Archive ouverte HAL-LAAS
Not a member yet
12189 research outputs found
Sort by
Advantage and Challenge of Electrical Critical Dimension Test Structures for Electroplated High Aspect Ratio Nano Structures (HARNS) on Insulating Materials
No full textInternational audienceElectrical Critical Dimension Test Structures (ECD-TS) have been applied for development of new technology: Metallic High Aspect Ratio Nano Structures (Metal-HARNS). The HARNS refer to electro mechanical structures scaled down to submicron width (as narrow as 260nm confirmed) while keeping its thickness (up to 1500nmconfirmed). Two major advantage and challenge have been confirmed through measurement: (1) insulating material showed critical advantage on ECD over Scanning Electron Microscope CD assessment and (2) seed conductive layer affected the measurement as leakage path
Finding the right regression testing method: a taxonomy-based approach
No full textInternational audienceWith numerous regression testing (RT) methods available in the literature, it is challenging to choose the right one for a specific context. Practitioners need support identifying suitable research. To this end, recent work has proposed a taxonomy. By mapping both the RT problem and existing solutions onto the taxonomy, practitioners should be able to determine which solutions are best aligned with their problem. Our work explores the practical relevance of this idea through an industrial case study. The context is the development of R&D projects at a major automotive company, in the domain of connected vehicles. We developed an RT problem solving approach based on the taxonomy. Following the approach, we characterized the RT problem, identified a set of 8 potentially relevant solutions from a set of 52 papers, and empirically evaluated their suitability. Our approach was successful, as we found effective RT methods among those selected using the taxonomy. One method, in particular, demonstrated remarkable robustness across various datasets, making it a strong recommendation for the industrial partner. However, this success came at the cost of difficulties due to unclear taxonomy elements, missing elements, and paper classification errors. We conclude that the taxonomy has practical value but would have to mature for easier applicability
Safety Monitoring of Machine Learning Perception Functions: a Survey
No full text25 pages, 2 figuresInternational audienceMachine Learning (ML) models, such as deep neural networks, are widely applied in autonomous systems to perform complex perception tasks. New dependability challenges arise when ML predictions are used in safety-critical applications, like autonomous cars and surgical robots. Thus, the use of fault tolerance mechanisms, such as safety monitors, is essential to ensure the safe behavior of the system despite the occurrence of faults. This paper presents an extensive literature review on safety monitoring of perception functions using ML in a safety-critical context. In this review, we structure the existing literature to highlight key factors to consider when designing such monitors: threat identification, requirements elicitation, detection of failure, reaction, and evaluation. We also highlight the ongoing challenges associated with safety monitoring and suggest directions for future research
Technologie des capteurs de gaz à semi-conducteurs
No full textGas microsensors are of great industrial interest due to their small size, low consumption, low cost and therefore deployable in a distributed network. Semiconductor gas sensors, although having limited performance compared to analyzers, are among the most commercialized sensors along with electrochemical and optical sensors.This article covers the operation of these sensors, the main materials used (metal oxides) with their detection mechanism as well as the design, manufacturing, characterization and calibration techniques. Current avenues of research around electronic noses and associated perspectives are also discussed.Les microcapteurs de gaz ont un très gros intérêt industriel de par leur faible encombrement, leur faible consommation, leur faible coût, ce qui permet de les déployer en réseau distribué. Les capteurs de gaz à oxydes métalliques semiconducteurs, bien qu’ayant des performances limitées par rapport à des analyseurs, font partie des capteurs les plus commercialisés avec les capteurs électrochimiques et optiques.Cet article couvre le fonctionnement de ces capteurs, les principaux matériaux utilisés avec leur mécanisme de détection, ainsi que les techniques de conception, de fabrication, de caractérisation et d'étalonnage. Sont également abordées les pistes de recherche actuelles autour des nez électroniques, ainsi que les perspectives associées
Caractérisation de détecteurs SPADs pour les télécommunications fibrées quantiques
No full textNational audienceLes télécommunications quantiques, basées sur le codage d'information sur des photons uniques ou des paires de photons, constituent un moyen parfaitement sûr pour la distribution de clefs de cryptage à plusieurs utilisateurs distants. Même si seule la clef de cryptage est transmise, les pertes de photons sont importantes et un débit suffisant est nécessaire pour la transmission. Les émetteurs dans ces systèmes sont des sources de photons uniques capables de délivrer des photons intriqués en temps, en fréquence ou en polarisation. Nous présentons dans cette communication un système de caractérisation de ce type de source basé sur des SPAD (Single Photon Avalanche Detector) en mode « gating » (fenêtrage), un laser pulsé et différentes solutions de filtrage. Le système est utilisé pour des premiers tests menés sur des résonateurs non-linéaires, susceptibles de générer des paires de photons par mélange quatre ondes, donc à partir d'une pompe à 1550 nm
Accélérer la recherche sur les nanomatériaux énergétiques : modèles de substitution basés sur les données pour les simulations coûteuses
No full textThis thesis explores the development and application of machine learning techniques, particularly surrogate modeling and adaptive sampling, to address the computational challenges associated with the simulation and optimization of nanothermite combustion. Traditional physical simulations of nanothermites are computationally intensive, limiting their exploitation for structural design through optimization. To overcome these limitations, this research focuses on building efficient surrogate models and implementing adaptive data generation strategies that significantly reduce computational costs while maintaining predictive accuracy. The first phase of this work involves benchmarking various machine learning algorithms to identify the most effective surrogate modeling approach for nanothermite combustion. Multilayer Perceptrons (MLPs) emerged as the optimal choice, achieving superior accuracy and computational efficiency compared to other models, including Gaussian Process Regression (GPR). These surrogates demonstrated the capability to approximate the nonlinear dynamics of combustion processes with remarkable precision, paving the way for their integration into adaptive sampling frameworks. The second phase introduces an Interest Region Bayesian Sampling (IRBS) methodology, designed to inject physical knowledge into the sampling process to generate training data efficiently by focusing on regions of interest within the design space. The research emphasizes two critical advancements: enabling parallelized evaluations of candidate designs to accelerate data collection and enhancing the sampling process through a novel figure of merit (FOM) formulation. This FOM leverages Kernel Density Estimation (KDE) for exploration, replacing the computationally expensive GPR uncertainty metric. The KDE-based approach not only improves sampling efficiency but also enhances the adaptability of the framework to diverse data distributions. Furthermore, the integration of MLPs as surrogates within the IRBS framework validated the methodology's scalability to higher dimensional design spaces. The results underscore the potential of adaptive sampling and surrogate modeling to revolutionize the design and optimization of complex materials. By reducing computational overhead and enabling rapid exploration of the design space, this work addresses a key bottleneck in materials science, particularly for energetic nanomaterials like thermites. Beyond its immediate application, the proposed framework lays the groundwork for extending machine learning techniques to multi-physics simulations and the optimization of other high-dimensional, computationally demanding systems.Cette thèse explore le développement et l'application de techniques d'apprentissage automatique, en particulier la modélisation par substitut (surrogate modeling) et l'échantillonnage adaptatif, pour relever les défis computationnels associés à la simulation et à l'optimisation de la combustion des nanothermites. Les simulations physiques traditionnelles des nanothermites sont extrêmement coûteuses en termes de calcul, ce qui limite leur exploitation pour la conception via l'optimisation. Pour surmonter ces limitations, cette recherche se concentre sur la construction de modèles substitutifs efficaces et la mise en œuvre de stratégies adaptatives de génération de données, permettant de réduire considérablement les coûts de calcul tout en maintenant une précision prédictive élevée. La première phase de ce travail consiste en une évaluation comparative de divers algorithmes d'apprentissage automatique afin d'identifier l'approche de modélisation par substitut la plus efficace pour la combustion des nanothermites. Les Perceptrons Multi-Couches (MLP) se sont révélés être le choix optimal, offrant une précision et une efficacité calculatoire supérieures par rapport à d'autres modèles, y compris la Régression par Processus Gaussiens (GPR). Ces modèles de substitution ont démontré leur capacité à approximer les dynamiques non linéaires des processus de combustion avec une précision remarquable, ouvrant la voie à leur intégration dans des cadres d'échantillonnage adaptatif. La deuxième phase introduit une méthodologie d'échantillonnage bayésien axée sur les régions d'intérêt (Interest Region Bayesian Sampling, IRBS), conçue pour injecter des connaissances physiques dans le processus d'échantillonnage afin de générer efficacement des données d'entraînement en se concentrant sur des régions spécifiques de l'espace de conception. Cette recherche met en avant deux avancées majeures : l'implémentation d'évaluations parallélisées des configurations optimales pour accélérer la collecte des données, et l'amélioration du processus d'échantillonnage à travers une nouvelle formulation d'une fonction de mérite (FOM). Cette fonction utilise une estimation de densité par noyau (Kernel Density Estimation, KDE) pour l'exploration, remplaçant la métrique coûteuse d'incertitude basée sur le GPR. L'approche basée sur le KDE améliore non seulement l'efficacité de l'échantillonnage, mais renforce également l'adaptabilité du cadre à des distributions de données diverses. De plus, l'intégration des MLP comme modèles substitutifs dans le cadre d'IRBS a validé la possibilité d'implémenter cette méthodologie aux espaces de conception de plus haute dimension. Les résultats mettent en lumière le potentiel de l'échantillonnage adaptatif et de la modélisation par substitut pour révolutionner la conception et l'optimisation de matériaux complexes. En réduisant les coûts de calcul et en permettant une exploration rapide de l'espace de conception, ce travail répond à une limitation majeure dans la science des matériaux, en particulier pour les nanomatériaux énergétiques tels que les nanothermites. Au-delà de ses applications immédiates, le cadre proposé établit une base pour l'extension des techniques d'apprentissage automatique à des simulations multi-physiques et à l'optimisation d'autres systèmes de haute dimension et exigeants en puissance de calcul
Theoretical investigations of Soliton solutions to the Lugiato-Lefever equation for Fabry-Perot resonators
No full textWe investigate existence of Soliton solutions to a variant of the Lugiato-Lefever equation (LLE) that describes wave propagation in Fabry-Perot (FP) resonators in the context of Kerr frequency comb generation. We first study conditions at which the FP-LLE, when damping is neglected, admits Soliton solutions and we provide an analytical expression for the Solitons. Then, we use a numerical continuation method to look for Solitons when the damping value is non-zero, which is the standard situation
Key Recovery from Side-Channel Power Analysis Attacks on Non-SIMD HQC Decryption
No full textInternational audienceHQC is a code-based cryptosystem that has recently been announced for standardization after the fourth round of the NIST post-quantum cryptography standardization process. During this process, the NIST specifically required submitters to provide two kinds of implementation: a reference one, meant to serve lisibility and compliance with the specifications; and an optimized one, aimed at showing the performance of the scheme alongside other desirable properties such as resilience against implementation misuse or side-channel analysis. While most side-channel attacks regarding PQC candidates running in this process were mounted over reference implementations, very few consider the optimized, allegedly side-channel resistant (at least, constant-time), implementations. Unfortunately, HQC optimized version only targets x86-64 with Single Instruction Multiple Data (SIMD) support, which reduces the code portability, especially for non-generalist computers. In this work, we present two power side-channel attacks on the optimized HQC implementation with just the SIMD support deactivated. We show that the power leaks enough information to recover the private key, assuming the adversary can ask the target to replay a legitimate decryption with the same inputs. Under this assumption, we first present a key-recovery attack targeting standard Instruction Set Architectures (ARM T32, RISC-V, x86-64) and compiler optimization levels. It is based on the well known Hamming Distance model of power consumption leakage, and exposes the key from a single oracle call. During execution on a real target, we show that a different leakage, stemming from to the micro-architecture, simplifies the recovery of the private key. This more direct second attack, succeeds with a 99% chance from 83 executions of the same legitimate decryption. While the weakness leveraged in this work seems quite devastating, we discuss simple yet effective and efficient countermeasures to prevent such a key-recovery
Single-beam Bose-Einstein condensate source: a demonstration of optical grating andmagnetic chip hybridization for onboard quantum sensors
No full textInternational audienceCompact, robust sub-recoil atomic sources are essential for onboardquantum sensors. We present a novel grating-based magnetic chip capable of generating rubidium Bose-Einstein condensates using a singlebeam, a step forward for onboard quantum sensors
Approximating the order 2 quantum Wasserstein distance using the moment-SOS hierarchy
No full textInternational audienceOptimal transport theory has recently been extended to quantum settings, where the density matrices generalize the probability measures. In this paper, we study the computational aspects of the order 2 quantum Wasserstein distance, formulating it as an infinite dimensional linear program in the space of positive Borel measures supported on products of two unit spheres. This formulation is recognized as an instance of the Generalized Moment Problem, which enables us to use the moment-sums of squares hierarchy to provide a sequence of lower bounds converging to the distance. We illustrate our approach with numerical experiments