Portail HAL UHA (Université de Haute-Alsace)
Not a member yet
17715 research outputs found
Sort by
Improving monocular visual odometry localization with wheel and inertial measurements
No full textInternational audienceGlobal Navigation Satellite System (GNSS) is a popular localization system for autonomous vehicles, but an onboard localization system is often required to take over when the signal degrades or is absent, for example in difficult conditions (tunnel, building, forest etc.). Among the existing possibilities, visual odometry is often selected. It does not need any prior knowledge, and is accurate enough if used for a limited period of time. However, it has also several drawbacks: low frequency (due to important processing time) and inconsistency (due to low illumination, glare conditions or moving objects). Moreover, when using a monocular camera, the localization is obtained up to a scale factor. These drawbacks can be solved using other onboard sensors. The most common sensors for autonomous vehicles are inertial measurement units (IMU), wheel encoders, or Lidars. In this paper, a comparison of visual, inertial and wheel odometries (VO, IO and WO respectively) is proposed. The goal is to precisely evaluate their performances on the estimation of the 2D pose (position and heading) of an autonomous platform. With this information, our goal is to increase frequency, and correct inconsistencies of our VO localization. First results correcting the scale and improving the accuracy will be demonstrated using a ground vehicle in outdoor environments
Ciprofloxacin (Zwitterion, Chloride, and Sodium Forms): Experimental and DFT Characterization by Vibrational and Solid-State NMR Spectroscopies
No full textInternational audienceCiprofloxacin (Cipro), a widely used fluoroquinolone antibiotic, exists in multiple protonation states, which influence its structural and spectroscopic properties. Despite its pharmaceutical relevance and concerns regarding its accumulation in the environment, a comprehensive characterization of its zwitterionic, cationic, and anionic solid forms remains limited, particularly in terms of their vibrational and nuclear magnetic resonance (NMR) spectral assignments. The focus of this study was to identify spectral signatures that differentiate Cipro and its cationic form (as a chloride salt) and anionic form (as a sodium salt). All samples were characterized in solid-state using X-ray diffraction, thermogravimetric analysis coupled with mass spectrometry, infrared and Raman spectroscopies, and cross-polarized magic-angle spinning (CP-MAS) solid-state NMR, with support from density functional theory (DFT) calculations. Both salts were synthesized in this study. The cipro sodium was isolated as a monohydrate salt, a previously unreported phase. Six spectral regions were identified to distinguish the Cipro zwitterion from its cationic and anionic forms by using vibrational spectroscopy. Both experimental 13 C CP-MAS solid-state NMR and theoretical analyses revealed pronounced chemical shifts induced by protonation and counterion interactions, which also differentiate the three forms. The analysis presented here provides clear fingerprints of the three Cipro forms, which can be used to support reference spectroscopic data, with direct implications in pharmaceutical formulations as well as for environmental studies
Automatisé apprentissage automatique avec métaheuristique
No full textAutomated Machine Learning (AutoML) assists in automating the entire machine learning pipeline and is widely used in different areas, especially medical imaging. In this thesis, multiple AutoML based evolutionary approaches were proposed for medical image classification and segmentation tasks. First, a genetic algorithm based on a variable-length encoding scheme was proposed for the medical image classification task. In this approach, a mono-objective fitness function combining multiple objectives was used. Second, different performance estimation strategies were incorporated into Neural Architecture Search (NAS) for the medical image classification task. In the first approach, a surrogate-assisted NAS approach was proposed in which surrogate models were used to predict the fitness of individuals. In the second approach, Zero-Cost (ZC) proxies were used to predict the fitness of individuals in 2D/3D NAS. The whole approach was divided into two stages. In the first stage, the architecture was searched by an evolutionary algorithm, and then the best data augmentation was searched by the automatic data augmentation approach. The robustness of medical images against adversarial attacks was also studied in terms of architectural aspects, and a robust NAS approach was proposed to search for robust architectures against adversarial attacks. Furthermore, ZC proxies were also used for the retinal vessel segmentation task. Moreover, a multi-objective NAS approach was proposed to find hardware-efficient architectures.L'apprentissage automatique automatisé (AutoML) aide à automatiser l'ensemble du pipeline d'apprentissage automatique et est largement utilisé dans différents domaines, en particulier l'imagerie médicale. Dans cette thèse, plusieurs approches évolutives basées sur AutoML ont été proposées pour les tâches de classification et de segmentation des images médicales. Tout d'abord, un algorithme génétique basé sur un schéma de codage à longueur variable a été proposé pour la classification des images médicales. Dans cette approche, une fonction de fitness mono-objectif combinant plusieurs objectifs a été utilisée. Ensuite, différentes stratégies d'estimation des performances ont été incorporées dans la recherche d'architecture neuronale (NAS) pour la classification des images médicales. Dans la première approche, une approche NAS assistée par des surrogate models a été proposée, ces derniers étant utilisés pour prédire la fitness des individus. Dans la deuxième approche, des Zero-Cost (ZC) proxies ont été utilisés pour prédire la fitness des individus dans NAS 2D/3D. L'approche a été divisée en deux étapes. Dans la première étape, l'architecture a été recherchée par un algorithme évolutif, puis la meilleure augmentation des données a été recherchée par l'approche d'augmentation automatique des données. La robustesse des images médicales contre les attaques adversariales a également été étudiée, et une approche NAS robuste a été proposée pour rechercher des architectures robustes. En outre, des ZC proxies ont été utilisés pour la segmentation des vaisseaux rétiniens. De plus, une approche NAS multi-objectifs a été proposée pour trouver des architectures économes en matériel
Rupture des filés de fibres : modèle analytique et approche expérimentale
No full textInternational audienceRupture des filés de fibres : modèle analytique et approche expérimental
Brèves réflexions sur le multilinguisme européen : au-delà du recul du français, une promesse à réinventer
No full textInternational audienc
Ethnography as a way of seeing: Understanding activist communication in polarized public spheres
No full textInternational audienceEthnography in Communication Studies is more than a research method. Grounded in long-term immersion, it allows researchers to grasp the communicational and cultural processes of audiences that often remain invisible to quantitative approaches. This presentation explores this perspective through fieldwork conducted within the activist action of NGOs defending the rights of migrants. It highlights how ethnography, within the field of Communication Studies, enables an understanding of participants’ worldviews and the socio-political dynamics that shape their experiences in polarized public spheres. By engaging with lived experiences and situated practices, ethnography opens a space for producing knowledge that is both empirically grounded and politically meaningful
High-Pressure Intrusion of Saline Solutions in Hydrophobic STT-Type Zeosil
No full textInternational audienceHigh-pressure intrusion of water and LiCl aqueous solutions at different concentrations in hydrophobic STT-type zeosil was studied for possible applications in absorption and storage of mechanical energy. The water is intruded at a pressure of 35 MPa and remains trapped in the pores after pressure release, which corresponds to bumper behavior with total energy absorption. The use of LiCl solution leads to a change in system behavior, regardless of the concentration investigated (10, 15, or 20 M). Its intrusion is mainly reversible, but a small part of the intruded liquid remains in the pores after the first intrusion–extrusion cycle, which corresponds to a mixed behavior of bumper and shock absorber. The intrusion pressure rises strongly with an increase in salt concentration and reaches 227 MPa for a LiCl 20 M solution; the stored energy of 27 J/g can be achieved. The characterization of STT-type zeosil before and after intrusion–extrusion tests by structural and physicochemical methods shows that silanol defects are formed both under the intrusion of water and LiCl solutions. The relationship between zeosil structure and intrusion–extrusion characteristics is discussed by comparing the results obtained with those of other structural types of zeosils
From Rigid to Elastic: Tailoring the Properties of Linear Poly(IBOA-co-2EHA) Copolymers Synthesized by UV Polymerization
No full textInternational audienceThis study systematically investigated poly(isobornyl acrylate (IBOA)-co-2-ethylhexyl acrylate (2-EHA)) linear random copolymers synthesized via free radical photopolymerization to develop polyacrylate materials with tunable mechanical properties. A series of homopolymers and five copolymer compositions with varying IBOA content were prepared to study the relationship between structure and properties, as well as their correlation with mechanical behavior. The resulting linear polyacrylates were analyzed using multiple techniques. Structural characterization by ¹H NMR and MALDI-TOF MS provided detailed information on the chemical environment, molar mass, monomer conversion rate, comonomer sequence distribution within the copolymer backbone, and end-group analysis. Size exclusion chromatography revealed molar mass distributions ranging from 10⁵ to 10⁶ g/mol. DSC analysis revealed composition-dependent glass transition temperatures and no crystalline melting transitions, confirming the amorphous nature of all copolymers. Tensile testing at room temperature revealed distinct mechanical properties: Poly(IBOA) (PIBOA) exhibited brittle, rigid characteristics, and Poly(2-EHA) (P2EHA) displayed flexible, elastic properties. These copolymers demonstrate tunable mechanical properties as a function of comonomer composition, combining the structural rigidity of PIBOA with the flexibility of P2EHA. This compositional control over mechanical behavior makes these copolymers promising candidates for applications including biomedical devices, protective coatings, adhesives, and flexible electronics, where tailored mechanical properties are essential.Cette étude a systématiquement examiné les copolymères linéaires aléatoires poly(acrylate d'isobornyle (IBOA)-co-acrylate de 2-éthylhexyle (2-EHA)) synthétisés par photopolymérisation radicalaire afin de développer des matériaux polyacrylates aux propriétés mécaniques ajustables. Une série d'homopolymères et cinq compositions de copolymères à teneur variable en IBOA ont été préparés afin d'étudier la relation entre la structure et les propriétés, ainsi que leur corrélation avec le comportement mécanique. Les polyacrylates linéaires obtenus ont été analysés à l'aide de plusieurs techniques. La caractérisation structurelle par ¹H RMN et MALDI-TOF MS a fourni des informations détaillées sur l'environnement chimique, la masse molaire, le taux de conversion des monomères, la distribution des séquences de comonomères dans le squelette du copolymère et l'analyse des groupes terminaux. La chromatographie d'exclusion stérique a révélé des distributions de masse molaire allant de 10⁵ à 10⁶ g/mol. L'analyse DSC a révélé des températures de transition vitreuse dépendantes de la composition et aucune transition de fusion cristalline, confirmant la nature amorphe de tous les copolymères. Les essais de traction à température ambiante ont révélé des propriétés mécaniques distinctes : le poly(IBOA) (PIBOA) présentait des caractéristiques de fragilité et de rigidité, tandis que le poly(2-EHA) (P2EHA) affichait des propriétés de souplesse et d'élasticité. Ces copolymères présentent des propriétés mécaniques ajustables en fonction de la composition du comonomère, combinant la rigidité structurelle du PIBOA et la flexibilité du P2EHA. Ce contrôle de la composition sur le comportement mécanique fait de ces copolymères des candidats prometteurs pour des applications telles que les dispositifs biomédicaux, les revêtements protecteurs, les adhésifs et l'électronique flexible, où des propriétés mécaniques sur mesure sont essentielles
Energy balance in women during polar trekking—The POWER study
No full textInternational audienceAbstract Polar expeditions pose a significant challenge, contributing to a substantial energy deficit. However, few data exist on nonprofessional individuals, and none of them have investigated the regulation of energy balance in female participants during an Arctic expedition. Eleven nonathlete female adults who reached the North Pole by ski in full autonomy were studied. Before and the day after the expedition, resting metabolic rate (RMR) was measured by indirect calorimetry, and fat mass (FM) and fat‐free mass (FFM) by bioelectric impedance. Total daily energy expenditure (TDEE) and activity‐related energy expenditure (AEE) were assessed with the doubly labeled water (DLW) during the expedition. Before and throughout the expedition, daily physical activity was evaluated using accelerometry and heart rate, while surface skin temperature was measured using an iButton placed on the chest. Additionally, fasting salivary cortisol concentration was measured throughout the expedition as a stress marker. The seven‐day trekking was associated with a TDEE of 18.67 [SD 1.72] MJ/d. Body mass decreased by 1.67 (SE 0.42) kg, mostly due to fat mass loss (−1.26 [0.39] kg), indicating a negative energy balance. These findings show that nonathlete females can greatly increase their AEE and, hence TDEE over a short time period. However, their food consumption was insufficient to meet their needs, thus leading to an energy deficit and body mass loss. Future studies should investigate whether this insufficient energy intake was due to a lack of available food or an inability to consume more, thereby deepening our understanding of energy balance regulation in extreme conditions
