Portail HAL Ensta
Not a member yet
11080 research outputs found
Sort by
Approche Hamilton-Jacobi pour les jeux différentiels avec des contraintes d’état et méthodes numériques d’apprentissage pour des problèmes de commande optimale
No full textThis thesis will focus on the study of a theoretical and numerical approach for the multi-objective control problems with state constraints. Multi-objective optimization is an important approach for modelling complex problems in order to analyse the balance between different criteria to minimize. Here, the approach that will be used is based on the theory of Hamilton-Jacobi equations. The goal is to introduce a new methodology to study the properties and compute the Pareto front for multi-objective problems using the value function of an optimal control problem.La thèse aura pour objectif d'étudier une approche théorique et numérique pour l'optimisation multi-objectif de trajectoires avec contraintes sur l’état. L'optimisation multi-objectif est une approche importante pour modéliser des problèmes complexes dans le but d’analyser le compromis entre différents critères à minimiser. Ici, l’approche qui sera utilisée est basée sur la théorie des équations de Hamilton-Jacobi. Le but est d’introduire une nouvelle méthodologie pour étudier les propriétés et calculer le front de Pareto pour les problèmes multi-objectif en utilisant la fonction valeur d’un problème de commande optimale
Feasibility interval and sustainable scheduling simulation with CRPD on uniprocessor platform
No full textInternational audienceThe use of hardware caches became essential in modern embedded systems to address the speed gap between processor and memory. In such systems, cache-related preemption delay (CRPD) may represent a significant proportion of task execution time. Addressing this delay in scheduling simulation of these systems stays an open and under-examined problem. Assumptions are often made to simplify the computation model used in simulation and capture the worst-case effect. Nevertheless, they can introduce situations in which scheduling simulation is considered not only pessimistic but also non-sustainable. In this article, we discuss the problem and propose a less pessimistic CRPD computation model that allows sustainable scheduling simulation regarding the capacity parameter. With the proposed model, a system that is schedulable with simulated worst-case execution times remains so when these parameters are reduced. These results improve the applicability of scheduling simulation in the early verification stage for systems with caches. Experiments conducted with our CRPD computation model show a 5% to 12% improvement of schedulability task set coverage and a 30% to 50% reduction of preemption cost with regard to existing CRPD computation models. An integration in a scheduling simulator and a performance evaluation are also realized for the proposed model
Assessing inter-annotator agreement from collaborative annotation campaign in marine bioacoustics
No full textInternational audienceIt is currently widely recognized that automated methods are crucial to help processing long-term recordings of marine bioacoustics. To evaluate the efficiency of such methods, it is essential to develop large-scale annotated datasets. However, besides being laborious and resource intensive, recent studies have suggested that such a task could also be highly subjective with the generation of annotator specific errors. In this work, we investigate the question of inter-annotator agreement from a multi-annotator annotation campaign performed on a marine bioacoustics dataset. After providing quantitative evidence of inter-annotator variability, we investigate potential sources on both the user annotation practice and the annotation data and task to better understand why and how such variability occurs. Our study reveals that the acoustic event type, the Signal-to-Noise Ratio of the acoustic event and the annotator profile are three examples of critical factors impacting the annotation results of a multi-annotator campaign
Naturally Constrained Online Expectation Maximization
No full textInternational audienceWith the rise of big data sets, learning algorithms must be adapted to piece-wise mechanisms to tackle large-scale calculations' time and memory costs. Furthermore, for most learning embedded systems, the input data are fed sequentially and contingently: one by one, and possibly class by class. Thus, learning algorithms should not only run online but cope with time-varying, non-independent, and non-balanced training data for the system's entire life. Online Expectation-Maximization is a well-known algorithm for learning probabilistic models in real-time, due to its simplicity and convergence properties. However, these properties are only valid in the case of large, independent and identically distributed samples. In this paper, we propose to constrain the online Expectation-Maximization on the Fisher distance between the parameters. After presenting the algorithm, we make a thorough study of its use in Probabilistic Principal Components Analysis. First, we derive the update rules, and then we analyze the effect of the constraint on major problems of online and sequential learning: convergence, forgetting and interference. Furthermore, we use several algorithmic protocols: iid vs sequential data, and constraint parameters updated step-wise vs class-wise. Our results show that this constraint increases the convergence rate of online Expectation-Maximization, decreases forgetting and slightly introduces positive transfer learning
Global optimization approach for the ascent problem of multi-stage launchers
No full textInternational audienceThis paper deals with a problem of trajectory optimization of the flight phases of a three-stage launcher. The aim of this optimization problem is to minimize the consumption of ergols that is need to steer the launcher from the Earth to the GEO. Here we use a global optimization procedure based on Hamilton-Jacobi-Bellman approach and consider a complete model including the transfer from the GTO to the GEO orbit. The Hamilton-Jacobi approach proposed here takes also into account parametric optimisation that appears in the flight phases. The work presented in this paper has been performed in the frame of CNES Launchers' Research and Technology progra
Dynamic response of shallow foundations on reinforced soil with rigid inclusions
No full textInternational audienc
Determination of Creep Behaviour of Adhesively Bonded Assembly -Application to Adhesively Bonded Steel Fasteners
No full textInternational audienceThe use of adhesively bonded assembly is clearly justified for the case of all-FRP structures. Yet, there are still some issues related to the appraisal of the durability for such assembly. This article presents some investigations led on adhesively bonded connections in the case of steel fasteners bonded to steel plate. The studied solution was developed by Cold Pad to propose alternative assembly to welding or bolting (Figure 1). This allows avoiding heavy on-site operations and is particularly well adapted to applications requiring cold working. In addition, it prevents from local steel material fragilization, residual stresses creation, or geometrical stress concentration that may decrease the structure's life expectancy. The understanding of creep behavior may help in a greater appraisal of long-term behavior of bonded solutions in other cases, such as all-FRP structures for instance. To be able to investigate the creep behavior of the developed solution, both experimental and numerical investigations were carried out. The experimental investigations were led on real scale assembly, at different stress levels, and under different load situations. The fastener may indeed be submitted to either predominant tension load, or shear load. These investigations led at different load levels allowed obtaining failure modes, repeatability, time to failure data, but also, local displacements evolution with time. The results revealed a non-linear evolution of those displacements closed to a Burger law. This model was thus chosen, and an analytical determination of creep parameters was realized. This was compared to finite element investigations to verify the adequacy of the proposed methodology. The good suitability of the modelling approach is demonstrated, and the dependency of the parameters with stress level is highlighted. In addition, the finite element investigations allow giving insight of internal stresses evolution during creep
Stability and Convergence Analysis of Time-domain Perfectly Matched Layers for The Wave Equation in Waveguides
No full textInternational audienceThis work is dedicated to the proof of stability and convergence of the Bérenger's perfectly matched layers in the waveguides for an arbitrary L ∞ damping function. The proof relies on the Laplace domain techniques and an explicit representation of the solution to the PML problem in the waveguide. A bound for the PML error that depends on the absorption parameter and the length of the PML is presented. Numerical experiments confirm the theoretical findings
