Archive ouverte HAL-LAAS
Not a member yet
12189 research outputs found
Sort by
Bose-Einstein condensate source on a optical grating-based atom chip for quantumsensor applications
No full textInternational audienceWe report the preparation of Bose-Einstein condensates (BECs) by integrating laser cooling with a grating magneto-optical trap (GMOT) and forced evaporation in a magnetic trap on a single chip. This new approach allowed us to produce a atom Bose-Einstein condensate of rubidium-87 atoms with a single laser cooling beam. Our results represent a significant advance in the robustness and reliability of cold atom-based inertial sensors, especially for applications in demanding field environments
Commande prédictive augmentée par diffusion pour l'évitement de collision dans un environnement robotique encombré
No full textNational audienceRobots that share workspaces with humans must be able to react safely to unforeseen changes in their environment, without sacrificing speed, precision, or compliance. While data-driven policies have greatly improved dexterity and generalization, they lack formal safety guarantees. In contrast, classical optimal control provides such guarantees, but often struggles to handle cluttered, dynamic environments in real time. This thesis aims to bridge this gap by combining the strengths of both paradigmsWe first integrate an hard distance-based constraint into a 100 Hz model predictive control loop for a torque-controlled 7 degrees-of-freedom manipulator. Using closed-form gradients, the controller guarantees that the robot remains outside a user-defined safety margin at every control step.However, relying solely on distance constraints becomes overly conservative in tight or cluttered spaces. To address this, we introduce a velocity damper constraint that regulates the robot's relative approach speed with respect to nearby obstacles. This constraint allows the robot to safely approach closer to obstacles and reach deeper into confined areas, such as boxes or shelves, without compromising safety. To reach real-time efficiency, we derive analytical gradients for ellipsoidal obstacle models.However, the real-time controllers we employ remain sensitive to local minima, particularly in complex and cluttered environments. To address this limitation, we propose a hybrid approach that combines learning-based trajectory generation with an optimization phase based on model predictive control. More specifically, we condition a diffusion model on object-centric representations extracted using a Slot Attention mechanism. This model is trained on densely populated synthetic scenes and is used to sample dynamically feasible, collision-free trajectories. The resulting trajectories serve as warm-starts for the optimal control solver, thereby improving convergence towards feasible solutions while preserving formal safety guarantees.The thesis delivers three main contributions: (i) a provably safe MPC architecture with closed-form derivatives, (ii) a velocity damper formulation capable of handling moving and cluttered environments, and (iii) a diffusion-augmented control pipeline that improves planning success in real-world scenes. All contributions have been released as open source and were validated on real hardware, such as the Franka Emika Panda and the KUKA LBR iiwa.Les robots qui partagent leur espace de travail avec des humains doivent être capables de réagir de manière sûre à des changements imprévus de leur environnement, sans compromettre leur vitesse, leur précision ou leur conformité. Si les approches fondées sur l’apprentissage automatique ont considérablement amélioré leur dextérité et leur capacité de généralisation, elles ne fournissent cependant aucune garantie formelle de sécurité. À l’inverse, les méthodes classiques de contrôle optimal offrent de telles garanties, mais peinent à gérer des environnements encombrés et dynamiques en temps réel. Cette thèse vise à combler cette lacune en combinant les avantages des deux paradigmes.Nous intégrons d’abord une contrainte stricte de distance dans une boucle de commande prédictive tournant à 100 Hz, appliquée à un bras manipulateur à 7 degrés de liberté contrôlé en couple. Grâce à des gradients analytiques, le contrôleur garantit que le robot reste en dehors d’une marge de sécurité spécifiée à chaque pas de contrôle.Cependant, l’utilisation exclusive d’une contrainte de distance se révèle trop conservatrice dans des environnements exigus ou fortement encombrés. Pour surmonter cette limitation, nous introduisons une contrainte d’amortissement de la vitesse relative entre le robot et les obstacles à proximité. Cette formulation permet au robot de s’approcher plus près des obstacles tout en maintenant une marge de sécurité, et d’atteindre des zones confinées, telles que l’intérieur d’un bac ou d’une étagère. Afin de garantir une exécution temps réel, nous dérivons les gradients analytiques de cette contrainte pour des obstacles modélisés sous forme ellipsoïdale.Toutefois, les contrôleurs temps réel que nous utilisons restent sensibles aux minima locaux, en particulier dans des environnements complexes et encombrés. Pour remédier à cette limitation, nous proposons une approche hybride qui combine la génération de trajectoires par apprentissage avec une phase d’optimisation par contrôle optimal. Plus précisément, nous conditionnons un modèle de diffusion sur des représentations d’objets extraites à l’aide d’un mécanisme de type Slot Attention. Ce modèle est entraîné à partir de scènes synthétiques densément peuplées, et permet d’échantillonner des trajectoires dynamiquement réalisables et exemptes de collisions. Les trajectoires générées servent alors d’initialisation (warm start) au solveur de contrôle optimal, facilitant sa convergence vers des solutions faisables tout en préservant les garanties de sécurité formelles.Cette thèse apporte trois contributions principales : (i) une architecture d'évitement de collision utilisant commande prédictive formellement sûre, (ii) une formulation de l'évitement de collision prenant en compte l’amortissement de la vitesse d'approche du robot aux obstacles, capable de gérer des environnements dynamiques, et (iii) une chaîne de traitement de contrôle augmentée par un modèle de diffusion, qui améliore significativement les performances du contrôleur dans des scènes encombrées. L’ensemble des contributions a été publié en open source et validé sur des robots réels tels que le Franka Emika Panda et le KUKA LBR iiwa
Stress-coupled spin state switching in a spin crossover composite modulates current in an organic semiconductor
No full textInternational audienceThe combination of spin-crossover (SCO) complexes with electrically conducting materials offers a promising route for developing stimuli-responsive electronics, yet the mechanism of charge transport modulation remains unexplored. Here, we investigate a bilayer heterostructure comprising silica-coated SCO nanoparticles [Fe(Htrz) 2 (trz)](BF 4 )@SiO 2 within a polyvinylpyrrolidone (PVP) matrix and organic semiconductors (OSCs), where mechanical stress generated by spin-state switching within the PVP:SCO layer modulates the conductance within the OSC layer. Through in situ piezo-resistivity characterization, we reveal a reversible conductance modulation in the OSC layer under hydrostatic pressure, providing a quantitative evaluation of pressure-induced stress sensitivity with the OSC layer. Crucially, the intrinsic properties of the SCO nanoparticles dictate key characteristics of the switching device such as the spin transition temperature and hysteresis width, enabling tunable and non-volatile memory behavior. Demonstrating robust switching over multiple thermal cycles-rooted in the intrinsic thermal stability of the SCO and validated by X-ray diffraction/optical spectroscopy analysis at elevated temperatures-this work lays the groundwork for a new class of stress-coupled spin-electronic systems, offering a potential route for the development of piezo-resistive sensors and adaptive memory devices
Towards a sustainable future: Advances and roadmap of activated carbon as a cost-effective bio-nanomaterial for electrochemical devices
No full textInternational audienc
Semidefinite hierarchies for diagonal unitary invariant bipartite quantum states
No full textWe investigate questions about the cone SEP of separable bipartite states, consisting of the Hermitian matrices acting on C^n ⊗ C^n that can be written as conic combinations of rank one matrices of the form xx * ⊗ yy * with x, y ∈ C^n . Bipartite states that are not separable are said to be entangled. Detecting quantum entanglement is a fundamental task in quantum information and a hard computational problem. We explore the Doherty-Parrilo-Spedaglieri (DPS) hierarchy of semidefinite conic approximations for SEP when the bipartite states have some additional structural properties: first, (i) for states with diagonal unitary invariance, and second (ii) for states with Bose symmetry. In case (i) we show that the DPS hierarchy can be block diagonalized, which, combining with its moment reformulation, leads to a substantially more efficient implementation. In case (ii), we give a characterization of the dual hierarchy, in terms of sums of squares of Hermitian complex polynomials, extending a known result in the generic case. It turns out that the completely positive cone CP, its dual cone COP, and their sums-of-squares based conic approximations K^(t) , play a central role in these two settings (i),(ii). We clarify these connections and test the block diagonal relaxations on classes of examples
Velocity Potential Field Modulation for Dense Coordination of Polytopic Swarms and Its Application to Assistive Robotic Furniture
No full textInternational audienceWe explore the use of a mobile furniture swarm that are intended to assist users with limited mobility in their daily indoor activities. We focus on the multi-robot coordination problem when a dense target pose configuration is required, such as in an apartment setting. In those cases, the convergence of one robot to the target can be significantly affected by neighboring robots with specific shapes. In this letter, we propose a solution, named Velocity Potential Field Modulation (VPFM), to deal with the dense coordination problem of a polytopic swarm in a decentralized manner. We adapt our method to assistive applications, such as room reconfigurations and facilitating indoor movement of wheelchair users. We evaluate the performance of our method in simulations and on real-world mobile furniture hardware, demonstrating its effectiveness and real-time performance
ETSI SmartM2M; oneM2M deployment guidelines and good practices
No full textThe full document is available onhttps://www.etsi.org/deliver/etsi_tr/103800_103899/103843/01.01.01_60/tr_103843v010101p.pdfThe present document describes how to complement the ETSI view by liaising with oneM2M technical bodies for consultation during the definition of the deployment scenarios. From the stakeholder point of view, it can help on the adoption of oneM2M through indicators such as performance evaluation to assess the performance of their proposed products. IoT platforms customers can define their specific deployment scenarios and evaluate the performance of a given IoT platform against their scenarios. Finally, the Open-Source Communities around oneM2M may benefit from these results by taking the output of the development of the TTF PoC into their roadmaps
Robotisation AFM pour le phénotypage mécanique des cellules
No full textInternational audienceL’étude des propriétés mécaniques des populations de cellules notamment de mammifères par AFM présente un enjeu majeur pour l’identification de sous phénotypes cellulaires d’origine mécanique [1,2]. Pour atteindre un nombre de cellules suffisant pour réaliser ce type d’étude, l’utilisation d’un AFM « manuel » ne suffit plus et il est nécessaire d’explorer les modalités d’automatisations de celui-ci [3]. Nous présenterons l’intégration d’algorithme de machine learning au service de la robotisation des me-sures AFM sur des centaines de cellules de mammifère.Références [1]Severac, C.; Proa-Coronado, S.; Formosa-Dague, C.; Martinez-Rivas, A.; Dague, E. Automation of Bio-Atomic Force Microscope Measurements on Hundreds of C. Albicans Cells. JoVE (Journal of Visualized Experiments) 2021, No. 170, e61315. https://doi.org/10.3791/61315.[2]Thomas - - Chemin, O.; Séverac, C.; Moumen, A.; Martinez-Rivas, A.; Vieu, C.; Le Lann, M.-V.; Trevisiol, E.; Dague, E. Automated Bio-AFM Generation of Large Mechanome Data Set and Their Analysis by Machine Learning to Classify Cancerous Cell Lines. ACS Appl. Mater. Interfaces 2024, 16 (34), 44504–44517. https://doi.org/10.1021/acsami.4c09218.[3]Thomas- -Chemin, O.; Janel, S.; Boumehdi, Z.; Séverac, C.; Trevisiol, E.; Dague, E.; Duprés, V. Advancing High-Throughput Cellular Atomic Force Microscopy with Automation and Artificial Intelligence. ACS Nano 2025. https://doi.org/10.1021/acsnano.4c07729
HyPlan: Hybrid Learning-Assisted Planning Under Uncertainty for Safe Autonomous Driving
No full textWe present a novel hybrid learning-assisted planning method, named HyPlan, for solving the collision-free navigation problem for self-driving cars in partially observable traffic environments. HyPlan combines methods for multi-agent behavior prediction, deep reinforcement learning with proximal policy optimization and approximated online POMDP planning with heuristic confidence-based vertical pruning to reduce its execution time without compromising safety of driving. Our experimental performance analysis on the CARLA-CTS2 benchmark of critical traffic scenarios with pedestrians revealed that HyPlan may navigate safer than selected relevant baselines and perform significantly faster than considered alternative online POMDP planners
Nonlinear modeling of AlN/GaN HEMT accounting for self-biasing effect during RF step stress: analysis and hard-SOA
No full textplateforme PROOF LAASInternational audienceIn this study, we investigate the non-linear (NL) behavior of AlN/GaN HEMT technologies under gain compression when submitted to 10 GHz single-tone RF-step stress, which is crucial for millimetre-wave power application robustness. We evaluate AlN/GaN transistors, targeting high-power amplifiers with operating frequency above 30 GHz. We present here an original method that includes, in a unique NL expression, the varying self-biasing effect caused by RF step-stress sequences. This methodology can be used as a tool for comparative analysis of technological variants and various transistor geometries post RF stress. The step-stresses are conducted on HEMT in saturated mode and in diode operation alone, to assess the electrical origins of defects and the critical Safe Operating Area (SOA) of these devices. We identify the mechanism of failure as stemming from the degradation of the Schottky gate when subjected to critical RF power levels, due to its constrained capacity to handle power signals exceeding 18 dBm. Furthermore, we highlight the remarkable RF robustness of this technology, achieving gain compression of around 10 dB without degradation