Portail HAL des publications du LIRMM
Not a member yet
13279 research outputs found
Sort by
A formal framework for the specification and verification of robotic skills composition for autonomous behaviors
No full textInternational audienceWith the goal to create autonomous systems that can accomplish complex tasks in dynamic environments, high-level software architectures are created using formal methods to describe the system with a set of elementary actions, named skills, necessary to accomplish their mission.However, these formalisms often lack a proper framework for designing autonomous missions as a composition of the specified skills, often relying on non-formal or semi-formal tools, such as behavior trees or UML/SysML-based tools, as well as being limited to specific elementary skill implementations. This makes the system less reliable and more prone to failures when placed in a dynamic, unknown environment. To fill this gap, we propose a formal framework to specify formal models for elementary skills and make them composable, called Skill Petri nets, and a set of composition patterns and their Petri net models. A tool will allow users to define these compositions using an extension of the Robot-Language DSL, from which the formal models and a controller code are generated. An industrial inspection mission use-case allowed to test the proposed framework.</p
Variability-Driven User-Story Generation using LLM and Triadic Concept Analysis
No full textInternational audienceA widely used Agile practice for requirements is to produce a set of user stories (also called "agile product backlog"), which roughly includes a list of pairs (role, feature), where the role handles the feature for a certain purpose. In the context of Software Product Lines, the requirements for a family of similar systems is thus a family of user-story sets, one per system, leading to a 3-dimensional dataset composed of sets of triples (system, role, feature). In this paper, we combine Triadic Concept Analysis (TCA) and Large Language Model (LLM) prompting to suggest the user-story set required to develop a new system relying on the variability logic of an existing system family. This process consists in 1) computing 3-dimensional variability expressed as a set of TCA implications, 2) providing the designer with intelligible design options, 3) capturing the designer's selection of options, 4) proposing a first user-story set corresponding to this selection, 5) consolidating its validity according to the implications identified in step 1, while completing it if necessary, and 6) leveraging LLM to have a more comprehensive website. This process is evaluated with a dataset comprising the user-story sets of 67 similar-purpose websites
FrontMod
No full textFrontMod is a Software/prototype for the migration of the "client-frontend" part of a software implemented in legacy technologies (GWT, WPF…) to recent technologies (Angular, react…)
Cost-Optimized and Highly Robust Latches Providing Complete Quadruple-Node-Upset Tolerance and Recovery With Algorithm based Verifications
No full textInternational audienceWith the aggressive shrinking of transistor feature sizes, nano-scale CMOS circuits are becoming more vulnerable to multi-node-upsets, e.g., triple-node-upsets (TNUs) as well as quadruple-node-upsets (QNUs), caused by the effects of harsh radiation. This paper first presents a QNU completely tolerant latch based on so-called TPDICE and the soft error interception module (SIM), namely QNUHL. We further propose an advanced latch, namely QNURHL, which mainly employs 24 interlocked Celements, to provide complete QNU recovery. Simulations based on the HSPICE tool demonstrate the complete QNU tolerance as well as the reasonable overhead for the QNUHL latch and also show that the QNURHL latch is recoverable from any possible QNU and can roughly save silicon area by 11% and power dissipation by 16%, compared to the QNU hardened latch (QRHIL) of the same type. Moreover, an algorithm-based method for the verifications of error tolerance and recovery of latches is proposed. The proposed method can simplify the verifications to demonstrate the capability of node-upset tolerance and/or recovery for latches
Weakly supervised segmentation of leaf symptoms in field conditions
No full textInternational audienceBackgroundCrop diseases can cause significant yield losses. Deep learning models forcomputer vision offers powerful tools to enhance human observation ofplant disease symptoms, for instance by using segmentation models tomark out foliar symptoms. However, the most common and effectivearchitectures rely on a fully supervised learning that requires numerous,costly and often unavailable, pixel-level annotated images.To overcomethis, we focus on weakly supervised segmentation [1]. The principle is togenerate segmentation masks from less informative annotations, such asimage-level labels, in order to train segmentation models with reducedannotation effort
PlantCLEF 2025: Advancing AI-based Multi-Species Plant Identification in Vegetation Quadrats for Supporting Environmental Law and Biodiversity Monitoring
No full textInternational audienceRecent developments in European environmental legislation, such as Directive (EU) 2024/1203 on environmental protection through criminal law and Regulation (EU) 2023/1115 on imported deforestation, highlight the need for robust and scalable methods to inventory and monitor plant communities. In this context, vegetation quadrats, rectangular ground plots widely used in ecological surveys, are central to biodiversity monitoring strategies, including forensic applications where reliable species-level identification can provide evidence for law enforcement and compliance monitoring.Traditional quadrat-based inventories are essential for tracking habitat quality, ecosystem change, and anthropogenic impacts. However, they are labor-intensive, costly, and rely on taxonomic expertise that is not always available at scale. This creates a gap between policy demands and operational monitoring capacity. PlantCLEF 2025 (Martellucci et al. 2025) directly addresses this challenge by introducing a new benchmark task: the automatic identification of all species present in high-resolution quadrat images, framing the problem as a multi-label classification challenge.</p
Spatial reconstruction from contact networks: A unified approach for dynamics and genomics
No full textSatellite meeting to ICIAM 2023 TokyoInternational audienceThe physical proximity of genomic sites to each other, within the 3D structure of a chromosome, can be experimentally measured in living cells and represented in the form of a matrix, called a contact map. These maps are very similar to recurrence plots of dynamical systems. Consequently, the same methods can be used for reconstructing the underlying spatial structure, either the chromosome 3D spatial structure or the attractor in the phase space. These methods involve two steps: (1) de- riving a complete distance matrix of the structure using graph distances on a contact network associated with the contact map; (2) reconstructing the structure from this distance matrix, using well-established methods from distance geometry or multidimensional scaling. We here review the different options for implementing the first step, according to the binary, graded or weighted nature of the contact map. We illustrate on three benchmarks (Lorenz model, white noise and EEG data) a novel spectral criterion reflecting the spatial dimension relevant in the second step
A cybernetic avatar system to embody human telepresence for connectivity, exploration, and skill transfer
No full textInternational audienceThis paper describes the cybernetic avatar system developed by Team JANUS for connectivity, exploration, and skill transfer: the core domains targeted by the ANA Avatar XPRIZE competition, for which Team JANUS was a finalist. We used as an avatar a humanoid robot with a human-like appearance and shape that is capable of reproducing facial expressions and walking, and built an avatar control system that allowed the operator to control the avatar through equivalent mechanisms of motion; that is, by replicating the upper-body movement with naturalness and by stepping to command locomotion. In this way, we aimed to achieve high-fidelity telepresence and managed to be well evaluated from the point of view of the operator during the competition. We introduce our solutions to the integration challenges and present experimental results to asses our avatar system, together with current limitations and how we are planning to mitigate them in future work
Fault Analysis through Body Bias Injection on the FLASH Memory Accelerator of a Microcontroller
No full textInternational audienceThe FLASH interface is a crucial component in modern Microcontrollers (MCUs), serving as an intermediary for transferring instructions between the processor and program memory. Previous studies have demonstrated the effectiveness of Electromagnetic Fault Injection (EMFI) and Laser Fault Injection (LFI) in disrupting the operation of FLASH accelerators, leading to instruction line replay and skip faults. However these studies are limited to the case of sequential code and to a single FLASH interface configuration of the target MCU. In this work, we present an investigation on the impact of Body Bias Injection (BBI) on the FLASH accelerator in a 32-bit MCU.The experiments confirm that BBI can similarly induce instruction line replay and skip faults. A detailed analysis of the fault manifestations under various operational configurations of the FLASH accelerator is provided. The study has also extended the fault model to the case of branch operation (non-sequential code).This research contributes a refined understanding of FLASH interface fault behavior under BBI, and highlights the security implication of the decorrelated design between the Program Counter and the FLASH interface
Run-time Energy-Efficiency Optimization for AI and HPC Workloads
No full textInternational audienceEffective power management is crucial for balancing high performance and environmental impact in the exascale era, particularly for datacenters dominated by massively parallel GPU systems due to the rise of AI. While many strategies rely on deep application knowledge, there is a growing need for application-agnostic approaches. We introduce a node-level power management runtime designed for regular applications, featuring minimal overhead and seamless deployment across any HPC/AI system. Our approach detects, at runtime, repetitive execution patterns via spectral analysis and then traces per-pattern energy consumption. A simple gradient-descent optimizer gradually adjusts the GPU frequency until the least per-pattern energy (i.e., maximum energy efficiency) is found. With this approach, we demonstrate up to a 15% reduction in energy consumption for equivalent computational tasks, with no overhead and minimal impact on execution time. This solution has been validated across a diverse range of AI applications, and we discuss the resulting energy savings