Archive ouverte de Centrale Lyon
Not a member yet
32420 research outputs found
Sort by
3rd International Workshop on Geographic Information Extraction from Texts (GeoExT 2025)
No full textInternational audienceA wealth of unstructured textual content holds geographic insights that are valuable across various domains. Despite recent significant advances in methods for extracting geographic information from texts, challenges remain regarding the reliability and applicability of these techniques. The Third GeoExT Workshop at ECIR 2025 is dedicated to addressing these challenges by fostering discussions on recent breakthroughs, novel approaches, and conceptual innovations in geographic information extraction. By facilitating in-depth discussions and networking among researchers and practitioners, the workshop aims to advance the field and inspire new solutions and applications
Electrical Characterization of Solderless Interconnections Dedicated to Pressure Assembly Power Modules
No full textInternational audienceThis paper presents the electrical characterization of solderless interconnection for the development of modular power modules with pressure assembly thanks to soft materials like metallic foams. This type of material can be used to connect two Printed Circuit Boards (PCB), and allows disassembling and changing the PCB if necessary. Electrical resistances of these soft materials are measured as a function of pressure and compared to classical interconnections like copper plates. With differences higher than 70% compared to copper plates, soft materials seem interesting for low resistance packaging interconnections
Sticky-PRE: A Sticky Proxy Re-Encryption Protocol for Persistent Vehicle Data Privacy
No full textStatus: Accepted. To be published soon.International audienceConnected vehicles generate a vast amount of dataand share it with external entities such as the cloud, neighbouringvehicles, Road-Side Units (RSUs), and other third-partyservices in a Vehicle to Everything (V2X) setting. This datais vulnerable and can lead to personal information leakageof vehicle owners, such as driving habits, travel routes, andidentity theft, among others. Moreover, with the implementationof the General Data Protection Regulation (GDPR), it becomesimperative to empower users with control over their data andthe ability to choose with whom they share it. To address thisobjective, we present a protocol that utilizes a combination ofsticky policies and a proxy re-encryption scheme. This protocolensures that user-defined access controls on the data persisteven when crossing organizational boundaries and addressesthe confidentiality, integrity, and accountability of vehicle data.Furthermore, we assess our protocol using a semi-honest threatmodel and analyze its vulnerabilities. Lastly, we perform aquantitative analysis of the data flow model to observe theperformance of the system
Comparing and Proposing Methods for Estimating Bilateral Migration Flows
No full textInternational audienceReliable data on international migration flows remain limited, despite their crucial importance for researchers—spanning fields such as economics, sociology, demography, and epidemiology—and for the formulation of effective public policies. Understanding the drivers of migration, predicting flow patterns, and assessing their consequences are therefore significant challenges. To tackle this issue, at least five synthetic estimators have been developed using bilateral migrant stock data from the World Bank and the United Nations \citep{abel2019}. However, comprehensive comparisons of these methods are scarce. In this paper, we evaluate and compare these approaches through simulations, while also demonstrating that some results can be theoretically established. After highlighting the importance of demographic adjustments across all methods and underscoring the need for a standardized framework, we introduce a new estimation method that combines stock differencing with migration rates, offering a balance of flexibility, simplicity, and accuracy. Our results suggest that this approach performs as well as the Pseudo-Bayes method and may even produce superior results if auxiliary data on return migration percentages between country pairs were available.Les données sur les flux migratoires entre paires de pays sont peu accessibles, malgré leur importance capitale pour les chercheurs -notamment en économie, sociologie et démographie -ainsi que pour l'élaboration de politiques publiques efficaces. Face à cette problématique, la seule solution disponible consiste à utiliser des estimations basées sur les données de stocks bilatéraux de migrants fournies par la Banque mondiale et les Nations Unies. Abel and Cohen (2019) ont recensé au moins cinq méthodes d'estimation possibles, mais aucune étude n'a encore effectué une comparaison exhaustive de ces méthodes. Dans cet article, nous évaluons et comparons ces approches à l'aide de simulations, tout en montrant que certains résultats peuvent être établis de manière théorique. Après avoir mis en évidence l'importance des ajustements démographiques pour chaque méthode, nous proposons une nouvelle approche d'estimation qui intègre le taux de retour dans la méthode de différence des stocks. Nos résultats montrent que cette méthode est comparable à la méthode Pseudo-Bayes, la plus couramment utilisée, et pourrait même offrir de meilleurs résultats si des données auxiliaires sur les pourcentages de retours migratoires entre paires de pays étaient disponibles
Losses Analysis of a Three-Phase Bidirectional Active Split Source Inverter for Traction Applications
No full textInternational audienceIn automotive applications, industrials use DC-DC boosting stages interfaced upstream of the Voltage Source Inverter (VSI) to increase the DC bus voltage. These topologies enable the motors to be driven with higher DC bus voltage than that delivered by the battery. Apparently, the use of extra DC-DC converter decreases the efficiency of the whole system. Recently a single-stage topology called the Split Source Inverter (SSI) was derived. This topology enables the boosting function. Unfortunatly, it exhibits very poor utilization of the DC bus voltage, thus preventing its use for traction applications. In previous work, the authors introduced a new single-stage topology called the Bidirectional Active Split Source Inverter (B-ASSI). It enables the boosting function with bidirectional power flow capability and a large utilization of the DC bus voltage. However, its efficiency was not thoroughly studied and comparison with conventionnal two-stage solution not conducted. This paper proposes an in-depth losses analysis of the B-ASSI. Analytical expression of the currents are proposed and validated by simulation. An extensive experimental validations are carry out on a permanent magnet synchronous machine (PMSM) test-bench. The derived expressions and proposed approadh enables a theoretical comparison of the B-ASSI to two-stage solutions and to the SSI but also giving insights about the design of the converter. This theoretical comparison is performed on a normalized driving cycle. It demonstrates that for the given design constraints the B-ASSI exhibits lower losses than the classical two-stage solution and the SSI for low-speed operating conditions corresponding to urban drive profiles
Density-valued symplectic forms from a multisymplectic viewpoint
No full textWe give and intrinsic characterization of multisymplectic manifolds that have the linear type of density-valued symplectic forms in each tangent space, prove Darboux-type theorems for these forms and investigate their symmetries
Ethical and Equitable Data Science: Bridging Social Justice and Technical Innovation ADBIS 2025 Doctoral Consortium Lecture
No full textInternational audienceThis paper summarises the fundamental background of the Doctoral Consortium titled "Ethical and Equitable Data Science". It introduces Freda, a methodology for designing ethical, frugal, and equitable data and algorithm-driven science. It bridges technical innovation with social justice by integrating data sovereignty, fairness-aware analytics, and community-in-the-loop infrastructure. Rooted in decolonial and feminist perspectives, Freda addresses transparency, accountability, and epistemic diversity through policy-aware Spark pipelines, federated learning, and negotiated resource dispatching. A case study illustrates how sovereignty-aware pipelines enable community control, minimise extractivism, and embed plural, justice-centred values into AI systems
VIRAL: VISION-GROUNDED INTEGRATION FOR REWARD DESIGN AND LEARNING
No full textInternational audienceThe alignment between humans and machines is a critical challenge in artificial intelligence today. Reinforcement learning, which aims to maximize a reward function, is particularly vulnerable to the risks associated with poorly designed reward functions. Recent advancements has shown that Large Language Models (LLMs) for reward generation can outperform human performance in this context. We introduce VIRAL, a pipeline for generating and refining reward functions through the use of multi-modal LLMs. VIRAL autonomously creates and interactively improves reward functions based on a given environment and a goal prompt or annotated image. The refinement process can incorporate human feedback or be guided by a description generated by a video LLM, which explains the agent's policy in video form. We evaluated VIRAL in five Gymnasium environments, demonstrating that it accelerates the learning of new behaviors while ensuring improved alignment with user intent. The source-code and demo video are available at: https://github.com/VIRAL-UCBL1/VIRAL and https://youtu.be/BBIB4bEVSjE.</div
Modèle numérique rapide de réseaux de Bragg sur fibre optique pour la mesure ultrasonore
No full textSHM : Structural Health Monitoring; GAPSUS - Acoustique Physique, Sous-Marine et Ultra-SonoreNational audienceLes rdB fibrés font l'objet d'un intérêt croissant depuis des années pour la surveillance des structures, favorisé par une faible intrusivité, de fortes capacités de multiplexage et une tenue en environnements difficiles. Ils peuvent notamment être utilisés pour des mesures ultrasonores. Les simulations en CSI sont utiles pour comprendre le comportement des structures, optimiser les configurations CSI et réaliser des analyses de sensibilité. La mise à disposition de modèles présentant un bon compromis entre rapidité d’exécution et précision de calcul est donc déterminante pour leur utilisation. Les éléments finis spectraux en espace avec schéma explicite en temps se sont révélés être une solution satisfaisante pour la modélisation de configurations CSI. Mais sans précaution adéquate, le maillage fin des fibres optiques, imposé par leur faible dimension, peut pénaliser l’efficacité d’une telle approche en termes de temps de calcul. Cette communication concerne une amélioration de la modélisation des rdB agissant comme récepteurs acoustiques. La stratégie proposée consiste en un couplage faible entre la structure et le rdB, qui néglige l'influence du capteur sur la propagation des ondes acoustiques dans la structure, permettant ainsi d’éviter sa prise en compte. Une transduction parfaite de la déformation mécanique de la structure au rdB est ainsi supposée. La réponse spectrale du rdB est calculée par une méthode matricielle après extraction de la déformation ultrasonore à la position du rdB. Le «edge filtering», technique expérimentale utilisée couramment pour réaliser les mesures ultrasonores via un rdB, est ensuite reproduite en extrayant la réponse à une longueur d’onde optique fixe en fonction du temps. Dans un second temps, des validations expérimentales sont présentées ; elles permettent notamment de montrer la capacité du modèle à retranscrire les caractéristiques de sensibilité des capteurs, avec un temps de calcul réduit par rapport aux approches précédentes
Modeling Magnetic Losses in Soft Ferromagnets: A Viscoelastic Approach Using Fractional Derivatives
No full textInternational audienceMagnetic losses in soft ferromagnetic materials (SFMs), particularly under high-frequency operation, pose significant challenges in modern electromagnetic applications. These losses result from complex magnetization dynamics that cannot be fully captured by classical empirical or semi-empirical models. Recent research has revealed that the magnetization process in SFMs exhibits viscoelastic characteristics, analogous to those observed in mechanical systems involving springs and dashpots. This document presents a summary of recent modeling efforts using fractional derivative operators to describe the viscoelasticity inherent in magnetic losses, with a focus on nanocrystalline cores.</div