Portail HAL Ensta
Not a member yet
11080 research outputs found
Sort by
Stochastic incremental mirror descent algorithms with Nesterov smoothing
No full textInternational audienceWe propose a stochastic incremental mirror descent method constructed by means of the Nesterov smoothing for minimizing a sum of finitely many proper, convex and lower semicontinuous functions over a nonempty closed convex set in a Euclidean space. The algorithm can be adapted in order to minimize (in the same setting) a sum of finitely many proper, convex and lower semicontinuous functions composed with linear operators. Another modification of the scheme leads to a stochastic incremental mirror descent Bregman-proximal scheme with Nesterov smoothing for minimizing the sum of finitely many proper, convex and lower semicontinuous functions with a prox-friendly proper, convex and lower semicontinuous function in the same framework. Different to the previous contributions from the literature on mirror descent methods for minimizing sums of functions, we do not require these to be (Lipschitz) continuous or differentiable. Applications in Logistics, Tomography and Machine Learning modelled as optimization problems illustrate the theoretical achievements
State-Constraint Transition: A Language for the Formal Specification of Dynamic Cyber-System Requirements
No full textInternational audienceExisting formal languages for the specification of self-adaptive cyber-physical systems focus on re-configuring the system-to-be depending on its current context, to satisfy the user’s requirements, that is by dynamically composing the software’s structure and behavior. While these approaches specify context-sensitive requirements, they rarely consider their run-time dynamic and scalable nature. The State-Constraint Transition (SCT) modeling language, introduced in this paper, provides an answer to the problems linked to the specification of dynamic requirements by introducing the concept of configuration states, in which requirements are translated into constraints. The expressiveness of existing approaches is thus extended, combining the ease of use of well-established notations, notably those based on characteristics, and those based on Finite-state Machines (FSM), with the computational power and expressiveness of the constraint programming approach. The paper briefly presents the results of the preliminary evaluation, which assesses the expressiveness, scalability, and domain independence of the SCT language
The interplay between Fano and Fabry–Pérot resonances in dual-period metagratings
No full textInternational audienceWe study theoretically and numerically the occurrence of Fano resonances in a metagrating made of slits with some symmetry breaking resulting in a dual period. At low frequency, a grating composed of long enough slits supports Fabry–Pérot resonances on which Fano resonances superimpose when the grating acquires dual period. The resulting spectrum exhibits flat-banded peaks interrupted by sharp dips with successions of perfect and zero transmissions. To model these scattering properties, homogenization theory is used resulting in an effective problem governing the solutions in the two, non-identical, slits, which are coupled through jump conditions at the grating interfaces. These jumps efficiently encode the effect of the evanescent field able to resonate in the radiative region due to the folding of the spoof plasmon polaritons branches. The model is validated with direct numerics and a local analysis allows us to characterize the resonant mechanisms
Algorithmes de perception autonomes pour une meute de robots sous-marins : stratégie de coordination à partir des images acquises par caméras
No full textIn recent years, multi-AUV systems are getting increasing attentions in the ocean exploration because of the potential advantages, such as high efficiency and high tolerance to the low-performance and low-cost sensors. Thus, formation control becomes a new research hotspot. Different to the land or aerial multi-agent systems, communication difficulty in the water is a big challenge to the multi-AUV system. Since the main acoustic communication is limited by bandwidth and noises, scholars begin paying attentions on the underwater optical communication. It can reduce the difficulty of acoustic communication. Its short working range can be extended by associating with multi-AUV systems, but the relevant studies are few. Therefore, the purpose of this work is to achieve the formation building based on the on-board camera for the multi-AUV system. First, we analyze the feasibility and effectiveness of underwater camera sensing information and study the formation building mechanism based on the local positions got from cameras. Then a coordination strategy based on the on-board cameras is proposed, including two parts: designing a robust planar pyramid formation and developing two kinds of formation control methods: local position-based control method and asynchronous discrete consensus-based formation control method with local information. Finally, the proposed new coordination strategy is tested in the software (Blender and Matlab) with multi-AUV systems and in the built limited-communication indoor environment (similar to the underwater environment) with NAO robots. After repeating simulations and experiments, the feasibility and stability of coordination strategy is verified.Au cours des dernières années, les systèmes multi-AUV ont été de plus en plus étudiés dans le domaine de l'exploration des océans en raison de leurs avantages, leur efficacité, leur faible performance et leur haute tolérance aux capteurs à faible coût. Le contrôle de formation est devenu un nouvel axe de recherche. Contrairement aux systèmes multiagents terrestres ou aériens, les difficultés de communication sous-marine constituent un défi majeur pour ces systèmes. Les communications acoustiques étant majoritairement limitées par la bande passante et le bruit, les communications optiques peuvent constituer une alternative. Leur courte portée peut être compensée par l’association de plusieurs robots. L’objectif de cette thèse s’est donc focalisé sur la formation en meute à l’aide des caméras fixées sur les robots. Premièrement, on a analysé la faisabilité et l’efficacité d’un traitement d’information basé sur les images acquises par les robots et on a étudié les mécanismes de construction de la formation basés sur des positions locales obtenues par traitement d’images. Ensuite, une stratégie de coordination comprenant deux volets a été proposée : la conception d’une formation pyramidale plane robuste et l’étude de deux méthodes de contrôle de la formation, une méthode de contrôle reposant sur des positions locales et une autre faisant intervenir un algorithme de consensus asynchrone discret. Les méthodes proposées sont évaluées en simulation sous Blender et Matlab ainsi qu’à travers des expérimentations en environnements intérieurs avec canaux de communication limités avec des robots NAO. Les simulations et les expériences répétées ont permis de vérifier la faisabilité et la stabilité de la stratégie de coordination proposée
Navigation sous-marine par la méthode des cycles stables
No full textNational audienceComment naviguer sans se perdre dans un environnement sous-marin où il n'existe pratiquement aucun point de repère pour la localisation et qu'aucun système de localisation externe n'est disponible ?Pour répondre à cette question, nous proposons le concept de "cycle stable" utilisé par de nombreux animaux pour la navigation et également par les anciens navigateurs. Le principe est de rebondir sur des isobathes (ou autre route maritime) de façon déterministe en répondant à une séquence précise. Cette séquence est déterminée de façon à ce que le robot visite la zone demandée et soit capable de revenir.Quelques expériences réelles montrent la faisabilité de cette approche pour la navigation sous-marine
Spatio-temporal constrained zonotopes for validation of optimal control problems
No full textInternational audienceA controlled system subject to dynamics with unknown but bounded parameters is considered. The control is defined as the solution of an optimal control problem, which induces hybrid dynamics. A method to enclose all optimal trajectories of this system is proposed. Using interval and zonotope based validated simulation and Pontryagin's Maximum Principle, a characterization of optimal trajectories, a conservative enclosure is constructed. The usual validated simulation framework is modified so that possible trajectories are enclosed with spatio-temporal zonotopes that simplify simulation through events. Then optimality conditions are propagated backward in time and added as constraints on the previously computed enclosure. The obtained constrained zonotopes form a thin enclosure of all optimal trajectories that is less susceptible to accumulation of error. This algorithm is applied on Goddard's problem, an aerospace problem with a bang-bang control
Modelling of the Thermophysical Properties of a-tocopherol + ethanol + n-pentane and a-tocopherol + gasoline mixtures
No full textInternational audienc
Modal estimation in underwater acoustics by data-driven structured sparse decompositions
No full textInternational audienceIn underwater acoustics, shallow water environments act as modal dispersive waveguides when considering lowfrequency sources. In this context, propagating signals can be described as a sum of few modal components, each of them propagating according to its own wavenumber. Estimating these wavenumbers is of key interest to understand the propagating environment as well as the emitting source. To solve this problem, we proposed recently a Bayesian approach exploiting a sparsity-inforcing prior. When dealing with broadband sources, this model can be further improved by integrating the particular dependence linking the wavenumbers from one frequency to the other. In this contribution, we propose to resort to a new approach relying on a restricted Boltzmann machine, exploited as a generic structured sparsity-inforcing model. This model, derived from deep Bayesian networks, can indeed be efficiently learned on physically realistic simulated data using well-known and proven algorithms
ARCH-COMP21 Category Report: Continuous and Hybrid Systems with Nonlinear Dynamics
No full textInternational audienceWe present the results of a friendly competition for formal verification of continuous and hybrid systems with nonlinear continuous dynamics. The friendly competition took place as part of the workshop Applied Verification for Continuous and Hybrid Systems (ARCH) in 2021. This year, 5 tools Ariadne, CORA, DynIbex, JuliaReach and Kaa (in alphabetic order) participated. These tools are applied to solve reachability analysis problems on five benchmark problems, two of them featuring hybrid dynamics. We do not rank the tools based on the results, but show the current status and discover the potential advantages of different tools
Multibeam outlier detection by clustering and topological persistence approach, ToMATo algorithm
No full textInternational audienceThe datasets acquired during hydrographic surveys contain outliers, i.e., soundings that do not describe the sea bottom. Many algorithms are developed to identify them. Here, we study unsupervised non-parametric algorithms with a densitybased approach. These algorithms make no assumption about the data and identify outliers as the data furthest away from their neighbors. We asses the ToMATo method developed by INRIA in 2009 to detect outlier soundings from multibeam echosounder data. This clustering algorithm combines a mode-seeking phase with a cluster merging phase using topological persistence. After the theoretical presentation of the ToMATo algorithm, we evaluate its performance on four data sets representing a wide variety of seabeds. We compare this method with the well-known DBSCAN and LOF algorithms. Finally, we suggest an application of the ToMATo algorithm to multibeam data acquired in extradetection mode, where topological persistence allows to form the most relevant clusters