as well as intensive EM computations. On the other hand, they require hyper-parameter tuning to achieve a good compromise between accuracy and generalization.

Federated Learning (FL) systems orchestrate the cooperative training of a shared Machine Learning (ML) model across connected devices. Recently, decentralized FL architectures driven by consensus have been proposed to enable the devices to share and aggregate the ML model parameters via direct sidelink communications. The approach has the advantage of promoting the federation among the agents even in the absence of a server, but may require an intensive use of communication resources compared to vanilla FL methods. This paper proposes a communication-efficient design of consensus-driven FL optimized for training of Deep Neural Networks (DNNs). Devices independently select fragments of the DNN to be shared with neighbors on each training round. Selection is based on a local optimizer that trades model quality improvement with sidelink communication resource savings. The proposed technique is validated on a vehicular cooperative sensing use case characterized by challenging real-world datasets and complex DNNs typically employed in autonomous driving with up to 40 trainable layers. The impact of layer selection is analyzed under different distributed coordination configurations. The results show that it is better to prioritize the DNN layers possessing few parameters, while the selection policy should optimally balance gradient sorting and randomization. Latency, accuracy and communication tradeoffs are analyzed in detail targeting sustainable federation policies.

Smart healthcare relies on artificial intelligence (AI) functions for learning and analysis of patient data. Since large and diverse datasets for training of Machine Learning (ML) models can rarely be found in individual medical centers, classical centralized AI requires moving privacy-sensitive data from medical institutions to data centers that process the fused information. Training on data centers thus requires higher communication resource/energy demands while violating privacy. This is considered today as a significant bottleneck in pursuing scientific collaboration across trans-national clinical medical research centers. Recently, federated learning (FL) has emerged as a distributed AI approach that enables the cooperative training of ML models, without the need of sharing patient data. This paper dives into the analysis of different FL methods and proposes a real-time distributed networking framework based on the Message Queuing Telemetry Transport (MQTT) protocol. In particular, we design a number of solutions for ML over networks, based on FL tools relying on a parameter server (PS) and fully decentralized paradigms driven by consensus methods. The proposed approach is validated in the context of brain tumor segmentation, using a modified version of the popular U-NET model with representative clinical datasets obtained from the daily clinical workflow. The FL process is implemented on multiple physically separated machines located in different countries and communicating over the Internet. The real-time test-bed is used to obtain measurements of training accuracy vs. latency trade-offs, and to highlight key operational conditions that affect the performance in real deployments.

Abstract In device-free radio frequency (RF) body occupancy inference systems, RF signals encode information (e.g., body location, posture, activity) about moving targets (not instrumented) that alter the radio propagation in the surroundings of the RF link(s). Such systems are now getting more attention as they enable flexible location-based services for new smart scenarios (e.g., smart spaces, safety and security, assisted living) just using off-the-shelf wireless devices. The goal of this paper is to set the fundamental signal processing methods and tools for performance evaluation of passive occupancy inference problems that leverage on the analysis of physical layer (PHY) channel state information (CSI) obtained from multiple antennas (spatial domain) and carriers (frequency domain) jointly. To this aim, we consider here a multiple-input multiple-output orthogonal frequency-division multiplexing (MIMO-OFDM) radio interface adopted in high-throughput WiFi networks such as IEEE 802.11n,ac. The proposed approach investigates at first relevant CSI features that are more sensitive to body presence; next, it proposes a space-frequency selection method based on principal component analysis (PCA). Considering an experimental case study with WiFi links, we show that the joint space- and frequency-domain processing of the radio signal quality indicators enable both detection and localization of two independent targets (i.e., human bodies) arbitrarily moving in the surroundings of the transmitter/receiver locations. Experiments are conducted using off-the-shelf WiFi devices configured to extract and process CSI over standard PHY preambles: performance analysis sets the best practices for system design and evaluation.

In this paper, we show the possibility of using the smartphone built-in cellular radio modem to track sudden changes in the environment around it, thus turning the cellphone into a radio-frequency (RF) virtual sensor. In particular, we demonstrate how to isolate anomalous RF patterns by applying time series modeling and analysis of downlink multi-cell radio signals. These RF anomalies may indicate a situation change, namely, a body or object(s), movement in the surrounding of the smart-phone. Unlike Wi-Fi and Bluetooth devices, that can be turned on and off according to the user demands, cellular radios are never really disconnected. Even in idle mode, they carry out continuous and autonomous measurements of the radio channel conditions, namely, the cellular signal quality (CSQ). This is performed in agreement with standardized cell reselection procedures. Body movements or scene changes in general in the surroundings of a cellular device are responsible for small CSQ fluctuations that can be isolated from normal network operations and classified accordingly. The validation of this unconventional RF sensing method is based on extensive measurement campaigns covering a period of one month, using up to four commercial off-the-shelf smartphones. As a practical application case study, we developed a real-time demonstrator that is able to detect body proximity events close to the device and discriminate other body-induced environmental changes in the surrounding of the smartphone. Usage of data analytics tools for passive sensing from cellular signals is a novel topic that shows great potential as paving the way to new applications and research opportunities.

In this paper, we discuss the problem of device-free localization and tracking, considering multiple bodies moving inside an area monitored by a wireless network. The presence and motion of non-instrumented subjects leave a specific footprint on the received Radio-Frequency (RF) signals by affecting the Received Signal Strength (RSS) in a way that strongly depends on people location. The paper targets specifically the modelling of the effects on the electromagnetic (EM) field, and the related inference methods. A multiple-body diffraction model is exploited to predict the impact of these bodies on the RSS field, i.e., the multi-body-induced shadowing, in the form of an extra attenuation w.r.t. the reference scenario where no targets are inside the monitored area. Unlike almost all methods available in the literature, that assume multi-body-induced shadowing to sum linearly with the number of people co-present in the monitored area, the proposed model describes also the EM effects caused by their mutual interactions. As a relevant case study, the proposed EM model is exploited to predict and evaluate the effects due to two co-located bodies inside the monitored area. The proposed real-time localization and tracking method, exploiting both average and deviation of the RSS perturbations due to the two subjects, is compared against others techniques available in the literature. Finally, some results, based on experimental RF data collected in a representative indoor environment, are presented and discussed.

Subject counting systems are extensively used in ambient intelligence applications, such as smart home, smart building and smart retail scenarios. In this paper, we investigate the problem of transforming an unmodified WiFi radio infrastructure into a flexible sensing system for passive subject counting. We first introduce the multi-dimensional channel features that capture the subject presence. Then, we compare Bayesian and neural network based machine learning tools specialized for subject discrimination and counting. Ensemble classification is used to leverage space-frequency diversity and combine learning tools trained with different channel features. A combination of multiple models is shown to improve the counting accuracy. System design is based on a dense network of WiFi devices equipped with multiple antennas. Experimental validation is conducted in an indoor space featuring up to five moving people. Real-time computing and practical solutions for cloud migration are also considered. The proposed approach for passive counting gives detection results with 99% average accuracy.

The adoption of dense wireless sensor networks in industrial plants is mandatorily paired with the development of methods and tools for connectivity prediction. These are needed to certify the quality (or reliability) of the network information flow in industrial scenarios which are typically characterized by harsh propagation conditions. Connectivity prediction must account for the possible coexistence of heterogeneous radio-access technologies, as part of the Industrial Internet of Things (IIoT) paradigm, and easily allow postlayout validation steps. The goal of this paper is to provide a practical evaluation of relevant coexistence problems that may occur between industrial networks employing standards such as Wireless HART IEC 62591, IEEE 802.15.4, and IEEE 802.11. A number of coexistence scenarios are experimentally tested using different radio platforms. For each case, experimental results are analyzed to assess tolerable interference levels and sensitivity thresholds for different configurations of channel overlapping. Finally, the problem of over-the-air spectrum sensing is investigated in real scenarios with heterogeneous industrial networks to enable a cognitive resource allocation that avoids intolerable interference conditions.

Wireless telemetry systems for remote monitoring and control of industrial processes are now becoming a relevant topic in the field of networked control. Wireless closed-loop control systems have stricter delay and link reliability requirements compared to conventional sensor networks for open-loop monitoring and call for the development of advanced network architectures. By following the guidelines introduced by recent standardization, this paper focuses on the most recent technological advances to enable wireless networked control for tight closed-loop applications with cycle times below 100 ms. The cooperative network paradigm is indicated as the key technology to enable cable replacing even in critical control applications. A cooperative communication system enables wireless devices placed at geographically separated locations to act as a virtual ensemble of antennas that creates a virtual multiple-antenna-distributed system. A proprietary link-layer protocol/based on the IEEE 802.15.4 physical layer has been developed and tested in an indoor environment characterized by non-line-of-sight (NLOS) propagation and dense obstacles. The measurements obtained from the testbed evaluate experimentally the benefits (and the limitations) of cable replacing in critical process control.

Wireless sensor networks for critical industrial applications are becoming a remarkable technological paradigm. Large-scale adoption of the wireless connectivity in the field of industrial monitoring and process control is mandatorily paired with the development of tools for the prediction of the wireless link quality to mimic network planning procedures similar to conventional wired systems. In industrial sites, the radio signals are prone to blockage due to dense metallic structures. The layout of scattering objects from the existing infrastructure influences the received signal strength observed over the link and thus the quality of service (QoS). This paper surveys the most promising wireless technologies for industrial monitoring and control and proposes a novel channel model specifically tailored to predict the quality of the radio signals in environments affected by highly dense metallic building blockage. The propagation model is based on the diffraction theory, and it makes use of the 3D model of the plant to classify the links based on the number and density of the obstructions surrounding each individual radio device. Accurate link classification opens the way to the optimization of the network deployment to guarantee full end-to-end connectivity with minimal on-site redesign. The link-quality prediction method based on the classification of propagation conditions is validated by experimental measurements in two oil refinery sites using industry standard ISA SP100.11a compliant devices operating at 2.4 GHz.