Autonomous driving applications demand large fields-of-view (FoVs) and high scan rates from microelectromechanical system (MEMS)-based light detection and ranging (LiDAR) systems. To meet these requirements, a multispot system is proposed using a single beam source and an MEMS structure for simultaneous beam steering and splitting. The system integrates an MEMS scanner with a diffractive optical element (DOE), achieving an 86 degrees H & times; 13 degrees V FoV at a 20-Hz frame rate. The electrostatic MEMS mirror is based on a gimbal architecture, with a reflective surface relief Dammann grating (1550-nm wavelength targeted) on the mirror surface to split the incident beam into five equally distributed beams, increasing five times the probing FoV for the same vertical mechanical deflection. Vertically asymmetric electrodes actuate the inner axis (horizontal FoV) at resonance, while the outer axis (vertical FoV) was optimized for quasi-static operation using staggered actuators. The device, fabricated on a 50-& micro;m-thick SOI wafer using a multilevel, self-aligned, dicing-free process, achieves an angular separation of the diffracted spots of 2.57 degrees, with a beam uniformity of 68%. This work demonstrates a combined beam steering and beam splitting MEMS mirror to simultaneously increase the FoV and scan line resolution, motivating their integration in next-generation multispot LiDAR.

Electromagnetic interference (EMI) remains a critical concern in cyber-physical systems (CPS), particularly in safety-critical applications where disturbance-vulnerability interactions can compromise system operation. Recent advances, such as risk-based electromagnetic compatibility management (RB-EMC) have reframed EMI as a lifecycle hazard, requiring systematic identification, evaluation, and control of electromagnetic vulnerabilities (EMVs). No fixed process exists for EMV discovery, leaving the identification stage highly case-dependent. This article addresses that gap by introducing A$^{2}$EM, an architecture-aware electromagnetic simulation-driven methodology for EMV discovery, diagnosis, and mitigation assurance in CPSs. Building on concepts from the sensor security domain, the approach models sensor readout circuits as structured chains of coupling and transformation stages, enabling systematic probing for latent failure modes. Brute-force and selective sweeps allow for iterative EMI identification, while targeted probing supports both failure diagnosis and validation of mitigation measures within the EMI control stage. A$^{2}$EM flexibly adapts to different points in the development lifecycle—from exploratory early-stage screening, to root cause analysis in mature designs, to reassessment of aging systems. By embedding EMV discovery and diagnosis into CPS design workflows, it provides a concrete procedural structure for RB-EMC, ensuring that risk management decisions are grounded in systematic architecture-based analyzes.

Accurate and reliable positioning is essential for a wide range of vehicular applications, including navigation, traffic management, Advanced Driver Assistance Systems (ADAS), and autonomous driving. Traditionally, vehicle localization has relied on egocentric approaches where each vehicle estimates its own position using onboard sensors. Since the introduction of Global Navigation Satellite System (GNSS), these methods have greatly advanced with multiple satellite constellations and correction techniques such as Real-Time Kinematic (RTK) and Precise Point Positioning (PPP). Yet, GNSS-based positioning remains challenged by signal obstructions in urban canyons, tunnels, and dense foliage. To overcome these limitations, sensor fusion combines data from inertial sensors, cameras, LiDAR, and radar to enhance accuracy and robustness, especially in GNSS-degraded conditions. Recently, the rise of vehicular communications has enabled a new paradigm: cooperative positioning. By exchanging data among vehicles and infrastructure, this approach mitigates individual sensor errors, improves accuracy, and supports shared situational awareness. This paper presents a comprehensive overview of vehicle localization, covering the transition from egocentric to cooperative positioning. We review key technologies and sensors, analyze cooperative architectures, and classify recent research, providing insights into current trends and future directions in vehicle localization.

Accurate recognition of road surface type and condition is essential for intelligent vehicle systems. While support vector machines (SVMs) and random forests (RFs) offer complementary strengths, their individual limitations in threshold stability and heterogeneous feature handling restrict their performance in complex environments. This article proposes a hybrid classifier, SVM-informed RF (SVM-RF), which integrates SVM margin information into the threshold selection and split generation of the RF training process, guiding tree construction toward the most discriminative regions of the feature space and stabilizing decision boundaries. The method is evaluated on a multisensor fusion dataset combining vibration, acceleration, and meteorological features across multiple road types and surface conditions. Results show that SVM-RF consistently outperforms standalone SVM and RF in accuracy and macro-F1-score, while maintaining low computational complexity suitable for real-time deployment. These findings demonstrate that SVM-RF provides an effective and lightweight solution for reliable road surface condition recognition.

Light detection and ranging (LiDAR) sensors play a critical role in enabling precise and reliable environmental perception for autonomous vehicles. However, handling the large amounts of data they generate presents a significant challenge. With the emergence of standards, such as the geometry based point cloud compression (G-PCC) standard, octrees have been used as a key data structure for compressing 3-D light detection and ranging (LiDAR) data. Despite their advantages, octrees often lead to high memory utilization and computational overhead, particularly when dealing with high-resolution datasets, limiting their utilization in systems with real-time requirements. This letter presents FOG-zip: a hardware-accelerated octree compression approach designed for embedded systems with limited resources. Experimental results demonstrate that FOG-zip achieves up to a 27.8% reduction in data size compared to the same uncompressed octree bitstream, while processing each frame within the frame rate limits of the sensor used.

The growing adoption of electric vehicles (EVs) and the rise of AI-assisted software-defined vehicle (SDV) designs increase the need for effective fault-tolerance mechanisms in automotive systems, as required by ISO 26262. Software Implemented Hardware Fault Tolerance (SIHFT) techniques, such as Control Flow Checking (CFC), strengthen robustness at the software level without additional hardware. This work evaluates a compilerlevel CFC algorithm with minimal application-level intrusion in a safety-critical ECU application, using a two-stage simulationbased methodology of Profiling and Hardening. Results show a detection rate of $16.53 \%$, a diagnostic coverage of $27.78 \%$, and a single-point fault metric of $\mathbf{5 6 . 5 9 \%}$ under program counter and register file faults. While below ASIL compliance thresholds, the findings highlight CFC as a complementary safety mechanism that enhances system resilience and demonstrates the importance of flexible, software-based fault-tolerance for addressing current and future safety challenges.

This article presents a highly tunable low-g microelectromechanical systems (MEMS) inertial switch. The switch is specifically designed as a low-power wake-up device for Internet-of-Things (IoT) applications. The designed inertial switch is sensitive to in-plane accelerations in all directions (omnidirectional). The proposed inertial switch has been modeled and simulated both analytically and using finite element analysis techniques. The switch prototype has been fabricated through an in-house optimized silicon-on-insulator (SOI) wafer process, which allows patterning on both the device and handle silicon layers. The fabricated inertial switch was experimentally tested using a closed-loop shaker system to measure acceleration thresholds and a tilt stage for electrical resistance measurements. This inertial switch demonstrates the ability to tune the acceleration threshold in a wide range of 0.25-2 g. This promising inertial sensor holds a significant potential for applications requiring a high dynamic range for acceleration threshold detection, such as mechanical wake-up functions in IoT devices.

Spatiotemporal modulation (STM) is used in Ultrasonic Mid-Air Haptics to create compelling tactile sensations. The STM can create perceptually distinct sensations. We specified the sensations of a palm-size pattern by varying the focal point's speed and pattern sampling rate. Three sensations were specified, named as Dynamic, Vibratory and Uniform. A selective identification study was conducted to evaluate if the sensations were recognizable to the perception when presented individually and simultaneously (combined stimuli). The results support the STM's specification and the selective recognition of the sensations was possible for some combinations.

Selecting features associated with patient-centered outcomes is of major relevance yet the importance given depends on the method. We aimed to compare stepwise selection, least absolute shrinkage and selection operator, random forest, Boruta, extreme gradient boosting and generalized maximum entropy estimation and suggest an aggregated evaluation. We also aimed to describe outcomes in people with chronic obstructive pulmonary disease (COPD). Data from 42 patients were collected at baseline and at 5 months. Acute exacerbations were the aggregated most important feature in predicting the difference in the handgrip muscle strength (dHMS) and the COVID-19 lockdown group had an increased dHMS of 3.08 kg (CI95 ≈ [0.04, 6.11]). Pack-years achieved the highest importance in predicting the difference in the one-minute sit-to-stand test and no clinical change during lockdown was detected. Charlson comorbidity index was the most important feature in predicting the difference in the COPD assessment test (dCAT) and participants with severe values are expected to have a decreased dCAT of 6.51 points (CI95 ≈ [2.52, 10.50]). Feature selection methods yield inconsistent results, particularly extreme gradient boosting and random forest with the remaining. Models with features ordered by median importance had a meaningful clinical interpretation. Lockdown seem to have had a negative impact in the upper-limb muscle strength.

Simulation-based Fault Injection (FI) is crucial for validating system behaviour in safety-critical applications, such as the automotive industry. The ISO 26262 standard’s Part 11 extension provides failure modes for digital components, driving the development of new fault models to assess software-implemented mechanisms against random hardware failures (RHF). This paper proposes a Fault Injection framework, QEFIRA, and shows its ability to achieve the failure modes proposed by Part 11 of the ISO 26262 standard and estimate relevant metrics for safety mechanisms. QEFIRA uses QEMU to inject permanent and transient faults during runtime, whilst logging the system state and providing automatic post-execution analysis. Complemented with a confusion matrix, it allows us to gather standard compliant metrics to characterise and evaluate different designs in the early stages of development. Comparatively to the native QEMU implementation, the tool only shows a slowdown of 1.4× for real-time microcontroller-based applications.

In the context of Industry 4.0, this paper explores the vital role of advanced technologies, including Cyber–Physical Systems (CPS), Big Data, Internet of Things (IoT), digital twins, and Artificial Intelligence (AI), in enhancing data valorization and management within industries. These technologies are integral to addressing the challenges of producing highly customized products in mass, necessitating the complete digitization and integration of information technology (IT) and operational technology (OT) for flexible and automated manufacturing processes. The paper emphasizes the importance of interoperability through Service-Oriented Architectures (SOA), Manufacturing-as-a-Service (MaaS), and Resource-as-a-Service (RaaS) to achieve seamless integration across systems, which is critical for the Industry 4.0 vision of a fully interconnected, autonomous industry. Furthermore, it discusses the evolution towards Supply Chain 4.0, highlighting the need for Transportation Management Systems (TMS) enhanced by GPS and real-time data for efficient logistics. A guideline for implementing CPS within Industry 4.0 environments is provided, focusing on a case study of real-time data acquisition from logistics vehicles using CPS devices. The study proposes a CPS architecture and a generic platform for asset tracking to address integration challenges efficiently and facilitate the easy incorporation of new components and applications. Preliminary tests indicate the platform’s real-time performance is satisfactory, with negligible delay under test conditions, showcasing its potential for logistics applications and beyond.

In this work a multilevel nanoimprint lithography (NIL) replication process was demonstrated to produce 1D MEMS mirrors employing vertical asymmetric comb -drive electrostatic actuation, in a 200 mm wafer SOI-based process. In comparison with a direct write laser (DWL) grayscale lithography step (which for the proposed layout requires around 40 h of exposure time per wafer), this NIL method greatly enhances fabrication throughput by reliably reproducing a master's multilevel topography onto the photoresist. This study describes the NIL master fabricated using grayscale lithography, the working stamp, and the replication micromachining processes. An extensive characterization of the morphology and topography of the intermediate working stamp is provided, along with an optimization study of the replica fabrication and the alignment procedure between the replica and the mirror substrate. The MEMS device pattern was effectively replicated, exhibiting electrode gaps of 3.66 mu m (grayscale process yields gaps of 3.5 mu m). Discrete photoresist levels of 1.53 mu m and 2.83 mu m were observed, with a misalignment to the preceding metal layer of 5 mu m and 10 mu m in the x and y directions, respectively. These deviations were found to be within the required alignment margin defined by the layout (20 mu m). The 1D MEMS mirror fabricated using this NIL process was successfully characterized using Scanning Laser -Doppler Vibrometry under atmospheric pressure conditions. The results obtained were in accordance with the theoretical design parameters, demonstrating that NIL can be successfully used as a fast, low-cost alternative lithography process to fabricate multilevel MEMS structures.

Active Noise Cancellation (ANC) systems are widely used to mitigate unwanted noises in several applications, such as automotive environments and high-end headsets. Multi-Channel (MC) ANC systems have shown promise in creating improved silent zones. Typically, these systems are implemented on FPGA platforms due to the systolic nature and granularity of optimization of these devices. This article describes the design, implementation, and evaluation of a wavelet-based MC ANC Filtered-x Normalized Least Mean Square (FxNLMS) on an FPGA platform.The use of wavelet transform enables the decomposition of complex noise signals into spectrally more compact signals (i.e., easier to process). In this work, for each decomposed signal, an independent NLMS is applied. The system implements 64 parallel NLMS with 1000 coefficients. Additionally, the static FIR filters employed for secondary and tertiary path estimations are of the 2047th order. The system adopts an integer arithmetic architecture and operates at a sampling rate of 47.97 kHz. To assess the performance of the wavelet-based approach, benchmark tests were conducted by comparing it against a similar implementation without the wavelet transform. The evaluation was performed using noise reduction (NR) tests with spectrally rich (20 Hz to 10 kHz) and high dynamic range noises. The experimental setup involved two error microphones and two secondary sources.The results show that the wavelet-based version has overall better performance than the traditional implementation, particularly in the higher frequency band of the spectrum (1 kHz to 8 kHz). For instance, in the case of city ambient noise (a realistic noise with high dynamic range), the relative NR achieved was 8.23 dB.To the authors’ knowledge, this is the first time that the implementation and field-test of a wavelet-based MC ANC on an FPGA platform was conducted. Moreover, the obtained results show that the novel approach is better in reducing complex noises than the traditional implementation – without wavelets.

In light of the increased concern about environmen-tal sustainability and the urgent need to reduce greenhouse gas emissions, methane (CH 4 ) detection plays a critical role regarding air quality and climate change. As the major component of natural gas, methane significantly contributes to global warming and serves as an ancestor to the creation of harmful atmospheric pollutants. Therefore, effective detection and mitigation of methane leaks are essential for promoting sustainable practices. The first step to achieve this relies on the creation of systems capable of reliably and rapidly detecting methane leaks into the environment. This paper explores solutions for methane detection leaks in IoT natural gas monitoring systems, with a focus on the integration of low-power sensors, particularly using Metal-Oxide-Semiconductor (MOS) technology. The research encompasses various aspects, including the development of a test plan, to ensure validation of sensor performance across diverse scenarios, aiming to simulate sensor behaviour under conditions closely resembling the environments in which they will be deployed.

Advancements in technology are propelling Cyber-Physical Systems (CPS) into crucial roles across various sectors, implying the need for stricter CPS safety and security measures as their deployment in safety-critical scenarios increases. Physical-to-cyber attacks are particularly alarming among emerging threats, targeting sensors and exposing significant vulnerabilities in CPS due to the inadequacy of current protection mechanisms. The development landscape for CPS also reveals other gaps such as the insufficiency of analog fault coverage and validation, and the threat of covert malicious circuit alterations by third-party outsourcing. This paper seeks to draw the scientific community’s attention to these topics from a unified perspective, presenting the main issues in the development of safety-critical CPS, along with a survey of related subjects. Subsequently, these topics are formally bridged with the introduction of this paper’s main contribution: the concept of Hardware Integrity Threats (HITs). This finding suggests that the community can develop countermeasures that are effective across the spectrum of these threats, which led to the formulation of two environment proposals to explore these design spaces. Our proposals incorporate co-simulation and simulation with Hardware-in-the-Loop (HiL) taking these safety and security validation concerns into account, as the exploration of these designs mostly entails hybrid analog hardware/software solutions.

Machine-Learning model implementation in Resource-Scarce Embedded Systems is becoming a standard in many systems and projects. This implementation allows systems to be less Cloud dependent and make decisions independently. As these systems’ reasoning becomes intricate with the Machine Learning model’s decision, an attack to change the Machine Learning model’s data structure can make the entire system misbehave, which in some solutions can be critical. Therefore, it is necessary to create low-overhead tools to flag any miscalculation or wrongdoing during the model’s inference phase. The following work presents a Machine Learning Health Monitoring system based on PCA and Control Charts to verify if the model’s inference function runs properly. The solution presents a reasonable flag rate and is implemented e ciently in a Resource-Scarce Embedded System due to its reduced memory and processing footprints.

Objectives: To compare positive airway pressure (PAP) adherence between patients with or without excessive daytime sleepiness (EDS) in mild, moderate and severe obstructive sleep apnea (OSA). Methods: Patients >18 years diagnosed with OSA in 2018 and 2019, without previous history of PAP usage and with adherence registration in the first medical consultation after treatment initiation, were included. EDS was defined as a score of >10 on the Epworth Scale. Patients were divided into two groups according to the adherence to PAP: "Adherent" if using the device for >4 h for >70% of the nights and "Nonadherent" otherwise. Simple and multiple logistic regression models for adherence were determined.Results: 321 patients were included, most male (64.2%), with mean age 56.56 years. Most patients had severe OSA (n = 159; 49.5%), and median AHI was 29.3/h [16.8; 47.5]. Being older or having a severe OSA resulted in an increased adherence (OR = 1.020, CI95% = [1.002; 1.039] and OR = 2.299, CI95% = [1.273; 4.191], respectively). In patients without EDS a statistically significant difference was found in adherence between those with severe OSA and both mild and moderate OSA categories (OR = 0.285, p = 0.023 and OR = 0.387, p = 0.026, respectively), with patients with severe OSA being adherent. There was no sta-tistical difference in adherence between patients with or without EDS (OR 1.083; p = 0.876), nor in the different degrees of severity in those with EDS.Conclusion: In our study there were no differences in PAP therapy adherence between patients with or without excessive daytime sleepiness. Older age and higher OSA severity resulted in higher adherence rates.& COPY; 2023 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

Satellite gravimetry applications require sub-ng acceleration resolution at frequencies below 100 mHz, demanding low-noise (mechanical-thermal, 1/f and electronic) and high-sensitivity MEMS sensors. The electrostatic pull-in time-based transduction enables very high sensitivities (75 μs/(μm/s <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> ) demonstrated) compared to direct capacitive transduction. The high-resolution requirement calls for large masses and compliant springs (at the limit of manufacturability), resulting in microstructures that require special encapsulation mechanisms. Hence, novel stoppers were designed to limit in- and out-of-plane displacements and improve robustness. Squeeze-film damping was tuned/maximized with dummy parallel-plates (to decrease the device quality factor) since the sensitivity of pull-in time-based operation strongly depends on it.

Recent concerns about real-time inference and data privacy are making Machine Learning (ML) shift to the edge. However, training efficient ML models require large-scale datasets not available for typical ML clients. Consequently, the training is usually delegated to specific Service Providers (SP), which are now worried to deploy proprietary ML models on untrusted edge devices. A natural solution to increase the privacy and integrity of ML models comes from Trusted Execution Environments (TEEs), which provide hardware-based security. However, their integration with heavy ML computation remains a challenge. This perspective paper explores the feasibility of leveraging a state-of-the-art TEE technology widely available in modern MCUs (TrustZone-M) to protect the privacy of Quantized Neural Networks (QNNs). We propose a novel framework that traverses the model layer-by-layer and evaluates the number of epochs an attacker requires to build a model with the same accuracy as the target with the information disclosed. The set of layers whose information makes the attacker spend less training effort than the owner training from scratch is protected in an isolated environment, i.e., the secure-world. Our framework will be evaluated in terms of latency and memory footprint for two ANNs built for the CIFAR-10 and Visual Wake Words (VWW) datasets. In this perspective paper, we establish a baseline reference for the results.

This simulation study explores the impact of different undesirable scenarios (e.g., collinearity, Simpson’s paradox, variable interaction, Freedman’s paradox) on feature selection and coefficients’ estimation using traditional methodologies, such as automatic selection (e.g., stepwise using Akaike information criterion and Bayesian information criterion) and penalized regression (e.g., least absolute shrinkage and selection operator (LASSO), elastic net, relaxed LASSO, adaptive LASSO, minimax concave penalty and smoothly clipped absolute deviation penalty, penalized regression with second-generation p-values). Specifically, we compare wrapper and embedded methods regarding the feature selection, coefficients’ estimation and models’ performance. Our results show that the choice of the methodology can affect the number and the type of selected features, as well as accuracy and precision of coefficients’ estimates. Furthermore, we find that the performance can also depend on the characteristics of the data.