Portail HAL ONERA
Not a member yet
12842 research outputs found
Sort by
A Transfer-Learning Approach for Cross-Subjects EEG Data Classification
No full textInternational audienceEEG data classification is a difficult task, especially when it comes to classifying between different individuals or different sessions. In these cases, it is necessary to use a so-called Domain Adaptation technique, but these can be tricky to implement and interpret. We propose a methodology based on the extension of the EEGNet network, decoupling the learning of subject-independent and subject-dependent parameters and from the optimized use of the available database. The result is a robust and easy to implement methodology that we named TL-EEGNet network
Hybridation optimisation-apprentissage pour la planification d'observations par un satellite
No full textInternational audienceLes enjeux liés aux missions d'observation de la Terre depuis l'espace sont multiples (gestion des risques, cartographie, défense...). Ces missions sont réalisées par des EOS (Earth Observation Satellites), des satellites en orbite basse effectuant des prises de vue d'endroits précis sur Terre. S'agissant de systèmes coûteux et complexes, il est naturel de chercher à optimiser leur usage en respectant notamment deux contraintes opérationnelles que sont premièrement les durées de transition nécessaires à la réorientation du satellite entre deux observations successives (ces durées étant time-dependent), et deuxièmement les fenêtres temporelles de visibilité pour chacune des zones d'intérêt. Le rôle d'un système de gestion de mission est alors de sélectionner des observations parmi l'ensemble des observations candidates et d'optimiser la séquence des observations datées à réaliser par le satellite, tout en maximisant un critère de qualité fonction de la récompense associée à chaque observation candidate. FIG. 1 -Portion d'orbite d'un satellite et mailles candidates pour une observatio
Interféromètres biréfringents pour l'imagerie hyperspectrale snapshot
No full textBirefringent interferometers in a Fourier transform imaging spectrometer offers two main advantages : compactness, as the interferometer is a single component that can be easily inserted into the optical path, and stability, since birefringent units are less prone to mechanical deformation compared to an amplitude-division interferometer with a beam splitter. However, birefringent plates and wedges introduce aberrations when convergent beam propagate through its. These aberrations affect not only the spatial quality of the image but also its spectral quality, particularly the contrast of interference fringes if the image spots of the ordinary and extraordinary paths do not perfectly overlap. We have shown that, for the simplest systems, the limiting aberration is the differential astigmatism between the ordinary and extraordinary paths. We then studied several interferometer architectures where aberrations between the two paths are compensated, enabling a bigger aperture and angle acceptance of the system without compromising the quality of the hyperspectral image. Alongside this theoretical study, we designed a snapshot hyperspectral camera, consisting of a microlens array, a birefringent interferometer, and a focal plane array. By selecting a polarizing detector, i.e., one with polarizers integrated at the pixel level, we were able to further simplify the system. This resulted in a snapshot camera capable of providing hyperspectral images of approximately 64x64 pixels with nearly 50 spectral bands between 525 nm and 850 nm. We successfully assembled this camera, characterized it in the laboratory, evaluated its performance, and eventually deployed it in the field.L'utilisation d'un interféromètre biréfringent dans un spectro-imageur à transformée de Fourier a deux avantages principaux : la compacité, puisque l'interféromètre se présente comme un simple composant à insérer dans le train optique, et la stabilité, les lames biréfringentes étant moins susceptibles de déformation mécanique qu'un interféromètre à division d'amplitude avec une lame séparatrice. Toutefois, les lames biréfringentes introduisent des aberrations quand elles sont traversées par un faisceau convergent. Ces aberrations affectent évidemment la qualité spatiale de l'image, mais aussi sa qualité spectrale, et en particulier le contraste des franges d'interférences si les taches images de la voie ordinaire et de la voie extraordinaire ne se superposent pas parfaitement. Nous avons montré que, pour les systèmes les plus simples, l'aberration limitante est l'astigmatisme différentiel entre les voies ordinaire et extraordinaire. Nous avons alors étudié plusieurs architectures d'interféromètres où les aberrations entre les deux voies sont compensées, permettant ainsi d'ouvrir le système sans perdre sur la qualité de l'image hyperspectrale. Parallèlement à cette étude théorique, nous avons conçu une caméra hyperspectrale snaspshot, formée de l'association d'une matrice de mini-lentilles, d'un interféromètre biréfringent, et d'un détecteur matriciel. En choisissant un détecteur polarisant, c'est-à-dire avec des polariseurs déposés au niveau du pixel, nous avons pu simplifier encore plus le système. Nous avons ainsi obtenu une caméra snapshot fournissant des images hyperpectrales d'environ 64x64 pixels et avec près d'une cinquantaine de bandes spectrales entre 525nm et 850nm. Nous avons pu assembler cette caméra, la caractériser en laboratoire, évaluer ses performances, et enfin la mettre en oeuvre sur le terrain
Conflict Management in a Distance to Prototype-Based Evidential Neural Network
No full textInternational audienc
Long term-extension of ICMEs and SIRs catalogs with deep learning and geomagnetic indices
No full textInternational audienceSpace weather event catalogs are essential tools for characterizing the near-Earth space environment. From a scientific standpoint, these catalogs provide extensive statistical insights into the physical properties of such events. Operationally, they support forecasting scenarios by offering a basis to assess the diverse impacts these events may have on the near-Earth space environment.Interplanetary Coronal Mass Ejections (ICMEs) and Stream Interaction Regions (SIRs) are two of the most significant drivers of space weather disturbances. Traditional catalogs of these large-scale solar wind structures are primarily built using in-situ measurements from L1 monitors like WIND, ACE, and DSCOVR. However, these datasets primarily cover the period after 1995, limiting the temporal scope of current catalogs.Conversely, geomagnetic indices have recorded Earth’s geomagnetic activity for several decades before the advent of the space era. These indices have been shown to respond differently to ICMEs and SIRs (e.g., Benacquista et al., 2017; Bernoux and Maget, 2020), making them a valuable resource for identifying these events in earlier periods.In this study, we adapt an existing deep learning-based method—originally developed for detecting ICMEs and SIRs using L1 solar wind data—to analyze geomagnetic index measurements. While the geomagnetic-based approach is inherently less precise than its solar wind counterpart, it successfully identifies time intervals likely associated with ICMEs or SIRsThis method is used to extend existing ICME and SIR catalogs back in time to cover the period from 1870 to 1995. Although the resulting extension is not exhaustive, it captures the most geoeffective events, offering a valuable dataset for long-term climatological studies of space weather. This work lays the groundwork for future research aimed at understanding historical space weather trends and their implications for Earth's near-space environment.This work was supported by both the FARBES (Forecast of Actionable Radiation Belts Scenarios) project, funded by the European Union's Horizon Europe research and innovation programme under grant agreement No 101081772 and ONERA internal fundings, through the federated research project PRF-FIRSTS
Measurement of bipolar charge distribution of lunar dust simulant under VUV irradiation
No full textInternational audienceUpcoming missions to the Moon represent new science opportunities and challenges. The electrostatic nature of the regolith combined with the solar wind makes it loft and adhere to almost any surface, which represents a threat for future manned and robotic missions. Understanding the charge state of the lunar soil under a representative environment is a key step towards ensuring safe lunar missions. While the global first order effect of exposure to the Sun's UV is to charge the soil positively, past experiments suggested that the transported dusts could be charged negatively. This counter-intuitive behavior was then supported by modeling, which explained the existence of negative charges but also predicted that of positively charged ones. To investigate the charging behavior of dust under a representative environment, we developed an experimental protocol based on a polarized sensitive sensor dedicated to the charge measurement of single dust grains with an accuracy of about 1 fC. The first set of measurements obtained with JSC-1A lunar dust simulants in high vacuum reveals the bipolar nature of lunar dust net charge in the regolith when exposed to UVs. Indeed, both positive and negative dusts were detected, supporting the complexity of the regolith charging processes suggested by the models
Surface à haute impédance à composite solide-liquide pour une métasurface fonctionnant en angle rasant
No full textNational audienc
A formal framework for the specification and verification of robotic skills composition for autonomous behaviors
No full textInternational audienceWith the goal to create autonomous systems that can accomplish complex tasks in dynamic environments, high-level software architectures are created using formal methods to describe the system with a set of elementary actions, named skills, necessary to accomplish their mission.However, these formalisms often lack a proper framework for designing autonomous missions as a composition of the specified skills, often relying on non-formal or semi-formal tools, such as behavior trees or UML/SysML-based tools, as well as being limited to specific elementary skill implementations. This makes the system less reliable and more prone to failures when placed in a dynamic, unknown environment. To fill this gap, we propose a formal framework to specify formal models for elementary skills and make them composable, called Skill Petri nets, and a set of composition patterns and their Petri net models. A tool will allow users to define these compositions using an extension of the Robot-Language DSL, from which the formal models and a controller code are generated. An industrial inspection mission use-case allowed to test the proposed framework.</p
Delay margin analysis of uncertain linear control systems using probabilistic mu
No full textInternational audienceMonte Carlo simulations have long been a widely used method in the industry for control system validation. They provide an accurate probability measure for sufficiently frequent phenomena but are often time‐consuming and may fail to detect very rare events. Conversely, deterministic techniques such as or IQC‐based analysis allow fast calculation of worst‐case stability margins and performance levels, but in the absence of a probabilistic framework, a control system may be invalidated on the basis of extremely rare events. Probabilistic ‐analysis has therefore been studied since the 1990s to bridge this analysis gap by focusing on rare but nonetheless possible situations that may threaten system integrity. The solution adopted in this paper implements a branch‐and‐bound algorithm to explore the whole uncertainty domain by dividing it into smaller and smaller subsets. At each step, sufficient conditions involving upper bound computations are used to check whether a given requirement–related to the delay margin in the present case–is satisfied or violated on the whole considered subset. Guaranteed bounds on the exact probability of delay margin satisfaction or violation are then obtained, based on the probability distributions of the uncertain parameters. The difficulty here arises from the exponential term classically used to represent a delay , which cannot be directly translated into the Linear Fractional Representation (LFR) framework imposed by ‐analysis. Two different approaches are proposed and compared in this paper to replace the set of delays . First, an equivalent representation using a rational function with unit gain and phase variations that exactly cover those of the original delays, resulting in an LFR with frequency‐dependent uncertainty bounds. Then, Padé approximations, whose order is chosen to handle the trade‐off between conservatism and complexity. A constructive way to derive minimal LFR from Padé approximations of any order is also provided as an additional contribution. The whole method is first assessed on a simple benchmark, and its applicability to realistic problems with a larger number of states and uncertainties is then demonstrated
