Archive ouverte de Centrale Lyon
Not a member yet
32420 research outputs found
Sort by
Yudovich theory under geometric regularity for density-dependent incompressible fluids
No full textThis paper focuses on the study of the density-dependent incompressible Euler equations in space dimension d = 2, for low regularity (i.e. non-Lipschitz) initial data satisfying assumptions in spirit of the celebrated Yudovich theory for the classical homogeneous Euler equations.We show that, under an a priori control of a non-linear geometric quantity, namely the directional derivative ∂X u of the fluid velocity u along the vector field X := ∇ ⊥ ρ, where ρ is the fluid density, low regularity solutions à la Yudovich can be constructed also in the non-homogeneous setting. More precisely, we prove the following facts:(i) stability: given a sequence of smooth approximate solutions enjoying a uniform control on the above mentioned geometric quantity, then (up to an extraction) that sequence converges to a Yudovich-type solution of the density-dependent incompressible Euler system;(ii) uniqueness: there exists at most one Yudovich-type solution of the density-dependent incompressible Euler equations such that ∂X u remains finite; besides, this statement improves previous uniqueness results for regular solutions, inasmuch as it requires less smoothness on the initial data
Volume Preserving Neural Shape Morphing
No full textInternational audienceShape interpolation is a long standing challenge of geometry processing. As it is ill-posed, shape interpolation methods always work under some hypothesis such as semantic part matching or least displacement. Among such constraints, volume preservation is one of the traditional animation principles. In this paper we propose a method to interpolate between shapes in arbitrary poses favoring volume and topology preservation. To do so, we rely on a level set representation of the shape and its advection by a velocity field through the level set equation, both shape representation and velocity fields being parameterized as neural networks. While divergence free velocity fields ensure volume and topology preservation, they are incompatible with the Eikonal constraint of signed distance functions. This leads us to introduce the notion of adaptive divergence velocity field, a constructioncompatible with the Eikonal equation with theoretical guarantee on the shape volume preservation. In the non constant volume setting, our method is still helpful to provide a natural morphing, by combining it with a parameterization of the volume change over time. We show experimentally that our method exhibits better volume preservation than other recent approaches, limits topological changes and preserves the structures of shapes better without landmark correspondences
Facilitate and scale up the creation of 3D meshes, 6D category-based datasets and grasping with generative models: genvegefruits3d
No full textInternational audienceDespite significant advances in 2D image, enabled by foundation models, progress in 3D understanding, particularly in 6D pose estimation and shape reconstruction, remains limited by the scarcity of datasets. In particular, in category 6D pose estimation, the high costs of real-world data collection have resulted in datasets with restricted categories, low diversity, and minimal instance variability. Recent methods have attempted to address this gap by leveraging synthetic image generation tools. However, these approaches are constrained by the limitations of available mesh datasets, which hinder the diversity, scalability, and inclusion of novel objects in generated samples. In this work, we propose a first automatic pipeline for the large-scale generation of 3D category-based datasets. Our approach uses 3D generative models guided by textual input to produce diverse and scalable datasets. To demonstrate its efficacy, we generated a new dataset named GenVegeFruits3D comprising 100 categories of fruits and vegetables, each containing over 1000 unique meshes. This significantly enhances the scale and diversity of existing category-based 3D datasets while reducing reliance on pre-existing 3D meshes. Additionally, we trained a 3D generative model, a 3D understanding model, and a grasping model, including on a real robotic setup. The dataset and code are available at: https://datasets.liris.cnrs.fr/genvegefruits3d-version1
Is Generative Artificial Intelligence Ready for Computational Social Science? ⋆
No full textInternational audienceIn this article, we analyse the potential of Large Language Models (LLMs) for social simulation by assessing their ability to: (a) make decisions aligned with explicit preferences; (b) adhere to principles of rationality; and (c) refine their beliefs to anticipate the actions of other agents. Through game-theoretic experiments, our results show that certain models, such as GPT-4.5 and Mistral-Small, exhibit consistent behaviours in simple contexts but struggle with more complex scenarios requiring anticipation of other agents' behaviour. Our study outlines research directions to overcome the current limitations of LLMs
Recommendation of TC RILEM TC 274-TCE: 3-point bending test procedure for earthen bricks—quality control of earth bricks for structural masonry by flexural strength
No full textInternational audienceBuilding with unstabilised earth bricks (EB) as units for structural masonry walls offers a sustainable alternative to mitigate resource depletion's negative impact and global warming. Using locally sourced earth for EB production is required to achieve these benefits. This leads to large variations in material composition. Consequently, quality control becomes critical to ensure adequate mechanical performance of the masonry units. In the framework of the RILEM TC 274-TCE, the characterisation of the flexural strength of bricks has been examined. A comprehensive inter-laboratory campaign involving seven laboratories across three countries was conducted, during which flexural tests were performed on 98 bricks. This article details the TC 274-TCE recommendations on the procedure to conduct a 3-point bending test on an EB. Key test parameters such as specimen pre-conditioning, setup boundary conditions, loading rate and determination of material flexural strength are discussed. The calculation of flexural strength was evaluated by comparing results obtained from the beam theory model and a solid mechanics approach, validated through finite element modelling. The findings confirm that the 3-point bending test is a reliable method to assess the mechanical performance of EB and to control their quality
Sobol’ Sequences with Guaranteed-Quality 2D Projections
No full textInternational audienceLow-discrepancy sequences, and more particularly Sobol’ sequences are gold standard for drawing highly uniform samples for quasi-Monte Carloapplications. They produce so-called (t, s)-sequences, that is, sequences of s-dimensional samples whose uniformity is controlled by a non-negative integer quality factor t. The Monte Carlo integral estimator has a convergence rate that improves as t decreases. Sobol’ construction in base 2 also provides extremely fast sampling point generation using efficient xor-based arithmetic. Computer graphics applications, such as rendering, often require high uniformity in consecutive 2D projections and in higher-dimensional projections at the same time. However, it can be shown that, in the classical Sobol’ construction, only a single 2D sequence of points (up to scrambling), constructed using irreducible polynomials x and x + 1, achieves the ideal t = 0 property. Reusing this sequence in projections necessarily loses high dimensional uniformity. We prove the existence and construct many 2D Sobol’ sequences having t = 1 using irreducible polynomials p and p +p +1. They can be readily combined to produce higher-dimensional low discrepancy sequences with a high-quality t = 1, guaranteed in consecutive pairs of dimensions. We provide the initialization table that can be directly used with any existing Sobol’ implementation, along with the corresponding generator matrices, for an optimized 692-dimensional Sobol’ construction. In addition to guaranteeing the (1, 2)-sequence property for all consecutive pairs, we ensure that t ≤ 4 for consecutive 4D projections up to 215 points
On the impact of the negative electrode in Graphite/LFP cells during first and second life aging experiments
No full textInternational audienceIn order to reuse batteries, aging models must describe cell aging in both first and second life to predict the remaining useful life. However, most of the aging studies on lithium-ion batteries are focused first life experiments from 100% to 70% of State of Health. The lack of data is a barrier to understand the aging and its impact on long-term user. This study aims to highlight the discrepancy of aging between cells in the same test condition and the mismatch between experimental results and aging models in a long-term aging study. Experimental tests are conducted on cylindrical 18650 cells in different cases of use: electric vehicle use, fast charge, reduce voltage window, partial or complete depth of discharge. Experiments were conducted over 2 years and help understand cell aging from new to end-of-life. The aging of the negative electrode is estimated and have a strong impact on the cell aging in long term use, regardless of the test case. In high C-rate cycling or with a complete depth of discharge, the impact of the negative electrode on the cell aging is significantly higher. An empirical method is presented to take into account the electrode aging in order to exhibit the impact of the electrode on the cell. The impact of this finding in tested on existing model and then discussed
L'application MOBILES : favoriser le partage d'expériences urbaines entre étudiants internationaux
No full textInternational audienceL'intégration des étudiants internationaux, enjeu majeur pour leur réussite académique, est fréquemment compromise par des difficultés telles que le choc culturel – se manifestant, par exemple, par des normes sociales déroutantes –, l'isolement social, intensifié par un manque de réseaux de soutien, et les barrières linguistiques, limitant leurs interactions et leurs apprentissages. Afin de favoriser leur engagement et leurs apprentissage, nous avons conçu MOBILES, une application innovante transformant l'exploration urbaine en un outil pédagogique. Elle incite les étudiants à explorer activement leur environnement et à partager leurs expériences, facilitant ainsi leur adaptation socioculturelle. Par des recommandations personnalisées, fondées sur leurs intérêts, MOBILES offre une expérience immersive qui renforce leur sentiment d'appartenance et établit un lien significatif entre intégration sociale et réussite académique, redéfinissant leur rapport à la communauté d'accueil
Humanidades digitais em tempos de crise: resistências, imaginários e formas de coconstrução
No full textInternational audienceThis article introduces the dossier “Humanidades digitais em tempos de crise: resistências, imaginários e formas de coconstrução”, situating digital humanities within contemporary contexts of historical convulsion, social crisis and radical technological transformation. Drawing on Latin American decolonial perspectives such as Améfrica Ladina and Abya Yala, the authors discuss how algorithmic systems co-construct social practices, reinforcing or subverting relations of power. They map two main strands of critical scholarship: analyses of platform capitalism, data extractivism, surveillance and epistemic violence, and studies foregrounding popular algorithmic cultures, digital activism and community-based appropriations of technology. The introduction then presents the dossier’s contributions, which examine gendered and racial asymmetries in science, hate speech against teachers, interfaces of digital newspaper archives, and feminist critiques of academic sexism and generative AI. Overall, the text argues for digital humanities and data science from the Global South as a situated, feminist and anti-racist project committed to epistemic justice and plural technological futures
Eukaryotic Ancestry in a Finite World
No full textInternational audienceFollowing genetic ancestry in eukaryote populations poses several open problems due to sexual reproduction and recombination. The history of extant genetic material is usually modeled backwards in time, but tracking chromosomes at a large scale is not trivial, as successive recombination events break them into several segments. For this reason, the behavior of the distribution of genetic segments across the ancestral population is not fully understood. Moreover, as individuals transmit only half of their genetic content to their offspring, after a few generations, it is possible that ghosts arise, that is, genealogical ancestors that transmit no genetic material to any individual. While several theoretical predictions exist to estimate properties of ancestral segments or ghosts, most of them rely on simplifying assumptions such as an infinite population size or an infinite chromosome length. It is not clear how well these results hold in a finite universe, and current simulators either make other approximations or cannot handle the scale required to answer these questions. In this work, we use an exact back-intime simulator of large diploid populations experiencing recombination that tracks genealogical and genetic ancestry, without approximations. We focus on the distinction between genealogical and genetic ancestry and, additionally, we explore the effects of genome structure on ancestral segment distribution and the proportion of genetic ancestors. Our study reveals that some of the theoretical predictions hold well in practice, but that, in several cases, it highlights discrepancies between theoretical predictions assuming infinite parameters and empirical results in finite populations, emphasizing the need for cautious application of mathematical models in biological contexts