Archive ouverte HAL-LAAS
Not a member yet
12189 research outputs found
Sort by
Verifying Properties of Binary Neural Networks Using Sparse Polynomial Optimization
No full text22 pages, 2 figures, 7 tablesInternational audienceThis paper explores methods for verifying the properties of Binary Neural Networks (BNNs), focusing on robustness against adversarial attacks. Despite their lower computational and memory needs, BNNs, like their full-precision counterparts, are also sensitive to input perturbations. Established methods for solving this problem are predominantly based on Satisfiability Modulo Theories and Mixed-Integer Linear Programming techniques, which are characterized by NP complexity and often face scalability issues. We introduce an alternative approach using Semidefinite Programming relaxations derived from sparse Polynomial Optimization. Our approach, compatible with continuous input space, not only mitigates numerical issues associated with floating-point calculations but also enhances verification scalability through the strategic use of tighter first-order semidefinite relaxations. We demonstrate the effectiveness of our method in verifying robustness against both and -based adversarial attacks
Comparison of 2 processes based on NIL and lift-off for the fabrication of thick nano antennas for solar PV-TE applications
No full textInternational audienceCombining photovoltaic cells with thermoelectric generators could enhance solar energy efficiency using a nanostructured photothermal interface based on metal nano antennas. Nanoimprint lithography and lift-off can produce them. We report here on the comparison of two processes we developed and on the successful 150-200 nm thick metal lift off
KeAD: knowledge- enhanced graph attention network for accurate a nomaly detection
No full textInternational audienceAnomaly detection has emerged as one of the core research topics to support workflow applications across domains. To differentiate anomalies from normal patterns of workflows, Graph Neural Networks (GNNs) models have been introduced. These models leverage time series data to construct graph structures, in order to explicitly capture task dependencies among industrial Internet of Things (IoT) devices, and thus, to identify deviations from predicted behaviours as anomalies. However, existing forecasting-based anomaly detection methods may not accurately detect certain anomalies, as they rely solely on historical sensory data while seldom considering the valuable information embedded in domain knowledge. To address this limitation, this paper proposes a Knowledge-enhanced graph attention-based Anomaly Detection (KeAD) method. Specifically, a knowledge-enhanced graph structure is constructed by incorporating domain-specific knowledge to represent spatio-temporal dependencies between IoT devices. Based on which, a knowledge-enhanced graph attention-based forecasting network is developed to predict the future behaviours of IoT devices. Anomalies, such as those caused by cyber-attacks in workflows, are detected by analyzing deviations from these predicted behaviours in conjunction with domain-specific knowledge. A case study is presented, along with extensive experiments conducted on publicly available datasets. Evaluation results demonstrate that KeAD outperforms the state-of-the-art techniques in terms of anomaly detection accuracy
Assessing Human Cooperation for Enhancing Social Robot Navigation
No full textInternational audienceSocially aware robot navigation is a planning paradigm where the robot navigates in human environments and tries to adhere to social constraints while interacting with the humans in the scene. These navigation strategies were further improved using human prediction models, where the robot takes the potential future trajectory of humans while computing its own. Though these strategies significantly improve the robot's behavior, it faces difficulties from time to time when the human behaves in an unexpected manner. This happens as the robot fails to understand human intentions and cooperativeness, and the human does not have a clear idea of what the robot is planning to do. In this paper, we aim to address this gap through effective communication at an appropriate time based on a geometric analysis of the context and human cooperativeness in head-on crossing scenarios. We provide an assessment methodology and propose some evaluation metrics that could distinguish a cooperative human from a non-cooperative one. Further, we also show how geometric reasoning can be used to generate appropriate verbal responses or robot actions
Human–Robot Interaction: Successes, Hurdles, and Remaining Challenges [Opinion]
No full textInternational audienceThe past decades have seen an increasing number of robots deployed in the vicinity of humans, from vacuum cleaners roaming in our living rooms, drones flying over our heads, to prostheses attached to our bodies. Today, global efforts are focused on designing the next generation of robots, which will be employed and function in close or direct interactions with lay users. We are no longer in the realm of factory robots used by well-trained practitioners. It is, hence, not conceivable that these robots can be programmed without a deep understanding of the social, ethical, and cultural rules that underpin human environments. Developing robots that are cognizant of the world that surrounds them has led to a wide range of efforts worldwide, all of which fall under the general field of human-robot interaction (HRI). We advocate for a research roadmap for HRI over the next two decades towards the development of robot systems capable of interacting with humans in a pertinent and helpful manner in any kind of environment
Optimisation combinatoire et programmation par contraintes pour les missions spatiales : transferts de données, observations scientifiques et ordonnancement des opérations
No full textNational audienceSpace missions involve increasingly complex systems operating in an environment where resources such as communication or power are limited. Designing and operating a spacecraft demands a huge investment of time and money—the lifetime of a space mission can span decades and such projects can cost up to a billion euros—and the slightest mistake can have dramatic consequences. Many decisions on the ground segment, particularly data transfer and observation scheduling, are currently handled manually or semi-manually by teams of experts in collaboration with scientists. This process is time-consuming, prone to suboptimal solutions, and can lead to missed scientific opportunities when optimal schedules cannot be found.In this thesis, we explore how combinatorial optimization can enhance the offline scheduling processes of space missions. More precisely, we study data transfer problems coming from the past Rosetta mission but we also define new scheduling problems concerning the under-development NIMPH nanosatellite mission and ESA's candidate mission M-MATISSE. We provide a complexity analysis for all these problems, proving that they are all NP-hard, and present new methods—primarily Constraint Programming, alongside Mixed-Integer Linear Programming and heuristics—to tackle them.We evaluate our models and heuristics on both real scenarios and realistic synthetic instances against the current approaches to solve the problems encountered in this thesis. Our results show that our approaches can provide high-quality and sometimes optimal solutions on these difficult problems, thus reducing the effort of human operators while increasing the science return of the missions.Les missions spatiales impliquent des systèmes de plus en plus complexes, évoluant dans un environnement où des ressources telles que les communications ou l’énergie sont limitées. Concevoir et opérer ces missions demande un investissement considérable en temps et en argent—la durée de vie d’une mission spatiale peut s’étendre sur plusieurs décennies, et de tels projets peuvent coûter jusqu’à un milliard d’euros—et la moindre erreur peut avoir des conséquences dramatiques. De nombreuses décisions relatives au segment sol, en particulier le transfert de données et la planification des observations, sont le plus souvent prises manuellement ou semi-manuellement par des équipes d’experts en collaboration avec l'équipe scientifique. Ce processus est long, produit des solutions sous-optimales, et peut amener à réduire le retour scientifique de la mission.Dans cette thèse, nous explorons comment l’optimisation combinatoire peut améliorer le processus de planification des missions spatiales. Plus précisément, nous étudions des problèmes de transfert de données issus de la mission Rosetta, aujourd'hui terminée, mais nous définissons également de nouveaux problèmes d’ordonnancement relatifs à la mission nanosatellite NIMPH, actuellement en développement, ainsi qu’à la mission candidate de l’ESA, M-MATISSE. Nous proposons une analyse de complexité pour l’ensemble de ces problèmes, en démontrant qu’ils sont tous NP-difficiles, et nous présentons de nouvelles méthodes principalement basées sur la programmation par contraintes, mais aussi sur la programmation linéaire en nombres entiers et sur des heuristiques pour les résoudre.Nous évaluons nos modèles et heuristiques sur des scénarios réels ainsi que sur des instances synthétiques réalistes, en les comparant aux approches actuellement utilisées pour résoudre les problèmes abordés dans cette thèse. Nos résultats montrent que nos méthodes permettent d’obtenir des solutions de très bonne qualité (parfois optimales) à ces problèmes difficiles, réduisant ainsi l’effort des opérateurs humains tout en augmentant le retour scientifique des missions
Inclusion constants for free spectrahedra with applications to quantum incompatibility
No full textInternational audienceBuilding on the matrix cube problem, inclusions of free spectrahedra have been used successfully to obtain relaxations of hard spectrahedral inclusion problems. The quality of such a relaxation is quantified by the inclusion constant associated with each free spectrahedron. While optimal values of inclusion constants were known in certain highly symmetric cases, no general method for computing them was available. In this work, we show that inclusion constants for Cartesian products of free simplices can be computed using methods from non-commutative polynomial optimization, together with a detailed analysis of the extreme points of the associated free spectrahedra. This analysis also yields new closed-form analytic expressions for these constants. As an application to quantum information theory, we prove new bounds on the amount of white noise that incompatible measurements can tolerate before they become compatible. In particular, we study the case of one dichotomic and one k-outcome measurement, as well as the case of four dichotomic qubit measurements
Design of Multimode Hybrid Plasmonic waveguide for Refractometry
No full textInternational audienceWe investigate a multimode Si 3 N 4 -based hybrid plasmonic waveguide measuring less than 2 µm in length. Increasing Au slot thickness enhances sensitivity, achieving 780 nm/RIU in water medium, enabling high-sensitivity optical biosensing in integrated plasmonic sensors
Average current mode control for a GaN-based single-phase 2-level DAB converter
No full textInternational audienceDual Active Bridge (DAB) converters are widely employed in electric vehicles charging equipment, both in Onboard Chargers (OBCs) and offboard charging infrastructure, either in Grid-to-Vehicle (G2V) or Vehicle-to-Grid (V2G) power flow. However, due to the unavoidable mismatches in device parameters, a DC bias current can bring the transformer core to saturation, compromising the converter performance. Developing a means with a dedicated current control to eliminate the parasite DC current part is therefore essential for the structure to avoid operating points in magnetic saturated regions. This work aims, through MATLAB/Simulink simulations, to design a current mode control law with its low steady-state error reducing the DC parasite current. Operating in high switching frequency, GaN or SiC devices could represent today the best option available on the market for enhancing power density, efficiency and dynamic performance of the converter
Interacting Kalman Filters for Linear Systems with Coupling Based on Empirical Covariances
No full textInternational audienceThe problem of designing estimators for stochastic linear systems with distributed observations is considered. Each observation process is associated to a node in an undirected graph, which is used to compute a local estimate at the node. The injection gain used in the filter at each node is obtained from the empirical covariance of all the estimates available at that node. This way the underlying idea comes from the theory of ensemble filtering and we analyze the evolution of the coupled covariances over the entire graph. After providing a detailed derivation of the evolution of covariance matrix, we observe that the drift term in the differential equation for the coupled covariances has some inherent stability structure in case of regular graphs, which leads to the fluctuations around the steady state. We provide an illustration of our algorithm on an academic example while comparing it with centralized and ensemble Kalman filters