Archive ouverte de l'ENAC
Not a member yet
3458 research outputs found
Sort by
Modelling and Hovering Stabilisation of a Free-Rotating Wing UAV
No full textInternational audienceWe propose a multibody model of a freewing UAV. This model allows obtaining simulations of the UAV's behaviour and, in the future, to design a control law stabilising the entire flight envelope (hovering and forward flight). We also describe the realisation of a prototype and a comparison of possible methods for estimating the UAV's states. With this prototype, we report on experimental hovering flights with a non-linear incremental dynamic inversion controller to stabilise the wing and a proportional derivative controller for the fuselage stabilization
ArGaze: An Open and Flexible Software Library for Gaze Analysis and Interaction
No full textGaze analysis has evolved into a mature technique with diverse applications enabling theinvestigation of various human activities and cognitive processes such as reading, visualattention, memory as well as real time gaze interaction. However, existing proprietarysoftware or librairies often lacks the flexibility needed to incorporate emerging gaze metricsor real time processing, limiting researchers to predefined methodologies.To address these limitations, we introduce ArGaze, an open and flexible softwarelibrary designed to provide a unified and modular approach to gaze analysis or gazeinteraction. ArGaze facilitates real-time and/or post-processing analysis for bothscreen-based and head-mounted eye tracking systems. By offering a wide array of gazemetrics and supporting easy extension to incorporate additional metrics, ArGazeempowers researchers and practitioners to explore novel analytical approaches efficiently.This paper first reviews existing software solutions and their functionalities beforedelving into the design principles and features of the ArGaze library. Subsequently, threeuse cases demonstrate ArGaze’s efficacy in processing eye data from screen-based andhead-mounted eye trackers, showcasing its versatility and applicability across differentresearch and practical contexts
Theoretical aspects of robust SVM optimization in Banach spaces and Nash equilibrium interpretation
No full textInternational audienceThere are many real life applications where data can not be effectively represented in Hilbert spaces and/or where the data points are uncertain. In this context, we address the issue of binary classification in Banach spaces in presence of uncertainty. We show that a number of results from classical support vector machines theory can be appropriately generalized to their robust counterpart in Banach spaces. These include the representer theorem, strong duality for the associated optimization problem as well as their geometrical interpretation. Furthermore, we propose a game theoretical interpretation of the class separation problem when the underlying space is reflexive and smooth. The proposed Nash equilibrium formulation draws connections and emphasizes the interplay between class separation in machine learning and game theory in the general setting of Banach spaces
Comparison of performance between PMS and trombone arrival route topologies in terminal maneuvering area
No full textInternational audienceThe contradiction between air traffic capacity and demand exerts congestion and delay, resulting in airspace operating at or above its capacity. This paper proposes a performance comparison of two arrival route topologies in Terminal Maneuvering Area (TMA) with a tailored optimization algorithm, including Point Merge System (PMS) and trombone paradigms. We aim at enhancing the capacity of arrival topologies and efficiently landing more aircraft through identifying which topological route is better suitable for current traffic demands. The problem involves integrating and sequencing arrival operations into the TMA. A mathematical model is developed for which entry times, entry speeds and flight routes are regarded as decision variables. The operation constraints of this optimization modelling correspond to detection of potential conflicts between consecutive aircraft at merging waypoints and on the same link segments, based on the safety separation standards. The objective function is set to minimize the delay and speed deviation for all flights in the optimized system. Because of the highly combinatorial nature of this problem and the non-deterministic polynomial hard (NP-hard), a tailored selective simulated annealing algorithm is proposed to optimally make decisions for the arrival traffic flow. Finally, the validation of our approach has been performed for Ataturk Airport in Turkey, by using 11 different operation scenarios of arrival demands. For lower demands, all aircraft can be positively scheduled, while for more than 80% extra scenarios based on the regular traffic demand, there exist some remaining conflicts for the situations. The results illustrate that, compared with the trombone procedure, PMS paradigm could enhance the arrival capacity and has better performance throughout the arrival procedure. Additionally, more flexible sequence position shift is possible in PMS, and it has the ability to accommodate high-density flight operating environment
A Deep Learning Approach To Predict General Aviation Traffic Counts
No full textInternational audienceGeneral Aviation traffic prediction is a major concern for Air Navigation Service Providers with a direct impact on air traffic flow and capacity management measures. This paper introduces a Deep Learning methodology using meteorological and calendar data to predict General Aviation traffic. The methodology is evaluated in great detail using historical data from the Nice Cote D'Azur Terminal Control Center sectors with an increase of the global prediction performance of 32% with Recurrent Neural networks-based models compared to current tools used in operation. Additional tools are finally proposed to analyze and attain an in-depth understanding of the predictions generated by the various models
HLoOP - Hyperbolic 2-space Local Outlier Probabilities
No full textInternational audienceHyperbolic geometry has recently garnered considerable attention in machine learning due to its capacity to embed hierarchical graph structures with low distortions for further downstream processing. This paper introduces a simple framework to detect local outliers for datasets grounded in hyperbolic 2-space referred to as HLoOP (Hyperbolic Local Outlier Probability). Within a Euclidean space, well-known techniques for local outlier detection are based on the Local Outlier Factor (LOF) and its variant, the LoOP (Local Outlier Probability), which incorporates probabilistic concepts to model the outlier level of a data vector. The developed HLoOP combines the idea of finding nearest neighbors, density-based outlier scoring with a probabilistic, statistically oriented approach. Therefore, the method consists in computing the Riemmanian distance of a data point to its nearest neighbors following a Gaussian probability density function expressed in a hyperbolic space. This is achieved by defining a Gaussian cumulative distribution in this space.The HLoOP algorithm is tested on the WordNet dataset yielding promising results. Code and data will be made available on request for reproductibility
Uncertainty Quantification for Conformal Patch Antenna Installed on Infinite Cylinder
No full textInternational audienceSmall antennas installed on large structures can be critical elements which need to be fully characterized radiation wise. Such an issue naturally arises for airborne antennas for example. This paper presents a crucial step in the modeling of such a radiation in a complex environment. Considering a patch antenna installed on an infinite cylinder, its radiation is analyzed while taking into account variability in the antenna parameters, as well as in the cylinder geometry. By the Polynomial Chaos Expansion of the analytical field, the resulting stochastic radiation pattern can be derived to quantify uncertainty and thus bound the possible realizations of the antenna radiation with respect to variable parameters.</div
Emerging trends in signal processing and machine learning for positioning, navigation and timing information: special issue editorial
No full textInternational audienceLocation-based services, safety-critical applications, and modern intelligent transportation systems require reliable, continuous and precise positioning, navigation and timing (PNT) information. Global Navigation Satellite Systems (GNSS) are the main source of positioning data in open sky conditions; however their vulnerabilities to radio interferences and signal propagation limit their use in challenging environments. Consequently, enhancing conventional GNSS-based PNT solutions to incorporate additional sensing modalities and exploit other available signals of opportunity has become necessary for continuous and reliable navigation.Articles in the special issue span detection methods, estimation algorithms, signal optimization, and the application of machine learning, providing comprehensive insights into enhancing navigation and positioning accuracy. PNT technologyPositioning, Navigation, and Timing (PNT) technologies form the backbone of many of today's most critical applications, spanning various sectors and impacting everyday life. Modern PNT solutions rely on a sophisticated interplay of diverse sensors and systems to deliver precise and reliable information about position, navigation, and time (Dardari, et al., 2015). In this introduction, we discuss the fundamental components that constitute these advanced PNT systems and explore their wide array of applications.One of the cornerstones of PNT technology is Global Navigation Satellite Systems (GNSS) (Teunissen and Montenbruck, 2017; Morton, et al., 2021). GNSS involves satellite constellations that provide timing and positioning signals to receivers on Earth. These signals are pivotal in numerous domains, including personal navigation, geolocation services, agriculture, and search and rescue operations. GNSS enables users to determine their exact location anywhere on the globe with remarkable accuracy and reliability. Complementing GNSS are Inertial Navigation Systems (INS) (Britting, 2010), which consist of accelerometers, gyroscopes, and magnetometers. These systems measure inertial forces to provide orientation, velocity, and position data. This technology is particularly essential for applications such as aircraft navigation, autonomous vehicles, marine navigation, and robotics, where continuous and precise navigation information is critical, even in the absence of external signals.Another key technology in modern PNT solutions is Light Detection and Ranging (LiDAR) (Royo and Ballesta-Garcia, 2019). LiDAR utilizes laser sensors to emit laser light and process the reflections to measure ranges. LiDAR is invaluable in autonomous vehicles, environmental monitoring, mapping, and geospatial analysis, providing highresolution, three-dimensional information about the surroundings. Computer vision and camera systems (Hartley and Zisserman, 2004) also play a crucial role in PNT. These systems are based on image sensors and advanced processing algorithms, which analyze the environment. This technology finds applications in augmented reality, robotics, autonomous vehicles, and surveillance, enabling machines to interpret and respond to visual information with high precision.Additionally, wired and wireless communication networks (Monge and Contractor, 2003) are integral to modern PNT solutions. Technologies such as 5G, Wi-Fi, and Bluetooth enable data exchange between devices and systems. While some of these networks are not explicitly designed for PNT, they can be utilized as Signals of Opportunity (SoO) to enhance dedicated PNT systems. Typical applications include the Internet of Things (IoT), smart cities, real-time traffic management, and remote sensing.In summary, modern PNT solutions represent an intricate blend of various advanced technologies. Each component-whether GNSS, INS, LiDAR, computer vision, or communication networks-plays a vital role in delivering the precise and reliable PNT information that underpins countless applications in our daily lives and across numerous industries. As these technologies continue to evolve, their integration and capabilities will expand, driving further innovation and transforming how we navigate and understand the world. Challenges in PNT technologyTraditional PNT methods, while revolutionary, have inherent limitations that necessitate the integration of diverse sensors and systems to achieve continuous and reliable information. Each of these traditional technologies, -such as GNSS, INS, LiDAR, computer vision, and communication networks -has its unique strengths and weaknesses, shaped by their inherent characteristics and environmental interactions. Understanding these limitations underscores the need for a multifaceted approach to PNT solutions.GNSS is susceptible to several limitations that can impair its accuracy and reliability. Signal obstructions and multipath effects, common in urban environments, forests, and indoor settings, lead to signal loss or degradation resulting in reduced positional accuracy and reliability. Additionally, GNSS signals are vulnerable to interference and jamming from other electronic devices or malicious users, which can lead to the loss or corruption of useful signals. Atmospheric conditions, particularly ionospheric and tropospheric delays, can also affect signal propagation, causing positioning errors, which are aggravated near the equator or during solar activity. Moreover, GNSS relies on the visibility of a sufficient number of satellites. In environments such as tunnels or urban canyons, the number of visible satellites can drastically reduce, hindering the system's ability to resolve the navigation problem.INS, while powerful, face challenges primarily related to drift over time. The principle of INS involves integrating acceleration and angular rates over time, leading to cumulative errors known as drifts. Without periodic calibration or correction from external sources, these errors can accumulate, reducing the system's accuracy. Furthermore, high-precision INS is complex and expensive, limiting their suitability for low-cost applications.LiDAR systems, though effective, encounter limitations due to environmental conditions and intrinsic characteristics. Weather conditions such as rain, fog, or smog can degrade LiDAR performance, reducing its accuracy. LiDAR also has limitations in range and resolution, making it difficult to detect distant or small objects. Additionally, high-quality LiDAR systems are often expensive and require significant power, posing challenges for low-cost or low-consumption applications.Computer vision and camera-based systems are dependent on the lighting conditions of the scene they capture. Low light or highly dynamic lighting conditions can lead to poor quality images, affecting the system's performance. Occlusion and field of view limitations can result in incomplete images, hindering accurate analysis. Moreover, image processing algorithms are computationally intensive, requiring substantial power and memory, which can be challenging for real-time or low-cost applications.Communication networks also face several limitations that impact PNT solutions. Latency and bandwidth constraints can affect real-time data transmission and processing, introducing delays and bottlenecks, especially in data-intensive applications. Network coverage can be inconsistent, particularly in underserved areas, reducing the availability and reliability of PNT solutions.The limitations of traditional PNT methods highlight the necessity of integrating a diverse array of sensors and systems. Each technology brings unique capabilities and constraints. By combining these technologies, we can mitigate individual weaknesses and enhance overall PNT performance. This multifaceted approach is essential to meet the growing demand for precise, reliable, and continuous PNT information across various applications and environments.</div
Performance-based multi-DME station selection
No full textInternational audienceWhile the Global Navigation Satellite System (GNSS) has established as the predominant source for delivering highprecision and high-integrity services in numerous aeronautical applications, its susceptibility to interference due to the low signal strength has necessitated the development of Alternative Positioning, Navigation, and Timing (A-PNT). Multiple Distance Measuring Equipment (Multi-DME) navigation emerges as a promising candidate for A-PNT. Station selection should be considered for multi-DME based A-PNT, due to the limited channel recourses, and the constrained processing capability of the on-board devices. It is also beneficial to limit DME interference to GNSS L5/E5a. However, signal sources decreasing may degrade the navigation performance. Station selection is to make a tradeoff between the number of selected stations, interference on GNSS signals and performance of navigation. We propose a performance-based method to address this problem. The proposed performance-based method for multi-DME is based on the Gradient Boosting Machine (GBM) algorithm. The comparison with the widely used quasi-smallest PDOP method shows that the proposed method can achieve the RNP 1 requirement with fewer DME stations and less GNSS L5/E5a signal degradation
Scientific Evaluation of the Impact of an Increase in the Retirement Age on the Cognitive Functions and Well-Being of Air Traffic Controllers (ATCOs)
No full textInternational audienceIn 2020, the Swiss Federal Council made it a strategic objective to encourage Skyguide (Switzerland’s private air navigation service provider) and social partners (HelvetiCA) to work together to raise the retirement age from the current 56/59 to at least 60. In this context, HelvetiCA and Skyguide agreed to carry out a scientific study (RAFA study) to assess the possible impact of this increase in age, in particular on the psychological well-being and cognitive performance of ATCOs. Two studies were carried out following a review of the literature. The first study aimed to identify the factors relating to working conditions, individual characteristics and coping strategies that may be affected by ageing and sought to assess whether these factors have an impact on the ability to perform operational tasks according to the demands and conditions of the job carried out by ATCOs. The second study aimed to assess the cognitive functions of ATCOs of different ages using a battery of neuropsychological tests to examine the impact of ageing on cognitive performance, a crucial aspect of air traffic control activity. After a 13-month study period, a final report containing 17 recommendations was submitted to HelvetiCA/Skyguide. Within the context of the Collective Labour Agreement agreed in January 2024, the social partners agreed to implement the recommendations to help HelvetiCA and Skyguide manage this change safely and efficiently, with a transparent programme