Portail HAL Ensta
Not a member yet
11080 research outputs found
Sort by
Apprentissage machine pour la propagation troposphérique
No full textInternational audienceDans ce résumé, nous proposons une méthode pour simuler la propagation longue distance basée sur un réseau U-Net. Nous nous focalisons ici sur la prise en compte du relief, et l'obtention du champ sur une verticale à la hauteur de l'émetteur, mais le cadre se veut général. Le réseau est entrainé sur des données artificielles, construite par la méthode split-step wavelet. Les reliefs synthétiques associés sont générés de façon aléatoire à l'aide d'un hypercube latin. Pour l'entrainement, une fonction perte basée sur une combinaison linéaire de la norme L2 et L1 est utilisée de façon à obtenir un estimateur robuste. Enfin des tests sont effectués sur des données IGN pour montrer la précision de la méthode. De plus, une stratégie de réglage fin sur les derniers blocs du réseau permet de l'adapter à d'autres types de reliefs
Training K-means on Embedded Devices: a Deadline-aware and Energy Efficient Design
No full textInternational audienceWith the surge in data production, Machine Learning techniques are now commonly used to build intelligent models. Traditionally, powerful platforms process data collected from endpoint devices. However, to address security threats and minimize communication traffic, models can be learned near endpoint devices, despite their resource shortage. K-means clustering is among the most common machine learning tasks used for embedded applications. Because the system is running on scarce resources, the learning process needs to obey a certain time limit. Even if current implementations of K-means have been optimized for embedded devices, they do not consider running within a predefined time budget. In this paper, we propose a deadline-aware and energy-efficient version of K-means called Embedded K-means (EK-means) 1 , that relies on two main ideas : (1) smartly select the right subset of data to train on to meet the deadline at the expense of the smallest clustering error possible ; (2) by dropping part of the data, slack times are identified and exploited opportunistically to apply Dynamic Voltage and Frequency Scaling techniques (DVFS) so as to decrease the energy consumption of the learning task. EK-means has been built on top of an I/O optimized version of K-means for embedded devices to maintain a low I/O proportion regardless of memory constraints. EK-means allows to cluster data while meeting more than 98% of the deadlines with a loss of 1.43% of clustering quality, and an energy reduction of up to 84.26%
A parallel two-way split-step wavelet method for the tropospheric propagation
No full textInternational audienceModeling the atmospheric long-range propagation is an important step in the development of observation, satellite, or communication systems. This paper presents a parallelized version of the two-way split-step wavelet method, designed to enhance computational efficiency. Numerical experiments in the UHF band are provided to highlight the benefits of the proposed method. In the studied cases, we observe a 25%-60% gain in terms of computation time, depending on the terrain taken into account. The proposed approach is also leveraged to generate a dataset for training a machine-learning model based on the U-Net architecture
All-optical Compton scattering at shallow interaction angles
No full textInternational audienceAll-optical Compton sources combine laser-wakefield accelerators and intense scattering pulses to generate ultrashort bursts of backscattered radiation. The scattering pulse plays the role of a small-period undulator (∼1 µm) in which relativistic electrons oscillate and emit X-ray radiation. To date, most of the working laser-plasma accelerators operate preferably at energies of a few hundreds of megaelectronvolts and the Compton sources developed so far produce radiation in the range from hundreds of kiloelectronvolts to a few megaelectronvolts. However, for such applications as medical imaging and tomography the relevant energy range is 10-100 keV. In this article, we discuss different scattering geometries for the generation of X-rays in this range. Through numerical simulations, we study the influence of electron beam parameters on the backscattered photons. We find that the spectral bandwidth remains constant for beams of the same emittance regardless of the scattering geometry. A shallow interaction angle of 30 • or less seems particularly promising for imaging applications given parameters of existing laser-plasma accelerators. Finally, we discuss the influence of the radiation properties for potential applications in medical imaging and non-destructive testing
Multipolar SAFT-VR Mie Equation of State: Predictions of Phase Equilibria in Refrigerant Systems with No Binary Interaction Parameter
No full textInternational audienc
Path Planning Algorithms For Unmanned Aerial Vehicle: Classification, Performance, and Implementation
No full textInternational audiencePath planning and obstacle avoidance form the basis of the UAV operations. The objective of an Unmanned Aerial Vehicle (UAV) is to navigate an optimal path towards its destination while ensuring the avoidance of the obstacles along the way. Several algorithms have been proposed by many researchers to achieve this objective. In this paper, we focused on Global pathplanning algorithms for UAVs with obstacle avoidance. We compare various algorithms by highlighting their characteristics, advantages, and limitations. In addition, this paper implements four of the most well-known methods that tackle environmental challenges. Our results offer practical insights and guidance for researchers seeking to develop more effective path planning algorithms for UAVs
Computing the Radar Cross-Section of Dielectric Targets Using the Gaussian Beam Summation Method
No full textInternational audienceComputing the Radar Cross-Section (RCS) of a given object is a topic of major importance for many applications, e.g., target detection and stealth technology. In this context, high-frequency asymptotic methods are widely used. In this article, we derive a Gaussian Beam Summation (GBS) method for both metallic and dielectric targets. The basic idea is to use the GBS method to compute the scattering far fields generated by the equivalent currents flowing on the surfaces. The validity of the proposed method is then investigated in the X-band. To accomplish this, the results obtained using this technique were compared to those obtained using other sufficiently accurate methods such as the ray tracing of FEKO and the ray asymptotic solution. As an example of the method’s accuracy, the GBS method was used to obtain the wave field in a homogeneous medium by fitting the results to a point source. In the same way, the method was used to compute the RCS of dielectric cuboids
ON SDEs FOR BESSEL PROCESSES IN LOW DIMENSION AND PATH-DEPENDENT EXTENSIONS
No full textInternational audienceThe Bessel process in low dimension (0 ≤ δ ≤ 1) is not an Itô process and it is a semimartingale only in the cases δ = 1 and δ = 0. In this paper we first characterize it as the unique solution of an SDE with distributional drift or more precisely its related martingale problem. In a second part, we introduce a suitable notion of path-dependent Bessel processes and we characterize them as solutions of path-dependent SDEs with distributional drift
Study of consecutive long-lived meter-scale laser-guided sparks in air
No full textInternational audienceWe study the creation and evolution of meter-scale long-lived laser-guided electric discharges and the interaction between consecutive guided discharges. The lifetime of guided discharges from a Tesla high voltage generator is first increased up to several milliseconds by the injection of additional current. The subsequent discharge evolution is measured by recording the electric current and by Schlieren and fluorescence imaging. A thermodynamic model of the gas evolution is developed to explain the discharge evolution. Finally, we analyze the succession of laser-guided discharges generated at 10 Hz