Portail HAL des publications du LIRMM
Not a member yet
13279 research outputs found
Sort by
Robust super-twisting-based disturbance observer for autonomous underwater vehicles: Design, stability analysis, and real-time experiments
No full textInternational audienceThis paper proposes a new observation-based proportional-derivative control method for robust trajectory tracking of autonomous underwater vehicles (AUVs). The proposed control scheme is designed based on a new observation-based nonlinear model that captures the dynamics and uncertainties of the AUV’s behavior. The proposed control method is formulated in such a way that it can handle system nonlinearities and uncertainties, making it robust to external disturbances and model uncertainties. The effectiveness of the proposed control method is demonstrated through extensive real-time experiments in a real-world AUV trajectory tracking scenario. The obtained results show that the proposed control method outperforms other control methods in the literature regarding trajectory tracking accuracy, robustness, and disturbance rejection. Overall, the proposed observation-based proportional-derivative control method can significantly improve the trajectory tracking performance of AUVs in real-world applications
Analyse et Modélisation de Réseaux d'Interconnexion Cohérent Puce-à-Puce
No full textThe slowdown of Moore's Law has led to a paradigm shift in High-Performance Computing (HPC), forcing computer architects to explore alternatives. This shift has given rise to chiplet-based System-on-Chip (SoC) architectures, where logic is distributed across multiple dies within a single package, instead of traditional monolithic single-chip multicore architectures. While this transition improves cost-efficiency and yields, it introduces new challenges in design, evaluation, and optimization. Chiplet-based designs present both challenges and opportunities, introducing varying intercore latencies while enabling post-tape-out heterogeneous integration and high-bandwidth on-package interconnects. New protocols like CCIX, UCIe, and OpenCAPI have emerged to facilitate communication between interconnected chips. This approach allows flexibility from single-chiplet CPUs for edge applications to multi-chiplet CPUs for high-performance cloud computing. Maintaining coherence in modern multicore systems is challenging, requiring frequent protocol design and validation. The extension of cache coherence to multiple chiplets significantly increases design complexity. While high-level simulation could address these challenges without time-intensive RTL designs or costly tape-outs, existing architecture simulators often lack the flexibility and capabilities needed for modern coherent interconnect protocols. Moreover, capable simulators are frequently proprietary, hindering academic research. This thesis explores the modeling of coherent interconnect protocols for chiplet-based systems. We address key limitations in existing solutions: current multi-chiplet simulations often use multiple parallel simulator instances with separate OS environments, poorly representing targeted architectures. Additionally, extended monolithic network-on-chip models often fail to maintain coherence across chips, and open-source platforms typically lack native support for multiple network-on-chip handling. Our contributions include: First, a methodology for modeling in modern, open-source system-level simulators, presenting gem5 with key aspects and debugging strategies. Second, an analysis of challenges in modeling coherent chip-to-chip interconnects, providing insights for simulating multi-chip architectures. Finally, we implement a chip-to-chip interconnect model in gem5, extending the Arm CHI protocol across multiple chips while maintaining coherence and enabling full-system simulation. This thesis provides both methodology for designing simulation models and a complete implementation of a coherent chip-to-chip interconnect model. Our contribution advances the accurate simulation of cache-coherent multi-chip architectures and aims to inspire future optimization work. Through open-source tools and methodologies, we seek to accelerate research in computer architecture, enabling more efficient computing systems.Le ralentissement de la loi de Moore a conduit à un changement de paradigme dans le calcul haute performance (HPC), forçant les architectes informatiques à explorer des alternatives. Ce changement a donné naissance aux architectures System-on-Chip (SoC) basées sur des chiplets, où la logique est distribuée sur plusieurs dies au sein d'un même package, remplaçant les architectures multicœurs monolithiques traditionnelles. Bien que cette transition améliore la rentabilité et les rendements, elle introduit de nouveaux défis en matière de conception, d'évaluation et d'optimisation. Les conceptions basées sur les chiplets présentent à la fois des défis et des opportunités, introduisant des latences variables entre les cœurs tout en permettant une intégration hétérogène post-tape-out et des interconnexions à haute bande passante. De nouveaux protocoles comme CCIX, UCIe et OpenCAPI ont émergé pour faciliter la communication entre les puces interconnectées. Cette approche permet une flexibilité allant des CPUs à chiplet unique pour les applications edge aux CPUs multi-chiplets pour le cloud computing haute performance. Le maintien de la cohérence dans les systèmes multicœurs modernes est un défi, nécessitant une conception et une validation fréquentes des protocoles. L'extension de la cohérence de cache à plusieurs chiplets augmente significativement la complexité de conception. Bien que la simulation puisse répondre à ces défis sans recourir à des conceptions RTL chronophages ou des tape-outs coûteux, les simulateurs d'architecture existants manquent souvent de flexibilité et de capacités nécessaires pour les protocoles d'interconnexion cohérents modernes. De plus, les simulateurs capables sont souvent propriétaires, entravant la recherche académique. Cette thèse explore la modélisation des protocoles d'interconnexion cohérents pour les systèmes basés sur les chiplets. Nous abordons les principales limitations des solutions existantes : les simulations multi-chiplets actuelles utilisent souvent plusieurs instances parallèles de simulateur avec des environnements OS séparés, représentant mal les architectures ciblées. De plus, les modèles network-on-chip monolithiques étendus échouent souvent à maintenir la cohérence entre les puces, et les plateformes open-source manquent généralement de support natif pour la gestion de multiple network-on-chip. Nos contributions incluent : premièrement, une méthodologie de modélisation dans les simulateurs système open-source modernes, présentant gem5 avec ses aspects clés et stratégies de débogage. Deuxièmement, une analyse des défis de la modélisation des interconnexions cohérentes puce à puce, fournissant des directives pour la simulation d'architectures multi-puces. Enfin, nous implémentons un modèle d'interconnexion puce à puce dans gem5, étendant le protocole Arm CHI à travers plusieurs puces tout en maintenant la cohérence et permettant une simulation full-system. Cette thèse fournit à la fois une méthodologie pour concevoir des modèles de simulation et une implémentation complète d'un modèle d'interconnexion cohérent puce à puce. Notre contribution fait progresser la simulation précise des architectures multi-puces cohérentes en cache et vise à inspirer de futurs travaux d'optimisation. Grâce aux outils et méthodologies open-source, nous cherchons à accélérer la recherche en architecture informatique, permettant des systèmes informatiques plus efficaces
Ongoing projects to understand and mitigate bycatch from the longline bluefin tuna fishery in the french mediterranean
No full textInternational audienceUnderstanding and mitigating bycatch is a major issue for the management of interactions between fisheries and marine ecosystems. In a context of biodiversity loss and climate change that impact marine fauna, research actions are needed to provide solutions for a sustainable future. The French longline fishery for Atlantic Bluefin tuna in the Gulf of Lion reports bycatch for different species. In order to provide solutions, several initiatives in collaboration with professional fishermen have been carried out since the beginning of this fishery in 2011. In the present paper, the different projects and programs developed in the Gulf of Lion are listed and explained. Different observation approaches were designed to characterize the fishery, collect data on bycatch, study the ecology and the post-release survival of these species, while other projects focused on innovative mitigation solutions. Preliminary results and work perspectives are presented.Entender y mitigar la captura fortuita es cuestión fundamental para la ordenación de las interacciones entre las pesquerías y los ecosistemas marinos. En un contexto de pérdida de biodiversidad y cambio climático que impacta en la fauna marina, se necesitan acciones de investigación para facilitar soluciones para un futuro sostenible. La pesquería de palangre francesa para el atún rojo del Atlántico en el golfo de León comunica captura fortuita de diferentes especies. Con el objetivo de aportar soluciones, desde el inicio de esta pesquería en 2011 se han llevado a cabo varias iniciativas en colaboración con pescadores profesionales. Este documento enumera y explica los diferentes proyectos y programas desarrollados en el golfo de León. Se diseñaron distintos enfoques de observación para caracterizar la pesquería, recopilar datos sobre captura fortuita, estudiar la ecología y la supervivencia posterior a la liberación de estas especies, mientras que otros proyectos se centraron en soluciones innovadoras de mitigación. Se presentan los resultados preliminares y las perspectivas de trabajo.La compréhension et l'atténuation des prises accessoires constituent un enjeu majeur de la gestion des interactions entre les pêcheries et les écosystèmes marins. Dans un contexte de perte de biodiversité et de changement climatique ayant un impact sur la faune marine, des actions de recherche sont nécessaires pour fournir des solutions pour un avenir durable. La pêche palangrière française ciblant le thon rouge de l'Atlantique dans le golfe du Lion déclare des prises accessoires de différentes espèces. Afin d'apporter des solutions, plusieurs initiatives en collaboration avec les pêcheurs professionnels ont été menées depuis le début de cette pêcherie en 2011. Dans le présent document, les différents projets et programmes développés dans le golfe du Lion sont énumérés et expliqués. Différentes approches d'observation ont été conçues pour définir la pêcherie, collecter des données sur les prises accessoires, étudier l'écologie et la survie suivant la remise à l'eau de ces espèces, tandis que d'autres projets se sont concentrés sur des solutions d'atténuation innovantes. Les résultats préliminaires et les perspectives de travail sont présentés dans ce document
Modelling the drivers of outbreak communication in online media news for improved event-based surveillance
No full textSource Agritrop Cirad (https://agritrop.cirad.fr/616019/)International audienceEpidemic intelligence (EI) practitioners at health agencies monitor various sources to detect and follow up on disease outbreak news, including online media monitoring. The Platform for Automated Extraction of Disease Information from the Web (PADI-web), developed in 2016 for the French Platform for Epidemiosurveillance in Animal Health (Platform ESA), monitors and collects outbreak-related news from online media, allowing users to detect and anticipate response to disease outbreaks. Given the mass number of outbreak-related news collected with PADI-web, we aimed to understand better what drives communication on outbreaks by the different online media sources captured by this tool to allow for a more targeted and efficient EI process by its users. We built a bipartite network of sources communicating on outbreaks of avian influenza (AI) and African swine fever (ASF) captured by PADI-web between 2018 and 2019 worldwide. We used an Exponential Random Graph Model (ERGM) to assess epidemiological, socioeconomic, and cultural factors that drive communication on disease outbreaks from the different online media sources. Our AI network comprised 969 communicated news (links) from 436 news reports from 212 sources describing 199 AI outbreaks. The ASF network comprised 1340 communicated news (links) from 594 news reports from 204 sources and 277 ASF outbreaks. The ERGM was fitted for each network. In both models, international organisations and press agency sites were more likely to communicate about outbreaks than online news sites (OR = 4.8 and OR = 3.2, p < 0.001 for AI; OR = 3.1 and OR = 4.7, p < 0.001 for ASF). Research organisations for AI (OR = 2.3, p < 0.001) and veterinary authorities for ASF (OR = 3.6, p < 0.001) were also more likely to be a source of information than online news sites. Our work identified the factors driving communication about animal and zoonotic infectious disease outbreaks in online media sources monitored by PADI-web. This information can guide EI practitioners and users of PADI-web to monitor specific sources based on their specialisation and coverage and the disease's epidemiological status. Our results also suggest that EI practitioners may use other means to collect EI information in countries and regions that are not well-represented in the data
Structural Properties of Entropic Vectors and Stability of the Ingleton Inequality
No full textWe study constrained versions of the Ingleton inequality in the entropic setting and quantify its stability under small violations of conditional independence. Although the classical Ingleton inequality fails for general entropy profiles, it is known to hold under certain exact independence constraints. We focus on the regime where selected conditional mutual information terms are small (but not zero), and the inequality continues to hold up to controlled error terms. A central technical tool is a structural lemma that materializes part of the mutual information between two random variables, implicitly capturing the effect of infinitely many non-Shannon--type inequalities. This leads to conceptually transparent proofs without explicitly invoking such infinite families. Some of our bounds recover, in a unified way, what can also be deduced from the infinite families of inequalities of Matúš (2007) and of Dougherty--Freiling--Zeger (2011), while others appear to be new
Calcul incrémental de l'ensemble des ensembles de périodes
No full text21 pages, 4 figures, 5 algorithms, 27 references, 1 related resource on Zenodo, 6 appendices; accepted for publication at 50th conference SOFSEM 2025International audienceÜberschneidungen zwischen Wörtern sind in vielen Bereichen der Informatik von entscheidender Bedeutung, beispielsweise im Code-Design, in der Stringologie und in der Bioinformatik. Ein sich selbst überschneidendes Wort zeichnet sich durch seine Perioden und Borders aus. Eine Periode eines Wortes ist die Startposition eines Suffixes von , das gleichzeitig ein Präfix von ist, und ein solches Suffix wird als Border bezeichnet. Jedes Wort der Länge n>0 hat eine Menge von Perioden, aber nicht alle Kombinationen von ganzen Zahlen sind Mengen von Perioden. Die Berechnung der Periodenmenge eines Wortes erfordert lineare Zeit in der Länge von . Wir befassen uns mit der Frage der Berechnung der Menge aller Periodenmengen von Wörtern der Länge . Obwohl Periodensätze charakterisiert wurden, gibt es keine Formel zur Berechnung der Kardinalität von (die exponentiell in ist), und der bekannte dynamische Programmieralgorithmus zur Aufzählung von leidet unter seiner Raumkomplexität. Wir stellen einen inkrementellen Algorithmus zur Berechnung von aus vor, der die Raumkomplexität reduziert, und anschließend einen konstruktiven Zertifizierungsalgorithmus, der für Verifizierungszwecke nützlich ist. Der inkrementelle Ansatz definiert eine Eltern-Kind-Beziehung zwischen Mengen in und , wodurch man die Dynamik von Periodenmengen und ihre interessanten statistischen Eigenschaften untersuchen kann. Darüber hinaus ist die Periodenmengen eines Wortes der Schlüssel zur Berechnung der Abwesenheitswahrscheinlichkeit von in zufälligen Texten. Daher ist die Kenntnis von nützlich, um die Bedeutung von Wortstatistiken, wie beispielsweise die Anzahl fehlender Wörter in einem zufälligen Text, zu bewerten.Overlaps between words are crucial in many areas of computer science, such as code design, stringology, and bioinformatics. A self overlapping word is characterized by its periods and borders. A period of a word is the starting position of a suffix of that is also a prefix , and such a suffix is called a border. Each word of length, say n>0, has a set of periods, but not all combinations of integers are sets of periods. Computing the period set of a word takes linear time in the length of . We address the question of computing, the set, denoted , of all period sets of words of length . Although period sets have been characterized, there is no formula to compute the cardinality of (which is exponential in ), and the known dynamic programming algorithm to enumerate suffers from its space complexity. We present an incremental approach to compute from , which reduces the space complexity, and then a constructive certification algorithm useful for verification purposes. The incremental approach defines a parental relation between sets in and , enabling one to investigate the dynamics of period sets, and their intriguing statistical properties. Moreover, the period set of a word is the key for computing the absence probability of in random texts. Thus, knowing is useful to assess the significance of word statistics, such as the number of missing words in a random text.Les chevauchements entre les mots sont essentiels dans de nombreux domaines de l'informatique, tels que la conception de code, l'algorithmique du texte et la bio-informatique. Un mot auto-chevauchant est caractérisé par ses périodes et ses bords. Une période d'un mot est la position de départ d'un suffixe de qui est également un préfixe , et un tel suffixe est appelé un bord de . Chaque mot de longueur, disons n>0, possède un ensemble de périodes, mais toutes les combinaisons d'entiers ne sont pas des ensembles de périodes. Le calcul de l'ensemble des périodes d'un mot prend un temps linéaire par rapport à la longueur de . Nous abordons la question du calcul de l'ensemble, noté , de tous les ensembles de périodes de mots de longueur . Bien que les ensembles de périodes aient été caractérisés, il n'existe aucune formule pour calculer la cardinalité de (qui est exponentielle en ), et l'algorithme de programmation dynamique connu pour énumérer souffre de sa complexité en espace. Nous présentons une approche incrémentale pour calculer à partir de , approche qui réduit la complexité en espace, puis un algorithme de certification constructif utile à des fins de vérification. L'approche incrémentale définit une relation parentale entre les ensembles dans et , ce qui permet d'étudier la dynamique des ensembles de périodes et leurs propriétés statistiques intrigantes. De plus, l'ensemble de périodes d'un mot est la clé pour calculer la probabilité d'absence de dans des textes aléatoires, ce qui souligne son importance pratique
Learning Compact Representations of Constraint Networks
No full textInternational audiencePassive constraint acquisition aims to learn constraint networks from examples of solutions and non-solutions. There typically exist many constraint networks that are consistent with a given set of examples, so the performance of an acquisition system is critically dependent on its ability to determine which network will generalize the best to unseen data. We introduce a framework for representing constraint networks in compressed form and present a novel method for constraint acquisition. Our method learns a constraint network that achieves a high compression ratio, with the idea that such networks are highly structured and therefore less prone to overfitting. Experiments demonstrate that this approach significantly reduces the number of examples needed for training and achieves a high accuracy on unseen data
Analysis, quantification and identification of in situ bioluminescence signals by an innovative sensor (CEMSOR2)
No full textInternational audienceBioluminescence, the light emitted naturally by marine organisms, is the main light source in the mesopelagic zone. Nearly 75% of marine organisms, from the surface to the deep sea, use this capability for communication with diverse ecological goals (predation, repulsion...). Bioluminescence detection thus offers an indirect way of tracking the presence, distribution and migrations of organisms. Such detection can lead for example to a better understanding of vertical migrations of organisms and consequently of a better quantification of the active carbon export in the mesopelagic ocean. However, current technologies still limit large deployments, and high frequency observations of in situ bioluminescence.To overcome these limitations, the CEMSOR2 project, led by several institutions (including LIRMM, MIO and IFREMER), aims to develop an innovative, low cost, compact, multi-instrumented sensor capable of measuring bioluminescence in situ. The CEMSOR2 is designed to be easily deployable on a wide range of vectors (such as underwater gliders, CTDs, buoys, trawls, living organisms). The sensor being easy to deploy will enable us to collect a wide range of bioluminescent data with high spatiotemporal resolution, while recording environmental and behavioral variables related to the organisms.Validation of this sensor relies on a series of tests in a controlled environment to verify its robustness under marine conditions (pressure, water), and to calibrate and characterize it. Field testing of the CEMSOR2 is an essential part of the project. Controlled experiments have been performed on several luminous species (Pennatula Rubra, Pteroides Griseum, and Veretillum Cynomorium). By hypothesizing that each species or individual emits distinct bioluminescent flashes, we mechanically stimulated these organisms with a water current in a dark room, detected their light signals with the sensor, and then analyzed their light emissions. These results are essential for calibrating the sensor and refining detection algorithms. This research highlights distinct light signatures for these species of cnidarians.Once deployed, a template script allows to analyse bioluminescence signals according to their spatio-temporal distribution in the water column. By classifying light peaks according to their characteristics (shape, intensity, duration), we aim to link these events to species behavior and environmental variables. The aim is to develop on-board algorithms to detect and process these signals within the sensor
Neurosymbolic AI for Natural Language Inference in French : combining LLMs and theorem provers for semantic parsing and natural language reasoning
No full textInternational audienceIn this article, we describe the first comprehensive neurosymbolic pipeline for the task of Natural Language Inference (NLI) for French, with the synergy of Large Language Models (CamemBERT) and automated theorem provers (GrailLight, LangPro). LLMs prepare the input for GrailLight by tagging each token with Part-of-Speech and grammatical information based on the Type-Logical Grammar formalism. GrailLight then produces the lambda-terms given as input to the LangPro theorem prover, a tableau-based theorem prover for natural logic originally developed for English. Currently, the proposed system works on the French version of SICK dataset. The results obtained are comparable to the ones on the English and Dutch versions of SICK with the same LangPro theorem prover, and are better than the results of recent transformers on this specific dataset. Finally, we have identified ways to further improve the results obtained, such as giving access to the theorem prover to lexical knowledge via a knowledge base for French
Fuzzy-Based Ensemble Method for Robust Concept Drift Detection in Multivariate Time Series
No full textInternational audienceConcept drift detection (CDD) is the general problem of identifying significant changes in streaming data distribution over time. Effective drift detection is important in industrial processes such as oil and gas exploration to mitigate financial losses, ensure personnel safety, and reduce environmental risks. However, current CDD methods face challenges in large-scale, multivariate datasets, where single drift detectors (DD) often fail to capture variable interdependencies. While ensemble drift detectors (EDD) are usually adopted to mitigate the adoption of a single DD, EDD may suffer when detections do not converge. This misalignment can cause voting mechanisms to neglect critical intervals with high detection rates. To address this issue, we propose a fuzzy ensemble drift detector (FEDD) that integrates unsupervised threshold voting with fuzzy logic to provide time tolerance and reconcile minor temporal misalignments in drift detection. FEDD is evaluated using the 3W dataset, a realistic public benchmark with rare undesirable real events in oil wells. The results demonstrate that FEDD outperforms existing approaches by improving detection robustness and coverage, ensuring more reliable drift detection in high-dimensional, noisy environments