In recent years, neural networks have emerged as data-driven tools to solve problems which were previously addressed with model-based methods. In particular, image processing has been largely impacted by convolutional neural networks (CNNs). Recently, CNN-based auto-encoders have been successfully employed for lossy image compression [1,2,3,4]. These end-to-end optimized architectures are able to dramatically outperform traditional compression schemes in terms of rate-distortion trade-off. The auto-encoder is composed of an encoder and a decoder both learned from the data. The encoder is applied to the input data to produce a latent representation with minimum entropy after quantization. The latent representation, derived through several convolutional layers composed of filters and activation functions, is multi-channel (the output of a particular filter is called a channel or a feature) and non-linear. The representation is then quantized to produce a discrete-valued vector. A standard entropy coding method uses the entropy model inferred from the representation to losslessly compress this discrete-valued vector. A key element of these frameworks is the entropy model. In earlier works [1,2,3], the learned representation was assumed independent and identically distributed within each channel and the channels were assumed independent of each other, resulting in a fully-factorized entropy model. Moreover, a fixed entropy model was learned once, from the training set, preventing any adaptation to the input image during the operational phase. The variational auto-encoder proposed in [4] proposed to use a hyperprior auxiliary network. This network estimates the hyper-parameters of the representation distribution, for each input image. Thus, it does not require the assumption of a fully-factorized model which conflicts with the need for context modeling. This variational auto-encoder achieves compression performance close to the one of BPG (Better Portable Graphics) at the expense of a considerable increase in complexity. However, in the context of on-board compression, a trade-off between compression performance and complexity has to be considered to take into account the strong computational constraints. For this reason, the CCSDS (Consultative Committee for Space Data Systems) lossy compression standard has been designed as a highly simplified version of JPEG2000. This work follows the same logic, however in the context of learned image compression. The aim of this paper is to design a simplified version of the variational auto-encoder proposed in [4] in order to meet the on-board constraints in terms of complexity while preserving high performance in terms of rate-distortion. Apart from straightforward simplifications of the transform (e.g. reduction of the number of filters in the convolutional layers), we mainly propose a simplified entropy model that preserves the adaptability to the input image. A preliminary reduction of the number of filters reduces the complexity by 62% in terms of FLOPs with respect to [4]. It also reduces the number of learned parameters with a positive impact on the memory occupancy. The entropy model simplification exploits a statistical analysis of the learned representation for satellite images, also performed in [5] for natural images. This analysis reveals that most of the features are well fitted by centered Laplacian distributions. The complex hyperprior model based on a non-parametric distribution of [4] can thus be replaced by a simpler parametric centered Laplacian model. The problem then amounts to a classical and simple estimation of a single parameter referred to as the scale. Our simplified entropy models reduces the complexity of the variational auto-encoder coding part by 22% and outperforms the end-to-end model proposed in [1] for the high target rates.

In the context of automatic and preventive condition monitoring of rotating machines, this paper revisits the demodulation process essential for detecting and localizing cracks in gears and bearings. The objective of the paper is to evaluate the performance of the well-known Hilbert demodulation by providing a quantified assessment in terms of signal processing. For this purpose, vibration test signals are simulated guided by the analysis of real-world measurements. The database comes from a natural wear experimentation on a test bench at an industrial scale and without any fault initiation. In the proposed simulated model, the amplitude modulation is designed in a physical approach in order to be able to set up the number of faulty teeth and their location. The impact of a limited spectral bandwidth filtering is quantified not only for the amplitude but also for the phase modulation estimations. The interactions between the amplitude and phase estimations are discussed. A focus is made on the analytic signal ambiguity due to the non-uniqueness of the amplitude estimation. This property induces an original investigation when demodulating the residual generated after a time synchronous averaging. Finally, as the objective is a continuous surveillance of a machine, results are given for a sequence of real-world measurements in order to visualize the fault evolution through the demodulation process.

The recent decades have seen an increasing interest in Medium Earth Orbit and Low Earth Orbit satellite constellations. However, there is little information on the delay variation characteristics of these systems and the resulting impact on high layer protocols. To fill this gap, this paper simulates a constellation that exhibits the same delay characteristics as the already deployed Iridium but considers closer bandwidths to constellation projects'. We identify five major sources of delay variation in polar satellite constellations with different occurrence rates: elevation, intra-orbital handover, inter-orbital handover, orbital seam handover and Inter-Satellite Link changes. We simulate file transfers of different sizes to assess the impact of each of these delay variations on the file transfer. We conclude that the orbital seam is the less frequent source of delay and induces a larger impact on a small file transfers: the orbital seam, which occurs at most three times during 24 hours, induces a 66% increase of the time needed to transmit a small file. Inter-orbital and intra-orbital handovers occur less often and reduce the throughput by approximately about 8% for both low and high throughput configurations. The other sources of delay variations have a negligible impact on small file transfer, and long file transfers are not impacted much by the delay variations.

Recently, the joint design of the GNSS message structure and the associated channel‐coding scheme have been investigated as a means to reduce the Time‐To‐First‐Fix (TTFF) and particularly the time to retrieve the Clock and Ephemerides Data (CED). In this context, a new method to co‐design the navigation message and the channel‐coding scheme structure is proposed in this paper. This new co‐design enables us to reduce the time to retrieve the CED while enhancing error‐correction capabilities under degraded channel conditions. In order to fulfill such requirements, some structured coding schemes are designed, which provide both maximum distance separable (MDS) and full diversity properties under a non‐ergodic channel assumption.

In this paper, we study the the complexity of packet localization at reception, for recent synchronous Random Access (RA) techniques based on the protocol ALOHA for satellite communications. The promising CRDSA (Contention Resolution Diversity Slotted ALOHA) offers better throughput, in comparison to the traditional slotted ALOHA protocols, thanks to the use of Successive Interference Cancellation (SIC) along with multireplica transmission. MARSALA (Multi-replicA decoding using corRelation baSed locALizAtion) is one of the many variants and enhancement schemes of CRDSA that have been proposed in the literature. It is applied to CRDSA each time a decoding deadlock situation is reached (when no packets can be retrieved by CRDSA). MARSALA first localizes the replicas of collided packets on a chosen reference time slot using correlations. Then it performs coherent signal combination of packet replicas prior to decoding. However, despite the good performance offered by MARSALA, its localization process adds a significant complexity to the receiver in terms of correlation operations. R-SPOTiT (Random Shared POsition Technique for Interfered random Transmissions) mitigates this complexity by introducing a shared information between the receiver and each of the transmitters, about all potential packets' locations on the frame, without any additional signaling overhead. We focus in this paper on the analysis of the total number of correlations which are needed to localize packets' replicas for both MARSALA and R-SPOTiT, with a single or with multiple Gold preambles. This should include preamble detection operations that are performed at CRDSA with a coarse and fine tracking. The results show that the most suitable system to use is the multi-preamble R-SPOTiT with two preambles.

As part of aircraft certification and optimization, wing bending and twist measurements are performed under various load cases (aircraft weight, speed, angle of attack, etc.) to validate and improve wing deformation models. Since these measurements are acquired during flight, their analysis requires to face strong environmental constraints. Indeed, the highly varying luminosity conditions, the presence of possible reflections or shadows, the vibrations and the deformations of the entire aircraft, are strong constraints that need to be considered carefully. Current approaches applied in Airbus are based on inertial measurement units installed inside the wing, or on photogrammetry-based solutions using calibrated sensors and retro-reflective targets located on the wings. These methods are not only highly intrusive, but also require time-consuming installation, calibration phases and dedicated flights to produce only sparse measurements. Moreover, the use of reflective targets on the wing has an impact on the wing aerodynamic, which should be avoided. In this paper, we investigate a new method for estimating wing deformations. This method adapts a photogrammetry approach classically used for reconstructing buildings or art structures to the aircraft environment. To this aim, we propose to use synchronous videos from high resolution cameras, which can be easily installed on the aircraft windows and on the vertical stabilizer. Appropriate features are extracted from the images acquired by these cameras, related to wing joints or reference points located on the aircraft wing. The system uses these features to autonomously recalibrate itself at each frame and provide a dense 3D reconstruction of the wing in the aircraft reference coordinate system. Some experiments conducted on real data acquired on Airbus aircrafts show that the proposed estimation method provide promising results.

Global Navigation Satellite Systems (GNSS) are the main source of position, navigation, and timing (PNT) information and will be a key player in the next-generation intelligent transportation systems and safety-critical applications, but several limitations need to be overcome to meet the stringent performance requirements. One of the open issues is how to provide precise PNT solutions in harsh propagation environments. Under nominal conditions, the former is typically achieved by exploiting carrier phase information through precise positioning techniques, but these methods are very sensitive to the quality of phase observables. Another option that is gaining interest in the scientific community is the use of large bandwidth signals, which allow obtaining a better baseband resolution, and therefore more precise code-based observables. Two options may be considered: (i) high-order binary offset carrier (HO-BOC) modulations or (ii) the concept of GNSS meta-signals. In this contribution, we assess the time-delay and phase maximum likelihood (ML) estimation performance limits of such signals, together with the performance translation into the position domain, considering single point positioning (SPP) and RTK solutions, being an important missing point in the literature. A comprehensive discussion is provided on the estimators’behavior, the corresponding ML threshold regions, the impact of good and bad satellite constellation geometries, and final conclusions on the best candidates, which may lead to precise solutions under harsh conditions. It is found that if the receiver is constrained by the receiver bandwidth, the best choices are the L1-M or E6-Public Regulated Service (PRS) signals. If the receiver is able to operate at 60 MHz, it is recommended to exploit the full-bandwidth Galileo E5 signal. In terms of robustness and performance, if the receiver can operate at 135 MHz, the best choice is to use the GNSS meta-signals E5 + E6 or B2 + B3, which provide the best overall performances regardless of the positioning method used, the satellite constellation geometry, or the propagation conditions.

Named Data Networking (NDN), an Information-Centric Network (ICN) architecture, is based on caching, multipath and multi-producers retrieving. These properties provide new opportunities for a single user to increase its Quality of Experience (QoE). However, handling multiple flows, each of them having its own multiple paths, is more complex. To tackle this challenge, we highlight three main principles a solution should include. Nodes should cooperate, supervise their output queues and, eventually, wisely manage the multipath capacities of NDN. These three elements are the core of our proposition : Cooperative Congestion Control (CCC). More than a solution, CCC is proposed as a framework where each principle could be implemented in multiple ways. The ultimate objective is to fairly distribute the flows on the network and maximize QoE of users. We choose basic algorithms in order to evaluate the overall framework. We evaluate our solution with simulations and compare their results with a theoretical model.

The derivation of tight estimation lower bounds is a key tool to design and assess the performance of new estimators. In this contribution, first, the authors derive a new compact Cramér–Rao bound (CRB) for the conditional signal model, where the deterministic parameter's vector includes a real positive amplitude and the signal phase. Then, the resulting CRB is particularised to the delay, Doppler, phase, and amplitude estimation for band-limited narrowband signals, which are found in a plethora of applications, making such CRB a key tool of broad interest. This new CRB expression is particularly easy to evaluate because it only depends on the signal samples, then being straightforward to evaluate independently of the particular baseband signal considered. They exploit this CRB to properly characterise the achievable performance of satellite-based navigation systems and the so-called real-time kinematics (RTK) solution. To the best of the authors’ knowledge, this is the first time these techniques are theoretically characterised from the baseband delay/phase estimation processing to position computation, in terms of the CRB and maximum-likelihood estimation.