Portail HAL ONERA
Not a member yet
12842 research outputs found
Sort by
Hybridizing Machine Learning and Optimization for Planning Satellite Observations
No full textInternational audiencePlanning the activities of an Earth observation satellite is a highly combinatorial task. It consists in regularly computing the sequence of observations to be performed by a satellite to collect images of candidate points of interest (POIs), while taking into account the time-dependent maneuvers required to point the satellite to the successive POIs. To solve such a recurrent optimization problem, we propose a novel approach that exploits offline learning techniques to approximate scheduling feasibility for sets of observation tasks. For this, we build a 0/1 neural network classifier whose inputs are related to the geographical positions of the POIs to be observed over the satellite orbit. We also learn hard capacity constraints limiting the number of observable POIs within orbit portions of various sizes. Finally, we introduce a hybrid algorithm, called HySSEO, optimizing the observation schedules based on a two-step process. The latter first searches for an optimal selection of POIs given the learned constraints, and then exploits this selection to bootstrap the search for an optimal schedule satisfying detailed time-dependent transition constraints. This hybrid optimization approach significantly improves the solution quality when compared to a standard scheduling approach
How Big is the Automaton? Certified Lower Bounds on the Size of Presburger DFAs
No full textInternational audienceLower bounds provide essential insights into the minimal computational resources required for algorithm execution. This paper focuses on logical theories, a domain where estimating resources is particularly difficult, and provides a novel, fully-automated method for computing lower bounds on memory usage, serving as a proxy for the computational resources required to perform logical reasoning. Specifically, the paper focuses on computing lower bounds on the size of the minimal deterministic finite automaton that encodes the solution set of a given Presburger arithmetic (also known as linear integer arithmetic) formula. The lower bounds are accompanied by independently verifiable certificates which also support a union-like operation that can be used to increase the computed bounds. We conducted an extensive empirical evaluation of our method using over 5 000 formulae from the quantifier-free fragment of Presburger arithmetic, sourced from the SMT-LIB repository. The results show that our method often produces lower bounds that are close to the actual size of the minimal deterministic finite automaton. Moreover, it succeeds in computing non-trivial bounds even for instances that are out of reach (by several orders of magnitude) for the existing state-of-the-art automata-based tools for solving Presburger arithmetic
A survey on multi-fidelity surrogates for simulators with functional outputs: unified framework and benchmark
No full textInternational audienceMulti-fidelity surrogate models combining dimensionality reduction and an intermediate surrogate in the reduced space allow a cost-effective emulation of simulators with functional outputs. The surrogate is an input–output mapping learned from a limited number of simulator evaluations. This computational efficiency makes surrogates commonly used for many-query tasks. Diverse methods for building them have been proposed in the literature, but they have only been partially compared. This paper introduces a unified framework encompassing the different surrogate families, followed by a methodological comparison and the exposition of practical considerations. More than a dozen existing multi-fidelity surrogates have been implemented under the unified framework and evaluated on a set of benchmark problems. Based on the results, guidelines and recommendations are proposed regarding multi-fidelity surrogates with functional outputs. Our study shows that most multi-fidelity surrogates outperform their tested single-fidelity counterparts under the considered settings. However, no particular surrogate is performing better on every test case. Therefore, the selection of a surrogate should consider the specific properties of the emulated functions, in particular the correlation between the low- and high-fidelity simulators, the size of the training set, and the local nonlinear variations in the residual fields
Action planning modulates perceptual confidence through action monitoring processes
No full textInternational audienceAbstract Dominant models of metacognition argue that the quality of sensory information determines perceptual confidence. However, recent accounts suggest that also motor signals contribute to confidence judgments. In line with this assumption, we conducted three pre-registered experiments to investigate the role of the motor preparation of perceptual decisions on confidence. Participants reported the orientation of a Gabor and indicated the level of confidence in their response. A visual cue, displayed before the Gabor, induced the planning of an action that could be congruent/incongruent with the response side and compatible/incompatible with the effector subsequently used to report the Gabor’s orientation. In the three experiments, we observed that confidence consistently increased when participants prepared spatially incongruent actions compared to congruent actions, irrespectively of the effector primed and independently of the correctness of their responses. In the third experiment, electroencephalography recordings (EEG) showed increased P2 amplitude for incongruent compared to congruent actions, suggesting that the planning of incongruent actions led to a larger involvement of early attentional resources required for response inhibition which in turn impacted post-decisional markers of confidence (Error Positivity). Taken together, these findings suggest that motor information might trigger action monitoring mechanisms susceptible to alter confidence in our decisions, implying that motor processes are not only the output, but also an input of the decision process. Significance Statement While virtually every decision we make leads to an action, the role of motor processes in decision making has been largely neglected. Our results show that retrospective confidence in a perceptual discrimination task is boosted when the motor execution is spatially incongruent with motor preparation, independently of the correctness of the response. Electroencephalography recordings indicate that this effect could be explained by a larger involvement of early attentional resources related to action monitoring, which has an impact on confidence computations. Taken together, these results suggest that motor processes might trigger action monitoring mechanisms susceptible to alter retrospective confidence in our decisions, implying that motor processes are not only the output, but also an input of the decision mechanisms
Synergy of multispectral and hyperspectral data for quantifying industrial aerosol emissions
No full textInternational audienc
Toward an Intelligible Vibrotactile Stimulation: A Gestalt-Based Perspective
No full textInternational audienceVibrotactile interfaces offer a promising alternative to vision and audition in critical environments, although they remain underutilized. This paper argues that current design strategies often focus on low-level parameters, and neglect perceptual organization, which in turn limits intelligibility and broader adoption. We propose a perspective with Gestalt theory as a relevant framework to structure vibrotactile feedback in a more robust manner. Reinterpreted through contemporary models such as Bayesian inference, Gestalt principles may inform the design of vibrotactile feedback due to their capacity to be rapidly and reliably processed, even under noisy or cognitively demanding conditions. We aim to provide future directions for the development of vibrotactile interfaces by promoting an approach grounded in holistic perceptual processing of information, with the goal to enhance situations handling, thereby improving the safety of the operator in critical contexts
E2E Gated-Mamba for cross-scenarios prognostics
No full textInternational audienceData-driven Prognostics and Health Management (PHM) models often struggle to generalize across diverse industrial settings, particularly under the “4Cs” challenges: cross-device, cross-prediction targets, cross-sequence lengths, and cross-input physics quantities. Existing approaches typically require retraining for each scenario, limiting their scalability and practicality. To address this, we propose Gated-Mamba, a novel end-to-end PHM backbone built on Selective State Space Models (SSMs) with dynamic gating. The architecture adaptively modulates memory evolution and information flow through input-conditioned discretization and gating mechanisms, enabling it to process variable-length sequences without retraining. Experimental results on mixed datasets — including bearings, batteries, and tool wear — demonstrate strong generalization, computational efficiency, and robustness to moderate sensor noise. Internal dynamics visualizations reveal that Gated-Mamba forms scenario-specific “thought waves,” enabling flexible adaptation to diverse degradation patterns. These capabilities make Gated-Mamba a scalable and interpretable solution for predictive maintenance in complex mechatronic systems
Évaluation de modèles aérodynamiques de faible à haute fidélité pour le flottement gyroscopique des hélices : comparaisons préliminaires avec des résultats expérimentaux
No full textInternational audienc
Recent advances on both numerical and experimental multipactor predictions
No full textInternational audienceCNES, the French Space Agency, has been studying space high power radio frequency (RF) effects -Multipactor, Corona and Passive Intermodulationfor many decades, starting from J. Sombrin Multipactor theory and models [1] to ongoing activities covering TRL 1 from 1 to 7 with our collaborators from academia, agencies, and industries. This paper intends to discuss recent advances related to Multipactor analysis, and present our way towards our main objective in this field: to improve modelling and experimental Multipactor predictions and the synergy between the two. We are studying electron emission physics to enhance our models and measurement methods on dielectric materials, their TEEY 2 , charge dynamics, treatment of secondary and backscattered electrons and the impact on Multipactor predictions. We are developing SPIS 3 to create a robust Multipactor modelling tools, dealing with dielectric materials and electron sources, while considering couplings with current reference software such as CST Studio, Spark3D and ANSYS Multipaction. We are also studying multipactor mitigation techniques based on surface treatments for both conductor and dielectric materials, and RF components design innovations to deal with current trends such as miniaturisation and high performances leading to high power density hence Multipactor risks. These studies align with the European roadmap on Multipactor theme [2] and complement ESA funded activities. We share the main goal as to give our community experimental and numerical tools to get better Multipactor predictions to, in fine, improve the reliability and performances of our high power RF systems. As a national agency we also engage on community awareness, explaining to various space communities that high power effects should be a common concern, and robust analyses should be better integrated in development plans and not wait for an anomaly and/or a major satellite loss to happen.</div