You will find here a list of my publications, last updated in June 2021. You can also find publication lists on my Google Scholar profile, as well as on Inria's HAL. My Ph.D. thesis is available, in French, on TEL.
Wireless full-duplex (FD) transceivers have become more commonplace in the recent years, enabled by improvements in hardware as well as in signal processing algorithms dedicated to self-interference (SI) mitigation. A key design target for any FD radio is to limit the analog-to-digital converter (ADC) quantization noise affecting the signal of interest (SoI), which results from the difference in dynamic range between the SI and the SoI at the input of the ADC. Indeed, when using conventional ADCs, since the SI power is much larger than that of the SoI, the SI spans most of the ADC’s dynamic range and the SoI becomes distorted by a large quantization noise. Reducing the power of the SI before the ADC, for example via analog domain SI cancellation, is so far the only means to mitigate this undesired effect. In this paper we consider the use of so-called modulo ADCs as a new tool to reduce the quantization noise that affects the SoI in the presence of SI. We demonstrate theoretically and numerically that by substituting the conventional ADCs with modulo-ADCs, which fold the analog input signal before analog-to-digital conversion, and by using appropriately designed analog gain control and digital-domain SI cancellation, one can reduce the quantization noise that affects the SoI. As such, the work in this paper provides a new technique for counteracting the detrimental effect of SI in analog-to-digital conversion and, thus, provides a new route to be explored in order to enhance the performance of FD radios.
Channel charting is a data-driven baseband processing technique consisting in applying unsupervised machine learning techniques to channel state information (CSI), with the objective of reducing the dimension of the data and extracting the fundamental parameters governing the distribution of CSI samples observed by a given receiver. In this work, we focus on neural network-based approaches, and propose a new architecture based on triplets of samples. It allows to simultaneously learn a meaningful similarity metric between CSI samples, on the basis of proximity in their respective acquisition times, and to perform the sought dimensionality reduction. The proposed approach is evaluated on a dataset of measured massive MIMO CSI, and is shown to perform well in comparison to the state-of-the-art methods (UMAP, autoencoders and siamese networks). In particular, we show that the obtained chart representation is topologically close to the geographical user position, despite the fact that the charting approach is not supervised by any geographical data.
Channel charting is a data-driven baseband processing technique consisting in applying unsupervised machine learning techniques to channel state information (CSI), with the objective of reducing the dimension of the data and extracting the fundamental parameters governing the distribution of CSI samples observed by a given receiver. In this work, we focus on neural network-based approaches, and propose a new architecture based on triplets of samples. It allows to simultaneously learn a meaningful similarity metric between CSI samples, on the basis of proximity in their respective acquisition times, and to perform the sought dimensionality reduction. The proposed approach is evaluated on a dataset of measured massive MIMO CSI, and is shown to perform well in comparison to the state-of-the-art methods (UMAP, autoencoders and siamese networks). In particular, we show that the obtained chart representation is topologically close to the geographical user position, despite the fact that the charting approach is not supervised by any geographical data.
We consider the use of deep neural networks (DNNs) in the context of channel state information (CSI)-based localization for Massive MIMO cellular systems. We discuss the practical impairments that are likely to be present in practical CSI estimates, and introduce a principled approach to feature design for CSI-based DNN applications based on the objective of making the features invariant to the considered impairments. We demonstrate the efficiency of this approach by applying it to a dataset constituted of geo-tagged CSI measured in an outdoors campus environment, and training a DNN to estimate the position of the UE on the basis of the CSI. We provide an experimental evaluation of several aspects of that learning approach, including localization accuracy, generalization capability, and data aging.
We present recent experiments on data-driven approaches for baseband processing in wireless communications systems. In particular, we focus on dimensionality reduction applied to channel state information (CSI) data, in the spirit of the channel charting approach introduced by Studer et al. in [1].
The majority of modern communication systems adopt quadrature amplitude modulation (QAM) constellations as transmission schemes. Due to their square structure, however, QAM do not provide satisfying protection to phase noise effects, as the number of constellation points grows, increasing at the same time their peak-to-average-power ratio. This requires an expensive power amplifier and an oscillator at the transmitter to guarantee low distortion, complicating the adoption of dense transmission schemes in practical high-data rate systems. In this paper, we construct a coded modulation scheme based on regular amplitude and phase shift keying modulations. We propose a novel multilevel coding labeling for the constellation points separating the amplitude and phase domains. We provide a novel multistage decoding scheme allowing for a low-complexity log-likelihood ratio calculation for soft-input decoding of component codes, along with a suitable rate design. Finally, we compare the proposed scheme with the state-of-the-art QAM constellations and optimize the constellations in the presence of phase noise.
We report on experimental results on the use of a learning-based approach to infer the location of a mobile user of a cellular network within a cell, for a 5G-type Massive multiple input, multiple output (MIMO) system. We describe how the sample spatial covariance matrix computed from the CSI can be used as the input to a learning algorithm which attempts to relate it to user location. We discuss several learning approaches, and analyze in depth the application of extreme learning machines, for which theoretical approximate performance benchmarks are available, to the localization problem. We validate the proposed approach using experimental data collected on a Huawei 5G testbed, provide some performance and robustness benchmarks, and discuss practical issues related to the deployment of such a technique in 5G networks..
We identify phase noise as a bottleneck for the performance of digital self-interference cancellers that utilize a single auxiliary receiver---single-tap digital cancellers---and operate in multipath propagation environments. Our analysis demonstrates that the degradation due to phase noise is caused by a mismatch between the analog delay of the auxiliary receiver and the different delays of the multipath components of the self-interference signal. We propose a novel multi-tap digital self-interference canceller architecture that is based on multiple auxiliary receivers and a customized Least-Mean-Squared (LMS) filtering for self-interference regeneration. Our simulation results demonstrate that our proposed architecture is more robust to phase noise impairments and can in some cases achieve 10~dB larger self-interference cancellation than the single-tap architecture.
As network deployments become denser, interference arises as a dominant performance degradation factor. To confront with this trend, Long Term Evolution (LTE) incorporated features aiming at enabling cooperation among different base stations, a technique termed as Coordinated Multi Point (CoMP). Recent field trial results and theoretical studies of the performance of CoMP schemes revealed, however, that their gains are not as high as initially expected, despite their large coordination overhead. In this paper, we review recent advanced Coordinated Beamforming (CB) schemes, a special family of CoMP that reduces the coordination overhead through a joint choice of transmit and receive linear filters. We focus on assessing their resilience to uncoordinated interference and Channel State Information (CSI) imperfections, which both severely limit the performance gains of all CoMP schemes. We present a simple yet encompassing system model that aims at incorporating different parameters of interest in the relative interference power and CSI errors, and then utilize it for the performance evaluation of the state-of-the-art in CB schemes. It is shown that blindly applying CB in all system scenarios can indeed be counter-productive.
We consider the use of statistical channel state information (CSI) in the context of a multi-user Massive MIMO cellular communication system. If prior information about per-user channel spatial covariance is available, properly chosen non-orthogonal pilot sequences can significantly reduce the amount of spectral resources dedicated to acquiring instantaneous CSI. In this article, we introduce a CSI acquisition architecture which leverages the use of non-orthogonal pilots. We evaluate the performance of this architecture under practical considerations, including non-ideal covariance estimation and tracking, CSI and feedforward quantization, and realistic LTE-like system and channel models.
We identify phase noise as a bottleneck for the performance of digital self-interference cancellers that utilize a single auxiliary receiver---single-tap digital cancellers---and operate in multipath propagation environments. Our analysis demonstrates that the degradation due to phase noise is caused by a mismatch between the analog delay of the auxiliary receiver and the different delays of the multipath components of the self-interference signal. We propose a novel multi-tap digital self-interference canceller architecture that is based on multiple auxiliary receivers and a customized Least-Mean-Squared (LMS) filtering for self-interference regeneration. Our simulation results demonstrate that our proposed architecture is more robust to phase noise impairments and can in some cases achieve 10~dB larger self-interference cancellation than the single-tap architecture.
The output of oscillators is usually not stable over time. In particular, phase variations—or phase noise—corrupts the oscillations. In this letter, we describe a circuit that is designed to average the phase noise processes and frequency offsets in frequency-matched oscillators. The basic circuit uses the independence of two phase noise processes to provide a cleaner oscillating output with a lower phase noise variance. We describe extensions of the circuit designed to average out more than two oscillators, as well as a single one through delay elements. In all the examples covered, we provide a theoretical analysis of the resulting phase noise process when the input phase noise processes follow a Wiener model.
With the advent of 5G, standardization and research are currently defining the next generation of the radio access. Considering the high constraints imposed by the future standards, disruptive technologies such as Massive MIMO are being proposed. At the heart of this process are wireless channel models, that now need to cover a massive increase in design parameters to consider, a large variety of frequency bands, and very heterogeneous deployments. This tutorial describes how channel models address this new level of complexity and which tools the community prepares to efficiently but accurately capture the upcoming changes in radio access design. We analyze the main drivers behind this overhaul of channel modeling tools, the challenges they pose to the community, and survey the current approaches to overcome them.
Modern cellular networks in traditional frequency bands are notoriously interference-limited especially in urban areas, where base stations are deployed in close proximity to one another. The latest releases of Long Term Evolution (LTE) incorporate features for coordinating downlink transmissions as an efficient means of managing interference. Recent field trial results and theoretical studies of the performance of joint transmission (JT) coordinated multi-point (CoMP) schemes revealed, however, that their gains are not as high as initially expected, despite the large coordination overhead. These schemes are known to be very sensitive to defects in synchronization or information exchange between coordinating bases stations as well as uncoordinated interference. In this article, we review recent advanced coordinated beamforming (CB) schemes as alternatives, requiring less overhead than JT CoMP while achieving good performance in realistic conditions. By stipulating that, in certain LTE scenarios of increasing interest, uncoordinated interference constitutes a major factor in the performance of CoMP techniques at large, we hereby assess the resilience of the state-of-the-art CB to uncoordinated interference. We also describe how these techniques can leverage the latest specifications of current cellular networks, and how they may perform when we consider standardized feedback and coordination. This allows us to identify some key roadblocks and research directions to address as LTE evolves towards the future of mobile communications.
As we move towards densified networks for mobile access, with a larger number of access points, the need for high performance wireless backhaul increases drastically. In this respect, line-of-sight (LoS) multiple-input, multiple-output (MIMO) communication is a promising technique which can enable spatial diversity gains in such links, without increasing the allocated bandwith or overstepping requirements in radiated power. In this paper, we show that under practical conditions, the LoS MIMO channel matrix with dual-polarized (DP) antennas is block-circulant with circulant blocks (BCCB). This fact allows one to blindly diagonalize the channel, even in the absence of channel side-information (CSI) at the transmitter. In this case we show that appropriate precoding steps can greatly improve the performance of non-linear decoding in the LoS MIMO channel.
As we move towards densified networks for mobile access, with a larger number of access points, the need for high performance wireless backhaul increases drastically. In this respect, line-of-sight (LoS) multiple-input, multiple-output (MIMO) communication is a promising technique which can enable spatial diversity gains in such links, without increasing the allocated bandwith or overstepping requirements in radiated power. In this paper, we show that under practical conditions, the LoS MIMO channel matrix with dual-polarized (DP) antennas is block-circulant with circulant blocks (BCCB). This fact allows one to blindly diagonalize the channel, even in the absence of channel side-information (CSI) at the transmitter. In this case we show that appropriate precoding steps can greatly improve the performance of non-linear decoding in the LoS MIMO channel.
Interference Alignment (IA) is technique that, in a large sense, makes use of the increasing signal dimensions available in the system through MIMO and OFDM technologies in order to globally reduce the interference suffered by users in a network. In this paper, we address the problem of downlink cellular networks, the so-called interfering broadcast channels, where mobile users at cell edges may suffer from high interference and thus, poor performance. Starting from the downlink IA scheme proposed by Suh \textit{et al.}, a new approach is proposed where each user feeds back multiple selected received signal directions with high signal-to-interference gain. A scheduler selects a subset of users to be served simultaneously, balancing between sum-rate performance and fairness. The performance of such an approach with an exhaustive search is shown to be good in small dimension systems, but may become untractable in dense network scenarios where many users send simultaneous requests. Therefore, we develop a sub-optimal scheduler that greatly decreases the complexity while preserving a near-optimal data rate gain. Simulation results highlight the efficiency of our proposed scheme, and show a significant spectral efficiency gain over the OFDM scheme in practical scenarios of interest.
LTE-advanced and beyond cells are required to serve an increasing capacity despite their densification and small cells deployment. Tri-sectorial BS with full-reuse (1×3×1) is usually admitted, where each sector behaves as an independent cell, to maximize the user (UE) capacity as well as to simplify the tedious planning phase. As a consequence, the whole time-frequency OFDMA frame is exploitable by each cell but edge users in turns pay the most and suffer from strong interference. The global capacity is anyhow affected when homogeneous traffic is effective (uniform rate condition).
The packet error rate (PER) is a metric of choice to compute the practical performance of communication systems experiencing block fading, e.g., fading processes whose coherence time is relatively slow when compared to the symbol transmission rate. For these types of channels, we derive a closed-form asymptotic expression which approximates the value of the PER for high signal-to-noise ratio (SNR). We also provide another approximation based on a unit-step formulation of the symbol error rate (SER). We show that the two approximations are related and may be derived from one another, thereby allowing us to obtain closed-form approximations of the block fading PER in both coded and uncoded systems. We then show how these approximations may be used in practice, through the derivation of a packet error outage (PEO) metric covering the case where the links experience shadowing on top of block fading, as well as asymptotically optimal power allocations in relay channels under a block fading hypothesis.
As 4G wireless networks are vastly and rapidly deployed worldwide, 5G with its advanced vision of all connected world and zero distance communications is already at the corner. Along with the super quality of user experience brought by these new networks, the shockingly increasing energy consumption of wireless networks has become a worrying economic issue for operators and a big challenge for sustainable development. Green Transmission Technologies (GTT) is a project focusing on the energy-efficient design of physical-layer transmission technologies and MAC-layer radio resource management in wireless networks. In particular, fundamental tradeoffs between spectrum efficiency and energy efficiency have been identified and explored for energy-efficiency-oriented design and optimization. In this article, four selected GTT solutions are introduced, focusing on how they utilize the degrees of freedom in different resource domains, as well as how they balance the tradeoff between energy and spectrum efficiency. On top of the elaboration of separated solutions, the GTT toolbox is introduced as a systematic tool and unified simulation platform to integrate the proposed GTT solutions together.
Cellular networks have been notoriously interference-limited systems in dense urban areas, where base stations are deployed in close proximity to one-another. Recently, a signal processing method called Interference Alignment has emerged, making use of the increasing signal dimensions available in the system through multiple-input multiple output (MIMO) and Orthogonal Frequency Division Multiplexing (OFDM) technologies. In this report, we review the state of the art of interference alignment since its foundation, and we detail algorithms and baseline comparisons to make when applying interference alignment schemes to downlink cellular networks. We also propose a number of research directions of interest which are not yet answered in the current literature.
The exponentially growing traffic in cellular networks induced standardization groups to focus on spectral efficiency (SE) and providers to densify their networks in crowded areas. The price to pay is a significant reduction of the energy efficiency (EE). As a result, more balanced EE-SE solutions have attracted significant interest lately in the wireless community. However, the Pareto optimal bound of this problem is so far not well established. This paper makes a step by defining precisely this bound in a typical cell where the interference-plus-noise distribution is known. An analytical EE-SE bound is derived considering an optimal superposition coding mode and also three sub-optimal time-frequency sharing approaches. The typical noise-plus-interference distribution used in this paper is obtained from Poisson distributed cellular network simulations validated by the Greentouch reference model, but the analytical results broadly apply to any other reference distribution.
While many theoretical and simulation works have highlighted the potential gains of cognitive radio, several technical issues still need to be evaluated from an experimental point of view. Deploying complex heterogeneous system scenarios is tedious, time consuming and hardly reproducible. To address this problem, we have developed a new experimental facility, called CorteXlab, that allows complex multi-node cognitive radio scenarios to be easily deployed and tested by anyone in the world. Our objective is not to design new software defined radio (SDR) nodes, but rather to provide a comprehensive access to a large set of high performance SDR nodes. The CorteXlab facility offers a 167 m2 electromagnetically (EM) shielded room and integrates a set of 24 universal software radio peripherals (USRPs) from National Instruments, 18 PicoSDR nodes from Nutaq and 42 IoT-Lab wireless sensor nodes from Hikob. CorteXlab is built upon the foundations of the SensLAB testbed and is based the free and open-source toolkit GNU Radio. Automation in scenario deployment, experiment start, stop and results collection is performed by an experiment controller, called Minus. CorteXlab is in its final stages of development and is already capable of running test scenarios. In this contribution, we show that CorteXlab is able to easily cope with the usual issues faced by other testbeds providing a reproducible experiment environment for CR experimentation.
ans cette communication, nous dérivons une approximation du taux d'erreur paquet dans les canaux à évanouissements quasi-statiques, où le temps de cohérence du canal est de l'ordre du temps d'émission d'un paquet. Pour différentes hypothèses quant aux modulations considérées et au modèle d'évanouissement du canal, nous obtenue une expression analytique bornant le gain de codage asymptotique du taux d'erreur paquet. Nous étendons ensuite cette étude aux canaux à relais en considérant plusieurs modèles de relayage, pour lesquels nous dérivons l'approximation asymptotique du taux d'erreur paquet de bout-en-bout. A l'aide de cette approximation, nous sommes en mesure de fournir une allocation de puissance optimale pour les différents modèles de canaux à relais, et nous montrons que cette allocation fournit un gain en performance sur toute la plage de rapport signal-sur-bruit.
Dans cette communication, nous montrons que sous une contrainte de puissance globale à répartir entre les noeuds, les bornes supérieures et inférieures decode-and-forward de la capacité d'un canal à relais cohérent sont les mêmes que celle d'un canal à relais non-cohérent équivalent. Cette relation d'équivalence existe pour les modèles full-duplex et half-duplex, où le relais ne peut écouter et transmettre simultanément. L'expression des bornes de capacité étant grandement simplifiée, il est possible en utilisant ce résultat d'extraire très simplement une répartition de puissance optimale sous forme analytique. Cette allocation nous amène à identifier un critère de sélection de relais original, dont nous évaluons les performances en le comparant à ceux utilisés dans la littérature.
Body Area Networks (BANs) exhibit a unique form of channel variations and traditional propagation models fail to adequately match the behavior of BAN communication links. We present here the results of a measurement campaign whose primary objective was to characterize the complete mesh of a BAN and simultaneous analyze the quality of every radio link between the different nodes. We performed several measurement campaigns with indoor and outdoor walking scenarios. The data harvested allows us to highlight symmetry issues in the communications, and we prove that it is due in part to hardware variations in the sensors. Furthermore, the simultaneous measurements of the link quality allows us to extract the correlation in their temporal evolution. We show that scenario-based link correlation matrices cannot be considered, but their evolution over time is stable enough to consider that practical protocols may estimate them with sufficient precision.
In this paper, we study the optimal power allocations in cooperative multiple access channels (CMACs), where we aim at maximizing the rate achievable by both sources simultaneously rather than the sum of achievable rates. Separating our study between the coherent and non-coherent case, we obtain closed-form expressions for the optimal power allocations w.r.t. the outer bounds of the capacity region, as well as decode-and-forward and non-cooperative inner bounds.We point out during our resolution that the general CMAC model behaves as a multiple access relay channel (MARC), where a “virtual” relay node is introduced to represent the cooperation between the sources. This equivalent model simplifies the original power allocation problem. We finally show that the general cut-set outer bound on the capacity region of the equivalent MARC matches exactly the tightest known outer bound on the capacity region of the original CMAC.
Relay channels have been extensively studied in the literature since the seminal paper by Cover and El Gamal. Nevertheless, characterizing the capacity of relay channels still presents open issues. While numerous works addressed this problem with constant powers or targeted the sum-rate optimization, computing the capacity in the case of a global power constraint was less studied. In this paper, we introduce the concept of virtual nodes to derive analytical expressions of the relay channel capacity as a function of the total power. This transformation leads to simple closed-form expressions of the upper bound and decode-and-forward (DF) lower bound on the capacity of the full- and half-duplex relay channels. The half-duplex study is separated into low and high signal-to-noise ratio (SNR) cases. The impact of these approximations is evaluated and found to achieve a large part of the maximal capacity in the worst case where the equivalent received SNR is neither low nor high, typically between 0-10dB.
In this paper, we aim at characterizing the gain induced by using relay channels in a linear network under both capacity constraint and realistic energy model. We express a general model based on a convex optimization problem. Then, we use numerical tools to obtain results on the outer and inner bounds of the capacity of the full and half duplex relay channel. We then further this study with more complex networks based on relay channels, especially networks formed by a linear chain of nodes. We describe the Pareto optimal solutions of the minimization problem with respect to the consumed energy and latency in such a linear network. From the simple case of the linear multi-hop network, we study the gains when implementing a linear chain of relay channels and compare these results to the simpler multi-hop transmission.
In this paper, we present preliminary results on achievables rates in half-duplex cooperative multiple access channels (CMAC).We show that the upper bound on the capacity of the half-duplex CMAC can be solved using convex optimization techniques. Under a Gaussian model, we study the maximal achievable rate by every node in the network. We propose a number of scenarios, encompassing existing and theoretical cooperation schemes. Using these ypotheses, we evaluate the performance of both a non-cooperative concurrent access and simple cooperative multi-hop or relaying schemes with respect to the upper bound. The performance is compared for the various scenarios, and we provide analyses of specific cases in order to illustrate how our framework may be used to answer targeted questions about the capacity of CMACs.
As 4G wireless networks are vastly and rapidly deployed worldwide, 5G with its advanced vision of all connected world and zero distance communications is already at the corner. Along with the super quality of user experience brought by these new networks, the shockingly increasing energy consumption of wireless networks has become a worrying economic issue for operators and a big challenge for sustainable development. Green Transmission Technologies (GTT) is a project focusing on the energy-efficient design of physical-layer transmission technologies and MAC-layer radio resource management in wireless networks. In particular, fundamental tradeoffs between spectrum efficiency and energy efficiency have been identified and explored for energy-efficiency-oriented design and optimization. In this article, four selected GTT solutions are introduced, focusing on how they utilize the degrees of freedom in different resource domains, as well as how they balance the tradeoff between energy and spectrum efficiency. On top of the elaboration of separated solutions, the GTT toolbox is introduced as a systematic tool and unified simulation platform to integrate the proposed GTT solutions together.
Body Area Networks (BAN) offer amazing perspectives to instrument and support humans in many aspects of their lives. In this paper, we consider on-body BAN nodes transmitting information towards a common sink, in a star topology. While this setup is usual in wireless networks, the high instability of the BAN radio channel and the proximity of the body make classical communication protocols inefficient. These networks are further constrained by the low transmission power required for both battery life and health concerns. Opportunistic cooperation techniques are of great interest in such environment to ensure reliable communications. In previous works, we studied simple opportunistic relaying schemes under independent BAN links, using a packet error outage criterion. In this paper, we introduce a more realistic case where shadowing variations around the body are now assumed strongly correlated. Generally speaking, there is a lack of definitive measurements and models for the shadowing correlation in multi-hop networks, while it can play a crucial role at the higher layers. Based on the measurement and simulation results of the French BANET project, we use the BAN context as an illustrative example to exhibit how shadowing correlations have a strong impact on relaying approaches performance.