Quand un satellite fonctionne dans des conditions normales, de nombreux paramètres sont enregistrés et transmis vers un centre de contrôle. Ces paramètres sont analysés régulièrement pour détecter la présence d’éventuelles anomalies. Lorsque ces anomalies sont détectées, le satellite doit les corriger et doit éviter que ces erreurs apparaissent ultérieurement. Dans ce contexte, plusieurs algorithmes de détection d’anomalies ont montré leur intérêt comme l’algorithme LoOP (Local Outlier Probability) qui fournit pour chaque vecteur de données une probabilité d’anomalies ou l’algorithme One-Class SVM qui permet de définir un domaine contenant la majorité des vecteurs de données (supposés normaux) et qui détecte les anomalies situées à l’extérieur de ce domaine. L’objectif de ce séminaire est de proposer des méthodes qui modifient l’algorithme de détection d’anomalies non supervisé à l'aide d'un retour utilisateur afin de diminuer les erreurs de classification (détections manquées/fausses alarmes).

This article investigates new strategies to compute accurate low-complexity Log Likelihood Ratio (LLR) values based on the Bayesian formulation under uncorrelated fading channels for both antipodal and CSK modulations when no Channel State Information (CSI) is available at the receiver. These LLR values are then used as input to modern error correcting schemes used in the data decoding process of last generation GNSS signals. Theoretical analysis based on the maximum achievable rate is presented for the different methods in order to evaluate the performance degradation with respect to the optimal CSI channel. Finally, Frame Error Rate (FER) simulation results are shown, validating the appropriate performance of the proposed LLR approximation methods.

In this article, we provide closed-form approximations of log-likelihood ratio (LLR) values for direct sequence spread spectrum (DS-SS) systems over three particular scenarios, which are commonly found in the Global Navigation Satellite System (GNSS) environment. Those scenarios are the open sky with smooth variation of the signal-to-noise ratio (SNR), the additive Gaussian interference, and pulsed jamming. In most of the current communications systems, block-wise estimators are considered. However, for some applications such as GNSSs, symbol-wise estimators are available due to the low data rate. Usually, the noise variance is considered either perfectly known or available through symbol-wise estimators, leading to possible mismatched demodulation, which could induce errors in the decoding process. In this contribution, we first derive two closedform expressions for LLRs in additive white Gaussian and Laplacian noise channels, under noise uncertainty, based on conjugate priors. Then, assuming those cases where the statistical knowledge about the estimation error is characterized by a noise variance following an inverse log-normal distribution, we derive the corresponding closed-form LLR approximations. The relevance of the proposed expressions is investigated in the context of the GPS L1C signal where the clock and ephemeris data (CED) are encoded with low-density parity-check (LDPC) codes. Then, the CED is iteratively decoded based on the belief propagation (BP) algorithm. Simulation results show significant frame error rate (FER) improvement compared to classical approaches not accounting for such uncertainty.

Recently, convolutional neural networks have been successfully applied to lossy image compression. End-to-end optimized autoencoders, possibly variational, are able to dramatically outperform traditional transform coding schemes in terms of rate-distortion trade-off; however, this is at the cost of a higher computational complexity. An intensive training step on huge databases allows autoencoders to learn jointly the image representation and its probability distribution, possibly using a non-parametric density model or a hyperprior auxiliary autoencoder to eliminate the need for prior knowledge. However, in the context of on board satellite compression, time and memory complexities are submitted to strong constraints. The aim of this paper is to design a complexity-reduced variational autoencoder in order to meet these constraints while maintaining the performance. Apart from a network dimension reduction that systematically targets each parameter of the analysis and synthesis transforms, we propose a simplified entropy model that preserves the adaptability to the input image. Indeed, a statistical analysis performed on satellite images shows that the Laplacian distribution fits most features of their representation. A complex non parametric distribution fitting or a cumbersome hyperprior auxiliary autoencoder can thus be replaced by a simple parametric estimation. The proposed complexity-reduced autoencoder outperforms the Consultative Committee for Space Data Systems standard (CCSDS 122.0-B) while maintaining a competitive performance, in terms of rate-distortion trade-off, in comparison with the state-of-the-art learned image compression schemes

Machine learning and data-driven algorithms have gained a growth of interest during the past decades due to the computation capability of the computers which has increased and the quantity of data available in various domains. One possible application of machine learning is to perform unsupervised anomaly detection. Indeed, among all available data, the anomalies are supposed to be very sparse and the expert might not have the time to label all the data as nominal or not. Many solutions exist to this unsupervised problem, but are known to provide many false alarms, because some scarce nominal modes might not be included in the training dataset and thus will be detected as anomalies. To tackle this issue, we propose to present an existing iterative algorithm, which presents potential anomaly to the expert at each iteration, and compute a new boundary according to this feedback using One Class Support Vector Machine.

The estimation of wing deformation is part of the certification of an aircraft. Wing deformation can be obtained from 3D reconstructions based on conventional multiview photogrammetry. However, 3D reconstructions are generally degraded by the variable flight environments that degrade the quality of 2D images. This paper addresses this issue by taking benefit from a priori knowledge of the wing mechanical behaviour. Specifically, mechanical limits are considered to regularize the bundle adjustment within the photogrammetry reconstruction. The performance of the proposed approach is evaluated on a real case, using data acquired on an aircraft A350-900.

The estimation of wing deformation is part of the certification of an aircraft. Wing deformation can be obtained from 3D reconstructions based on conventional multiview photogrammetry. However, 3D reconstructions are generally degraded by the variable flight environments that degrade the quality of 2D images. This paper addresses this issue by taking benefit from a priori knowledge of the wing mechanical behaviour. Specifically, mechanical limits are considered to regularize the bundle adjustment within the photogrammetry reconstruction. The performance of the proposed approach is evaluated on a real case, using data acquired on an aircraft A350-900.

Critical aeronautical communications are a major issue for flight safety. For a long time, these have relied solely on voice, which is transmitted via an analog communication system. Given the growth in air traffic, this mean of communication has reached saturation and moreover, it has sometimes shown its limits in terms of understanding voice messages, hence the need to find an alternative method. The development of communication technologies based on digital signals allows text messages to be exchanged over a long distance. Initially reserved for noncritical airline operations, it was quickly adopted for communications between the pilot and the air traffic controller, in order to offload the dedicated radio channel. This is known as Data Link. This system, included in a more global infrastructure called the ATN/OSI, has the double advantage of relieving congestion on the frequencies used, but also of limiting the misunderstanding of certain messages. The next evolutions of this aeronautical communication system based on the IP suite and called ATN/IPS is under development. It will have to solve certain problems by proposing new communication technologies and innovative network solutions that can adapt to the increase in critical air data traffic. In this thesis, we address several issues related to the development of ATN/IPS. The first one concerns the network mobility of the aircraft. Indeed, the ATN/IPS will gather several operators, each providing their own subnetworks composed of one or more access methods. Given the limited range of some of them, an aircraft necessarily needs to use several of them during a flight. A handover is triggered as soon as an aircraft connects to a new ground station, which in some cases requires a change in routing to the aircraft. We propose to combine and adapt two mobility protocols, PMIPv6 and LISP, to guarantee continuity of critical data transmission while minimizing the impact on the avionics architecture and the radio communication channel. Our solution is compared to a standard IP mobility solution in a simulated network environment and specifically developed under OMNeT++. The results show that our approach reduces the handover delay, while lightening the signaling traffic on the radio channel. Moreover, in order to propose the best aircraft connectivity, we propose an automation of the selection of the best links in the multilink and ATN/IPS context. Typically, multilink algorithms (or link selection) are split into three parts : collecting link information, deciding which links to use, and using the new links. As the mobility solution proposed in this thesis is also compatible with multilink, we are interested in the first two steps. We propose to use an active method to probe the links and estimate their quality. This approach has the advantage of being independent of the underlying communication technologies. We then compare three estimation methods based on round trip delay and evaluate the performance of each of them. The first method is based on threshold determination, the second is based on a probabilistic model and the third uses supervised learning. This learning-based method makes it possible to estimate the link over time with good precision. Finally, we propose a link selection algorithm in the case where the primary link no longer meets the quality of service requirements.

This letter studies a new expectation maximization (EM) algorithm to solve the problem of circle, sphere and more generally hypersphere fitting. This algorithm relies on the introduction of random latent vectors having a priori independent von Mises-Fisher distributions defined on the hypersphere. This statistical model leads to a complete data likelihood whose expected value, conditioned on the observed data, has a Von Mises-Fisher distribution. As a result, the inference problem can be solved with a simple EM algorithm. The performance of the resulting hypersphere fitting algorithm is evaluated for circle and sphere fitting.

Critical aeronautical communications are a major issue for flight safety. For a long time, these have relied solely on voice, which is transmitted via an analog communication system. Given the growth in air traffic, this mean of communication has reached saturation and moreover, it has sometimes shown its limits in terms of understanding voice messages, hence the need to find an alternative method. The development of communication technologies based on digital signals allows text messages to be exchanged over a long distance. Initially reserved for noncritical airline operations, it was quickly adopted for communications between the pilot and the air traffic controller, in order to offload the dedicated radio channel. This is known as Data Link. This system, included in a more global infrastructure called the ATN/OSI, has the double advantage of relieving congestion on the frequencies used, but also of limiting the misunderstanding of certain messages. The next evolutions of this aeronautical communication system based on the IP suite and called ATN/IPS is under development. It will have to solve certain problems by proposing new communication technologies and innovative network solutions that can adapt to the increase in critical air data traffic. In this thesis, we address several issues related to the development of ATN/IPS. The first one concerns the network mobility of the aircraft. Indeed, the ATN/IPS will gather several operators, each providing their own subnetworks composed of one or more access methods. Given the limited range of some of them, an aircraft necessarily needs to use several of them during a flight. A handover is triggered as soon as an aircraft connects to a new ground station, which in some cases requires a change in routing to the aircraft. We propose to combine and adapt two mobility protocols, PMIPv6 and LISP, to guarantee continuity of critical data transmission while minimizing the impact on the avionics architecture and the radio communication channel. Our solution is compared to a standard IP mobility solution in a simulated network environment and specifically developed under OMNeT++. The results show that our approach reduces the handover delay, while lightening the signaling traffic on the radio channel. Moreover, in order to propose the best aircraft connectivity, we propose an automation of the selection of the best links in the multilink and ATN/IPS context. Typically, multilink algorithms (or link selection) are split into three parts : collecting link information, deciding which links to use, and using the new links. As the mobility solution proposed in this thesis is also compatible with multilink, we are interested in the first two steps. We propose to use an active method to probe the links and estimate their quality. This approach has the advantage of being independent of the underlying communication technologies. We then compare three estimation methods based on round trip delay and evaluate the performance of each of them. The first method is based on threshold determination, the second is based on a probabilistic model and the third uses supervised learning. This learning-based method makes it possible to estimate the link over time with good precision. Finally, we propose a link selection algorithm in the case where the primary link no longer meets the quality of service requirements.