Portail HAL des publications du LIRMM
Not a member yet
13279 research outputs found
Sort by
Deep Learning for Species Recognition under High Uncertainty: Application to jellyfish images
No full textDoctora
A 4 foils hydrofoiler regulation solution and experiments
No full textInternational audienceThis paper introduces the Black Pearl, a novel 4-foil electric hydrofoiler, by presenting its design, control structure, and experimental findings. The innovative aspect of our approach lies in the redundant design of the sustentation structure, which utilizes four independent foils to control three degrees of freedom: roll, pitch, and altitude above the water. This redundancy improves actuation efficiency, significantly reducing the risk of actuation saturation. Moreover, the explicit management of this redundancy ensures robust actuation performance, even in scenarios where one foil becomes locked at a fixed angle. The paper also presents experimental trials conducted on the Black Pearl prototype, as well as the proposal and simulation testing of novel guidance laws designed to enhance the hydrofoiler's flight performance
QAOA : une nouvelle manière de chercher des bonnes solutions et de les trouver
No full textNational audienc
Towards the 30 by 30 Kunming-Montreal global biodiversity framework target : optimising graph connectivity in constraint-based spatial planning
No full textIJCAI : International Joint Conference on Artificial Intelligence AI and Social Good, Montréal, CAN, 16-/08/2025 - 22/08/2025International audienceThe Kunming-Montreal Global Biodiversity Framework aims to protect 30% of terrestrial, inland water, marine, and coastal ecosystems worldwide, and ensuring that at least 30% of these areas are under effective restoration by 2030. Maintaining and restoring ecological connectivity between natural habitats and protected areas is a key feature of this target. Achieving it will require effective and inclusive spatial planning supported by appropriate decision-support tools. Most spatial planning models address budget as an objective and connectivity as a constraint, formulating problems with Steiner trees. In many real-world cases, such as landscape-scale restoration planning, this formulation is inappropriate when environmental managers seek to optimise connectivity under a budget constraint. This problem was previously addressed with Constraint Programming (CP) and graph variables, but the current approach is severely limited in terms of spatial resolution. In this article, we formalise this problem as the budget-constrained graph connectivity optimisation problem. Based on a real case study: the restoration of forest connectivity in New Caledonia, we illustrate why ``naive' CP approaches are inefficient. In response, we provide a preprocessing method based on Hanan grids which preserves the existence of at least one optimal solution. Finally, we assess the efficiency of our approach in the New Caledonian case study
Exploring the Effectiveness of Machine Learning and Deep Learning Techniques for EEG Signal Classification in Neurological Disorders
No full textInternational audienceNeurological disorders are among the leading causes of both physical and cognitive disabilities worldwide, affecting approximately 15% of the global population. This study explores the use of machine learning (ML) and deep learning (DL) techniques in processing Electroencephalography (EEG) signals to detect various neurological disorders, including Epilepsy, Autism Spectrum Disorder (ASD), and Alzheimer's disease. We present a detailed workflow that begins with EEG data acquisition using a headset, followed by data preprocessing with Finite Impulse Response (FIR) filters and Independent Component Analysis (ICA) to eliminate noise and artifacts. Furthermore, the data is segmented, allowing the extraction of key features such as Bandpower and Shannon entropy, which improve classification accuracy. These features are stored in an offline database for easy access during analysis, to be then applied for both ML and DL models, systematically testing their performance and comparing the results to prior studies. Hence, our findings show impressive accuracy, with the random forest model achieving 99.85% accuracy in classifying autism vs. healthy subjects and 100% accuracy in distinguishing healthy individuals from those with dementia using Support Vector Machines (SVM). Moreover, deep learning models, including Convolutional Neural Networks (CNN) and ChronoNet, demonstrated accuracy rates ranging from 92.5% to 100%. In conclusion, this research highlights the effectiveness of ML and DL techniques in EEG signal processing, offering valuable contributions to the field of brain-computer interfaces and advancing the potential for more accurate neurological disease classification and diagnosis
Online matching with delays and stochastic arrival times
No full textA preliminary version appeared in the Proceedings of the 22nd International Conference on Autonomous Agents and Multi-agent Systems (AAMAS) 2023 pp. 976–984.International audienceThis paper presents a new research direction for the Min-cost Perfect Matching with Delays (MPMD), a problem introduced by Emek et al. (STOC'16). In the original version of this problem, we are given an n-point metric space, where requests arrive in an online fashion. Our goal is to minimize the matching cost for an even number of requests. However, contrary to traditional online matching problems, a request does not have to be paired immediately at the time of its arrival. Instead, the decision of whether to match a request can be postponed for time t at a delay cost of t. For this reason, the goal of the MPMD is not only to minimize the distance cost of the generated matching but to minimize the overall sum of distance and delay costs. Interestingly, it is proved that in the standard case of the adversarially generated requests, no online algorithm can achieve a competitive ratio better than O(log n/ log log n) (Ashlagi et al., APPROX/RANDOM'17).Here we consider a stochastic version of the MPMD problem where the input requests follow a Poisson arrival process. For such problem, we show that the above lower bound can be improved by presenting two deterministic online algorithms which, in expectation, are constant competitive, i.e., the ratio between the expected costs of the output matching and the optimal offline solution is bounded by a constant. The first one is a simple greedy algorithm that matches any two requests once the sum of their delay costs exceeds their connection cost, i.e., the distance between them. The second algorithm builds on the tools used to analyze the first one in order to obtain even better performance guarantees. This result is rather surprising as the greedy approach cannot achieve a competitive ratio better than O(m log 1.5+ε ) in the adversarial model, where m denotes the number of agents. Finally, we prove that it is possible to obtain similar results for the general case when the delay cost follows an arbitrary positive and non-decreasing function, for the asymmetric distance case, as well as for the MPMD variant with penalties to clear pending requests.</p
Combining Open Data and Formal Reasoning for Autonomously Controlled Spreading near Water Bodies
No full textNational audienceApplying fertilizers and pesticides near bodies of water poses significant environmental risks, primarily due to the potential for chemical runoff to contaminate aquatic ecosystems. To avoid this, regulations establish prohibition zones based on environmental and application-specific parameters, such as terrain slope, wind speed, precipitation, and the type and composition of substances used. This paper presents an autonomous robotic system that was developed to comply with these regulations while maximizing usable agricultural land. The robot scans its environment with sensors, including LiDAR, to measure features such as the distance to nearby bodies of water and the slope of the ground underneath. The InteGraal reasoning framework uses samples of regulations encoded in machine-readable RDF formats (using PAM vocabularies) and sensor observations modeled by the Semantic Sensor Network Ontology (SSNO) to make real-time decisions about where to stop or resume spraying. We extend existing vocabularies to include fertilizer-specific regulations, ensuring a comprehensive, semantically rich decision-support system for autonomous farms
Are those URIs so cool? URI Resolution and Content Negotiation in Ontology Repositories
No full textDemo accepted at ISWC 2025 but not published in the corresponding proceedings because of a copyright issue: https://ceur-ws.org/Vol-4085/International audienceThe Semantic Web relies on persistent, resolvable, and negotiable URIs. However, many ontology URIs fail to meet these expectations, whether for identifying the ontologies themselves or the entities they contain. This paper presents an analysis of more than 1900 URIs from two major ontology repositories (aka. Semantic Artefact Catalogues) -AgroPortal and BioPortal-revealing that 54% of ontology URIs are not resolvable and that 92% do not support HTTP content negotiation. In response, we introduce a suite of tools and infrastructure enhancements developed within AgroPortal to diagnose and address URI management challenges at no cost for ontology developers and end users. Our approach offers a standalone diagnostic tool, the generation of "twin URIs" that support resolution and negotiation, and mechanisms to enable HTTP redirection. This system applies both to ontology URIs and to content URIs within ontologies. For the latter, our infrastructure returns -in four syntaxes-only the RDF statements directly related to the resource in question, not the complete source file. We demonstrate the entire system through the AgroPortal web interfaces and services (http://agroportal.lirmm.fr)
GAMMA_OEIS: un programme Python pour calculer l'ensemble des ensembles de périodes de mots de longueur n, ainsi que l'ensemble des ensembles de périodes mourant à la longueur n.
No full textSoftware GAMMA_OEISProgram to compute both the sets of periods sets and the set of dying period setsWe provide a Python program called =incremental_gamma_dead_period_sets.py= to compute and from , where- the set of all period sets of all words of length - the set of dying period set for words of length .Both and are combinatorial sets related to the possible combinations of self-overlaps in finite words/strings (which are synonymous terms meaning sequences of symbols).For precise definitions and bibliographic reference, please consult the README file (pdf)
Young apple tree development under agroforestry radiative conditions: a multi-scale morphological and architectural dataset
No full textData Availability: The multi-scale tree graphs are permanently available on a https://github.com/openalea/GAFAM21 repository and are assigned a permanent doi: 10.5281/zenodo.14253433. The spreadsheet tables of variables extracted from MTGs are provided as Supporting Information of this article.International audienceAgroforestry is a major adaptation and mitigation strategy facing climate warming, but its agronomic viability depends on actual plant responses to shade conditions. Growing fruit trees under dominant trees may reduce the risks related to extreme climatic events, such as frost or heat waves. Nonetheless, except for some sciaphilous plants such as coffee or cacao, their physiological and architectural responses to agroforestry conditions are little known, especially in temperate climate. We present a dataset describing the architecture and morphology of 45 young apple trees, acquired in two consecutive years, along a radiative gradient, as in three growing conditions of an agroforestry plot: i) the open field, and ii) between and iii) along rows of dominant walnut trees. The data is stored as standard multi-scale tree graphs (MTG) that allow to store the topology, geometry and attributes of the plant at different scales. It includes plant traits at three topological scales: whole tree, growth unit and the internode. The traits include organ fate (latent, vegetative, floral bud and bud extinction sites); length and an estimate of the leaf area of growth units; diameter, zenith and azimuth angles of second order branches. The number of leaves, flowers, fruits and fruit drops is also counted on a sample of ten, possibly apical, flower buds per tree. The dataset includes ancillary measurements on sampled shoots, used to derive allometric relationships between shoot length and leaf area; and an estimate of the radiation reaching each apple tree during the vegetative season. The multi-scale description and the different light growing conditions characterizing the digitized trees allows to investigate relationships between the shade-related agroforestry environment and the apple tree morphological and architectural plasticity, during the early tree development, from the internode to the whole tree