Portail HAL des publications du LIRMM
Not a member yet
13279 research outputs found
Sort by
Implementation of a Virtual Dental Assistant based on GPT: Part 2
No full textInternational audienceIn the rapidly evolving landscape of artificial intelligence (AI), the development of specialised virtual assistants for dentistry represents a significant advance in improving clinical practice and education. In this second article, we focus on the practical implementation of a dental GPT assistant and describe the methodology used in its creation, the customisation processes performed, and the iterative modifications based on our user feedback and performance evaluations. The integration of AI into the dental practice supports the clinician in making informed decisions by providing real-time access to relevant information.[…
Comprehensive Management of Cocaine-Induced Midline Destructive Lesions: A Young-IfOS Consensus
No full textInternational audienceBackground: Recreational nasal cocaine use (RNCU) presents a significant challenge for rhinologists due to cocaine-induced midline destructive lesions (CIMDLs). This clinical consensus statement (CCS) offers guidelines for diagnosing, assessing, and managing both proven and suspected cases of CIMDL (including those without a prior RNCU history). It aims to support clinicians in addressing these complex cases effectively.Methods: An international, multidisciplinary panel of 18 specialists employed a three-round modified Delphi-method survey to evaluate statements covering CIMDL management issues such as definition, clinical evaluation and diagnosis, initial management approaches, and surgical management of complications and reconstructions. This study primarily targets otorhinolaryngologists.Results: Out of 44 evaluated statements, 20 achieved strong consensus, 20 reached consensus, 3 approached near-consensus, and 1 failed to achieve consensus. Consensus-covered areas included the definition of CIMDL, clinical evaluations, first-line management, and management of complications. However, reconstructive techniques remained a contentious topic.Conclusions: In the absence of extensive data, this CCS establishes a management framework for CIMDL, significantly bridging a knowledge gap. It highlights the need for standardized assessments, multidisciplinary cooperation, and customized follow-up care for patients with CIMDL. Considering the widespread use of cocaine, physicians should consistently consider the possibility of RNCU when encountering chronic inflammatory lesions in the sinonasal tract
Practical Full Key Recovery on a Google Titan Security Key
No full textInternational audienceThis chapter presents a practical case study of side-channel analysis. The work studies the security of the Google Titan Security Key1 (a hardware security token for two-factor authentication) and shows that its secure element, the NXP A700x chip, is susceptible to a side-channel attack (through the observation of its local electromagnetic (EM) activity). Given physical access to a Google Titan Security Key for around 10 hours, this allows us to retrieve a user-specific secret key (there is one key for each remote account) and therefore to clone the security device
A class of locally recoverable codes over finite chain rings
No full textInternational audienceLocally recoverable codes deal with the task of reconstructing a lost symbol by relying on a portion of the remaining coordinates smaller than an information set. We consider the case of codes over finite chain rings, generalizing known results and bounds for codes over fields. In particular, we propose a new family of locally recoverable codes by extending a construction proposed in 2014 by Tamo and Barg, and we discuss its optimality. The principal issue in generalizing fields to rings is how to handling the polynomial evaluation interpolation constructions
A Nonlinear MPC for Physical Human-Aerial Robot Interaction in Collaborative Transportation Tasks
No full textInternational audienceAerial robots are transitioning from traditional surveillance and monitoring roles to more advanced tasks involving physical interaction. Despite this progress, physical Human-Aerial Robot Interaction remains largely underexplored due to the complexity and stability-related issues of such platforms. This paper introduces a novel control framework that enables an aerial platform to cooperatively transport an object with a human operator. The control approach is built on a nonlinear model predictive control (NMPC), integrating the dynamic models of the human, the aerial robot, and the transported object. To ensure safe and robust physical interaction, the NMPC is combined with a compliant controller. Additionally, our controller prioritizes forward motion over lateral movements to accommodate the human's natural direction of motion. We validate this framework through indoor flight experiments, demonstrating how a human operator and a fully actuated hexarotor can effectively collaborate to transport a bar. The results highlight the aerial robot's ability to assist the human during physical transportation tasks, enhancing efficiency and comfort
Advent of code, jour 7
No full textNational audienceL'advent of code, littéralement le calendrier de l'avent du code est une mini fictionnécessitant de résoudre deux énigmes par jour, du premier au 25 décembre, la seconde énigmedu jour étant toujours une variante un peu plus difficile de la première. L'énigme du 7 décembre2024 m’a semblé un sujet intéressant pour parler d’algorithmique dans les colonnes de notremagazine préféré
From underwater to drone: A novel multi-scale knowledge distillation approach for coral reef monitoring
No full textInternational audienceDrone-based remote sensing combined with AI-driven methodologies has shown great potential for accurate mapping and monitoring of coral reef ecosystems. This study presents a novel multi-scale approach to coral reef monitoring, integrating fine-scale underwater imagery with medium-scale aerial imagery. Underwater images are captured using an Autonomous Surface Vehicle (ASV), while aerial images are acquired with an aerial drone. A transformer-based deep-learning model is trained on underwater images to detect the presence of 31 classes covering various coral morphotypes, associated fauna, and habitats. For aerial analysis these predictions are refined (some classes are merged, others are retained, while some are removed) resulting in a final set of 12 ecological categories that serve as annotations for training a second model applied to aerial images. The transfer of information across scales is achieved through a weighted footprint method that accounts for partial overlaps between underwater image footprints and aerial image tiles. The results show that the multi-scale methodology successfully extends fine-scale classification to larger reef areas, achieving a high degree of accuracy in predicting coral morphotypes and associated habitats. The method showed a strong alignment between underwater-derived annotations and ground truth data, reflected by an AUC (Area Under the Curve) score of 0.9251. This shows that the integration of underwater and aerial imagery, supported by deep-learning models, can facilitate scalable and accurate reef assessments. This study combines multi-scale imaging and AI to provide scientific information on coral reef monitoring and conservation. Our approach leverages underwater and aerial imagery, aiming for the precision of fine-scale analysis while extending it to cover a broader reef area
Adaptive STA Altitude Control of Hybrid Aerial-Underwater Vehicles
No full textInternational audienceHybrid Aerial-Underwater Vehicles (HAUVs) represent an advanced type of robotic systems capable to operate across multiple environments. This feature gives them enhanced versatility in movement capabilities, allowing a wide range of potential applications. However, these vehicles also pose significant challenges, including unmodeled dynamics, parametric uncertainties arising from variations in environmental density, and external disturbances, among others. This paper presents a novel adaptive gains Super Twisting Algorithm (STA) to address the tracking control problem. The proposed adaptive controller effectively manages the transition between operating environments without relying on techniques such as Hierarchical or Gain-Scheduling-based controllers, which necessitate prior knowledge of operational gains for each medium. Finally, numerical simulations are provided to validate the performance of the proposed controller
M1.9 Third Synchronisation Workshop Report
No full textThe Synchronisation Force workshops, annual series held between 2022 and 2024, brought together various stakeholders to periodically assess the progress in the FAIR-IMPACT focus areas. This Milestone concerns the 2024 workshop series. This Milestone demonstrates that FAIR-IMPACT, specifically Task 1.3 Synchronisation with EOSC Partnership and relevant projects and initiatives, organises an annual workshop series to monitor progress in FAIR practices across a wide range of EOSC initiatives and to disseminate such developments among the representatives of these initiatives.
SkiNet: un outil orienté utilisateur pour l'analyse de skills robotiques par des réseaux de Petri
No full textInternational audienceDespite the rise in applications of formal methods for the design and analysis of robotic systems, their integration into existing frameworks remains a challenge. Part of the problem is due to the important changes in the software architecture of robotic design tools needed to accommodate for formal methods, but another challenge is the inherent difficulty for non-experts to use formal methods. In this work, we present SkiNet, a user-oriented tool which aims to assist in the design of skill-based robotic systems using a formal model based on Petri nets. The main idea for SkiNet is to interpret the user-written specifications and properties, into elements from the formal model, and vice versa. This allows the user to be more independent when diagnosing and validating the specified behavior, without having to directly manipulate the backend formal tools. An example of the verification process of robotic skills specifications using SkiNet is given to showcase this paradigm