Seminar

Ce stage explore l’utilisation des Graph Neural Networks (GNNs) pour résoudre conjointement la détection d’activité et l’estimation de canal dans un contexte de communication massive IoT et un protocole de type Grant-Free Random Access. L’objectif est d’étudier la capacité des GNNs à modéliser ce problème et de concevoir une architecture adaptée à ces deux tâches simultanées.

This paper studies indirect reciprocity in the infinitely repeated helping game with random pairings, where the donors control the level of costly benefit the recipients receive. As the cost of help (i.e., cost of cooperation) is privately monitored, cooperative equilibrium generally depends on the players’ beliefs about the amount of help individually provided, thus leading to a coordination problem. We solve this coordination problem in two steps. First, we characterize a class of grim-type strategies that allow sustaining cooperative equilibrium when the cost of cooperation is endogenous. Second, we demonstrate that, in the absence of external coordination, the pursuit of efficiency and risk minimization associated with cooperation is sufficient to induce players to adopt a common, unique equilibrium strategy.

This presentation explores the emerging role of quantum and quantum-inspired optimization techniques in addressing complex combinatorial problems in next-generation 6G wireless communication systems. Focusing on two innovative use cases—fluid-antennas MIMO and 1-bit reconfigurable intelligent surface (RIS)-aided index modulation—it highlights how unconventional computing paradigms like Coherent Ising Machines (CIM) and quantum annealing can efficiently tackle binary-constrained optimization tasks that traditional methods struggle with. Numerical results demonstrate significant performance gains and open pathways for robust, scalable solutions in advanced wireless network design.

Semantic communication aims to convey meaning for effective task execution, but differing latent representations in AI-native devices can cause semantic mismatches that hinder mutual understanding. In this talk we discuss a novel approach to mitigating latent-space misalignment in multi-agent AI-native semantic communications. In a downlink scenario, we consider an access point (AP) communicating with multiple users to accomplish a specific AI-driven task. Our method implements a protocol that shares a semantic pre-equalizer at the AP and local semantic equalizers at user devices, fostering mutual understanding and task-oriented communication while considering power and complexity constraints. To achieve this, we employ a federated optimization for the decentralized training of the semantic equalizers at the AP and user sides. Numerical results validate the proposed approach in goal-oriented semantic communication, revealing key trade-offs among accuracy, communication overhead, complexity, and the semantic proximity of AI-native communication devices.

With the advent of 5G, new higher frequency ranges, the millimter wave (mmWave) bands, were allocated for cellular communication. The deployment of these bands brings new challenges for the design of radio font-ends, such as increased power comsumption. In this talk, an overview of mmWave massive multiple-input multiple-output (MIMO) antenna systems and a methodology for deriving a realistic power consumption model will be presented.

The efficient transmission of short frames is a prerequisite for the effectiveness of a wireless Internet of Things (IoT) network. In classical systems, each message is preceded by a preamble to help in detecting the arrival of the frame and simplify the demodulation operations. To avoid the bandwidth loss introduced by this preamble, we propose to study a new type of frame called Quasi-Cyclic Short Packet (QCSP). The whole QCSP frame can be considered first as a preamble to simply perform the detection and synchronization functions, then as an encoded codeword to correct the transmission errors. A QCSP frame is based on the association of a Cyclic Code Shift Keying (CCSK) modulation with a non-binary error correction code. This work studies the QCSP frame reception problem by combining theoretical aspects with the definition and evaluation of algorithms. We first studied a detection algorithm adapted to QCSP frames. A theoretical model, validated by Monte-Carlo simulation, allows us to fully characterize the proposed algorithm. Then, we develop different time, frequency and phase synchronization algorithms. In particular, we propose to add an over-modulation of CCSK symbols to remove the time synchronization ambiguity at the symbol level. In addition, the non-binary code structure is also used to help the time and phase synchronizations. We formalized the QCSP frame parameter optimization problem as a trade-off between detection, synchronization and decoding performance. Finally, Software Defined Radio (SDR) modules allow us to experimentally validate the theoretical contributions of the work. It is thus possible to transmit 360 bits of information at a very low signal-to-noise ratio (-12 dB) with a transmission time reduced by 23% compared to the use of a classical frame.

Handheld mobile devices, such as cellular phones and navigation receivers, are inherently power and hardware constrained due to their battery-operated nature and compact design. Nevertheless, features like mobility and background applications necessitate continuous data exchange, demanding continuous timing and carrier synchronization. A significant power-consuming component in the receiver chain is radio frequency (RF) analog hardware, particularly the multi-bit analog-to-digital converter (ADC). Power consumption in an ADC is proportional to F_s 2^b , where F_s is the sampling frequency and b is the number of bits per sample. Consequently, power consumption can be significantly reduced by employing one-bit samples. This work explores the feasibility of achieving successful carrier synchronization using only one-bit samples. We exploit Fourier expansion techniques to develop a theoretical framework for carrier parameter estimation at high signal-to-noise ratios (SNR). Based on this theory, we propose a low-complexity estimation technique suitable for high SNR scenarios. Furthermore, we utilize approximation techniques to develop an alternative technique that operates efficiently at low SNR as well.

The theme of this talk is the implementation of extremely miniaturized wireless transceivers for mobile sensor networks. With this goal in mind, I will discuss the design of the single-chip mote, a custom integrated circuit wireless testbed. The goal of the single-chip mote was to remove a radio’s crystal oscillator, thereby dramatically reducing the physical size and energy consumption of a wireless communicating device. Efforts were made to co-integrate the antenna on the chip, further reducing the device’s overall size. Removing a crystal from a radio has a major downside: the radio relies on an inaccurate and imprecise notion of both frequency and time. I will present a number of different network-based calibration strategies that show competitive performance even against crystal-enabled transceivers. Finally, I will discuss future challenges regarding reconfigurability, optimal modulations, and new strategies for chip-to-chip calibration.

Continual learning is a key paradigm for adapting machine learning models to non-stationary environments without forgetting previously learned information (a phenomenon known as catastrophic forgetting). In this seminar, we explore its applications in dynamic wireless channel charting, where evolving radio environments demand real-time adaptation of machine learning models for efficient pseudo-localization. Wireless channel charting involves applying dimensionality reduction techniques to channel state information (CSI) collected during the operation of wireless communication systems. Due to hardware constraints, it is desirable to limit the amount of information stored for that purpose; therefore, we consider online sample selection to perform channel charting using streaming CSI data. Specifically, we focus on dimensionality reduction via triplet contrastive learning, and study several suitable unsupervised replay-based continual learning strategies.

Large-scale MIMO systems have emerged as a cornerstone for next-generation wireless communication networks. While extended research has been conducted on signal processing and transceiver design in these systems, the fundamental Shannon capacity limit remains elusive in many settings, particularly in the presence of system non-linearities. In this talk, we explore the connection between statistics and communication, and introduce a novel approach that leverages information-theoretic asymptotics from Bayesian statistics to derive the Shannon capacity of such systems. We reveal the critical role of the Fisher information and Jeffreys’ prior in this characterization, and demonstrate how to apply this method to derive the asymptotic capacity of various channel models. Examples include the MIMO channels with 1-bit ADC, clipping, phase noise, and imperfect channel state information.

In this presentation i will present MOCSID, a synthetic multi-cell outdoor channel state information (CSI) dataset generated with Sionna ray tracer, designed to mimic a campus scenario with pedestrian mobility and high base station density (10 within a 625 m x 535 m area). MOCSID captures realistic signal propagation characteristics, including line-of-sight and non-line-of-sight conditions, path loss, shadowing, and multipath effects. User mobility is modeled using probabilistic roadmap path finder, with emphasis on spatial consistency to support channel charting research.

Motivated by recursive learning in Markov Decision Processes, this talk addresses the problem of best-arm identification in bandit settings where each arm’s reward is drawn from a multinomial distribution with a known support. We analyze and compare the performance of various strategies, including LUCB, under two scenarios: (1) when the support knowledge is not utilized, and (2) when the support knowledge is leveraged. In the first scenario, classical results from the single-parameter exponential family (SPEF) are employed. In the second, where the estimation of a full probability distribution is required, we incorporate deviation bounds (Hoeffding and Bernstein) on each dimension independently, followed by the Empirical Likelihood method (EL-LUCB) applied to the joint probability vector. The effectiveness of these approaches is validated through simulations across scenarios with varying structural complexities, showcasing the benefits of incorporating support knowledge.

TBA

In this presentation, I’m going to present a novel ultra-low power method of indoor localization of smartphones based on zero-energy-devices beacons instead of active wireless beacons, and the evaluation of detection probability under a strict false alarm probabilityn using Neyman-Pearson approach.

Cette thèse explore l’application des réseaux de neurones à impulsions (Spiking Neural Networks, SNNs) pour les communications sans fil dans le contexte de l’Internet des Objets (IoT), où la gestion de l’énergie constitue un défi majeur. Les Wake-up Radios (WuRs) sont des dispositifs qui permettent aux nœuds IoT de rester en veille avec une consommation d’énergie minimale, en ne s’activant que lors de la réception de signaux spécifiques. Cette étude propose l’utilisation des SNNs comme WuR, le réseau de neurones étant chargé de reconnaître des séquences d’activation spécifiques dans un flux de bits, permettant ainsi l’activation du nœud concerné. Une contribution majeure de cette thèse est l’introduction du modèle “Saturating Leaky Integrate and Fire” (SLIF), qui incorpore un phénomène bio-inspiré appelé “Saturation Synaptique” pour créer un filtre temporel sensible aux intervalles entre impulsions (Inter-Spike Timing). L’étude approfondit les paramètres du modèle SLIF, explore différentes topologies de réseaux, et démontre la pertinence de cette approche pour la reconnaissance de séquences temporelles analogiques. Les résultats établissent les bases pour l’utilisation des réseaux neuromorphiques dans les dispositifs IoT à très faible consommation d’énergie, offrant des perspectives prometteuses pour la conception de WuRs capables de traiter des signaux temporels de manière efficace et économe en énergie.

Detecting active users in a non-orthogonal multiple access (NOMA) network poses a significant challenge for 5G/6G applications. Traditional algorithms tackling this task, relying on classical processors, have to make a compromise between performance and complexity. However, a quantum computing based strategy called quantum annealing (QA) can mitigate this trade-off. In this paper, we first propose a mapping between the AUD searching problem and the identification of the ground state of an Ising Hamiltonian. Then, we compare the execution times of our QA approach for several code domain multiple access (CDMA) scenarios. We evaluate the impact of the cross-correlation properties of the chosen codes in a NOMA network for detecting the active user’s set.

Federated Learning (FL) enables multiple entities with local datasets to collaboratively train machine learning models without requiring data centralization. Most existing FL approaches assume that local datasets are either pre-collected or sampled I.I.D from an unknown distribution. However, this assumption can be too limiting in many scenarios, such as crowd sensing, where sensors continuously collect user data, or model-free reinforcement learning, where agents interact with an evolving environment and train their policy to maximize the reward. In this work, we relax the I.I.D. assumption by modeling each client’s data as coming from an ergodic Markov chain, which converges to their local distribution. We then analyze the sample and communication complexity of two classical FL algorithms—Minibatch SGD and Local SGD. Our findings show that, under this Markovian framework, both algorithms achieve linear speedup with respect to the number of clients, and their communication complexities are unchanged, given that they perform more local computation.

Covert communication, also known as communication with low probability of detection, refers to a scenario where a legitimate receiver and a eavesdropper share the same channel while the sender wishes to prevent the eavesdropper from guessing whether a communication is ongoing or not. In this scenario, the transmitter and receiver share a secret key which gives the receiver the advantage needed in order to decode. It can be shown that the maximum message length that can be transmitted reliably and covertly scales like the square root of the number of channel uses. In this work, we extend previous results for Gaussian channels and study the fundamental limits of covert communications over general memoryless additive-noise channels. Under mild integrability assumptions, we find a general upper bound on the square-root scaling constant. Furthermore, we show that under some additional assumptions, this upper bound is tight. We also provide upper bounds on the length of the secret key required to achieve the optimal scaling.

Many optimization problems are inherently multi-objective. To address them, we formalize Jacobian descent (JD), a direct generalization of gradient descent for vector-valued functions. Each step of this algorithm relies on a Jacobian matrix consisting of one gradient per objective. The aggregator, responsible for reducing this matrix into an update vector, characterizes JD. While the multi-task learning literature already contains a variety of aggregators, they often lack some natural properties. In particular, the update should not conflict with any objective and should scale proportionally to the norm of each gradient. We propose a new aggregator specifically designed to satisfy this. Emphasizing conflict between objectives, we then highlight direct applications for our methods. Most notably, we introduce instance-wise risk minimization (IWRM), a learning paradigm in which the loss of each training example is considered a separate objective. On simple image classification tasks, IWRM exhibits promising results compared to the direct minimization of the average loss. The performance of our aggregator in those experiments also corroborates our theoretical findings. Lastly, as speed is the main limitation of JD, we provide a path towards a more efficient implementation.

In a brief but important 1974 paper, L. R. Welch gave a family of bounds on the maximum modulus of inner products between distinct vectors in a set of m unit vectors in an n-dimensional space. It was noted that these bounds have implications in the design of sequences having desirable correlation properties for multichannel communication systems. Over the years, motivated by several applications, sets that attain the Welch bound have been sought after. On the other hand, mutually unbiased bases (MUBs) are a primitive used in quantum information processing. From a mathematical standpoint, the existence and construction of maximal sets of MUBs in all dimensions pose important open questions. Maximal sets of MUBs turn out to be the same objects as complex projective 2-designs. This talk will discuss these various mathematical objects in connection to sets that attain the Welch bounds.