Portail HAL des publications du LIRMM
Not a member yet
13279 research outputs found
Sort by
Robust Bipedal Walking with Closed-Loop MPC: Adios Stabilizers
No full textMedia material :https://youtu.be/hfg7dTpuqu4We present a new walking control scheme based onthe dynamics of the inverted pendulum. Our scheme includes re-planning the step locations and step timings, feet force control,and a walking pattern generation that is closed-loop thanks tofeedback in the state of the real humanoid robot pendulum(CoM position/speed and ZMP). No additional control policyis used to maintain the static and dynamic balance of thehumanoid. We experimented this framework on five differenthumanoid robots over multiple disturbances including suddenpushes during walking or in a static state and by achievinglocomotion over uneven and compliant grounds
LifeCLEF 2025 Teaser: Challenges on Species Presence Prediction and Identification, and Individual Animal Identification
No full textInternational audienceAccurate identification, monitoring, and understanding of species distribution is important for biodiversity conservation, invasive species control, understanding climate change, and ecosystem management. Current methodologies for species identification, animal re-identification, and large-scale population monitoring are both resource-intensive and technically complex, posing significant challenges for widespread implementation. This highlights a need for automated, scalable solutions to enhance efficiency and accuracy. Since 2011, the LifeCLEF lab has driven progress in this field by organizing annual challenges to promote innovation in biodiversity informatics. The 2025 edition introduces five – one new, and four continued – data-driven tasks aimed at addressing current challenges in species recognition: (i) AnimalCLEF: multi-species individual animal identification, (ii) BirdCLEF: bird species identification in soundscape recordings, (iii) FungiCLEF: few shot classification with rare fungi species, (iv) GeoLifeCLEF: multi-modal species prediction using remote sensing and large-scale biodiversity data, and (v) PlantCLEF: multi-species plant identification in vegetation plot image
Life Cycle Assessment of Edge Data Centers: Case Study in Presence of Renewable Energy and Refurbished Servers
No full textInternational audienceAs the demand for computing power increases, the ecological footprint of data centers becomes a critical concern. Evaluating and mitigating this trend is essential to ensure a sustainable future for digital technologies. This paper addresses the issue by comparing the environmental impacts of a solar-powered edge data center (SDC) with refurbished servers and a conventional edge data center (CDC). The comparison is conducted using a multicriteria life cycle assessment for both data center types, focusing on indicators such as climate change, depletion of abiotic resources, acidification, ionizing radiation, and particle emissions. Our results indicate that, under a reference data center design scenario, SDC achieves a reduction in environmental impact ranging from 11% to 60% compared to CDC. This advantage is maintained when scaling up both data centers. Further reductions, up to 92% for some indicators, are possible by leveraging SDC characteristics such as increased solar energy use, grid energy integration, and extended compute node lifetimes. Our analysis shows that the environmental overhead of replacing components during server refurbishment in SDC is minimal, ranging from 1% to 6%, thereby reinforcing the sustainability and reliability of SDC. This comparative study illustrates the potential for edge data centers to yield positive environmental outcomes. Finally, we discuss the limitations and potential improvements of this study, as well as the challenges associated with deploying SDC and CDC in specific environments and for particular tasks
Adapting a global plant identification model to detect invasive alien plant species in high-resolution road side images
No full textInternational audienceEarly detection of invasive alien plant species is crucial for addressing their environmental impact. Recent advancements in vehicle-mounted equipment enable automatic analysis of high-resolution images to detect invasive plants along roadsides, a primary vector for their spread. Deep learning technologies show promise for processing this data efficiently, but the choice of approach significantly affects both computational and human resource costs. Object detection and segmentation methods require costly annotations, making them impractical for scaling to the thousands of invasive species worldwide. In contrast, multi-label classification, i.e. to predict all species present in the image, is less demanding but still challenging to implement without many annotated images for numerous species. However, large datasets from citizen science platforms such as Pl@ntNet or iNaturalist offer rich visual data for classifying individual plant species. In this article, we assess whether large plant identification models trained on such data can be leveraged for species detection in high-resolution images. Specifically, we explore two approaches: a multi-label classification model and a tiling-based model, using a vision transformer from the Pl@ntNet platform. We evaluate these models on high-resolution roadside images, both using a pre-trained model without fine-tuning and after applying fine-tuning. Our findings indicate that the tiling approach significantly outperforms other methods without fine-tuning and shows a slight advantage when fine-tuning is applied, demonstrating significant potential for detecting thousands of species without task-specific adaptation
Towards estimating the proportion of dead and missing vines at the field level
No full textInternational audienceThis work aimed to improve practices for estimating dead and missing vines, particularly when these estimations are made via sampling. Vine mortality is a well-known phenomenon that many studies have attempted to better explain, but only a few have focused on estimating the number of dead and missing vines at the field-level, which remains a challenging task for grapegrowers. The present article aimed to determine to what extend an increased an increased sampling effort improves estimation accuracy and thus help practitioners define a sample size adapted to vineyard properties and the accuracy expected by the grower. The first part of the study investigated whether vineyard properties (year of planting and variety) and available ancillary data (soil resistivity and fraction of vegetation cover [Fcover]) can provide a priori information on the proportion of dead and missing vines and guide the sampling strategy. The analysis was based on an exhaustive dataset created by individually mapping 14,199 dead and missing vines across 29 fields in a 20 ha vineyard. In this vineyard, regression models showed an increase in dead and missing vines of around 2.2 % per year from the tenth year onwards. Plantation year, variety and vigour (assessed by Fcover) were identified as providing valuable prior information on the proportion of dead and missing vines within the fields. The second part of the study focused on characterising the estimation error values that can result from these estimations. It focuses on sampling estimation, highlighting to what extent field properties (such as the number of dead and missing vines and their spatial autocorrelation) affect the accuracy of estimates. The results showed that the same sampling protocol can result in different accuracies from one field to another. Thus, they underline the importance of leveraging prior information, such as available field data, to tailor sampling efforts accordingly. Based on these results, this article introduces new guidelines to facilitate the estimation of the proportion of dead and missing vines in vineyards
Link between the Birth-Death process and the Kingman Coalescent — Applications to Phylogenetic Epidemiology
No full textInternational audienceThe two most popular tree models used in phylogenetics are the birthdeath (BD) and the Kingman coalescent (KC). These two models differ in several respects, notably: (i) population size is random in the BD versus fixed in the KC, (ii) the BD makes assumptions about the way samples are collected, while the KC conditions on the number of samples and the collection times, thus bypassing the need to describe the sampling procedure. These two models have been applied to different contexts: the BD in macroevolutive studies of clades of species, and the KC for populations. The exception is the field of phylogenetic epidemiology which uses both models. It then asks the question of how such different models can be used in the same context. In this paper, we study large-population limits of the BD, in a search for a mathematical link between the BD and the KC. We show that the KC is the large-population limit of a BD conditioned on a given population trajectory, and we provide the formula for the parameter θ of the limiting KC. This formula appears in earlier studies, but the present article is the first to show formally how the correspondence arises as a large-population limit, and that the BD needs to be conditioned for the KC to arise. Besides these fundamentally mathematical results, we demonstrate how our findings can be used practically in phylogenetic inference. In particular, we propose a new method for phylogenetic epidemiology, ensuing from our results. We conjecture that this new method, used in conjunction with auxiliary data, should allow to estimate important epidemiological parameters (e.g. the prevalence and the effective reproduction number), in a way that is robust to the data-generating model and the sampling procedure. Future studies will be needed to put our claims to the test
A constant-factor approximation for weighted bond cover
No full textInternational audienceThe WFD for a class of graphs asks, weighted graph , for a minimum weight vertex set such that The case when is minor-closed and excludes some graph as a minor has received particular attention but a constant-factor approximation remained elusive for WFD. Only three cases of minor-closed are known to admit constant-factor approximations, namely , and . We study the problem for the class of -minor-free graphs, under the equivalent setting of the problem, and present a constant-factor approximation algorithm using the primal-dual method. For this, we leveragea structure theorem implicit in [Joret, Paul, Sau, Saurabh, and Thomassé, SIDMA'14] which states the following: any graph containing a -minor-model either contains a large two-terminal , or contains a constant-size -minor-model, or a collection of pairwise disjoint connected sets that can be contracted simultaneously to yield a dense graph. In the first case, we tame the graph by replacing the protrusion with a special-purpose weighted gadget. For the second and third case, we provide a weighting scheme which guarantees a local approximation ratio. Besides making an important step in the quest of (dis)proving a constant-factor approximation for WFD, our result may be useful as a template for algorithms for other minor-closed families
Evaluating K-Mer Transformations for Cache Coherence and Uniformity
No full textK-mer indexes are crucial tools for comparing assembled and unassembled datasets across various applications. While probabilistic indexes are cost-efficient, exact k-mer indexes are essential for users requiring precise results or access to specific matching k-mer sequences. Static indexes often use SPSS to minimize memory usage by "assembling" k-mers, while dynamic indexes employ quotienting techniques to avoid preprocessing, becoming more memory-efficient as the k-mer set grows. However, the non-uniform distribution of k-mer sequences due to biological phenomena can negatively impact performance. To address this issue, k-mer transformations have been proposed to achieve a uniform distribution. Recently, optimizing shared prefixes of successive k-mers has been suggested to reduce cache misses and improve throughput. This study implements and evaluates known k-mer transformations, assessing their impact on distribution uniformity and prefix similarity. The goal is to identify the most effective transformations for specific use cases, thereby enhancing the efficiency and applicability of k-mer indexes. The benchmark bijecthash index is available as an open-source C++ library under the AGPL3 license at https://github.com/cagret/bijecthash. It is designed to be user-friendly and includes detailed instructions for adding custom transforms
Generating realistic artificial Human genomes using adversarial autoencoders
No full textA publicly available human genome serves as both a valuable resource for researchers and a potential risk to the individual who provided the genome. Many actors with selfish intentions could exploit it to extract information about the donor’s health or that of their relatives. Recent efforts have employed artificial intelligence models to simulate genomic data, aiming to create synthetic datasets with scientific merit while preserving patient anonymity. However, a major challenge arises in dealing with the vast amount of data that constitutes a complete human genome and the resources required to process it. We have developed a dimension reduction method that combines artificial intelligence with our knowledge of in vivo mutation association mechanisms. This approach enables the processing of large amounts of data without significant computational resources. Our genome segmentation follows chromosomal recombination hotspots, closely resembling mutation transmission mechanisms. Training data is sourced from the 1000 Genomes Project, which catalogues over 2500 genomes from diverse ethnic groups. Variational autoencoders, utilising neural networks, serve as an extension to the generative model. Wasserstein Generative Adversarial Networks (WGAN) are a benchmark among generation methods for various data types. After optimisation of our data simulation strategy our pipeline allows the generation of a simulated population meeting several essential criteria. It demonstrates good diversity, closely resembling that found in the reference dataset. It is plausible, as newly generated combinations of mutations do not disrupt the linkage disequilibria found in humans. It also preserves donor anonymity by synthesising combinations of reference genomes that are distant from reference samples
Control of Underactuated Mechanical Systems: Stabilisation and Limit Cycle Generation
No full textInternational audienceControl of Underactuated Mechanical Systems: Stabilization and Limit Cycle Generation Clearly and simply explains stabilization and stable limit cycle generation in the field of underactuated mechanical systems (UMS). It explores control design challenges and demonstrates concepts through real-time experiments.This book is organized into three parts:•Part I: General context and case study •Part II: Control solutions for the stabilization problem•Part III: Control solutions for stable limit cycle generationThe first part of this book introduces the state of the art related to underactuated mechanical systems, including some examples of systems reported in the literature, it will introduce the reader to the concept of stabilization and limit cycle generation. Then it focuses on the description of the inertia wheel inverted pendulum.The second part is devoted to the problem of stabilization; where various control solutions are presented and discussed, as well as their validation through numerical simulations and real-time experiments. The final part addresses the problem of stable limit cycle generation, where three proposed control solutions are detailed, as well as their related validation through different case studies