Portail HAL Ensta
Not a member yet
11080 research outputs found
Sort by
The Potential of Metaverse Technology in E-Learning: Case of Engineering Students
No full textInternational audienceMetaverse technology integrates virtual and augmented reality, has significantlyimpacted many industries, and opened up new opportunities for educators andlearners alike. This article focuses on its potential to transform e-learning, especiallyin engineering education, and highlights the importance of understandingengineering students' attitudes toward adopting new technologies. This study shedslight on the potential of e-learning in general, and the metaverse in specific, toengage and motivate students. We conducted a quantitative online survey (n=120) to collect data from engineering students. The analysis of collected data explores and evaluates the students' awareness and acceptance of the metaverse in e-learning. Our results demonstrate that engineering students have a good awareness, a positive attitude, and motivation towards using new technologies and highlight a good opportunity for the metaverse to enhance engineering students’ online interactions and participation compared with traditional e-learning methods. We have identified several challenges and opportunities in using the metaverse in e-learning, including the need for new competencies, specialized hardware and software, and data privacy and security concerns. The paper concludes with recommendations for future research, emphasizing the benefits, e-learning's potential, and challenges of the metaverse in e-learning
Propagation des ondes dans des milieux quasi-périodiques
No full textThe goal of this thesis is to develop efficient numerical methods for the solution of the time-harmonic wave equation in quasiperiodic media, in the spirit of methods previously developed for periodic media. The goal is to use as in quasiperiodic homogenization the idea that an elliptic PDE with quasiperiodic coefficients can be interpreted as the cut of a higher-dimensional PDE which is elliptically degenerate, but with periodic coefficients. The periodicity property allows to use adapted tools, but the non-elliptic aspect makes the mathematical and numerical analysis of the PDE delicate. One application concerns transmission problems between periodic half-spaces (typically photonic crystals) when (1) the interface does not cut the periodic half-spaces in a direction of periodicity, or (2) when the periodic media have noncommensurate periods along the interface.L'objectif de la thèse est de développer des méthodes numériques originales pour la résolution de l'équation des ondes en régime harmonique dans des milieux quasi-périodiques, dans l'esprit des méthodes développées précédemment pour des milieux périodiques. L'idée est d'utiliser comme dans des travaux d'homogénéisation quasi-périodique le fait que l'étude d'une EDP elliptique avec des coefficients quasi-périodiques se ramène à l'étude d'une EDP elliptiquement dégénérée en dimension supérieure, mais dont les coefficients sont périodiques. Le caractère périodique permet d'utiliser des outils adaptés, mais le caractère non-elliptique rend toutefois l'analyse mathématique et numérique de l'EDP délicate. Une des applications étudiées dans ce manuscrit concerne des problèmes de transmission entre des demi-plans périodiques (typiquement des cristaux photoniques) quand (1) l'interface ne coupe pas les demi-plans périodiques dans une direction de périodicité, ou (2) quand les milieux périodiques n'ont pas des périodes commensurables le long de l'interface
Laser powder bed fusion processing of AD730® alloy
No full textThe Additive Manufacturing (AM) ability of AD730® alloy by laser powder bed fusion (PBF-LB) is investigated for comparison with the cast & wrought AD730® alloy developed by Aubert & Duval for gas turbine applications. Several designs of experiments were needed to identify stable manufacturing conditions and build-up strategy in terms of low defect rate and microstructure soundness. In particular, the laser power, the scan speed, the layers' thickness, the hatching distance, and the scanning strategy were determined to successfully produce crack-free samples. The PBF-LB AD730® alloy exhibits columnar grains along the build direction. The resultant texture is close to a singlecrystalline-like cube texture. Upon annealing (solution annealing + single aging treatment), such a strong texture is preserved despite the formation of a network of fine transgranular secondary g' precipitates. Room temperature tensile properties are encouraging with yield stress lying between the coarse and fine grain microstructures of cast & wrought AD730®, for both orientations. An anisotropic mechanical behavior is obtained with higher properties for specimens printed along X and Y orientations than along Z. Finally, no evidence of premature failure due to micro-cracks has been detected in the overall test specimens
Representation of Irregularly Sampled Time Series with Generative Language Models for Classification and Transfer Learning: a Case Study in Activities of Daily Living Recognition
No full textInternational audienceHuman Activity Recognition plays a crucial role in various fields, including healthcare and smart homes. With the increasing prevalence of smart homes equipped with ambient sensors, there is a growing interest in leveraging artificial intelligence techniques to understand and recognize human activities within these environments. However, the irregular and noisy nature of data collected by ambient sensors presents unique challenges. To address these challenges, we propose the use of a pre-trained embedding trained on sensor activation sequences, specifically an algorithm based on an architecture similar to GPT, to improve the classification performance of activities of daily living in smart homes. Additionally, we investigate the concept of transfer learning between smart homes, utilizing the knowledge gained from one environment to enhance activity recognition in another. The results demonstrate that the GPT-like-based approach outperforms other algorithms in terms of accuracy and balanced accuracy across multiple datasets. The findings also highlight the potential of transfer learning, with GPT-like pre-trained embeddings from a clean and large dataset showing promising results in various scenarios
Relaxed-inertial proximal point algorithms for problems involving strongly quasiconvex functions
No full textInternational audienc
Propagation des ondes dans les guides partiellement enfouis : résolution du problème direct et imagerie par méthode de type échantillonnage
No full textThis work is about the non destructive testing of partially buried or immersed slendered structures such as a steel cable partially buried in concrete or a steel plate partially immersed in liquid sodium. Such structures can be seen as the junction of two closed waveguides. In order to perform computing, the open part of the structure is truncated in the transverse direction with PMLs. As a result, a partially buried waveguide can be treated as the junction of two closed waveguides, in one of which the propagation of waves is governed by an equation involving complex coefficients due to the presence of the PMLs. This observation has lead us to tackle first the simpler case of the junction of two closed acoustic waveguides. For this simple case, we have proposed a strategy to solve the inverse problems based on the one hand on the introduction of the so-called reference fields, which are the total field response of the structure without defects to an incident field coming frome both half-guides, and on the other hand on the use of the reciprocity of the Green function of the structure without defect. Following this strategy, we have obtained an efficient modal formulation of the LSM which has enabled us to retrieve defects. In this simple case, we have taken advantage of the completeness of the modes to analyze the forward and inverse problems. The loss of the completeness of the modes in the half-guide truncated in the transverse direction with PMLs has led us to study the forward problem with Kondratiev theory. The tools introduced for the junction of two closed waveguides have been adapted to solve the inverse problem. Finally, we have tackled the more complex, but more realsitic case of an elastic waveguide partially immersed in a fluid. For this difficult case, we have developped adapted computing tools adapted and extended the tools introduced before solving the inverse problem.Ce travail de thèse porte sur le contrôle non destructif de structures élancées partiellement enfouies ou immergées, par exemple un câble d'acier partiellement enfoui dans du béton ou une plaque d'acier partiellement immergée dans du sodium liquide. Ces structures peuvent être vues comme la jonction d'un guide fermé et d'un guide ouvert. Pour effectuer des calculs, nous avons tronqué transversalement la partie ouverte de la structure avec des PML finies. Un guide partiellement enfoui peut alors être traité comme la jonction de deux guides fermés, dont la propagation des ondes dans l'un des guides est régie par une équation impliquant des coefficients complexes liés à la présence des PML. Ce constat nous a amené à commencer par traiter dans un premier temps le cas plus simple de la jonction de deux guides acoustiques fermés. Pour ce cas simple, nous avons proposé une démarche de résolution du problème inverse adaptée aux jonctions de guides d'ondes fermés. Elle repose d'une part sur l'introduction des champs de référence, qui sont les réponses de la structure totale sans défaut à un mode provenant d'un des deux demi-guides, et d'autre part sur l'utilisation de la relation de réciprocité de la fonction de Green de la structure sans défaut. Suivant cette démarche, nous avons obtenu une formulation modale efficace de la LSM qui nous a permis d'identifier des défauts. Dans ce cas simple, nous avons tiré parti de la complétude des modes pour analyser les problèmes direct et inverse. Dans un second temps, nous avons traité le cas d'un guide acoustique partiellement enfoui. La perte de complétude des modes dans le demi-guide tronqué transversalement avec des PML nous a amenée à étudier le problème direct à l'aide de la théorie de Kondratiev. Les outils introduits pour la jonction de deux guides fermés ont été ensuite adaptés à la résolution du problème inverse. Dans un troisième temps, nous avons abordé le cas plus réaliste, mais plus complexe, d'un guide élastique partiellement immergé dans un fluide. Pour ce cas difficile, nous avons développé des outils de simulation adaptés et étendus les outils introduits précédemment pour résoudre le problème inverse
Using Confounded Data in Latent Model-Based Reinforcement Learning
No full textInternational audienceIn the presence of confounding, naively using off-the-shelf offline reinforcement learning (RL) algorithms leads to sub-optimal behaviour. In this work, we propose a safe method to exploit confounded offline data in model-based RL, which improves the sample-efficiency of an interactive agent that collects and learns from online, unconfounded data. First, we import ideas from the well-established framework of do-calculus to express model-based RL as a causal inference problem, thus bridging the gap between the fields of RL and causality. Then, we propose a generic method for learning a causal transition model from offline and online data, which captures and corrects the confounding effect using a hidden latent variable. We demonstrate that our method is correct and efficient, in the sense that it attains better generalization guarantees thanks to the confounded offline data (in the asymptotic case), regardless of the confounding effect (the offline expert's behaviour). We showcase our method on a series of synthetic experiments, which demonstrate that a) using confounded offline data naively degrades the sample-efficiency of an RL agent collecting and learning from online data; b) using confounded offline data correctly improves its sample-efficiency
Goal Space Abstraction in Hierarchical Reinforcement Learning via Set-Based Reachability Analysis
No full textInternational audienceOpen-ended learning benefits immensely from the use of symbolic methods for goal representation as they offer ways to structure knowledge for efficient and transferable learning. However, the existing Hierarchical Reinforcement Learning (HRL) approaches relying on symbolic reasoning are often limited as they require a manual goal representation. The challenge in autonomously discovering a symbolic goal representation is that it must preserve critical information, such as the environment dynamics. In this paper, we propose a developmental mechanism for goal discovery via an emergent representation that abstracts (i.e., groups together) sets of environment states that have similar roles in the task. We introduce a Feudal HRL algorithm that concurrently learns both the goal representation and a hierarchical policy. The algorithm uses symbolic reachability analysis for neural networks to approximate the transition relation among sets of states and to refine the goal representation. We evaluate our approach on complex navigation tasks, showing the learned representation is interpretable, transferrable and results in data efficient learning
Eco-evolutionary Dynamics of Non-episodic Neuroevolution in Large Multi-agent Environments
No full textInternational audienceNeuroevolution (NE) has recently proven a competitive alternative to learning by gradient descent in reinforcement learning tasks. However, the majority of NE methods and associated simulation environments differ crucially from biological evolution: the environment is reset to initial conditions at the end of each generation, whereas natural environments are continuously modified by their inhabitants; agents reproduce based on their ability to maximize rewards within a population, while biological organisms reproduce and die based on internal physiological variables that depend on their resource consumption; simulation environments are primarily single-agent while the biological world is inherently multi-agent and evolves alongside the population. In this work we present a method for continuously evolving adaptive agents without any environment or population reset. The environment is a large grid world with complex spatiotemporal resource generation, containing many agents that are each controlled by an evolvable recurrent neural network and locally reproduce based on their internal physiology. The entire system is implemented in JAX, allowing very fast simulation on a GPU. We show that NE can operate in an ecologically-valid non-episodic multi-agent setting, finding sustainable collective foraging strategies in the presence of a complex interplay between ecological and evolutionary dynamics