Portail HAL des publications du LIRMM
Not a member yet
13279 research outputs found
Sort by
Optimization-Based TDA for CDPRs with Elastic Cables: Twice Continuously Differentiable Cable Tensions
No full textInternational audienceThe need for C 2 -continuous cable tensions profiles arises in torque-based control of Cable-Driven Parallel Robots (CDPRs), when considering cable elasticity, as discontinuous profiles can hinder smooth motor torques and overall system performance. To tackle this, we propose a novel Tension Distribution Algorithm (TDA) formulated as an optimization problem. The method incorporates a 9th-degree polynomial trajectory generation to ensure C 2 -class continuity in desired motions while generating tension profiles that respect system constraints and avoid abrupt changes. Using an inverse quadratic penalty-based cost function, the approach guarantees regularity and smoothness, as demonstrated through rigorous mathematical proofs. Simulation results validate the method on a 6-DoF CDPR with eight actuating cables. This contribution lays the groundwork for achieving enhanced trajectory tracking and precise control in future applications
Distances Between Formal Concept Analysis Structures
No full textInternational audienceIn this paper, we study the notion of distance between the most important structures of formal concept analysis: formal contexts, concept lattices, and implication bases. We first define three families of Minkowski-like distances between these three structures. We then present experiments showing that the correlations of these measures are low and depend on the distance between formal contexts
Impact of deep learning and post-processing algorithms performances on biodiversity metrics assessed on videos
No full textPublisher: Public Library of ScienceInternational audienceAssessing the escalating biodiversity crisis, driven by climate change, habitat destruction, and exploitation, necessitates efficient monitoring strategies to assess species presence and abundance across diverse habitats. Video-based surveys using remote cameras are a promising, non-invasive way to collect valuable data in various environments. Yet, the analysis of recorded videos remains challenging due to time and expertise constraints. Recent advances in deep learning models have enhanced image processing capabilities in both object detection and classification. However, the impacts on models’ performances and usage on assessment of biodiversity metrics on videos is yet to be assessed. This study evaluates the impacts of video processing rates, detection and identification model performance, and post-processing algorithms on the accuracy of biodiversity metrics, using simulated remote videos of fish communities and 14,406 simulated automated processing pipelines. We found that a processing rate of one image per second minimizes errors while ensuring detection of all species. However, even near-perfect detection (both recall and precision of 0.99) and identification (accuracy of 0.99) models resulted in overestimation of total abundance, species richness and species diversity due to false positives. We reveal that post-processing model outputs using a confidence threshold approach (i.e., to discard most erroneous predictions while also discarding a smaller proportion of correct predictions) is the most efficient method to accurately estimate biodiversity from videos
Deep-Plant-Disease Dataset Is All You Need for Plant Disease Identification
No full textInternational audienceDeep learning models have emerged as a promising alternative to conventional approaches for plant disease identification, a critical challenge in agricultural production. However, the existing plant disease datasets are insufficient to address the complexities of realworld agricultural scenarios, such as multi crop disease, unseen, few-shot, and domain shift adaptation. Additionally, the lack of standardized evaluation protocols and benchmark datasets hinders the fair evaluation of models against these challenges. To bridge this gap, we introduce Deep-Plant-Disease, the largest and most diverse dataset with novel text data designed to enhance model generalization in multi crop disease identification. We revisit and reformulate the task by establishing a standardized evaluation framework that supports consistent benchmarking and guides future research. Through experiments, we further validate the robustness and adaptability of models trained on our dataset, highlighting their effective transferability to real-world agricultural challenges
FFT-Based Anomaly Detectors: Cutoff Frequency Adjustment and SMA-Based Approach
No full textCet article est une extension de la communication https://hal-lirmm.ccsd.cnrs.fr/lirmm-04683135v1/International audienceThis article presents a method for anomaly detection in time series based on the Fast Fourier Transform (FFT) using high-pass filtering. In addition to five existing strategies for determining the cutoff frequency (TF, AF, CAF, BSF, CBSF), a novel approach called SMAF is introduced. SMAF combines spectral analysis with adaptive smoothing using the Simple Moving Average, enabling the detection of high-frequency anomalies without requiring the inverse transform. The experiments employ the Yahoo Webscope dataset and the Numenta Anomaly Benchmark (NAB), providing a comprehensive evaluation. FFT-based approaches are compared to traditional statistical techniques (FBIAD and ARIMA) and machine learning methods (LSTM, ELM, and SVM). The results show that FFT-based methods outperform both statistical and machine learning techniques in terms of F1 score, precision, accuracy, and execution time. Among them, SMAF achieves the highest precision and the lowest execution time, reinforcing the potential of FFT-based filtering for efficient and accurate anomaly detection in time series
Leveraging Large Language Models for Time Series Prediction on Low-Frequency Data
No full textInternational audienceTime series prediction is critical in domains such as economics, industry, and agriculture. Real-world scenarios often involve challenges like low data frequency, high variability, and non-repetitive patterns. Traditional statistical models and machine learning approaches, including Long Short-Term Memory (LSTM) networks, underperform in low-data contexts due to overfitting risks, intensive training requirements, and the lack of benchmarks tailored to such scenarios. Large language models (LLMs) have emerged as tools for time series forecasting, leveraging their ability to generalize and capture temporal dependencies and patterns across datasets without requiring extensive task-specific feature engineering. This study investigates the potential of time series foundation models (TSFMs), specifically Lag-Llama and Chronos, on low-frequency datasets by comparing zero-shot prediction across one-step-ahead and multi-step-ahead approaches. Our findings evaluate predictive error, robustness, efficiency, and applicability, showing how TSFMs address these limitations and enhance forecasting in data-scarce scenarios
AgroforestAR: A mobile app for visualizing Agroforestry systems in Augmented Reality
No full textReferences:Lemiere L, Jaeger M, Gosme M, Subsol G (2023) Combinatorial Maps, a New Framework to Model Agroforestry Systems. Plant Phenomics 5:0120. https://doi.org/10.34133/plantphenomics.0120Rafflegeau S, Gosme M, Barkaoui K, et al (2023) The ESSU concept for designing, modeling and auditing ecosystem service provision in intercropping and agroforestry systems. A review. Agron Sustain Dev 43:43. https://doi.org/10.1007/s13593-023-00894-9International audienceAgroforestry is gaining more and more attention from researchers and practitioners in temperate areas, but it remains a vague concept for most of the public. This is because the renewal of interest for agroforestry systems is quite recent, demonstration sites are rare, and trees are still young, and therefore not very visible. The aim of AgroforestAR is to allow visualizing what an agroforestry system could look like on a given piece of land (including in your garden!). It uses the augmented reality capabilities already available in most smartphones, to superimpose, on the view seen by the phone’s or tablet’s camera, trees aligned along a line defined by the user by walking from one side of the piece of land to the other side. The user can then choose the tree species (among 5 available species currently) and size, as well as different distances between tree lines and different distances between trees along the line. The user can choose between four seasons, which will affect sun elevation, and for deciduous species also canopy leafiness, in order to visualize tree shade projection at different times of day. The app is freely available on Apple (https://eneo.fr/agroforestAR_ios) and Android (https://eneo.fr/agroforestAR_android) app stores. To use it, stand at the bottom-left corner of the plot, open the app, “scan” the ground around you to detect the soil surface, and click on the dotted area where you want to plant the first tree. Then walk to the top-left corner of the plot, and click where you want to plant the last tree in the row. Four rows of trees are automatically placed to your right. Beyond the use as an awareness-raising tool for the public, this app could be used in the future to help farmers decide between several possible options for their agroforestry project. Therefore, in the future, we intend to add the possibility to download more complex agroforestry patterns, using the ESSU concept (Rafflegeau et al. 2023) and combinatorial maps (Lemiere et al. 2023) to represent complex agroforestry systems. Thus, an advisor could design one or several alternative systems, send a download code to the farmer, who could then visualize the different options directly in their own fields. The following step will then to link this tool with prediction models to visualize the production of ecosystem services
Biomechanically consistent real-time action recognition for human-robot interaction
No full textThis paper presents a novel framework for real-time human action recognition in industrial contexts, using standard 2D cameras. We introduce a complete pipeline for robust and real-time estimation of human joint kinematics, input to a temporally smoothed Transformer-based network,for action recognition. We rely on a new dataset including 11 subjects performing various actions, to evaluate our approach. Unlike most of the literature that relies on joint center positions (JCP) and is offline, ours uses biomechanical prior, eg. joint angles, for fast and robust real-time recognition. Besides, joint angles make the proposed method agnostic to sensor and subject poses as well as to anthropometric differences, and ensure robustness across environments and subjects. Our proposed learning model outperforms the best baseline model, running also in real-time, along various metrics. It achieves 88% accuracy and shows great generalization ability, for subjects not facing the cameras. Finally, we demonstrate the robustness and usefulness of our technique, through an online interaction experiment, with a simulated robot controlled in real-time viathe recognized actions
Unbalanced Triangle Detection and Enumeration Hardness for Unions of Conjunctive Queries
No full textInternational audienceWe study the enumeration of answers to Unions of Conjunctive Queries (UCQs) with optimal time guarantees. More precisely, we wish to identify the queries that can be solved with linear preprocessing time and constant delay. Despite the basic nature of this problem, it was shown only recently that UCQs can be solved within these time bounds if they admit free-connex union extensions, even if all individual CQs in the union are intractable with respect to the same complexity measure. Our goal is to understand whether there exist additional tractable UCQs, not covered by the currently known algorithms.As a first step, we show that some previously unclassified UCQs are hard using the classic 3SUM hypothesis, via a known reduction from 3SUM to triangle listing in graphs. As a second step, we identify a question about a variant of this graph task that is unavoidable if we want to classify all self-join-free UCQs: is it possible to decide the existence of a triangle in a vertex-unbalanced tripartite graph in linear time? We prove that this task is equivalent in hardness to some family of UCQs. Finally, we show a dichotomy for unions of two self-join-free CQs if we assume the answer to this question is negative.In conclusion, this paper pinpoints a computational barrier in the form of a single decision problem that is key to advancing our understanding of the enumeration complexity of many UCQs. Without a breakthrough for unbalanced triangle detection, we have no hope of finding an efficient algorithm for additional unions of two self-join-free CQs. On the other hand, a sufficiently efficient unbalanced triangle detection algorithm can be turned into an efficient algorithm for a family of UCQs currently not known to be tractable
Диагональ Кантора и другие рассуждения
No full textInternational audienceA short popular brochure in Russian addressed to high school students and undergraduates and explaining what is known as Cantor diagonal argument (Baire theorem, sometimes even "forcing"). Several simple example (non-countability of real numbers, isomorphism of dense sets, immune sets, etc.) are explained