Search
Talk
Interaction protocoles de transport/fiabilisation MAC pour service mobile par satellite
Seminars of TeSA, Toulouse, October 19, 2017.
L’accès à Internet par satellite permet de relier certains lieux isolés, où le déploiement d’un réseau terrestre est impossible. L’utilisation de constellations de satellites pose cependant des problèmes spécifiques à ce type de réseau, comme des délais longs et variables, ou des forts taux d’erreurs, notamment entre le dernier satellite sur le chemin du message et le récepteur mobile au sol (canal LMS, Land Mobile Satellite). Pour palier ce canal difficile, des mécanismes de fiabilisation ont été introduits au niveau des couches liaison et physique. Ces mécanismes allient l’utilisation de codes FEC et des retransmissions afin d’optimiser l’utilisation du canal LMS. En ce qui concerne la couche transport, le protocole le plus couramment utilisé est TCP, mais ce dernier fonctionne très mal avec les mécanismes de fiabilisation du canal LMS. Lors de cette présentation, nous verrons pourquoi les performances de TCP s’effondrent dans le cas de communications par satellite, puis nous verrons les solutions proposées dans cette thèse pour améliorer les performances de TCP.
Networking / Space communication systems
Conference Paper
Impact of Delayed Acknowledgment on TCP Performance over LEO Satellite Constellations
In Proc. Fifth Federated and Fractionated Satellite Systems Workshop, Toulouse, France, November 2-3, 2017.
This paper aims at quantifying the impact of a default TCP option, known as Delayed Acknowledgment (DelAck), in the context of LEO satellite constellations. Satellite transmissions can suffer from high channel impairments, especially on the link between a satellite and a ground gateway. To cope with these errors, physical and link layer reliability schemes have been introduced, at the price of an increase of the end-to-end delay seen by the transport layer (e.g. TCP). Although DelAck is used to decrease the feedback path load and for overall system performance, the use of this option conjointly with satellite link layer recovery schemes might increase the delay and might be counterproductive. To assess the impact of this option, we drive simulation measurements with two well-deployed TCP variants. The results show that the performance gain depends on the variant used and that this option should be carefully set or disabled as a function of the network characteristics. DelAck has a negative impact on TCP variants which are more aggressive such as TCP Hybla, and should be disabled for these versions. However, it shows benefits for TCP variants less aggressive such as NewReno.
Networking / Space communication systems
Approximate Maximum Likelihood Estimation Using a 3D GNSS Simulator for Positioning in MP/NLOS Conditions
nIn Proc. Institute of Navigation Global Navigation Satellite Systems (ION GNSS), Portland, USA, Septembre 25-29, 2017.
Recent trends in Global Navigation Satellite System (GNSS) applications in urban environments have led to a proliferation of studies in this field that seek to mitigate the adverse effect of non-line-of-sight (NLOS) phenomena. However, these methods reduce the availability of positioning in deep urban conditions. For such harsh urban settings, this paper proposes a methodology of constructive use of NLOS signals, instead of their elimination. We propose to compensate for the NLOS errors using a 3D GNSS simulator to predict the measurements bias and integrate them as observations in the estimation method. We investigate a novel GNSS positioning technique based on measurement similarity scoring of an array of position candidates. We improve this technique using an estimation of the uncertainty on the bias prediction by 3D modeling. Experiment results using real GNSS data in a deep urban environment confirm the theoretical sub-optimal efficiency of the proposed approach, despite it intensive computational load.
Signal and image processing / Localization and navigation
Making H-ARQ Suitable for a Mobile TCP Receiver over LEO Satellite Constellations
In Proc. 9th EAI International Conference on Wireless ans Satellite Systems (PSATS), Oxford, United Kingdom, September 14-15, 2017.
This paper investigates strategies to carry out delay tolerant services over LEO satellite constellations for mobile receiver. In this context, LEO constellations are characterized by important delay variations where propagation impairments are mostly localized on the Land Mobile Satellite (LMS) channel (i.e. on the last hop). To cope with this issue, distinct reliability schemes can be introduced at the physical or link layers. Although their capacity to cope with transmission errors has been demonstrated, these recovery schemes may induce a high jitter that could severely damage TCP's internal timers and reliability schemes. As a matter of fact, transport and link layers’ reliability schemes exhibit a clear discrepancy. Following temporal traces representing the delay between a mobile terminal and the last hop satellite from a LEO constellation, we assess how HARQ mechanisms impact on the RTO based retransmission and the duplicate acknowledgments of TCP. Based on ns-2 simulations, we propose a layer-2 buffer that let both link and transport layers to conjointly perform. Our evaluations show an end-to-end data rate increase and more generally illustrate the benefit of re-ordering packets at the link layer when link-layer erasure coding recovery mechanisms are used conjointly with TCP.
Networking / Space communication systems
Cardiac Motion Estimation in Ultrasound Images using Spatial and Sparse Regularizations
In Proc. EEE International Conference on Image Processing (ICIP), Beijing, China, September 17-20, 2017.
This paper investigates a new method for cardiac motion estimation in 2D ultrasound images. The motion estimation problem is formulated as an energy minimization with spatial and sparse regularizations. 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 in terms of motion estimation and strain accuracy and compared with state-of-the-art algorithms using a dataset of realistic simulations. These simulation results show that the proposed method provides very promising results for myocardial motion estimation.
Signal and image processing / Earth observation
High-Resolution Spectral Image Reconstruction based on Compressed Data Fusion
In Proc. EEE International Conference on Image Processing (ICIP), Beijing, China, September 17-20, 2017.
Signal and image processing / Earth observation
Un algorithme de reconstruction dans le cas d’un échantillonnage non ponctuel, périodique ou non
In Proc. Groupement de Recherche en Traitement du Signal et des Images (GRETSI), Juan-Les-Pins, France, September 5-8, 2017.
Les composants électroniques destinés à l’échantillonnage haute fréquence des signaux ont des temps de réponse susceptibles de dégrader les performances des algorithmes de reconstruction. Dans cet article, on considère un signal échantillonné par un composant modélisé par un filtre linéaire invariant activé à des instants potentiellement irréguliers et qui forment une suite vérifiant la condition de Landau (généralisation de la condition de Nyquist). On étudie les conséquences des imperfections des échantillonneurs dans la reconstruction du signal et dans l’estimation de l’enveloppe et de la phase.
Signal and image processing / Space communication systems
Sur l’égalisation fréquentielle des modulations à phase continue
In Proc. Groupement de Recherche en Traitement du Signal et des Images (GRETSI), Juan-Les-Pins, France, September 5-8, 2017.
Dans cet article, nous proposons d’étudier différents algorithmes d’égalisation pour signaux à Modulation de Phase Continue (continuous phase modulation, CPM) dans le domaine fréquentiel à l’aide d’une représentation polyphase du signal. Par cette approche unifiée, nous montrons l’équivalence analytique de ces égaliseurs sous certaines hypothèses liées au canal de transmission, les liens, ainsi que leurs limitations dans certains cas.
Digital communications / Aeronautical communication systems and Space communication systems
Tissue Motion Estimation using Dictionary Learning : Application to Cardiac Amyloidosis
In Proc. IEEE International Ultrasonics Symposium (IUS), Washington, September 6-9, 2017.
Cardiac strain estimation from ultrasound images is an efficient tool for the diagnosis of cardiac diseases. This study focuses on cardiac amyloidosis, a pathology characterized by non-specific early symptoms such as the increased wall thickness. Recent clinical studies have demonstrated that patients with cardiac amyloidosis present an apex-to-base gradient longitudinal strain pattern, i.e., a normal strain in apex and abnormally lower values for base segments. Existing cardiac motion estimation methods belong to three categories based on optical flow, speckle tracking and elastic registration. To overcome the ill-posedness of motion estimation, they use local parametric models (e.g., affine) or global regularizations (e.g., B-splines). The objective of this study is to evaluate a recently proposed cardiac motion estimation method based on dictionary learning on patients subjected to cardiac amyloidosis.
Signal and image processing / Earth observation
Une approche distribuée asynchrone pour la factorisation en matrices non-négatives - Application au démélange hyperspectral
In Proc. 26eme Colloque sur le Traitement du Signal et des Images (GRETSI), Juan-les-Pins, France, September 5-8, 2017.
Le démélange d’images hyperspectrales est un exemple particulier du problème de factorisation en matrices non-négatives (NMF) qui consiste à identifier les signatures spectrales d’un milieu imagé ainsi que leurs proportions dans chacun des pixels. Toutefois, le nombre important de pixels composant ces images peut s’avérer contraignant en termes de mémoire ou de temps de calcul, ce qui peut motiver le développement de techniques d’estimation distribuées (entre plusieurs processeurs et/ou plusieurs machines). Dans cette perspective, nous proposons une approche distribuée du problème de démélange, basée sur de récentes avancées en optimisation distribuée asynchrone inspirées de l’algorithme proximal alternating linearized minimization (PALM). L’intérêt d’une estimation asynchrone par rapport à une procédure synchrone est illustré dans ce contexte sur des données synthétiques.
Signal and image processing / Earth observation
ADDRESS
7 boulevard de la Gare
31500 Toulouse
France