Portail HAL Ensta
Not a member yet
11080 research outputs found
Sort by
CETIROISE : a cetacean passive acoustic observatory in a French Marine Natural Park
No full textInternational audienceThe CETIROISE project (funded under the French economic recovery plan) aimed at monitoring cetacean presence in the Iroise sea (Britanny, France) using passive acoustic monitoring. Targeted species were porpoise, delphinids (particularly common and bottlenose dolphins) and large baleen whales. From May 2022 until May 2023, 7 sites were equipped in the northern sector of the Iroise Marine Natural Park (PNMI). Two coastal moorings (20m depth) and five deepest moorings (100m depth) were each equipped with two recorders: the F-POD (Chelonia Limited), an echolocation-click logging device, and the Sylence LP (RTSys), a broadband recorder sampling at 128 kHz. Delphinid whistles were detected with PAMGuard’s whistle and moan detector while delphinid and porpoise clicks were directly assessed from the FPOD detections. Manual annotation was performed on the web-based annotation platform APLOSE. The annotations were used to assess delphinid clicks and whistles detection performances and also to monitor the low frequency baleen whale calls on the noisy low frequency band. Results highlight the continuous yearly presence of delphinids, especially at the farthest sites from the coast. But these results also point out the need to develop tools to be able to discriminate delphinids species from their sounds. Porpoise were recorded to a lesser extent, but they also appear to dwell in this region during most of the year. Among baleen whales, only a few minke whale calls were found in the data during the first analysed months. These findings supplement the picture drawn by other observation technics and support the development of the PNMI conservation plan
Palladium-mediated S-arylation and S-allylation for late-stage peptide functionalisation and protein conjugation
No full textInternational audienceThe chemoselective functionalization of peptides and proteins through manipulation and modification of the integral components within a complex polyfunctional system represents a key area of research for the elucidation of biological mechanisms and generation of potential pharmaceuticals. Messaoudi and co-workers recently reported the palladium-catalyzed Buchwald-Hartwig-Migita (BHM) coupling of thiosugars as well as aryl- and alkyl-thiols with various iodoarenes at room temperature. The coupling involves the use of the Pd G3 XantPhos as a precatalyst and exhibits a high functional group tolerance. [1] The synthetic utility of this method was subsequently expanded by using semi-aqueous conditions to perform various conjugation reactions on unprotected para-iodophenylalanine (pIPhe) containing peptides, with various glycosyl thiols. Herein, we report the extension of this BHM coupling beyond thioglycoconjugation. Additionally, we report the first example of a carbon-sulfur bond-forming Tsuji-Trost reaction with a peptide or protein as substrate. Strikingly, the Tsuji-Trost S-allylation and the BHM S-arylation have very different requirements in terms of Pd-based catalytic systems and therefore could be employed in tandem as orthogonal reactions, for either one-pot or sequential syntheses. We applied these novel methodologies to the labeling of biologically relevant peptides and proteins, demonstrating impressive functional group tolerance, high yields, and finally in cellulo evaluation of some of theconjugate
Fusion d’information multi-source en monde ouvert - Vulnérabilités du processus et gestion des conflits
No full textIn the context of close collaboration between separate information systems, the means used to merge so many sources of information make isolated entities function as a whole. This is the case in the maritime domain, where naval information systems communicate and act on the basis of uncertain information that is generally incomplete, heterogeneous, imprecise and vulnerable. Thanks to its ability to skillfully handle uncertainty in all its forms, Dempster-Shafer theory is a subtle and effective mathematical framework for achieving this fusion process. What is more, the more complex a system is, the more conflictual the information it exchanges is likely to be. So, what about the possible values of the state of a system as complex as that of a ship ? Furthermore, if vulnerable information is subject to corruption, what is its impact and how does it propagate in an information system ? In this thesis, we are interested in the collaborative aspect of information conveyed in a complex system, in how information is merged and in understanding the origin of conflict in information. The theoretical core of our research is based on the definition of a conflict measure whose evolution is non-monotonic, and of a new information fusion rule that is robust to corruption. We carry out our research in the context of the open-world : a paradigm where the presence of conflict can call into question our a priori knowledge, and the representation of possibilities it entails.Dans le cadre d'étroite collaboration entre différents systèmes d'information séparés, les moyens mis en oeuvre pour la fusion d'autant de sources d'information font fonctionner des entités isolées comme un tout. C'est le cas du domaine maritime, où les systèmes d'information navals communiquent et agissent sur la base d'information incertaine généralement incomplète, hétérogène, imprécise et vulnérable. Par sa capacité à gérer habilement l'incertitude sous toutes ses formes, la théorie de Dempster-Shafer est un cadre mathématique subtil et efficace pour réaliser ce processus de fusion. En outre, plus un système est complexe, plus l'information qu'il échange est susceptible d'être conflictuelle. Aussi, qu'en est-il des possibles valeurs de l'état d'un système aussi complexe que celui d'un navire ? Par ailleurs, si l'information vulnérable est sujette à corruption, quelle est son impact et comment se propage-t-elle dans un système d'information ? Dans ce mémoire de thèse, nous nous intéressons à l'aspect collaboratif de l'information véhiculée dans un système complexe, à la façon dont l'information est fusionnée et à comprendre l'origine du conflit dans l'information. Le coeur théorique de notre recherche repose sur la définition d'une mesure de conflit dont l'évolution est non-monotone et d'une nouvelle règle de fusion de l'information, robuste en cas de corruption. Nous portons notre recherche dans le cadre du monde ouvert : un paradigme où la présence de conflit peut remettre en cause notre connaissance a priori, et la représentation des possibles qu'elle lui incombe
CO2 hydrates crystallization kinetic parametric study: effect of impeller type
No full textInternational audienceCO2 hydrate slurry is a promising phase change material for cold storage and distribution issues, due to its high latent heat (about 500 kJ.kg-1 of water, higher than that of ice - 333 kJ.kg-1) and wide melting temperature range suitable for air conditioning application. While the thermodynamic properties of CO2 hydrates are now well established, knowledge of crystallization kinetics phenomena is still a challenge. Trying to understand and control the formation of gas hydrates is a key factor, since their discovery in pipeline plugs. The goal here, for cooling application, is to increase the rate of CO2 hydrate formation. Most laboratory-scale hydrate reactors are equipped with pressure and temperature probes, and mass balance on CO2 allows formation kinetic studies, but with assumptions on CO2 concentration in the liquid phase and the hydration number.The present work investigates the kinetics of CO2 hydrate crystallization for two different types of stirrers, on one hand a three pitched blade stirrer and on the other hand a hollow shaft eight blade Rushton turbine. We use a jacketed stirred batch reactor, with a specially developed sensor (a thermopile) to determine experimentally heat balance on the cooling jacket. The mass fraction of crystallized hydrate during the time is determined directly from this heat balance. Those results are compared with mass balance approach. To highlight the effect of stirrer type on mass transfer, especially between vapour and liquid phases, an additional experimental study to determine vapour/liquid mass transfer coefficient, kLa, were achieved. We conclude about the global performance for heat and mass transfers of both stirrer, and give general recommendation for the choice of stirrer type and scale-up process
HeROsim: An Allocation and Scheduling Simulator for Evaluating Serverless Orchestration Policies
No full textInternational audienc
Vers une interaction personnalisée entre l’humain et le véhicule en mode multimodal
No full textThis thesis explores the advancement of intelligent vehicles that are increasingly personalized and capable of seamless multi-modal human-machine interaction (HMI). The research addresses key challenges in understanding driver preferences, enhancing vehicle dynamic control, and developing interfaces that foster a reliable and trustworthy driving experience.The study begins by investigating the impact of human factors, including emotions and personality traits, on driving behavior and road safety. Through driving simulations and empirical studies, it highlights how individual differences—such as age, experience, and personality—affect driving performance, particularly under emotional stress. These insights are crucial for developing adaptive models that improve the interaction between humans and vehicles. The research then examines the influence of narcissism on driving behavior, particularly in car-following scenarios under various weather conditions. By integrating personality psychology with driving behavior analysis, the study underscores the importance of systems that can dynamically adjust to individual drivers' traits, enhancing both safety and the overall driving experience.An innovative approach is presented through the development of a haptic interface designed to improve multi-modal takeover requests (TOR) in smart vehicles. The Pneumatic Tactile Silicon Sac (PTSS) interface, which integrates visual, auditory, and haptic feedback, is shown to significantly enhance driver takeover performance, and build greater trust in autonomous systems. Further exploration focuses on self-sensing and virtual sensor technologies, proposing a unified framework for vertical state estimation that incorporates uncertainty, thus advancing vehicle dynamic control and ensuring greater stability and safety.Finally, the thesis explores the critical role of trust in autonomous vehicles through a detailed analysis of real-world driving scenarios, identifying key moments when drivers take over control. This study emphasizes the continuous effort needed to build and maintain trust in self-driving technologies.In conclusion, this research contributes to the ongoing journey towards more personalized vehicles and sophisticated multi-modal human-machine interactions. By integrating psychological insights with advanced engineering solutions, the study advances the adaptability, safety, and trustworthiness of future intelligent vehicles, laying the groundwork for more personalized and user-centered transportation systems.Cette thèse explore l'avancement des véhicules intelligents, de plus en plus personnalisés et capables d'une interaction homme-machine (HMI) multi-modale fluide. La recherche aborde les défis liés à la compréhension des préférences des conducteurs, à l'amélioration du contrôle dynamique des véhicules et au développement d'interfaces pour une conduite fiable et digne de confiance. L'étude examine l'impact des facteurs humains, comme les émotions et les traits de personnalité, sur le comportement au volant et la sécurité routière. Par des simulations de conduite et des études empiriques, elle montre comment les différences individuelles—âge, expérience, personnalité—affectent les performances de conduite, surtout sous stress émotionnel. Ces informations sont cruciales pour développer des modèles adaptatifs améliorant l'interaction humain-véhicule. La recherche se penche ensuite sur l'influence du narcissisme dans des scénarios de suivi sous diverses conditions météorologiques, soulignant l'importance des systèmes capables de s'ajuster dynamiquement aux traits individuels des conducteurs. Une approche innovante est présentée avec le développement d'une interface haptique pour améliorer les demandes de reprise de contrôle (TOR) dans les véhicules intelligents. L'interface Pneumatic Tactile Silicon Sac (PTSS), intégrant retours visuels, auditifs et haptiques, améliore significativement la performance de reprise de contrôle du conducteur et renforce la confiance dans les systèmes autonomes. L'exploration se poursuit avec les technologies d'auto-perception et de capteurs virtuels, proposant un cadre unifié pour l'estimation de l'état vertical intégrant l'incertitude, améliorant ainsi le contrôle dynamique et la stabilité des véhicules .Enfin, la thèse explore le rôle crucial de la confiance dans les véhicules autonomes à travers une analyse des scénarios réels, identifiant les moments clés où les conducteurs reprennent le contrôle. Cette étude souligne l'effort continu pour construire et maintenir la confiance dans les technologies de conduite autonome. En conclusion, cette recherche contribue au développement de véhicules plus personnalisés et d'interactions homme-machine plus sophistiquées. En intégrant des connaissances psychologiques avec des solutions d'ingénierie avancées, l'étude fait progresser l'adaptabilité, la sécurité et la fiabilité des véhicules intelligents de demain, jetant ainsi les bases de systèmes de transport plus personnalisés et centrés sur l'utilisateur
Allocation et placement dynamiques sur ressources hétérogènes pour le cloud serverless
No full textThe serverless model is a cloud paradigm in which hardware resources are not reserved but allocated on demand, triggered by incoming user requests. Clients are billed based on their actual resource usage. In return, the responsibility for resource allocation and task placement lies with the provider. This orchestration mechanism can lead to latency overhead and degrade the quality of service for users. We focused on the provider’s ability to guarantee quality of service given an infrastructure comprising heterogeneous hardware resources (CPU, GPU, FPGA, DLA), in the context of two deployments: a deepfake detection application, based on stateless functions, and an intrusion detection application, relying on chains of functions that communicate intermediate results. We proposed an orchestration policy that optimizes the deployment of applications composed of a single function, minimizing penalties on quality of service. We also modeled complex applications, taking into account the dependencies between the functions they comprise, aiming to reduce energy consumption and task disaggregation across compute nodes. Finally, we developed a simulator to represent such environments and to evaluate and compare different cloud orchestration strategies.Le modèle serverless est un paradigme pour le cloud dans lequel les ressources matérielles ne sont pas réservées, mais allouées à la demande, lors de la réception de requêtes utilisateur. La facturation des clients s’effectue sur la base de l’utilisation réelle des ressources. En contrepartie, la responsabilité de l’allocation des ressources et du placement des tâches incombe au fournisseur. Ce mécanisme d’orchestration peut induire un surcoût en latence et dégrader la qualité de service pour les utilisateurs. Nous nous sommes intéressés à la capacité pour le fournisseur à garantir la qualité de service étant donnée une infrastructure dotée de matériel hétérogène (CPU, GPU, FPGA, DLA), dans le cadre de deux déploiements : une application de détection de deepfake, basée ur des fonctions sans état, et une application de détection d’intrusions, reposant sur des chaînes de fonctions communiquant des résultats intermédiaires. Nous avons proposé une politique d’orchestration qui optimise le déploiement d’applications composées d’une unique fonction, en minimisant les pénalités sur qualité de service. Nous avons également modélisé des applications complexes, en prenant en compte les dépendances entre fonctions qui les composent, pour chercher à réduire la consommation énergétique et la désagrégation des tâches sur les nœuds de calcul. Enfin, nous avons développé un simulateur permettant de représenter de tels environnements, et d’évaluer et de comparer différentes stratégies d’orchestration pour le cloud
Studies of antibacterial activity (in vitro and in vivo) and mode of action for des-acyl tridecaptins (DATs)
No full textInternational audienceTridecaptins comprise a class of linear cationic lipopeptides with an N-terminal fatty acyl moiety. These 13-mer antimicrobial peptides consist of a combination of d- and l-amino acids, conferring increased proteolytic stability. Intriguingly, they are biosynthesized by non-ribosomal peptide synthetases in the same bacterial species that also produce the cyclic polymyxins displaying similar fatty acid tails. Previously, the des-acyl analog of TriA1 (termed H-TriA1) was found to possess very weak antibacterial activity, albeit it potentiated the effect of several antibiotics. In the present study, two series of des-acyl tridecaptins were explored with the aim of improving the direct antibacterial effect. At the same time, overall physico-chemical properties were modulated by amino acid substitution(s) to diminish the risk of undesired levels of hemolysis and to avoid an impairment of mammalian cell viability, since these properties are typically associated with highly hydrophobic cationic peptides. Microbiology and biophysics tools were used to determine bacterial uptake, while circular dichroism and isothermal calorimetry were used to probe the mode of action. Several analogs had improved antibacterial activity (as compared to that of H-TriA1) against Enterobacteriaceae. Optimization enabled identification of the lead compound 29 that showed a good ADMET profile as well as in vivo efficacy in a variety of mouse models of infection
Development and distribution of an in situ experimental wind turbine noise database
No full textInternational audienceThe PIBE project is the first French collaborative research project on wind turbine noise. One of its objectives is to quantify the experimental dispersion observed in situ on acoustic quantities/metrics and influencing parameters, and to compare it with the numerical dispersion estimated using acoustic propagation models. A long-term experimental campaign was carried out over consecutive 410 days around a wind farm comprising eight 80-metre-high and 90-metre diameter turbines. Acoustic sensors (five Class I sound level meters) were deployed on site, recording third-octave spectra and overall A- and Z-weighted values of the Leq indicator and fractile indices (L10, L50 and L90), at distances from 325 m to 1400 m from the wind turbines row. In addition, three 3D ultrasonic anemometers and a Lidar system were used to measure meteorological variables influencing long-range sound propagation (e.g. wind and temperature verical profiles). Besides, train and aircraft passing close to the site were inventoried to identify periods of high background noise. All experimental data have been processed into an ElasticSearch database. An interactive web application has been developed in R Shiny to facilitate processing, visualization and analysis of the experimental database
nférence statistique de la profondeur de la colonne d’eau par analyses radiométrique et géométrique de l’imagerie spectrale
No full textThe complexities of mapping shallow water bodies, particularly coastal areas, have long been studied, yet a focused understanding of Water Column Depth (WCD) uncertainties remains notably lacking. While accurate measurement of these uncertainties is crucial for reliable bathymetric charts and safe navigation, current research has primarily focused on WCD estimation without considering associated uncertainties. Traditional approaches, whether radiometric or geometric analyses of spectral imagery, face distinct limitations in optically complex waters and featureless seabeds, respectively. Our work introduces a likelihood-based inference framework designed for robust estimation of WCD uncertainties through two main objectives. First, we develop a radiometric inferential approach based on Radiative Transfer (RT) modeling with Semi-Analytical (SA) techniques, enabling detailed analysis of uncertainties and inherent biases. Second, we establish a geometric inferential approach using stereo-photogrammetric triangulation to evaluate WCD uncertainties from a geometric perspective. These frameworks are validated through realistic simulations, focusing on understanding and quantifying WCD uncertainties. Our findings confirm the effectiveness of the likelihood-based inference framework in quantifying uncertainties for both radiometric and geometric WCD estimations. This research makes a significant contribution to passive remote sensing and marine science by providing reliable, comprehensive methods for assessing uncertainties in shallow water mapping.La cartographie précise des eaux peu profondes, particulièrement en zones côtières, reste un défi majeur, notamment en ce qui concerne les incertitudes dans la mesure de la profondeur de la colonne d’eau (WCD). Cette précision est essentielle tant pour la navigation que pour la recherche, qui s’est jusqu’ici principalement concentrée sur l’estimation de la WCD sans considérer pleinement les incertitudes associées. Les méthodes traditionnelles, qu’elles soient radiométriques ou géométriques, présentent chacune leurs limitations spécifiques. Notre travail propose un cadre d’inférence basé sur la vraisemblance pour estimer rigoureusement ces incertitudes. L’approche se développe selon deux axes : une analyse radiométrique fondée sur la modélisation du transfert radiatif avec des techniques semi-analytiques, et une approche géométrique basée sur la triangulation stéréophotogrammétrique. Cette double méthodologie permet d’évaluer les biais et les erreurs inhérents à chaque approche. Le développement de ces cadres d’inférence dans les domaines radiométrique et géométrique est validé par des simulations réalistes, permettant une compréhension approfondie des incertitudes dans l’estimation de la WCD. Nos résultats démontrent l’efficacité de cette approche pour quantifier les incertitudes dans les deux types d’estimations. Cette recherche apporte une contribution significative à la télédétection passive et aux sciences marines en proposant des méthodes fiables pour évaluer les incertitudes dans la cartographie des eaux peu profondes