Portail HAL des publications du LIRMM
Not a member yet
13279 research outputs found
Sort by
A Legislation-Aware Robotic Framework for Autonomous Fertilization Near Protected Water Bodies
No full textInternational audiencePutting autonomous mobile robots into practice typically requires the integration of background information such as considering and including both legal as well as regulatory aspects into the process. This document presents a novel AI framework for such an integration. In particular, the framework features (I) a declarative formalization of environmental regulations for an integration into explainable processes using AI, (II) a learning-based method to fuse realtime sensor data from the mobile robot into an implicit spatial map, which can be queried for environmental measurements, and (III) a logical knowledge base, which leverages reasoning to detect and explain violations of the legislation. The resulting process is validated on simulated data as a proof-of-concept. Its context is autonomous fertilization in the agricultural domain near protected water bodies
An In-depth Analysis of the Linguistic Characteristics of Science Claims on the Web and their Impact on Fact-checking
No full textInternational audienceWeb claims, seen as assertions shared on the web and eligible for fact-checking, are at the heart of online discourse. They have been studied extensively on a variety of downstream tasks such as fact-checking, claim retrieval, bias detection, argument mining or viewpoint discovery. On the other hand, claims originating from scientific publications have also been the subject of several downstream NLP tasks. However, research carried out so far has yet to focus on scientific web claims, which are scientific claims made on the web (e.g., on social media and news articles). The process of detecting and fact-checking a claim from the web can be very different depending on whether the claim is scientific or not, thus making it crucial for the developed datasets, methods, and models to make a distinction between the two. With this work, we aim at understanding what makes this distinction necessary, by understanding the linguistic differences between scientific and non-scientific claims on the web, and the impact those differences have on existing downstream tasks. To do so, we manually annotate 1,524 web claims from established benchmarks for fact-checking-related tasks, and we run statistical tests to analyze and compare linguistic features of each group. We find that scientific claims on the web use more analytical speech, but also use more sentiment-related speech, more expressions of physical motion, and have distinct parts of speech (PoS) and punctuation styles. We also conduct experiments showing that BERT-based language models perform worse on scientific web claims by up to 17 F1 points for several downstream tasks. To understand why, we develop a novel methodology to map predictive tokens of language models to explainable linguistic features and find that language models fail to detect a specific subset of predictive features of scientific web claims. We conclude by stating that language models aimed at studying scientific web claims ought to be trained on scientific web discourse, as opposed to being trained only on generic web discourse or only on scientific text from scientific publications
Acyclic, star and injective colouring: A complexity picture for H-free graphs
No full textInternational audienceA (proper) colouring is acyclic, star, or injective if any two colour classes induce a forest, star forest or disjoint union of vertices and edges, respectively. The corresponding decision problems are Acyclic Colouring, Star Colouring and Injective Colouring. We give almost complete complexity classifications for Acyclic Colouring, Star Colouring and Injective Colouring on H-free graphs (for each of the problems, we have one open case). Moreover, we give full complexity classifications if the number of colours k is fixed, that is, not part of the input. From our study it follows that for fixed k, the three problems behave in the same way, but this is no longer true if k is part of the input. To obtain several of our results we prove stronger complexity results that in particular involve the girth of a graph and the class of line graphs of multigraphs
Low-Rank Equilibrium Propagation: An Online Incremental Learning Architecture for Analog-Based Hardware Accelerators
No full textInternational audienceEdge machine learning is emerging as a fundamental technology to address the rapid growth of smart devices and sensors. However, deploying deep learning at the edge remains difficult due to limited memory and computational capacity. Among analog processing solutions, Equilibrium Propagation (EP) has emerged as a promising alternative to backpropagation, offering the potential for significant gains in embedded systems. Yet, its practical implementation remains challenging.In this work, we introduce a novel method to address the challenges of EP by proposing a low-rank approximation of the EP algorithm. This approach enables efficient, end-to-end online incremental training on analog neuromorphic accelerators, which is essential for handling the aging of analog devices, imprecise programming of memristors, and other hardware imperfections.While EP offers a unified forward and backward process, its standard implementation requires storing gradients for each device, leading to substantial overhead in both area and power. Our Low-Rank Equilibrium Propagation (LOREP) scheme mitigates the need for high-precision gradient storage and reduces read–write cycles by approximating gradients with low-rank factors. Experimental results on two popular datasets show that LOREP recovers 2–3% of the lost accuracy while requiring less than 5% of the gradient capacitors needed for full-scale training, highlighting its potential for deployment in resource-constrained environments
FAIRfest: Celebrating advancements of FAIR solutions in EOSC - post event report
No full textFAIRfest: A Festival of FAIR Solutions in the European Open Science LandscapeFAIR-IMPACT and FAIRCORE4EOSC partners were pleased to invite the broad European research community to FAIRfest, a festival that celebrated advancements in FAIR (Findable, Accessible, Interoperable, and Reusable) solutions within the European Open Science research landscape. Held in The Hague (The Netherlands) alongside the 19th International Digital Curation Conference (IDCC25), the event brought together European and international participants to explore the latest achievements in FAIR: Persistent Identifiers and Knowledge Graphs for Findability, Semantic Artefacts for Accessibility, technical and legal aspects for Interoperability, and certification, metrics, and guidelines for Reusability. The festival featured a vibrant Marketplace where adopters and implementers showcased FAIR-enabling solutions, tools, methodologies, and practices through stands, coffee tables, and posters. Attendees engaged directly with experts, asked questions, and learned from those already applying FAIR in real research contexts. Demonstrations of FAIRCORE4EOSC components were presented, and participants heard implementation stories from teams supported by the FAIR-IMPACT programme. The event also offered opportunities to meet FAIR Champions, Ambassadors, and interoperability specialists, creating a dynamic environment for exchange and collaboration. More info and a video of the FAIRfest can be found on the event webpage: https://www.fair-impact.eu/events/fair-impact-events/fairfest-celebrating-advancements-fair-solutions-eos
D1.4 - FAIR-IMPACT Sustainability Plan: Expanding FAIR solutions across EOSC - D1.4
No full textFAIR-IMPACT: Expanding FAIR solutions across EOSC ran from June 2022 to May 2025 and included 28 partners from 11 countries. The project sought to build on the successful practices, policies, tools, and technical specifications arising from FAIRsFAIR, other Horizon 2020 projects and initiatives, and from the EOSC Association Task Forces, such as the one on FAIR metrics and data quality. By enabling the FAIR principles for EOSC across scientific communities, stakeholder groups, and research outputs at a European, national, and institutional level, FAIR-IMPACT improved access to and management of FAIR research outputs and stimulated the development and uptake of a wide range of innovative services. By translating and implementing FAIR solutions across domains, the project also contributed to multi-disciplinary scientific cooperation. With its focus on increasing FAIRness, FAIR-IMPACT improved public trust in, and reproducibility of, science, transforming the way researchers share and exploit research outputs. The ultimate aim of this was to support better quality, validation, and higher research productivity. This report details the sustainability planning for the 19 project outputs identified as important for EOSC and the global research community.These outputs are categorised as Key Exploitable Results (KER), Key Outputs (KO), and Key Functions (KF), representing a wide range of recommendations, guidelines, policies, best practices, frameworks, tools, and services. 1. Key Exploitable Result (KER): an identified main interesting result6 which has been selected and prioritised due to its high potential to be made useful and derive benefits downstream in the value chain of a product, process or solution, or act as an important input to policy, further research, or education. 2. Key Output (KO): other important project result not selected as a KER which requires sustainability measures during and after the project. 3. Key Function (KF): function performed by the project which requires sustainability measures during and after the project.A summary of the outputs and their sustainability measures is listed in the table below. Whether maintained as static resources or carried forward by project partners and/or community stakeholders, these outputs will help shape policy development and best practices for a viable and FAIR EOSC
Interactive pipeline for mandible reconstruction surgery planning using fibula free flap
No full textInternational audiencePurpose: Mandible reconstruction surgery using fibula free flap is a long and expensive process requiring extensive surgical experience. Indeed, the planning stage, mandible shaping, and therefore osteotomy positioning on the fibula are tedious, often done by hand, and can take months. This delay is unacceptable when mandible deterioration is caused by a time-sensitive disease such as cancer. In this paper, we propose an interactive pipeline for an easy-to-use and time-efficient surgical planning tool tailored to be used directly by the surgeon.Methods: From CT scans of patient’s mandible and fibula, we propose to register a cutting structure to the mandible and to segment and mesh the fibula; then, respecting anatomical constraints (mandible curvature, flap size, vessel preservation, etc.), we generate a surgery plan. Next, in a 3D interactive environment, the surgeon can intuitively shape the mandible by cutting, moving, and modifying bone fragments nondestructively. This stage allows surgeons to express their expertise, and the resulting cutting plane positions are then sent to a robot serving as a cutting guide for the surgery.Results: We demonstrate the efficiency of our method through patient-specific surgery planning for two different pathologic cases. We show our results are comparable to a commercial solution away from cutting guides design.Conclusion: Our proposed pipeline allows for a patient-specific precise planning and to cut down the preoperative planning phase of the mandible reconstruction surgery from days to minutes
Data-driven fuzzy logic control method for improved USV path planning
No full textInternational audienceThis paper proposes a data-driven fuzzy logic control method to improve the Dynamic Window Approach (DWA) for path planning of Unmanned Surface Vehicles (USVs). The proposed method aims to reduce the subjectivity introduced by human factors in the parameter settings of conventional fuzzy logic control. A data-driven dataset was constructed by extracting parameters from conventional fuzzy logic control algorithms, and a fuzzy neural network was employed to derive a new fuzzy logic controller from this dataset. The resulting controller exhibits a more rational distribution of variables compared to conventional fuzzy logic controllers, demonstrating the superiority of controllers generated by neural networks. Numerical simulations show that proposed the data-driven USVs path planning method offers improvements in terms of path length, navigation time, and reduced turning angles
Optimal Bounds for Dissatisfaction in Perpetual Voting
No full textInternational audienceIn perpetual voting, multiple decisions are made at different moments in time. Taking the history of previous decisions into account allows us to satisfy properties such as proportionality over periods of time. In this paper, we consider the following question: is there a perpetual approval voting method that guarantees that no voter is dissatisfied too many times? We identify a sufficient condition on voter behavior ---which we call 'bounded conflicts' condition---under which a sublinear growth of dissatisfaction is possible. We provide a tight upper bound on the growth of dissatisfaction under bounded conflicts, using techniques from Kolmogorov complexity. We also observe that the approval voting with binary choices mimics the machine learning setting of prediction with expert advice. This allows us to present a voting method with sublinear guarantees on dissatisfaction under bounded conflicts, based on the standard techniques from prediction with expert advice
Advanced robust desired compensation adaptive control for parallel robots: From concept to real-time validation
No full textInternational audienceIn this paper, a novel robust adaptive controller is proposed for Parallel Kinematic Manipulators (PKMs). The objective of this research is to merge a model-based adaptive controller with a robust non-model-based controller. The key contribution lies in the redesign of the feedback component of the Desired Compensation Adaptive Law (DCAL) controller using the Robust Integral of Sign of Error (RISE) scheme. This new design enhances the tracking performance compared to the DCAL controller by incorporating robust terms. These terms effectively compensate for non-modeled phenomena that are not handled by the model-based adaptive compensation in the original controller. Moreover, the inclusion of the integral of the error sign motivates the adaptation law redesign to minimize the feedback action instead of combined errors. The stability analysis of the proposed control solution is performed based on Lyapunov's theory. To demonstrate the superiority of the proposed controller, a comparative study is conducted through various real-time experimental scenarios in different operating conditions involving a PKM testbed