HAL-CentraleSupelec
Not a member yet
77624 research outputs found
Sort by
Integral Active Disturbance Rejection Control for Microgrid Load Frequency Regulation under Cyberattacks
No full textInternational audienc
RAPSOVA Meets Spatial Reuse: Ensuring Control-Plane Communication Efficiency in a Wi-Fi-Empowered Industry 4.0 Context
No full textInternational audienceIndustry 4.0 increasingly relies on Wi-Fi as the primary wireless fabric for coordinating massive numbers of sensors and robots in highly dynamic settings, especially where traditional wired infrastructures are impossible to deploy. In turn, Wi-Fi 7 and forthcoming Wi-Fi 8 standards aggressively target orders-of-magnitude gains in throughput, determinism, and robustness. Nevertheless, achieving efficient spatial reuse and effective load balancing are among the challenges that remain elusive. This paper introduces a novel Federated Deep Reinforcement Learning (DRL)-based spatial reuse scheme. Access Points (APs) learn policies for dynamically allocating Clear Channel Assessment (CCA) thresholds to control their own access to channels as well as that of their associated robots and sensors, i.e., Stations (STAs). We investigate the coexistence of this scheme with RAPSOVA, a DRL-based load-balancing framework that learns optimal STA-to-AP association policies to balance load on the APs. Both mechanisms require control-plane exchanges of local training updates to construct global policies. However, in large-scale, highly dynamic deployments, such exchanges risk saturating the control plane. Thus, we propose selectively transmitting only "good" training updates. We formulate this as a multi-objective multi-armed bandit (MOMAB) problem, aiming to trade off control-plane overhead against learning performance. Through extensive simulations, we validate the proposed spatial reuse scheme, its integration with RAPSOVA, and investigate the effects of heterogeneous AP placement and STAs' mobility on the MOMABs-based approach.</div
Enhancing urban data exploration: Layer Toggling and Visibility-Preserving Lenses for multi-attribute spatial analysis
No full textInternational audienceThis manuscript proposes two novel interaction techniques for visualization-assisted exploration of urban data, namely, Layer Toggling and Visibility-Preserving Lenses. The former mitigates visual overload by organizing information into distinct layers while enabling multi-layer comparisons through controlled overlays. The technique supports focused analyses without sacrificing spatial context and enables users to quickly switch between layers through a dedicated physical button interface. Visibility-Preserving Lenses, on the other hand, dynamically adapt their size and transparency so that users can effectively examine dense spatial regions and temporal attributes in detail. Both techniques support urban data exploration and improve prediction.Exploring urban data is essential for understanding complex phenomena related to crime, mobility, and residents’ behavior and equally important is the ability to predict and explain how they evolve over time, supporting informed urban planning and policymaking. However, navigating urban data in all their complexity is challenging, often resulting in cognitive overload, loss of spatial context, and excessive visual clutter due to the many layers that must be examined simultaneously. Although layered visualizations aim to mitigate those challenges, they face limitations with occlusion and effortless comparisons across data layers. Additionally, interaction methods are typically confined to mouse-based controls, limiting the fluidity of dynamic exploration.The visualization tool was validated through a comprehensive user study that measured user performance, cognitive load, and interaction efficiency across multiple devices. Using real-world data from São Paulo, including mobility patterns, climate conditions, and crime statistics, the way the approach enhances both exploratory and analytical tasks is demonstrated. The results also show how users perform when playing with different interactive devices, providing guidelines for future developments and improvements
A deep learning framework for spectrophotometric quantification of key microalgal pigments
No full textInternational audienceThis article presents a deep learning framework that links microalgae extract spectra to the quantification of individual pigments. To do so, it relies on Convolution Neural Network architectures. First, architectures from the literature were implemented and challenged. While showing good results for most pigments, they failed to predict adequately the zeaxanthin concentration (present in a very low amount, MAPE above 20%). Consequently, a specific network was designed. Upon finetuning, it reached 14.6% MAPE on the validation set. In addition to network architecture, data augmentation and preprocessing were explored. The results show that data augmentation by derivation alone (without extra preprocessing) yields the best results. Finally, the correlation between training dataset size and performance was investigated. Using the newly introduced learning curve tool, it was possible to evaluate the best achievable performance (3.10 to 8.57% MAPE) and convergence rate (approximately square root to quadratic, pigment-dependent) for the major pigments
Multi-criteria and multi-stage environmental study of Pl@ntnet service for the year 2024
No full textIn this study, we focus our investigation on Pl@ntNet, a citizen science platform, which re- lies on Artificial Intelligence (AI) models to identify plant species. Pl@ntNet provides a large-scale infrastructure supporting millions of users in over 200 countries. At this stage of deployment, and with years of experience developing the platform, Pl@ntNet is committed to understanding the environmental impacts of its identification service and contributing to the search for reduction opportunities. Our investigations assess the associated environmental impacts of Pl@ntNet for the year 2024. We based our approach on multi-criteria LCA, considering multiple impact type and the different life-cycle phases
Tracking solutions of time-varying variational inequalities
No full textTracking the solution of time-varying variational inequalities is an important problem with applications in game theory, optimization, and machine learning. Existing work considers time-varying games or time-varying optimization problems. For strongly convex optimization problems or strongly monotone games, these results provide tracking guarantees under the assumption that the variation of the time-varying problem is restrained, that is, problems with a sublinear solution path. In this work we extend existing results in two ways: In our first result, we provide tracking bounds for (1) variational inequalities with a sublinear solution path but not necessarily monotone functions, and (2) for periodic time-varying variational inequalities that do not necessarily have a sublinear solution path-length. Our second main contribution is an extensive study of the convergence behavior and trajectory of discrete dynamical systems of periodic time-varying VI. We show that these systems can exhibit provably chaotic behavior or can converge to the solution. Finally, we illustrate our theoretical results with experiments
MARL2GRID-TR: A multi-agent RL benchmark in power grid operations
No full textInternational audienceImproving power grid operations is essential for enhancing flexibility and accelerating grid decarbonization. Reinforcement learning (RL) has shown promise in this domain, most notably through the Learning to Run a Power Network (L2RPN) competition series, but prior work has primarily focused on single-agent settings, neglecting the often decentralized, multi-agent nature of grid control. We fill this gap with MARL2GRID-TR, the first multi-agent RL (MARL) benchmark for grid topology and redispatching, developed in collaboration with transmission system operators. Built on RTE France's high-fidelity simulation platform, our benchmark supports decentralized control across substations and generators, with configurable agent scopes, observability settings, expert-informed heuristics, and safety-critical constraints. The benchmark includes a suite of realistic scenarios that expose key challenges, such as coordination under partial information, longhorizon objectives, and adherence to hard physical constraints. Empirical results show that current MARL methods struggle under these real-world conditions. By providing a standardized, extensible platform, we aim to advance the development of scalable, cooperative, and safe learning algorithms for power grids
Monitoring morphometric drift in lifelong learning segmentation of the spinal cord
No full textInternational audienceMorphometric measures derived from spinal cord segmentations can serve as diagnostic and prognostic biomarkers in neurological diseases and injuries affecting the spinal cord. For instance, the spinal cord cross-sectional area can be used to monitor cord atrophy in multiple sclerosis and to characterize compression in degenerative cervical myelopathy. While robust, automatic segmentation methods to a wide variety of contrasts and pathologies have been developed over the past few years, whether their predictions are stable as the model is updated using new datasets has not been assessed. This is particularly important for deriving normative values from healthy participants. In this study, we present a spinal cord segmentation model trained on a multisite (n = 75 sites, 1,631 participants) dataset, including 9 different MRI contrasts and several spinal cord pathologies. We also introduce a lifelong learning framework to automatically monitor the morphometric drift as the model is updated using additional datasets. The framework is triggered by an automatic GitHub Actions workflow every time a new model is created, recording the morphometric values derived from the model's predictions over time. As a real-world application of the proposed framework, we employed the spinal cord segmentation model to update a recently introduced normative database of healthy participants containing commonly used measures of spinal cord morphometry. Results showed that (i) our model performs well compared with its previous versions and existing pathology-specific models on the lumbar spinal cord, images with severe compression, and in the presence of intramedullary lesions and/or atrophy achieving an average Dice score of 0.95 ± 0.03; (ii) the automatic workflow for monitoring morphometric drift provides a quick feedback loop for developing future segmentation models; and (iii) the scaling factor required to update the database of morphometric measures is nearly constant among slices across the given vertebral levels, showing minimum drift between the current and previous versions of the model monitored by the framework. The code and model are open source and accessible via Spinal Cord Toolbox v7.0.</div
A Proximal Approach for Stain Separation and Normalization of Whole-Slide Histopathological Images
No full textInternational audienceStain variability in histopathological images presents a major obstacle to developing accurate and reliable computer-aided diagnosis systems. In this paper, we introduce a novel stain normalization framework based on a stain separation proximal algorithm. Specifically, we first propose a new stain separation method that extracts individual stain components from a given stained image. Subsequently, we perform stain normalization by aligning the separated source stains with those of a target reference image. Experimental results demonstrate that the proposed approach effectively reduces inter-image color variations while preserving the essential structural details of the tissue.</div
Direction-of-Arrival Estimation of Coherent Sources with Leaky-Wave Antennas using Spatially Filtered Interpolation
No full textInternational audienceWith their frequency-beam scanning behavior, leaky-wave antennas (LWAs) are promisingsolutions to develop accurate and cost-effective direction-of-arrival (DoA) estimationsystems. However, DoA estimators such as MUSIC face challenges with coherent sourcesdue to the non-Vandermonde LWA steering matrix. Leveraging the unique radiation propertiesof LWAs, this paper first divides the entire field of view into several angular sectors,and then introduces a robust and accurate sectorized spatially-filtered interpolation (SFI)method to transform the LWA steering matrix into a Vandermonde matrix in each sectorwhile minimizing the issue of out-of-sector interference. The proposed method allowsthus the estimation of DoAs of coherent sources with LWAs. The simulation results showthat the DoAs of multiple coherent sources across the entire field-of-view, regardless theirangular sector, can be correctly estimated. The performance of the proposed method isshown be close to the Cramér-Rao Bound