HAL-Université de Bretagne Occidentale
Not a member yet
75442 research outputs found
Sort by
Internet of Vehicles via Rate-Splitting Multiple Access With Antenna and RIS Partitioning
No full textInternational audienceIntegrating rate-splitting multiple access (RSMA) and reconfigurable intelligent surfaces (RIS) into vehicular communication systems offers significant potential for enhancing quality of service but also introduces challenges in radio resource allocation due to increased system complexity and the heterogeneous mobility patterns of vehicles. In this paper, we propose a vehicle-to-infrastructure (V2I) communication system that leverages a partitioned antenna array and RIS elements for reduced computational complexity, with RSMA serving multiple vehicles and users therein. To mitigate inter-vehicular interference, we employ a zero-forcing technique that decouples precoding at the BS and decoding at the user equipment (UE) into a zero-forcing factor and a precoding factor. For sum-rate maximization, we develop a block coordinate descent (BCD) algorithm, incorporating a dynamic partitioning scheme that divides the BS antenna array and RIS elements into subplanes. Each subplane is dedicated to a single vehicle for transmitting private streams to UEs therein, along with common streams for all UEs across all vehicles. The passive beamforming for each RIS subplane is optimized using the Riemannian conjugate gradient (RCG) method, while precoding at the BS subarray is optimized using a weighted minimum mean square error (WMMSE) approach. Simulation results and comparisons with benchmarks demonstrate that the proposed partitioning scheme achieves robust performance in terms of system sum rate while significantly reducing computational complexity
Automatic Extraction of Timing Models for WCET Estimation From a High-Level Synthesis Flow
No full textInternational audienceReal-time, domain-specific processors require faithful timing models for WCET analysis. However, existing models are typically hand-crafted from sparse documentation, making them error-prone and difficult to maintain. This work aims to automatically extract WCET timing models from single-issue in-order processor pipelines generated by High-Level Synthesis (HLS). By deriving timing models directly from the SpecHLS intermediate representation, the models are faithful by construction. Experimental results show that our timing-model extraction process generalizes across diverse RISC-V core variants and yields WCET estimates within 0.48% on average of those from a handcrafted model, on the Mälardalen WCET benchmarks
Recent advances and perspectives on N(2) fixation by microbial bioelectrochemical systems
No full textInternational audienceThis review covers recent advances in the fixing of dinitrogen in microbial bioelectrochemical systems (BES) where bacteria release or accept electron to/from electrodes for their respiratory metabolism, either directly or indirectly. We discuss how BES may be interesting platforms for producing ammonium or biomass from N(2) fixation. The potential for N(2)-fixation in BES is first discussed with a focus on possible metabolism and different mechanism that may lead to an increase of fixed dinitrogen. We then review recent examples where dinitrogen is fixed at the cathodes of BES, either by pure cultures of hydrogenotrophic and/or diazotrophic bacteria using cathodic H(2) or reduced redox mediators as the electron, or by mixed enriched consortia. A section is then devoted to the special case of nitrogen fixation at anodic microbial electrode where organic matter oxidation also occurs. Finally, a comparison of the reported current performance of nitrogen fixation in BES with other biotic (anerobic digestion) or abiotic (Haber-Bosch process, electrochemical N(2) reduction) is provided together with a perspective on possible optimization and application of this emerging microbial electrochemical and technological process
Nonparametric simulation of multivariate extreme events via spectral bootstrap
No full textInternational audienceInference in extreme value theory (EVT) relies on a limited number of extreme observations, making estimation challenging. To address this limitation, we propose a nonparametric simulation scheme, the multivariate extreme events spectral bootstrap simulation procedure, relying on the spectral representation of multivariate generalized Pareto-distributed random vectors. Unlike standard bootstrap methods, our approach preserves the joint tail behavior of the data and generates additional synthetic extreme data, thereby improving the reliability of inference. We demonstrate the effectiveness of our procedure on the estimation of tail risk metrics, under both simulated and real data. The results highlight the potential of this method for enhancing risk assessment in high-dimensional extreme scenarios
I/O patterns modeling of HPC applications with call stacks for predictive prefetch
No full textInternational audienceModern high-performance computing (HPC) storage systems use heterogeneous storage technologies organized in tiers to find a compromise between capacity, performance, and cost. In these systems, prefetching is a common technique used to move the right data at the right moment from a slow to a fast tier to improve overall performance while using the costly high-performance tier only when needed. Effective prefetching requires precise knowledge of the application I/O patterns. This knowledge can be extracted through the source code, I/O tracing tools or I/O functions call stacks. State-of-the-art solutions based on the latter approach mainly focus on applications with regular I/O profiles to avoid scalability issues due to the grammar-based techniques used. In this paper, we present an approach based on I/O call stacks that models POSIX and STDIO I/O patterns for both regular and irregular applications, thanks to the use of directed graphs. We present different models usable for prefetching. Our models were used to predict the next I/O call stack on five real HPC applications with a prediction accuracy of up to 98%. Compared to the state-of-the-art Omnisc'IO, they incurred up to 120x lower model overhead (334 ns vs. 45 μs on LAMMPS) and had a model size 10x to 15x smaller (463 B vs. 7 kB on LQCD)
Using a comprehensive donor and host dataset to decipher the heritability and genetic architecture of pearl traits in the black-lipped oyster (Pinctada margaritifera)
No full textInternational audienc
Attachment insecurity, adverse childhood experiences (ACEs), and suicidality in French residential-care adolescents: a gender-differentiated study
No full textInternational audienceAbstract Background Suicidality is alarmingly prevalent among adolescents placed in residential child welfare facilities, often as a consequence of early adverse childhood experiences (ACEs) and disrupted attachment relationships. Although these vulnerabilities are well established, the gender-specific mechanisms underlying suicidality in institutionalized youth remain poorly understood. Clarifying how trauma exposure and attachment insecurity interact with mental health symptoms is critical to inform targeted prevention. Methods In a cross-sectional study, 98 adolescents aged 12–17 years (54 girls, 44 boys; M = 14.34, SD = 2.08) living in French residential care completed validated self-report instruments assessing ACEs, attachment security, depressive and anxiety symptoms, and suicidality. Descriptive statistics, gender comparisons, and multivariate logistic regressions were used to identify predictors of suicidality, with all predictors standardized prior to entry. Results One-third of participants (33%) reported suicidal ideation or at least one suicide attempt. Emotional and physical abuse were the most frequent ACEs. Cumulative ACEs and attachment insecurity were independently associated with suicidality, and both correlated with heightened anxiety and depressive symptoms. Gender-stratified analyses showed that suicidality in girls was primarily linked to maternal alienation and emotional dysregulation, whereas in boys it was more strongly related to cumulative trauma exposure and depressive symptoms. Conclusions Findings highlight suicidality as a major concern in residential care and identify two complementary risk pathways: adversity-related and attachment‐related. Trauma-informed and attachment-based approaches—supported by systematic screening and the integration of mental health professionals within child welfare systems—may enhance early detection and individualized care. While contextualized in the French system, these mechanisms likely generalize across jurisdictions, underscoring the global need for gender-sensitive, relationally focused suicide prevention
Fonctions narratives de ha, ma, setu dans cinq contes merveilleux recueillis par François- Marie Luzel
No full textInternational audienc
Real-time Robotic Needle Insertion In Deformable and Moving Structure Using Learning-by-Example Method
No full textInternational audienceThis paper presents an innovative and practical method for robotic needle steering in radio-frequency ablation (RFA) to treat cancer. One of the main challenges in this process is that tissue shifts and deforms during needle insertion, making it difficult to predict the needle's path in real-time accurately. Inverse finite element (iFE) simulations have been used to address this problem. While these methods are accurate, they often require further refinement for effective time performance in real-world robotic systems. This is because when incorporating the method with a real robot, there can be a delay in command execution. To address this challenge, we propose a machine learning based solution that learn from offline simulations, moving the intensive calculations required by iFE methods to an offline training stage and predicting tissue deformation online with less computation time. Our network was trained on data from numerous simulated needle insertions to capture the interactions between insertion forces, tissue properties, and the resulting motion. Once trained, the model produces predictions almost instantaneously, making it suitable for real-time applications. We validated the approach by steering the needle in simulated deformable and moving gel to compare it with numeric-based methods, and then performing needle steering within a reconstructed human body, which involves multiple structures and also integrates the robot's dynamics. The results demonstrated that the developed networks achieved slightly better accuracy in the first scenario, while also executing faster, resulting in improved performance with the robot's dynamics. These findings show that our method is a promising advancement toward real-time guidance systems for needle-based medical procedures
Deep Learning for Palm Tree Health Assessment: UAV-Based Segmentation in the Figuig Region of Morocco
No full textInternational audienceThis study addresses the challenge of classifying healthy and unhealthy date palm trees within the Figuig oasis region of Morocco using high-resolution Unmanned Aerial Vehicle (UAV) imagery and deep learning. Traditional methods like ground surveys are often time-consuming, costly, and subjective for large areas, while satellite-based remote sensing may lack the spatial resolution to assess individual tree health accurately. Our UAV-based deep learning approach aims to overcome these limitations by providing improved scalability, objectivity, and spatial precision. Recognizing that the 'unhealthy' status observable in top-down UAV imagery represents a visually complex aggregation of symptoms-potentially caused by various stressors prevalent in the Figuig oasis such as Bayoud disease (caused by Fusarium oxysporum f.sp. albedinis) and drought stress-our annotations focused on classifying trees exhibiting these general visual signs of poor health rather than specific causal agents. We therefore focused on semantic segmentation for pixel-level classification. High-resolution RGB orthomosaics were acquired via UAV, processed into tiles, and manually annotated to create a dataset distinguishing three classes: healthy palm, unhealthy palm, and background. This dataset, comprising 296 tiles derived from an initial set and split into training (70%), validation (20%), and testing (10%), was used to train and evaluate U-Net and DeepLabV3+ models implemented from scratch. Quantitative evaluation on the unseen test set demonstrated promising performance: the DeepLabV3+ model achieved a Macro Average F1-score of 82.06%, slightly outperforming the U-Net model's score of 81.28%. Both models showed strong capability in identifying background and healthy palms. However, accurately segmenting the diverse 'unhealthy' class remained the most significant challenge. This potentially highlights inherent difficulties in differentiating subtle or varied stress symptoms from aerial RGB data alone, and may also reflect potential limitations of these architectures in fully capturing fine-grained textural variations or distinguishing between visually similar stress responses without additional input (e.g., multispectral data). Despite these challenges, the findings underscore the potential of integrating UAV technology with custom deep learning models for practical, large-scale palm health status assessment in precision agriculture, offering a marked improvement over less scalable or lower-resolution traditional techniques. While limitations related to dataset diversity and the inability to distinguish specific stressors from RGB data exist, this research provides a foundation for developing AI-driven tools to support timely crop management decisions and promote sustainable date palm cultivation. Future work may focus on enhancing model robustness, incorporating complementary data sources (like thermal or multispectral imagery), and investigating model architectures better suited for subtle feature extraction.</div