Archive ouverte HAL-LAAS
Not a member yet
12189 research outputs found
Sort by
40 ans de développements en lithographie électronique : mémoires d’un dinosaure
No full textInternational audiencePersonal experience presentation over 40 years on Electron Beam Lithography (EBL) through its development history which led to the purchase of a 100keV EBL system at CNRS-LAA
Averaged Tracking Controllability of Parameter-Dependent Linear Systems and Volterra Integral Equations
No full textIn this paper, we explore linear parameter-dependent control systems, analyzing conditions under which an output variable, averaged over the parameters, can effectively track a prescribed trajectory within a finite time horizon. We characterize average tracking controllability through criteria that may be subtle to verify due to their infinitedimensional nature. However, in certain special cases, such as the single-input and single-output scenarios, we have applied an approach that we believe is novel, coming from the Volterra theory of integral equations, to derive criteria that lend themselves to straightforward verification. More interestingly, based on the system's architecture, this approach offers us, almost free of charge, the type of trajectories that can be reached exactly or approximately
Soliton self-injection locking to a fiber Fabry-Perot resonator with sub-100 mW pump power
No full textInternational audienceWe report the observation of soliton self-injection locking (SIL) to fiber Fabry-Perot (FFP) resonator with a laser pump power of less than 100 mW. The locking process is studied analytically with a model of a Fabry-Perot laser interacting with an external FFP and the modelling results are compared to the experimental ones. The model is then extended to the case of a nonlinear FFP modifying, due to the tilt of the resonance with self-phase modulation, the frequency response of the laser. Experimentally, we obtain a large locking range with direct access to primary combs, chaotic combs and soliton while keeping a stable laser lock
Retour en force en commande prédictive
No full textModel-predictive control is an appealing framework to control robots due to its ability to reason over the future and to update decisions online based on sensory information. Its effectiveness in generating automatically complex motions is now well recognized by the community. Yet, it remains limited in tasks involving controlled physical interaction with the environment. Indeed, predicting the evolution of the contact interaction supposes the knowledge of dynamic models that are in practice either inaccessible or unsuited for optimization purposes. An outstanding question remains to make MPC really useful in complex interaction tasks: how to include force feedback information in model-predictive controllers? In this thesis, we propose the first complete answer to this question. A novel paradigm that enables to include systematically measured efforts into the optimal control loop is presented. Our contributions are three-fold. First, we propose the first hardware demonstrations of closed-loop nonlinear MPC running at high frequencies on torque-controlled robots. Second, we expose the inherent inability of the classical MPC paradigm to achieve force feedback control, and propose new methodologies based on state model augmentation and online estimation to overcome this limitation. Third, we also propose fundamental contributions in numerical optimal control with hard constraints and experimental contributions in force estimation. All our contributions are supported by proofs, simulations and hardware experiments.Le contrôle prédictif (Model Predictive Control, MPC) constitue un cadre méthodologique particulièrement prometteur pour le contrôle des robots, grâce à sa capacité à anticiper l’évolution du système et à adapter les décisions en temps réel à partir des retours sensoriels. Son efficacité dans la génération automatique de mouvements complexes est aujourd’hui largement reconnue au sein de la communauté scientifique. Toutefois, son application reste limitée dans les tâches nécessitant une interaction physique contrôlée avec l’environnement. En effet, la prédiction de l’évolution des contacts suppose la connaissance de modèles dynamiques précis, qui s’avèrent en pratique souvent inaccessibles ou mal adaptés aux exigences de l’optimisation en ligne. Une question centrale demeure donc pour rendre le MPC pleinement opérationnel dans les tâches d’interaction complexe : comment intégrer de manière rigoureuse l’information issue du retour d’effort dans les contrôleurs prédictifs ? Cette thèse propose une réponse complète à cette problématique. Nous introduisons un nouveau paradigme permettant l’intégration systématique des efforts mesurés au sein de la boucle de contrôle optimal. Nos contributions se déclinent en trois axes principaux. Premièrement, nous présentons les premières démonstrations expérimentales d’un MPC non linéaire en boucle fermée fonctionnant à haute fréquence sur des robots commandés en couple. Deuxièmement, nous mettons en évidence les limitations structurelles du cadre MPC classique vis-à-vis du contrôle par retour d’effort, et proposons des méthodologies innovantes reposant sur l’augmentation du modèle d’état et l’estimation en ligne pour dépasser ces limitations. Troisièmement, nous apportons des contributions fondamentales en contrôle optimal numérique sous contraintes strictes, ainsi que des avancées expérimentales en estimation d’efforts. L’ensemble de nos travaux est validé par des preuves théoriques, des simulations et des expérimentations sur plateformes robotiques réelles
Etalonnage elasto-géométrique et dynamique de robots anthropomorphes.
No full textNational audienceThis thesis focuses on the development of geometric calibration and dynamic identification methods for anthropomorphic robotic systems, with a particular emphasis on mobile manipulators and humanoid robots. The work addresses key challenges in improving the operational accuracy and performance of these complex robotic systems.The thesis begins by highlighting the importance of accurate geometric and dynamic models for anthropomorphic robots, which are crucial for precise control and manipulation tasks. It then outlines the unique challenges posed by these systems, including their complex kinematic structures, floating bases, and the need to maintain balance during calibration procedures.A comprehensive literature review is presented, covering existing approaches to robot calibration and identification. This review reveals gaps in current methodologies, particularly for whole-body calibration of humanoid robots and the treatment of non-geometric effects in mobile manipulators.The thesis then presents two main case studies:1. Mobile Manipulator TIAGo: The work on TIAGo introduces a novel, comprehensive approach to improving its operational accuracy. Key contributions include:- A revisited methodology for geometric calibration using both external measurement systems and embedded sensors.- Detailed investigation and modeling of the suspension system in the mobile base, addressing an often-overlooked aspect of mobile robot accuracy.- Development of an advanced backlash model that accounts for gear clearance, shaft flexibility, and encoder misalignment.- Integration of these models into a complementary framework, resulting in significant improvements in end-effector positioning accuracy (up to 57% reduction in RMSE).2. Humanoid Robot TALOS: For TALOS, the thesis presents a novel approach to whole-body geometric calibration without relying on external sensors. Key aspects include:- Development of a plane-constrained calibration method using 3-point contacts between the robot's end-effector and a flat surface.- Introduction of the Information Ranking algorithm for selecting Optimal Calibration postures (IROC), which efficiently determines the minimal set of calibration postures needed.- Experimental validation showing significant improvements in positioning accuracy after calibration.A major contribution of the thesis is the development of FIGAROH (Free Identification of Geometrical and dynAmic parameters of RObots and Humans), an open-source Python toolbox. FIGAROH provides a unified framework for geometric calibration and dynamic identification, incorporating state-of-the-art methods and making them accessible to non-experts. Key features of FIGAROH include:- Automatic generation of optimal exciting postures and motions for calibration and identification.- Implementation of various least-squares methods for parameter estimation.- Comprehensive model validation tools.- Support for a wide range of robotic systems, including serial manipulators, mobile manipulators, humanoids, and even human motion models.The thesis demonstrates the effectiveness of these methods through extensive experimental validation on both TIAGo and TALOS robots, as well as other robotic platforms. Results show significant improvements in positioning accuracy and dynamic model fidelity across various scenarios.The work concludes by discussing the implications of these advancements for the field of robotics, particularly in enhancing the performance of anthropomorphic robots in real-world applications. It also outlines future research directions, including the potential for online calibration techniques and the integration of geometric and dynamic parameter identification.Cette thèse porte sur le développement de méthodes de calibration géométrique et d'identification dynamique spécifiques pour les systèmes robotiques anthropomorphes. L'objectif principal est d'améliorer la précision et les performances opérationnelles de ces systèmes complexes.La thèse commence par souligner l'importance d’utiliser des modèles géométriques et dynamiques précis pour le contrôle et la manipulation des robots anthropomorphes. Elle aborde ensuite les défis spécifiques liés aux structures cinématiques complexes de ces robots à leur base flottante.Une revue de la littérature révèle les lacunes des approches actuelles, notamment en ce qui concerne l’étalonage corps complet de manière efficace des robots humanoïdes et le traitement des effets non géométriques dans les manipulateurs mobiles. Deux études de cas sont présentées :1. Manipulateur Mobile TIAGo :- Méthodologie de calibration géométrique revisitée, utilisant des systèmes de mesure externes et des capteurs embarqués.- Modélisation détaillée du système de suspension de la base mobile.- Développement d'un modèle avancé de jeu mécanique, intégrant les jeux articulaires, la flexibilité de l'arbre moteur et les erreurs d'alignement des codeurs.- Intégration de ces modèles dans un nouveau modèle, permettant une amélioration significative de la précision du positionnement de l'effecteur (jusqu'à 57% de réduction de la RMSE).2. Robot Humanoïde TALOS :- Approche innovante de calibration géométrique corps complet sans capteurs externes, utilisant un contact à trois points entre l'effecteur et une surface plane.- Algorithme de sélection des postures optimales de calibration.- Validation expérimentale montrant des améliorations significatives après calibration.Un apport majeur de la thèse est FIGAROH, un outil open-source en Python, offrant un cadre unifié pour la calibration géométrique et l'identification dynamique, accessible même aux non-experts. FIGAROH propose :- La génération automatique de postures et mouvements optimaux pour la calibration.- La mise en œuvre de diverses méthodes de moindres carrés pour l'estimation des paramètres.- Des outils complets de validation de modèle.- Un support pour une large gamme de systèmes robotiques.Les méthodes développées ont été validées expérimentalement sur les robots TIAGo et TALOS, et montreent des améliorations notables en précision des modèles dynamiques.La thèse conclut en discutant les implications de ces avancées pour le domaine de la robotique, notamment pour améliorer les performances des robots anthropomorphes, et propose des axes de recherche futurs
Introduction of the book "Hybrid and Networked Dynamical Systems"
No full textInternational audienceThis chapter serves as introduction to the book and to the contributions therein. We provide some relevant background on hybrid and networked dynamical systems and summarize the contribution of each chapter, while illustrating their synergies and connecting themes. The broad spectrum of the contributed chapters is organized into three main thematic areas: Networked Systems; Hybrid Techniques; and Emerging Trends and Approaches for Analysis and Design
Density functional theory calculations of surface thermochemistry in Al/CuO thermite reaction
No full textInternational audienceThis paper investigates the thermochemistry of the heterogeneous Al/CuO thermite reaction through density functional theory calculations. We examine the interactions of atomic Al, Cu, O, as well as O2, AlO, Al2O, AlO2, Al2O2 molecular species, with Al(111), Cu(111), and Al2O3 (γ and amorphous) surfaces, all of which being condensed phase products during the thermite reaction. Al(111) exhibits a very high reactivity, characterized by adsorption energies ranging from 3 to 5.3 eV for atomic Al, Cu, O, and from 4 to 9.5 eV for all molecular species. This reactivity is attributed to barrierless molecular decomposition, followed by the spatial spreading of adsorbate species across the surface facilitated by hot adatom migration processes. The Al 2 O 3 surface also exhibits extremely high reactivity, with adsorption energies of 4.5 and 9.4 eV for atomic Cu and Al, respectively. Additionally, absorption energies range from 7 to 15 eV for condensation of AlxOy suboxides. Al-rich suboxides, namely Al 2 O and Al 2 O 2 , show the greatest adsorption energy with -15.05 eV for Al 2 O, against 6.52 eV for AlO 2 . In contrast, O and O 2 exhibit no reactivity on Al 2 O 3 surfaces exhibiting oxidation states being superior or equal to Al III . Finally, Cu(111) surface exhibits much lower reactivity compared to Al(111) and Al 2 O 3 , with absorption energies ranging from 2 to 3.5 eV for Al, O and Cu atoms. Although energetic, molecular AlxOy suboxides show non dissociative adsorption on Cu(111). This findings point to different modes of oxide nucleation on these surfaces, pleading for planar nucleation and growth onto Al(111), while being more difficult and localised onto Cu(111). They renew our understanding of the thermite reaction chemistry, quantitatively differentiating the various type of heterogeneous reactions and their implication on the overall reaction. They also provide valuable data for higher-level diphasic simulations of the computational fluid dynamics, aiming to achieve predictive capability.</div
Experimental UAV Flights to Collect Data Within Cumulus Clouds
No full textInternational audienceThis article presents the deployment of micro UAVs to study the evolution of cumulus clouds. The dynamic nature of the environment, the difficult weather conditions, the long distances, and the limited flight performances of micro UAV systems make such missions quite challenging. After describing the missions constraints and objectives, the system's main component is depicted: it is the definition of adaptive flight patterns that allow the UAVs to track the areas of interest in the clouds autonomously, using real-time sensor readings. The system architecture is then presented, considering the information feedback and decision process made by the operators and scientists on the ground and the necessary flight autonomy of the UAVs. The complete system has been deployed during an international meteorological campaign on Barbados island, and extended with extra test flights afterward. Details of the operational organization and the achieved flights are reported. The lessons learned during the field campaign revealed the strength and weaknesses of the proposed system and possible improvements are discussed. The collected data has contributed to a better understanding of cloud evolution, demonstrating that a tight coupling of the sensors and the flight control system is a crucial point for extending the performances of UAV systems in atmospheric science
NetGlyph: Representation Learning to generate Network Traffic with Transformers
No full textInternational audienceNetwork security has been a significant concern in recent years. Due to the rising number and complexity of cyberattacks, Machine Learning (ML) models have been proposed to enhance intrusion detection. However, training these models requires extensive data, which is challenging to collect. To tackle this issue, previous work has focused on the generation of synthetic network traffic data using generative neural networks. Considering network traffic as a sequence of packets that contains continuous, discrete, and binary features, we propose a novel approach to learn a discrete representation of network traffic using Vector-Quantized Variational Autoencoders (VQ-VAE). In this paper, we adapt this model to learn how to represent network flows as a sequence of discrete tokens, called NetGlyphs. We evaluate the model on a dataset of Command & Control flows and compare performances to another model that uses a continuous representation. We show that our model is able to, reconstruct the data accurately and better preserve the original distribution. We also find promising results on network traffic generation using a state-of-the-art Transformer model to generate new NetGlyphs sequences that can be decoded back into real network traffic.</div