HAL Portal UDL Université de Lyon
Not a member yet
327890 research outputs found
Sort by
Friction on Demand: A Generative Framework for the Inverse Design of Metainterfaces
No full textInternational audienceDesigning frictional interfaces to exhibit prescribed macroscopic behavior is a challenging inverse problem, made difficult by the non-uniqueness of solutions and the computational cost of contact simulations. Traditional approaches rely on heuristic search over low-dimensional parameterizations, which limits their applicability to more complex or nonlinear friction laws. We introduce a generative modeling framework using Variational Autoencoders (VAEs) to infer surface topographies from target friction laws. Trained on a synthetic dataset composed of 200 million samples constructed from a parameterized contact mechanics model, the proposed method enables efficient, simulation-free generation of candidate topographies. We examine the potential and limitations of generative modeling for this inverse design task, focusing on balancing accuracy, throughput, and diversity in the generated solutions. Our results highlight trade-offs and outline practical considerations when balancing these objectives. This approach paves the way for near-real-time control of frictional behavior through tailored surface topographies
Optimal sub-Gaussian variance proxy for truncated Gaussian and exponential random variables
No full textInternational audienceThis paper establishes the optimal sub-Gaussian variance proxy for truncated Gaussian and truncated exponential random variables. The proofs are based initially on reducing each distribution to their standardized versions. Geometrically, for the normal distribution, our argument consists of fitting a parabola to another parabola-looking function, which emerges from its moment generating function. For the exponential case, we show that the optimal variance proxy is the unique solution to a pair of equations and then provide this solution explicitly. Moreover, we demonstrate that truncated Gaussian variables exhibit strict sub-Gaussian behavior if and only if they are symmetric, meaning their truncation is symmetric with respect to the mean. Conversely, truncated exponential variables are shown to never exhibit strict sub-Gaussianity
The Verification Gap in AI Accessibility: A Meta-Analysis of Architectural Paradigms and Agency Deficits
No full textInternational audienceAI-powered accessibility systems have shifted from transcription to interpretation, creating a verification paradox: blind and low-vision (BLV) users, who most rely on AI-generated narratives, lack perceptual redundancy to detect errors such as hallucinations or semantic drift, resulting in an agency deficit. Through a systematic meta-analysis of 17 empirically evaluated AI accessibility systems (2019--2025), we examine how architectural choices shape verification affordances. We introduce the Verification-Cognition Balance Index (VCBI), a composite metric quantifying trade-offs between validation rigor, user agency, complexity, and risk. Our analysis uncovers weak overall balance across the field, with transformer-based systems showing higher risk density than classical computer vision with no gain in validation rigor. Only hybrid specialized architectures—constraining AI to narrow semantic subtasks—achieve good balance in our corpus, though this finding is preliminary given the small number of such systems evaluated. We argue for a necessary paradigm shift in evaluation from system accuracy to user verifiability. Our findings suggest that for high-stakes accessibility, task-specific constrained AI outperforms general-purpose models in epistemic safety and user empowerment
The sense of agency in near and far space: where do we stand?
No full textInternational audienceModern technology frequently places the consequences of our actions at a distance (e.g. remote surgery, smart-home control, virtual reality). Does spatial distance between an action and its outcome weaken the sense of agency (SoA) - the feeling of control over one’s actions and consequences? Two recent studies, by Jenkins and Obhi and Mariano et al., answered “yes,” reporting stronger temporal binding (TB) in near than far space and interpreting this as greater implicit agency. A third study - our own work with a similar paradigm - found no distance effect. Here we (i) provide a rigorous side-by-side methodological comparison of the three studies, (ii) argue why a direct test to establish a distance modulation of TB (the Near - Far difference of the Active - Passive delta) should be performed in order to reach meaningful conclusions, and (iii) report new reanalyses of our data and direct tests on the two target studies. Overall, current evidence does not support a distance effect on SoA. Our reassessment provides alternative explanations that converge with available evidence suggesting that distance may influence temporal interval perception, but that effect is independent of action intention and therefore of agency. Public Significance Statement: In our increasingly connected world, we often interact with devices and influence events that are physically distant from us - like controlling smart appliances remotely or engaging in virtual reality experiences. Does the distance between us and the effects of our actions change how much we feel in control of the outcomes of our actions? Recent studies have suggested that we feel less responsible when action outcomes appear in far space compared to near space. However, methodological limitations in these studies, as well as our own results with a similar paradigm, challenge the validity of this claim. We aim to contribute to the crucial debate on the role of spatial distance on humans’ feeling of responsibility by providing a respectful though critical analysis of recent findings and offering recommendations for future research
Comment les choix de design d'interface mène à des effets environnementaux indirects par l'intensification des usages: Une évaluation environnementale empirique des applications de messagerie
No full textInternational audienceIn sustainable HCI, the 'Cornucopian Paradigm' describes how the expansion of digital services goes hand in hand with the growth of digital infrastructure and its environmental footprint. Yet, the mechanisms driving this paradigm are not well understood at the design level. As a case study, we investigate how messaging apps factor in the cornucopian paradigm by conducting a feature analysis of 17 apps complemented with 12 user interviews. We identify four scale-up mechanisms that explain how design contributes to the cornucopian paradigm. These mechanisms all contribute to use intensification, i.e., digital practices becoming more intensive and leading to more data intensity, to infrastructure expansion, and indirectly to device obsolescence and replacement. Our analysis of use intensification builds upon a set of recurring cornucopian design patterns we identified in messaging apps. To mitigate the cornucopian paradigm, we propose a set of moderation heuristics to guide design decisions.Dans le champ de l'IHM soutenable (SHCI), le « paradigme de la corne d'abondance » décrit comment l'expansion des services numériques s'accompagne de l'expansion des infrastructures numériques et de leur impact environnemental. Pourtant, les mécanismes sous-jacents à ce paradigme restent mal compris au niveau du design. À titre d'étude de cas, nous analysons comment les applications de messagerie s'intègrent à ce paradigme en réalisant une analyse des fonctionnalités de 17 applications, complétée par 12 entretiens utilisateurs. Nous identifions quatre mécanismes d'accroissement qui expliquent comment le design contribue au paradigme de la corne d'abondance. Ces mécanismes contribuent tous à l'intensification des usages (quand les pratiques numériques deviennent plus intensives, entraînant une augmentation du volume de données), à l'expansion des infrastructures et, indirectement, à l'obsolescence et au remplacement des appareils. Notre analyse de l'intensification des usages s'appuie sur un ensemble de design patterns récurrents liés au paradigme de la corne d'abondance, que nous avons identifiés dans les applications de messagerie. Afin d'atténuer les effets de ce paradigme, nous proposons un ensemble d'heuristiques de modération pour guider les décisions de design numérique
Genomic insights into adaptation to eco-regional and cultural variables across human populations from North, Central and Southeast Asia
No full textNatural selection has been extensively studied in humans, providing many examples of how climate, diet, and pathogens can translate into local selective pressures and thus phenotypic diversity across populations. However, most studies have focused either on a global scale, with a limited number of populations per geographic area, or on a very local scale. To complement these approaches, we collected and studied a large genomic dataset at a continental scale covering North, Central, and mainland Southeast Asia, and consisting of 863 individuals from 25 culturally diverse populations. We aimed to decipher the selective pressures across the Asian continent, comparing populations from different geographic areas and having contrasted subsistence strategies, using both intra-populations (iHS) and inter-populations (Fst) statistics. We detected both local and continental-wide signals, with some geographically and culturally complex patterns emerging when being compared to the literature. Interestingly, among the regions detected as significant only in mainland Southeast Asia, we found two of them clearly pointing to selective pressures associated with immunity. The first one, detected with iHS only, contains three peptidoglycan 1 .</div
Identity observers design for a class of port-Hamiltonian systems
No full textWe study structure-preserving observer design problem for nonlinear port-Hamiltonian systems. We propose an identity observer in the original coordinates with an affine outputinjection term to preserve the explicit port-Hamiltonian representation. Within the convergent-dynamics framework, we derive explicit gain conditions that guarantee tunable exponential rates for state reconstruction. We treat practically relevant subclasses with constant Hessian or structure matrices and extend the results to the two combined cases. We then analyze the structurepreservation and passivity properties of the observer dynamics, the induced error system, and the system-observer cascade, illustrating how this analysis paves the way for observer-based controller design. The results are illustrated on two examples, including rotor-velocity reconstruction for a permanent-magnet synchronous motor from current measurements and an electricalcircuit case study
Impact of Surgical Masks on Newborns' Spontaneous Face Processing Skills
No full textInternational audienceABSTRACT At birth, newborns prefer upright faces over other stimuli, indicating that they already pay specific attention to facial internal features, with expectations about their featural configuration. The study investigated whether surgical mask wearing in maternity wards due to COVID‐19 disrupts newborns' face processing. Using a visual preference paradigm, newborns saw paired images of the same face. In Condition 1, both faces were upright, one masked, one unmasked. No preference was found. In Condition 2, newborns saw an upright masked face versus the same masked face but inverted (i.e., upside down), showing a preference for the upright version. Findings suggest that masked and unmasked static faces equally attract newborns' attention and that surgical masks do not disrupt their face configuration processing. It has implications for recommendations to the general public, especially individuals in contact with newborns