Les systèmes dits Sensor Webs sont des middlewares informatiques assurant la communication entre les capteurs et les applications. En tant que véritables médiateurs, la popularité de ces systèmes n’a cessé de grandir depuis l’apparition des tout premiers capteurs. Plus récemment, l’émergence de nouveaux paradigmes tels que l’Internet des Objets (IoT) a complètement révolutionné les systèmes basés sur les capteurs en général. Parmi eux, les Sensor Webs ne dérogent pas à cette règle et doivent maintenant répondre à de nouveaux défis, notamment en termes d’intégration, de Qualité des Observations (QoO) et d’adaptation système. Dans le cadre de ma thèse, j'ai proposé une nouvelle génération de Sensor Webs capables d’adapter la QoO distribuée de manière autonomique et spécifique à chaque application (QASWS). Premièrement, j'introduirai un framework générique destiné aux chercheurs et développeurs souhaitant concevoir leur propre solution QASWS. Dans un deuxième temps, je montrerai comment j'ai utilisé ce framework afin de développer une plateforme d’intégration pour l’évaluation de la QoO à la demande (iQAS). Après avoir évalué ses performances, je présenterai trois cas d’utilisation pour la plateforme iQAS. Finalement, je conclurai ma présentation en imaginant l’apport de certains paradigmes transverses vis-à-vis de la QoO dans un futur proche.

Les réseaux de capteurs type LPWAN (grande autonomie, grande couverture, faible débit) sont en plein essor. Ces réseaux, qui ont pour but de connecter des milliers de capteurs sans fils de la manière la plus efficace possible, seraient un maillon indispensable à l’internet des objets tel qu’il est imaginé dans les années à venir. Dans le but d’augmenter la couverture de ces services, un lien satellite est envisagé. Cette thèse étudie les contraintes liées à l’utilisation d’un lien satellite unidirectionnel dans un système terrestre étendu, sans modifications du matériel, supposé bas-coût. Après une rapide présentation du milieu de l’IoT et des solutions LPWAN terrestres, nous présenterons la solution et le système choisis, compatible LTE NB-IoT. L’utilisation d’un lien satellite limite le système en termes de bilan de liaison et d’interférences. On cherchera alors à quantifier ces interférences et à obtenir les performances d’un tel réseau.

For future satellite-to-ground communications link, very high throughput might be achievable at a reasonable cost assuming the use of existing single mode components developed for fiber technologies (optical detectors and amplifiers, MUX/DEMUX...). The influence of atmospheric turbulence degrades the injection efficiency of the incoming wave into single mode components. This leads to signal fading and channel impairments. Several mitigation strategies are considered to prevent them. The use of adaptive optics (AO) should contribute to reduce substantially the criticality of the fading at the expense of potentially complex and expensive systems if very high stability of the injection is requested. The use of appropriate numerical mitigation techniques (coding+interleaving) can help to relax the specifications and cost of AO systems but could lead among others to unmanageable buffer size. Thus the specification of AO correction and interleavers/forward error codes should be addressed jointly. A model to evaluate the channel capacity in terms of outage probability and packet error rate has been developed that jointly takes into account partial correction by AO and channel interleaving. It includes the capability to evaluate coded transmission performance over the correlated FSO channel in satellite-to-ground scenarios. This model is presented here and confronted to numerical simulations for several distinct correction cases. The interdependence of AO correction with numerical mitigation techniques is investigated.

Built upon the Internet of Things (IoT), the Internet of Everything (IoE) acknowledges the importance of data quality within sensor-based systems, alongside with people, processes and Things. Nevertheless, the impact of many technologies and paradigms that pertain to the IoE is still unknown regarding Quality of Observation (QoO). This paper proposes to study experimental results from three IoE-related deployment scenarios in order to promote the QoO notion and raise awareness about the need for characterizing observation quality within sensor-based systems. We specifically tailor the definition of QoO attributes to each use case, assessing observation accuracy within Smart Cities, observation rate for virtual sensors and observation freshness within post-disaster areas. To emulate these different experiments, we rely on a custom-developed integration platform for the assessment of QoO as a service called iQAS. We show that QoO attributes should be used to specify what is an observation of “good quality”, that virtual sensors may have specific and limiting capabilities impacting QoO and that network QoS and QoO are two complementary quality dimensions that should be used together to improve the overall service provided to end-users

The design of a new GNSS signal is always a trade-off between improving performance and increasing complexity, or even between improving different performance criteria. Position accuracy, receiver sensitivity (acquisition, tracking or data demodulation thresholds) or the Time-To-First-Fix (TTFF) are examples of those GNSS receivers performance criteria. Within the framework of Galileo 2nd Generation (G2G), adding a new signal component dedicated to aid the acquisition process on E1 can help to improve performance of GNSS receivers with respect to these criteria as it was shown in [1]. In order to create this new component, various aspects such as the spreading modulation, the data navigation content, the channel coding or the Pseudo-Random Noise (PRN) codes must be studied. To this end, this paper firstly proposes the study of new spreading modulations, and secondly, we investigate on PRN codes that can be well suited to the proposed Acquisition-Aiding signal.

The well-known conventional Least Squares (LS) and Extended Kalman Filter (EKF) are one of the most widely used algorithms in science and particularly in localization with GNSS measurements. However, these estimators are not optimal when the GNSS measurements become contaminated by non-Gaussian errors including multipath (MP) and non-line-of-sight (NLOS) biases. On the other hand, this kind of ranging measurements errors occurs generally in urban areas where GNSS-based potitionning applications require more accuracy and reliability. In this paper, we use additional information of the environment consisting of bias prediction from a 3D model and a GNSS simulator to exploit constructively NLOS measurements. We use this 3D GNSS simulator to predict lower and upper bounds of these biases. Then, we integrate this information in the position estimation problem by considering these biases as additive error and exploiting the bounds to end-up with a constrained state estimation problem that we resolve with existing Constrained Least Squares (CLS) and Constrained EKF (CEFK) algorithms. Experimental results using real GPS signals in Down-Town Toulouse show that the proposed estimator is capable of improving the positioning acuracy compared to conventional algorithms. Theoretical conditions have been established to determine the acceptable bias prediction error allowing better positioning performance than conventional estimators. Tests are conducted then to validate these conditions and investigate the influence of the bias prediction error on the localization performance by proposing new acuracy metrics.

Hyperspectral unmixing is a blind source separation problem which consists in estimating the reference spectral signatures contained in a hyperspectral image, as well as their relative contribution to each pixel according to a given mixture model. In practice, the process is further complexified by the inherent spectral variability of the observed scene and the possible presence of outliers. More specifically, multi-temporal hyperspectral images, i.e., sequences of hyperspectral images acquired over the same area at different time instants, are likely to simultaneously exhibit moderate endmember variability and abrupt spectral changes either due to outliers or to significant time intervals between consecutive acquisitions. Unless properly accounted for, these two perturbations can significantly affect the unmixing process. In this context, we propose a new unmixing model for multitemporal hyperspectral images accounting for smoothtemporalvariations,construedasspectralvariability,and abrupt spectral changes interpreted as outliers. The proposed hierarchical Bayesian model is inferred using a Markov chain Monte-Carlo (MCMC) method allowing the posterior of interest to be sampled and Bayesian estimators to be approximated. A comparison with unmixing techniques from the literature on synthetic and real data allows the interest of the proposed approach to be appreciated.

This paper introduces a new method for cardiac motion estimation in 2-D ultrasound images. The motion estimation problem is formulated as an energy minimization, whose data fidelity term is built using the assumption that the images are corrupted by multiplicative Rayleigh noise. In addition to a classical spatial smoothness constraint, the proposed method exploits the sparse properties of the cardiac motion to regularize the solution via an appropriate dictionary learning step. The proposed method is evaluated on one data set with available ground-truth, including four sequences of highly realistic simulations. The approach is also validated on both healthy and pathological sequences of in vivo data. We evaluate the method in terms of motion estimation accuracy and strain errors and compare the performance with state-of-the-art algorithms. The results show that the proposed method gives competitive results for the considered data. Furthermore, the in vivo strain analysis demonstrates that meaningful clinical interpretation can be obtained from the estimated motion vectors.