Archive ouverte de l'ENAC
Not a member yet
3458 research outputs found
Sort by
The Effects of LCCs Subsidies on the Tourism Industry *
No full textThis paper studies the relationship between air transportation, tourist flows, and subsidies to Low Cost Carriers (LCCs), a policy used by many national and local governments to stimulate tourist arrivals. To test the policy empirically, we use a two-stage empirical model. In the first stage, we estimate a structural model applied to air transport, and in the second stage, we link passenger arrivals to regional tourism flows. In this way, we use exogenous shocks (subsidies to LCCs) in airline supply to analyze the causal link with tourist arrivals. This model is applied to tourist flows from European regions to Italian regions from 2016 to 2018. Our counterfactual analyses consider two regimes for implementing subsidies to LCCs, following the literature coming from Oates (1993, 1999) contributions: a centralized, uniform policy for all regions and a decentralized policy in which subsidies are adopted by a single region. Our simulations reveal that subsidies to LCCs are effective in stimulating tourism, and that a centralized regime is more effective than a decentralized one. In fact, the latter generates externalities in regions that do not implement the subsidy, making the decentralized policy economically sub-optimal and unsustainable
Voronoi diagrams and Simulated Annealing for airspace block optimization
No full textInternational audienceThis paper investigates the design of airspace blocks using Voronoi diagrams and simulated annealing. This approach aims to optimize the layout of airspace blocks by minimizing the complexity gap between them. The algorithm is tested with different complexity metrics. By using Voronoi diagrams, which partition the airspace into regions around specified points, and simulated annealing, which iteratively refines solutions to find near-optimal configurations, the algorithm provides a systematic method for airspace design. The study focuses specifically on the French airspace, providing a real-world application of the proposed methodology. Through experimentation and evaluation, the algorithm demonstrates its ability to generate airspace block configurations that balance complexity. This research contributes to ongoing efforts in airspace management and optimization by providing insights and techniques for designing airspace structures that meet the evolving needs of air traffic control systems
Slot Allocation in a Multi-airport System under Flying Time Uncertainty
No full textInternational audienceSlot allocation in a single airport aims to maximize the utilization of airport-declared capacity under operational and regulation constraints, while that in a multi-airport system (MAS) has to take airspace capacity into account. This is due to the fact that the conflict of using the limited capacity of certain departure/arrival fixes in the terminal airspace could induce unnecessary flight delays. The uncertainty of flying times between the airport and congested fixes makes it even more complicated for slot allocation in a MAS. Traffic flow may exceed capacity when the flying times of flights change. In this paper, the authors propose an uncertainty slot allocation model for a MAS (USAM). The objective of the model is to minimize the total displacement of slot requests in the MAS while considering all of the capacity constraints, as well as the uncertainty of flying time. The constraints of departure/arrival fixes are formulated as chance constraints, and then the Lyapunov theorem is applied for reformulation. The USAM is applied in the MAS of the Guangdong-Hong Kong-Macao Greater Bay Area (GBA). Specifically, the impact of the uncertainty of flying times from five airports to airspace fix YIN is investigated. Results show that the total displacement would increase if the uncertainty of flying time was considered. The optimized schedule using the USAM, however, is more robust and can satisfy capacity constraints under various scenarios
Unscented Kalman Filter using Optimal Quantization
No full textInternational audienceThis paper presents a novel approach to deal with nonlinear filtering by augmenting an Unscented Kalman Filter (UKF) with an Optimal quantization algorithm, named OQ-UKF. The Unscented Kalman Filter uses a sigma-point based method to approximate the distribution of an unknown random variable onto which is applied a nonlinear transformation, providing a cloud of evolving points. However, the generation of these socalled sigma-points is done by a deterministic algorithm which needs tuning in order to accurately capture the distribution of the estimate. This tuning is often problem-dependent due to nonlinearities and sometimes not optimal. We propose to fuse an UKF with Optimal quantization whose objective is to find the best approximation of the density of a random variable. The designed OQ-UKF is described in this paper, and its performance is evaluated for some relevant practical problems, such as pose estimation of a two-dimensional mobile robot.</div
Estimation des paramètres de multitrajets pour des signaux GNSS par apprentissage profond et application en environnement de simulation 3D
No full textThe Global Navigation Satellite System (GNSS) signal reception system is vulnerable to the multipath phenomenon. Ubiquitous in urban environments, it can lead to a significant deterioration in receiver positioning performances. Despite the development of numerous techniques for detecting or reducing its effects such as the Fast Iterative Maximum-Likelihood Algorithm (FIMLA) or particle filters, reducing the multipath impact within a GNSS receiver remains a difficult objective. In this thesis, we propose to apply a Machine Learning algorithm to eliminate the multipath component by estimating four key parameters: its delay, Doppler frequency, amplitude and phase. We will use data in the form of labeled images. These images are representations of the I and Q channels at the receiver correlator outputs. The signals will be synthetic and the multipaths generated according to two different protocols: randomly or collected from a GNSS signal propagation simulator, capable of reproducing situations conducive to multipaths (e.g. simulation of reception in an urban canyon). The information of the signal and its parasitic replica is distributed between the I and Q images constructed by a correlator output generator with a number of outputs flexible in order to generate images with different resolutions. In order to carry out the multipath parameters estimation on I and Q images, a complex task due to the multichannel aspect of the problem, we will use Convolutional Neural Networks (CNN). Used repeatedly on classification problems, they have become a reference in image analysis via Machine Learning algorithms. Although they are less widely used in regression, they nevertheless show good estimation capabilities, as in the case of age estimation or for the estimation of body positions in 3D space. Here we introduce a new convolution neural network model combining the ability of convolution layers to automatically extract highly informative features with a soft-labeling technique that improves the CNN robustness during its generalization phase. The idea behind soft labelling is to reduce the algorithm trust on labels in order to make the regression task easier, given the complexity of image data. Labels are then seen as probability distributions with their means equal to the labels. Inspired by previous work by Imani et al. (2018), we use a loss function specifically tailored to measure dissimilarities between the new labels and their prediction with a Kullback-Leibler divergence. The new CNN model was tested on different image datasets with different noise levels and image sizes, in order to test the algorithm's estimation qualities and robustness. The results show a clear improvement in CNN estimation accuracy under all the conditions tested, using the soft labelling method compared with "conventional" regression on a continuous variable. Additionally, the performance gain provided by the soft labeling method with distributions means that we can work with input image sizes reduced from 80×80 to 20×20 pixels.Le système de réception de signaux Global Navigation Satellite System (GNSS) présente une vulnérabilité face au phénomène de multitrajet. Événement fréquent en milieu urbain, il peut mener à une forte détérioration des performances de positionnement du récepteur. Malgré le développement de nombreuses techniques à des fins de détection ou de réduction de ses effets comme le Fast Iterative Maximum-Likelihood Algorithm (FIMLA) ou les filtres particulaires, la diminution de l'impact du multitrajet au sein d'un récepteur GNSS reste un objectif difficile à atteindre. Dans cette thèse, nous nous proposons d'appliquer un algorithme de Machine Learning pour éliminer la composante du multitrajet par l'estimation de quatre paramètres déterminants : son retard, sa fréquence Doppler, son amplitude et sa phase. Nous utiliserons des données sous forme d'images labellisées. Ces images sont les représentations des canaux I et Q aux sorties du corrélateur. Les signaux seront synthétiques et les multitrajets générés selon deux protocoles différents : aléatoirement ou collectés à partir d'un simulateur de propagation de signaux GNSS, capable de reproduire des situations propices aux multitrajets (ex : simulation de réception dans un canyon urbain). Les informations sur le signal et sa réplique parasite sont réparties parmi les images I et Q des sorties de corrélateur construites par un générateur de sorties de corrélateur. Le nombre de sorties calculées par le générateur est flexible afin de générer différentes résolutions d'images. Dans le but d'effectuer la tâche de régression des paramètres de multitrajets sur les images I et Q, complexe par l'aspect multicanal du problème, nous utiliserons des réseaux de neurones à convolution ou Convolutional Neural Network (CNN). Utilisés à maintes reprises sur des problèmes de classification, ils sont devenus une référence dans l'analyse d'images via des algorithmes de Machine Learning. Malgré une utilisation moins démocratisée en régression, ils montrent pourtant de bonnes capacités d'estimation comme pour l'estimation d'âge ou l'estimation de positions d'un corps dans un espace 3D. Ici, nous introduisons un nouveau modèle de réseau de neurones à convolution combinant la capacité d'extraction automatique des points d'intérêts hautement informatifs par les couches de convolution et une technique de soft labelling améliorant la robustesse du CNN lors de son étape de généralisation. L'idée du soft labelling est de diminuer la confiance de l'algorithme envers les labels afin de rendre la tâche de régression plus aisée, étant complexe pour des données sous forme d'images. Les labels sont alors vus comme des distributions de probabilités d'espérances égales aux labels. Inspirés de travaux antérieurs effectués par Imani et al. (2018), nous utilisons une fonction perte spécifiquement adaptée à la mesure de dissimilarités entre les nouveaux labels et leur prédiction avec la divergence de Kullback-Leibler. Le nouveau modèle de CNN a été testé sur différents datasets d'images présentant différents niveaux de bruits et différentes tailles d'images afin de tester les qualités d'estimation de l'algorithme ainsi que sa robustesse. Les résultats montrent dans toutes les conditions testées une claire amélioration de la précision d'estimation par CNN avec la méthode de soft labelling par rapport à une régression " classique " effectuée sur une variable continue. De plus, le gain de performances apporté par la méthode de soft labelling par distribution (Distribution loss) permet de travailler avec des tailles d'images en entrée réduites de 80×80 à 20×20 pixels
A theoretical study of a radiofrequency wave propagation through a realistic turbulent marine atmospheric boundary layer based on large eddy simulation
No full textInternational audienceThis study aims at modeling and investigating the impact of realistic turbulence inhomogeneity on radiowave propagation by utilizing large eddy simulations (LES). An up-to-date version of the X-LES method for generating realistic turbulent phase screens is first introduced. It combines atmospheric simulations with the classical Tatarski statistical modeling. This method naturally incorporates vertical turbulence inhomogeneity into phase screens at scales resolved by LES. It is then extended to sub-grid scales by weighting statistically generated phase variations with the vertical profile of the turbulent structure constant extracted from atmospheric data. This method is applied to replicate turbulence of a classical tropical marine atmospheric boundary layer. The impact of the generated medium on the propagation of a 10 GHz spherical wave emanating from a Gaussian aperture is analyzed through a statistical study of log-amplitude profiles performed for three different source altitudes. Results first show that contrary to a classical homogeneous turbulence modeling, log-amplitude profiles resulting from the propagation into inhomogeneous turbulence exhibit a statistical heterogeneity strongly dependent of source altitude. Furthermore, classical stochastic phase screen generation from a homogeneous Von-Kármán Kolmogorov spectrum seems to give a statistically significant underestimation of the actual impact of turbulence compared to the X-LES method
JetStick: Using Air Jet Tactile Feedback to Prevent Aircraft Stalls
No full textInternational audienceHaptic enhanced flight sticks have been used to guide pilots for avoiding aircraft stalls. However, they do not inform the pilot about the aircraft flyingstatus during critical flying phases. We present JetStick, a flight stick that provides air jet feedback to help pilots avoiding dangerous situations in critical flightphases. Our preleminary results show that our device can provide tactile information about aerodynamic phenomenon outside the cockpit. This could be useful incontext such aerobatics or training flights
Real Time Aircraft Atypical Approach Detection for Air Traffic Control
No full textInternational audienceIn the rapidly evolving field of air traffic control, the need for innovative methods to improve safety and efficiency hasnever been greater. This paper presents a real-time method for detecting atypical aircraft approaches, a significant leap from traditional off-line methods. Our approach uses real-time data analysis and extends an existing off-line model to dynamically assess aircraft energy behaviour and trajectory during the critical approach and landing phases. Unlike its predecessors, our methodology provides instantaneous detection and resolution capabilities that are criticalto ensure safety and deal with the expected growth in air traffic. By providing air traffic controllers with real-time insight into aircraft approach trajectories, our methodology aims to improve situational awareness and reduce the risk of approach-related incidents
Strategic Path Change Maneuvers for Weather Obstacle Avoidance in Aviation
No full textWeather avoidance algorithms play a crucial role in significantly enhancing aircraft safety during flight operations, particularly in the presence of severe weather conditions. This paper presents a novel obstacle avoidance strategy based on the use of alternative paths to circumvent obstacles during the cruise phase. The objective of this study is to assess the benefits of using strategic information to address tactical avoidance issues. Firstly, the new strategy is simulated in a static and then dynamic space populated with weather obstacles. Subsequently, the proposed strategy is compared to a classical avoidance maneuver in several dynamic simulations, varying the size of the obstacles and the detection range of the radar. The results show that including strategic information on dynamic rerouting can be of significant benefit to both the pilot and air traffic controller, providing a supportive decision-making tool for bad weather avoidance
Air Traffic Flow Management et le problème des traînées de condensation
No full textInternational audienc