We address the problem of partitioning a network of nano-satellites to distribute fairly the network load under energy consumption constraints. The study takes place in a context where this swarm of nano-satellites orbits the Moon and works as, but not limited to, a distributed radio-telescope for low-frequency radio interferometry. During an interferometry mission, each nano-satellite collects observation data, then shares them with the other swarm members to compute a global image of space. However, the simultaneous transmission of large volumes of data can cause communication issues by overloading the radio channel, leading to potential packet loss. In this context, we investigate three division algorithms based on graph sampling techniques. We prove that random walk-based algorithms overall perform the best in terms of conservation of graph properties and fairness for group sizes down to 10% of the original graph.

In the last decade, it has been quickly recognized that backhauling Low Power Wide Area Networks (LPWAN) through Low Earth Orbit (LEO) satellites paves the way to the development of novel applications for a truly ubiquitous Internet of Things (IoT). Among LPWAN communications technologies, Narrowband IoT (NB-IoT) does not suffer from interference by other concurrent technologies since it works on a licensed frequency spectrum. At the same time, thanks to its medium access scheme based on contention resolution and resource allocation, NB-IoT is a key enabler for the specific market slice of IoT applications requiring a good level of reliability. In the architectural configuration analyzed throughout this contribution, an NB-IoT low power User Equipment (UE) can communicate with a LEO satellite equipped with an Evolved Node B (eNB) for a time limited to the visibility window of that satellite from the UE position on the Earth. However, the Doppler effect inherent to the time-varying relative speed of the eNB needs to be dealt with additional resources. The solutions proposed until now are non-trivial, thus making the use of NB- IoT for ground-to-satellite communications still expensive and energetically inefficient. Timely, this contribution proposes a procedure for a UE to infer the future values of the Doppler shift from the beacon signals so that frequency pre-compensation can be easily applied in the following interactions during the visibility time. The presented simulation results show that a UE needs to listen to about 10 beacon signals in 1 second to accurately and robustly predict the Doppler curve, thus enabling a lightweight (and eventually truly energy-efficient) implementation of NB-IoT over ground-to-satellite links.

This paper aims to characterize the estimation precision at the output of the GNSS receiver tracking stage. We define an original statistical modelling of the GNSS tracking loop, which can then be exploited by an optimal linear Kalman Filter (KF) in order to obtain an analytical expression of the steady-state regime. The latter is designed to encompass dynamic information of the GNSS receiver. Two observation models are of interest: the first one considers the propagation delay and Doppler parameters, and the second one also including the Doppler rate, i.e., the acceleration, which is known to be relevant for high dynamics scenarios and can easily be included into the acquisition step. Within this context, the steady-state asymptotic performance of the tracking stage is obtained by solving an algebraic discrete Riccati equation. In both cases, simulation results are provided to show the validity of the proposed approach and the resulting steady-state performance.

Global Navigation Satellite Systems (GNSS) are a key player in a plethora of applications. For navigation purposes, interference scenarios are among the most challenging operation conditions, which clearly impact the maximum likelihood estimates (MLE) of the signal synchronization parameters. While several interference mitigation techniques exist, a theoretical analysis on the GNSS MLE performance degradation under interference, being fundamental for system/receiver design, is a missing tool. The main goal of this contribution is to introduce a mathematical tool to evalute the effect of any type of interference on any GNSS signal. Regarding such tool, we provide closedform expressions of the misspecified Cram´er-Rao (MCRB) bound and estimation bias, for a generic GNSS signal corrupted by an interference. The proposed expressions are used to analyze the GNSS performance degradation induced by the distance measuring equipment (DME) system.

The aviation Minimum Operational Performance Standard defines the SBAS protection levels as the product of the estimated standard deviation of the positioning error and a scaling factor called K-factor. The K-factor depends on the time window of interest and on the correlation between errors in the time window. The K-factors defined in aviation are difficult to generalize to other specifications in other domains, such as rail and maritime applications. This article presents a simple formula to calculate the K-factor for any value of integrity risk and time interval. The resulting K-factor is shown to be mathematically rigorous under the hypothesis of Gaussian error distribution but without any assumption on the correlation structure of the successive position estimates. The Gaussian assumption can be relaxed and replaced by overbounding with a Gaussian distribution with a very good approximation. This formula can be used in any GNSS application where integrity is needed.

This work derives compact closed-form expressions of the misspecified Cramér–Rao bound and pseudo-true parameters of time-delay and Doppler for a high dynamics signal model. Those expressions are validated by analyzing the mean square error (MSE) of the misspecified maximum likelihood estimator. A noteworthy outcome of these MSE results is that, for some magnitudes of acceleration and signal-to-noise ratios, neglecting the acceleration is beneficial in the MSE sense. The variance performance improvement is obtained at the cost of a systematic error in the true parameter estimation. This can be seen as a specific case of the trade-off between bias and variance. Neglecting the acceleration can improve the Doppler estimation when the error induced on the misspecified model is less than the variance increase due to including an extra parameter to estimate. Then, for some non-zero acceleration magnitudes and short integration times, the Doppler estimation using a misspecified model outperforms a correctly specified model in the MSE sense.

In ranging-based applications, ignoring the presence of multipath often leads to a bias upon the estimated range, which actually originates from misspecified estimation problem because the assumed data signal model, here without multipath, is not equal to the true one. Such misspecification also results in an error covariance matrix around the biased estimates, so-called pseudotrue parameters, that differs from the Cramér–Rao bound applied to the true model. This error covariance matrix can be lower bounded by a misspecified Cramér–Rao bound (MCRB). In this work, a closed-form expression of the MCRB under multipath conditions is proposed, which only depends on the baseband signal samples and both delay, Doppler and complex amplitude pseudotrue parameters. These MCRB expressions are fundamental (i) to understand and characterize the impact of multipath conditions when not taken into account, (ii) for system/signal design, and (iii) to derive new robust estimators. The proposed MCRBs are validated for a representative navigation signal, comparing the resulting bounds with the mean square error obtained by the misspecified maximum likelihood estimator with respect to the pseudotrue parameters.

This thesis focuses on the problem to compute an optimal sending rate as a function of the network state. The challenge is to assess which kind of sensing is needed to achieve this goal. Historically, this job was done by the TCP congestion control mechanism. However, we do not restrict this thesis to TCP but to any kind of protocol that needs to adapt its sending rate whatever the service is (i.e., reliable with retransmissions or not). Most congestion control variants are used to monitor the delay and the losses to compute their sending rate (for instance, within a congestion window for TCP; a rate control for BBR [7] or certain multimedia applications such as Skype [11]). Both metrics (i.e., delay and loss event) have shown to be under-exploited [46] or too limited when considering only the loss ratio. In this thesis, we seek to rethink these metrics and how they can be better used to compute the optimal sending rate. In this context, we propose to investigate the use of a Deep Learning (DL) algorithm that seems particularly relevant for congestion control tasks. We also focused on improving this DL algorithm to ease its deployment and usage in real-life scenarios.

Cette thèse se concentre sur le problème du calcul d’un taux d’envoi optimal en fonction de l’état du réseau. Historiquement, ce travail a été effectué par le mécanisme de contrôle de congestion TCP. La plupart des variantes de contrôle de congestion sont utilisées pour surveiller le retard et les pertes afin de calculer leur taux d’envoi. Dans cette thèse, nous cherchons à repenser ces métriques et comment elles peuvent être mieux utilisées pour calculer le taux d’envoi optimal. Dans ce contexte, nous proposons d'étudier l’utilisation d’un algorithme de Deep Learning qui semble particulièrement pertinent pour les tâches de contrôle de congestion. Nous nous sommes également concentrés sur l’amélioration de cet algorithme de Deep Learning afin de faciliter son déploiement et son utilisation dans des scénarios réels.