arXiv subjects

Stefano Savazzi

Publications and source records attributed to Stefano Savazzi.

At least 19 recordsLinked to original sources

arXiv2026

EM Informed Holographic Imaging via Unrolled Deep Networks

Smart Radio Environments (SREs) are a foundational paradigm for the Internet of Everything (IoE), in which dense, phase-coherent antenna arrays form a shared infrastructure for integrated sensing and communication (ISAC). Radio-Frequency (RF) holography is a core sensing building block for SREs: it reconstructs a volumetric map of the electromagnetic (EM) scattering scene from phase-sensitive field measurements, an infrastructural snapshot from stray RF radiation, without dedicated sensors. To make reconstructions accurate and rapidly adaptable, this paper adopts algorithm unrolling, in which the iterations of a classical holographic solver become the layers of a compact, trainable deep network that preserves the EM physical interpretation and learns only a few parameters from limited data. Building on the unrolled Iterative Shrinkage-Thresholding Algorithm (ISTA), namely Learned ISTA (LISTA), this paper first proposes a Weighted LISTA (W-LISTA) that preserves the EM forward model and learns a spatially-varying regularization that steers the sparsity prior towards target shapes consistent with the deployment. Second, the Low-Rank Weighted LISTA (LoRaW-LISTA) applies a low-rank adaptation (LoRa) of the holographic operator to compensate for model mismatch from linearized EM approximations. Both methods are validated on full-wave EM simulations and on a 2.45 GHz indoor campaign with human-body phantoms, improving accuracy and resolution over baselines. Combining the spatially-varying regularization of W-LISTA with the LoRa adaptation of the EM model yields superior reconstruction where classical iterative solvers fail. The proposed tools are rapidly adaptable building blocks for the SRE sensing layer, whose reconstructions, up to body-shape imaging, remain privacy-preserving by design.

eess.SP↗

arXiv2026

A Body-of-Revolution Human Model for RF Sensing with Measurement-Driven Calibration for Indoor Environments

Model training for Device-Free Localization (DFL) and Radio-Frequency (RF) sensing systems heavily relies on large-scale datasets, which are costly and time-consuming to obtain through measurements across different environments and sensing configurations. Lightweight yet physically consistent propagation models are therefore critical for efficient generation of realistic RF sensing data. This paper presents an RF sensing prediction approach for indoor environments based on a Body of Revolution (BoR) human model. A fast 2.5-Dimensional Finite Element Method (2.5-D FEM) is proposed for computing the scattering fields of a human-like BoR model under the excitation of a vertical polarized dipole. Through comparisons, the proposed BoR model is shown to preserve scattering characteristics close to 3-D human bodies while yielding a smaller computational cost compared to a simple cylindrical model. A measurement-driven background-field modeling approach is further introduced for practical indoor applications, accounting for the complex propagation effects of indoor environments implicitly. Comparing with measurements of a typical indoor DFL scenario, the proposed approach achieves approximately 85% prediction accuracy and reproduces the spatial Received Signal Strength Indicator (RSSI) variations observed in practice, proving its potential for RF sensing prediction and large-scale database generation at a fraction of the computational cost required for full-wave simulations.

eess.SP↗

arXiv2026

WiFi-Based People Counting Using Beam-Steerable Antennas: A Test-bed Study

Ubiquitous perception through RF signals is a pivotal opportunity for future technology: it enables personalized services such as smart living, remote healthcare, automated logistics or interaction through free-space gestures. The ubiquity of Wi-Fi and cellular networks presents a promising platform for the development of innovative sensing tools. Future standards will also introduce dedicated sensing features which, for example, will allow routers to work as frequency modulated continuous wave radios targeting radar applications. Most of the current chip designs support ad-hoc firmware for CSI extraction with MIMO arrangements of the transmitter (TX) and receiver (RX) antennas and OFDM subcarriers. The CSI describes the phase shift and amplitude attenuation of multiple propagation paths on each subcarrier. The latest IEEE 802.11be standard (Wi-Fi 7) offers a wider subcarrier bandwidth of 160MHz (up to 320MHz), providing at least 120 usable pilot subcarriers for CSI or CIR estimation. Additionally, Wi-Fi signals have been recently exploited to track daily human movements and behaviors, while Wi-Fi signal variations have been shown to differ between different people and can consequently be used for their re-identification.

eess.SP↗

arXiv2026

On the Tradeoffs of On-Device Generative Models in Federated Predictive Maintenance Systems

Federated Learning (FL) has emerged as a promising paradigm for preserving client data ownership and control over distributed Internet of Things (IoT) environments. While discriminative models dominate most FL use cases, recent advances in generative models -- such as Variational Autoencoders (VAE), Generative Adversarial Networks (GAN), and Diffusion Models (DM) -- offer new opportunities for unsupervised anomaly detection in time series analysis, with relevant applications in predictive maintenance (PdM) in critical industrial infrastructures. In this work, we present a comprehensive analysis of VAEs, GANs, and DMs in the context of federated PdM. We analyze their performance and communication overhead under both full and partial federation setups, where only subsets of model components are shared. Building on this analysis, the paper proposes a novel taxonomy for federated generative models that formalizes partial component sharing as a principled mechanism for model personalization. Our experiments over a real-world time series dataset reveal distinct trade-offs in model utility, stability, and scalability, especially in heterogeneous and bandwidth-constrained FL settings. For the evaluated GAN-based configurations, full federation improves training stability relative to independent local training, although the model remains less robust than the VAE- and DDPM-based alternatives. For DMs, however, partial federation -- especially decoder sharing -- can outperform full federation in bandwidth-constrained, non-IID settings.

cs.LG↗

arXiv2026

Fast Full-Wave Simulation of Indoor RSS Maps for Pre-Measurement Validation in Device-Free Localization

Human localization is gaining momentum in security, healthcare, logistics, and smart spaces applications. While global navigation systems are unreliable indoor, device-free (a.k.a. passive) localization methods that exploit human-induced perturbations of radio propagation can be effectively used. This paper investigates the use of a compact full-wave electromagnetic (EM) setup as a fast and reliable tool to simulate indoor Wi-Fi propagation for human sensing. The goal is to provide a practical baseline for validating simplified propagation models, such as diffraction-based descriptions, and to reduce the need for costly measurement campaigns. Two-dimensional attenuation maps from received signal strength are generated and compared in controlled environments, focusing on attenuation statistics and interference patterns. The simulations reproduce the main spatial features, though discrepancies remain due to simplified material characterization. Diffraction-aware refinements are proposed to mitigate these effects. Overall, the approach provides an efficient pre-measurement reference to support device-free system design and to guide experimental planning.

eess.SP↗

arXiv2026

A Unified Approach to Human-Scale Blockage and Scattering Analysis in Sub-THz Propagation With Application to RF Sensing

RF sensing exploits phase-sensitive measurements of stray electromagnetic (EM) fields from wireless devices across various frequency bands to detect EM blockage and to reconstruct and map the surrounding environment in 2D/3D. Although blockage effects caused by objects or human motion are well-studied in ISM bands and frequencies up to 60~GHz, there is a significant lack of research for frequencies above 100~GHz. The paper proposes a unified signal processing framework for RF sensing in the sub-THz D-band (105--175~GHz), explicitly integrating EM blockage and scattering as a single process through the birth-death dynamics of multipath components (MPCs). The framework extracts, associates, and classifies MPCs from angle-delay measurements using statistically grounded detection and classification, enabling human-scale sensing from a single radio link. The modeling and classification of MPCs, along with large-scale EM parameters, are demonstrated through an indoor measurement campaign using multiple test targets. Experimental results show that newly formed, attenuated, and suppressed MPCs can be reliably identified with millimeter-scale delay resolution. Static object localization achieves average positioning errors of $8-20$~cm depending on range and material, while passive human localization yields errors of 12-17cm at 0.5m and 26-30cm at 2m, respectively. The proposed framework demonstrates that accurate sensing and localization are feasible at sub-THz frequencies using a single link.

eess.SP↗

arXiv2026

Efficient 2.5-D FEM-Based Scattering Analysis of the Human Body for RF Sensing

Model training for Device-Free Localization (DFL) and Radio-Frequency (RF) sensing heavily relies on large-scale datasets, which are difficult, expensive, and time-consuming to obtain through measurements. This paper proposes a fast 2.5-dimensional Finite Element Method (2.5-D FEM) for computing the scattering fields of a Body of Revolution (BoR) human model under the excitation of a z-directed dipole. The proposed method can evaluate the effect of human micro-movements through the statistical characteristics of the Received Signal Strength Indicator (RSSI). The numerical accuracy and the practical applicability of the proposed method are validated through comparisons with full-wave simulations and indoor RF sensing experiments. The simulation results show agreement with the experimental measurements, demonstrating that the method is a reliable tool for evaluating micro-movement-induced statistical variations. The proposed method provides a practical and efficient means for generating large-scale, labeled RF training datasets, thereby accelerating the development of indoor localization tools as well as the calibration and tuning of tomographic reconstruction methods.

eess.SP↗

arXiv2025

RF sensing with dense IoT network graphs: An EM-informed analysis

Radio Frequency (RF) sensing is attracting interest in research, standardization, and industry, especially for its potential in Internet of Things (IoT) applications. By leveraging the properties of the ElectroMagnetic (EM) waves used in wireless networks, RF sensing captures environmental information such as the presence and movement of people and objects, enabling passive localization and vision applications. This paper investigates the theoretical bounds on accuracy and resolution for RF sensing systems within dense networks. It employs an EM model to predict the effects of body blockage in various scenarios. To detect human movements, the paper proposes a deep graph neural network, trained on Received Signal Strength (RSS) samples generated from the EM model. These samples are structured as dense graphs, with nodes representing antennas and edges as radio links. Focusing on the problem of identifying the number of human subjects co-present in a monitored area over time, the paper analyzes the theoretical limits on the number of distinguishable subjects, exploring how these limits depend on factors such as the number of radio links, the size of the monitored area and the subjects physical dimensions. These bounds enable the prediction of the system performance during network pre-deployment stages. The paper also presents the results of an indoor case study, which demonstrate the effectiveness of the approach and confirm the model's predictive potential in the network design stages.

eess.SP↗

arXiv2025

Gesture-Based Robot Control Integrating Mm-wave Radar and Behavior Trees

As robots become increasingly prevalent in both homes and industrial settings, the demand for intuitive and efficient human-machine interaction continues to rise. Gesture recognition offers an intuitive control method that does not require physical contact with devices and can be implemented using various sensing technologies. Wireless solutions are particularly flexible and minimally invasive. While camera-based vision systems are commonly used, they often raise privacy concerns and can struggle in complex or poorly lit environments. In contrast, radar sensing preserves privacy, is robust to occlusions and lighting, and provides rich spatial data such as distance, relative velocity, and angle. We present a gesture-controlled robotic arm using mm-wave radar for reliable, contactless motion recognition. Nine gestures are recognized and mapped to real-time commands with precision. Case studies are conducted to demonstrate the system practicality, performance and reliability for gesture-based robotic manipulation. Unlike prior work that treats gesture recognition and robotic control separately, our system unifies both into a real-time pipeline for seamless, contactless human-robot interaction.

cs.RO↗

arXiv2025

Device-Free Localization with Multiple Antenna Receivers: Simulations and Results

Device-Free Localization (DFL) is a passive radio method able to detect, estimate, and localize targets (e.g., human or other obstacles) that do not need to carry any electronic device. According to the Integrated Sensing And Communication (ISAC) paradigm, DFL networks exploit Radio Frequency (RF) devices, used for communication purposes, to evaluate also the excess attenuation due to targets moving in the monitored area, to estimate the target positions and movements. Several target models have been discussed in the literature to evaluate the target positions by exploiting the RF signals received by networked devices. Among these models, Electromagnetic (EM) body models emerged as an interesting research field for excess attenuation prediction using commercial RF devices. While these RF devices are usually single-antenna boards, the availability of low-cost multi-antenna devices e.g. those used in WLAN (Wireless Local Area Network) scenarios, allow us to exploit array-based signal processing techniques for DFL applications as well. Using an array-capable EM body model, this paper shows how to employ array-based processing to improve angular detection of targets. Unlike single-antenna devices that can provide only attenuation information, multi-antenna devices can provide both angular and attenuation estimates about the target location. To this end, simulations are presented and preliminary results are discussed. The proposed framework paves the way for a wider use of multi-antenna devices based, for instance, on WiFi6 and WiFi7 standards.

eess.SP↗

arXiv2025

Bayesian Federated Learning for Continual Training

Bayesian Federated Learning (BFL) enables uncertainty quantification and robust adaptation in distributed learning. In contrast to the frequentist approach, it estimates the posterior distribution of a global model, offering insights into model reliability. However, current BFL methods neglect continual learning challenges in dynamic environments where data distributions shift over time. We propose a continual BFL framework applied to human sensing with radar data collected over several days. Using Stochastic Gradient Langevin Dynamics (SGLD), our approach sequentially updates the model, leveraging past posteriors to construct the prior for the new tasks. We assess the accuracy, the expected calibration error (ECE) and the convergence speed of our approach against several baselines. Results highlight the effectiveness of continual Bayesian updates in preserving knowledge and adapting to evolving data.

cs.LG↗

arXiv2024

A Federated Learning Platform as a Service for Advancing Stroke Management in European Clinical Centers

The rapid evolution of artificial intelligence (AI) technologies holds transformative potential for the healthcare sector. In critical situations requiring immediate decision-making, healthcare professionals can leverage machine learning (ML) algorithms to prioritize and optimize treatment options, thereby reducing costs and improving patient outcomes. However, the sensitive nature of healthcare data presents significant challenges in terms of privacy and data ownership, hindering data availability and the development of robust algorithms. Federated Learning (FL) addresses these challenges by enabling collaborative training of ML models without the exchange of local data. This paper introduces a novel FL platform designed to support the configuration, monitoring, and management of FL processes. This platform operates on Platform-as-a-Service (PaaS) principles and utilizes the Message Queuing Telemetry Transport (MQTT) publish-subscribe protocol. Considering the production readiness and data sensitivity inherent in clinical environments, we emphasize the security of the proposed FL architecture, addressing potential threats and proposing mitigation strategies to enhance the platform's trustworthiness. The platform has been successfully tested in various operational environments using a publicly available dataset, highlighting its benefits and confirming its efficacy.

cs.CY↗

arXiv2024

On the Impact of Data Heterogeneity in Federated Learning Environments with Application to Healthcare Networks

Federated Learning (FL) allows multiple privacy-sensitive applications to leverage their dataset for a global model construction without any disclosure of the information. One of those domains is healthcare, where groups of silos collaborate in order to generate a global predictor with improved accuracy and generalization. However, the inherent challenge lies in the high heterogeneity of medical data, necessitating sophisticated techniques for assessment and compensation. This paper presents a comprehensive exploration of the mathematical formalization and taxonomy of heterogeneity within FL environments, focusing on the intricacies of medical data. In particular, we address the evaluation and comparison of the most popular FL algorithms with respect to their ability to cope with quantity-based, feature and label distribution-based heterogeneity. The goal is to provide a quantitative evaluation of the impact of data heterogeneity in FL systems for healthcare networks as well as a guideline on FL algorithm selection. Our research extends beyond existing studies by benchmarking seven of the most common FL algorithms against the unique challenges posed by medical data use cases. The paper targets the prediction of the risk of stroke recurrence through a set of tabular clinical reports collected by different federated hospital silos: data heterogeneity frequently encountered in this scenario and its impact on FL performance are discussed.

cs.LG↗

arXiv2024

A Carbon Tracking Model for Federated Learning: Impact of Quantization and Sparsification

Federated Learning (FL) methods adopt efficient communication technologies to distribute machine learning tasks across edge devices, reducing the overhead in terms of data storage and computational complexity compared to centralized solutions. Rather than moving large data volumes from producers (sensors, machines) to energy-hungry data centers, raising environmental concerns due to resource demands, FL provides an alternative solution to mitigate the energy demands of several learning tasks while enabling new Artificial Intelligence of Things (AIoT) applications. This paper proposes a framework for real-time monitoring of the energy and carbon footprint impacts of FL systems. The carbon tracking tool is evaluated for consensus (fully decentralized) and classical FL policies. For the first time, we present a quantitative evaluation of different computationally and communication efficient FL methods from the perspectives of energy consumption and carbon equivalent emissions, suggesting also general guidelines for energy-efficient design. Results indicate that consensus-driven FL implementations should be preferred for limiting carbon emissions when the energy efficiency of the communication is low (i.e., < 25 Kbit/Joule). Besides, quantization and sparsification operations are shown to strike a balance between learning performances and energy consumption, leading to sustainable FL designs.

eess.SP↗

arXiv2024

Full-wave EM simulation analysis of human body blockage by dense 2D antenna arrays

Recently, proposals of human-sensing-based services for cellular and local area networks have brought indoor localization to the attention of several research groups. In response to these stimuli, various Device-Free Localization (DFL) techniques, also known as Passive Localization methods, have emerged by exploiting ambient signals to locate and track individuals that do not carry any electronic device. This study delves into human passive indoor localization through full-wave electromagnetic simulations. For the scope, we exploit simulations from the commercial tool FEKO software that employs the Method of Moments (MoM). In particular, we collect and analyze the electric field values in a scenario constituted by a dense 2D/3D deployment of receivers in the presence of an anthropomorphic mobile target. The paper describes in detail the collected dataset and provides a first analysis based on a statistical approach. Possible use cases are also investigated through examples in the context of passive localization, sensing, and imaging.

eess.SP↗

arXiv2024

On the impact of the antenna radiation patterns in passive radio sensing

Electromagnetic (EM) body models based on the scalar diffraction theory allow to predict the impact of subject motions on the radio propagation channel without requiring a time-consuming full-wave approach. On the other hand, they are less effective in complex environments characterized by significant multipath effects. Recently, emerging radio sensing applications have proposed the adoption of smart antennas with non-isotropic radiation characteristics to improve coverage.This letter investigates the impact of antenna radiation patterns in passive radio sensing applications. Adaptations of diffraction-based EM models are proposed to account for antenna non-uniform angular filtering. Next, we quantify experimentally the impact of diffraction and multipath disturbance components on radio sensing accuracy in environments with smart antennas.

eess.SP↗

arXiv2024

Physics-informed generative neural networks for RF propagation prediction with application to indoor body perception

Electromagnetic (EM) body models designed to predict Radio-Frequency (RF) propagation are time-consuming methods which prevent their adoption in strict real-time computational imaging problems, such as human body localization and sensing. Physics-informed Generative Neural Network (GNN) models have been recently proposed to reproduce EM effects, namely to simulate or reconstruct missing data or samples by incorporating relevant EM principles and constraints. The paper discusses a Variational Auto-Encoder (VAE) model which is trained to reproduce the effects of human motions on the EM field and incorporate EM body diffraction principles. Proposed physics-informed generative neural network models are verified against both classical diffraction-based EM tools and full-wave EM body simulations.

eess.SP↗

arXiv2024

Compressed Bayesian Federated Learning for Reliable Passive Radio Sensing in Industrial IoT

Bayesian Federated Learning (FL) has been recently introduced to provide well-calibrated Machine Learning (ML) models quantifying the uncertainty of their predictions. Despite their advantages compared to frequentist FL setups, Bayesian FL tools implemented over decentralized networks are subject to high communication costs due to the iterated exchange of local posterior distributions among cooperating devices. Therefore, this paper proposes a communication-efficient decentralized Bayesian FL policy to reduce the communication overhead without sacrificing final learning accuracy and calibration. The proposed method integrates compression policies and allows devices to perform multiple optimization steps before sending the local posterior distributions. We integrate the developed tool in an Industrial Internet of Things (IIoT) use case where collaborating nodes equipped with autonomous radar sensors are tasked to reliably localize human operators in a workplace shared with robots. Numerical results show that the developed approach obtains highly accurate yet well-calibrated ML models compatible with the ones provided by conventional (uncompressed) Bayesian FL tools while substantially decreasing the communication overhead (i.e., up to 99%). Furthermore, the proposed approach is advantageous when compared with state-of-the-art compressed frequentist FL setups in terms of calibration, especially when the statistical distribution of the testing dataset changes.

cs.LG↗