The proliferation of constrained Internet of Things (IoT) devices requires scalable over-the-air (OTA) firmware update mechanisms. Conventional approaches, predominantly based on one-to-one communication, such as Bluetooth Low Energy (BLE) Device Firmware Update (DFU), suffer from a scalability bottleneck that makes fleet-wide updates inefficient and energy-intensive. To address this limitation, this paper proposes FragCast, a broadcast-based OTA dissemination protocol that leverages the one-to-many communication paradigm of BLE advertising channels. The proposed solution implements an application-layer protocol that transports firmware fragments within custom-formatted BLE advertising packets. A detailed simulation framework was developed to analyse the protocol performance under different scanning profiles and fleet sizes of up to 100 devices. The proposed mechanism was also experimentally validated on Arduino UNO R4 Wi-Fi boards equipped with ESP32-S3 modules. Results show that the broadcast architecture enables efficient fleet-wide dissemination, with discovery latency remaining nearly constant as the number of receivers increases. In the evaluated configuration, a fleet of 100 devices can be updated in approximately 40 s using the Aggressive scanning profile and about 90 s using an energy-efficient iOS-like profile. These findings demonstrate that BLE advertising provides a practical and scalable foundation for OTA firmware dissemination in large fleets of constrained IoT devices.

The sub-THz spectrum offers numerous advantages, including massive multiple-input multiple-output (MIMO) technology with large antenna arrays that enhance spectral efficiency (SE) of future systems. Hybrid precoding (HP) thus emerges as a cost-effective alternative to fully digital precoding regarding complexity and energy consumption. However, sub-THz frequencies introduce hardware challenges, particularly phase noise (PN) from local oscillators (LOs). We analyze PN impact on MIMO systems using HP, leveraging singular value decomposition and common LO architecture. We adopt the Gaussian PN (GPN) model, recognized as accurate for describing PN behavior in sub-THz transmissions. We derive a lower bound on achievable SE and provide closed-form bit error rate expressions for quadrature amplitude modulation (QAM), specifically 4-QAM and 16-QAM, under high-SNR and strong GPN conditions. These analytical results are validated through Monte Carlo simulations. We show that GPN can be effectively counteracted with a single pilot symbol in single-user MIMO systems, unlike single-input single-output systems where mitigation proves infeasible. Simulation results compare conventional QAM against polar-QAM tailored for GPN-impaired systems. Finally, we introduce perspectives for further improvements in performance and energy efficiency.

This paper addresses the peak-to-average power ratio (PAPR) reduction of Orthogonal Frequency Division Multiplexing (OFDM) waveforms under unit modulus constraint for Integrated Sensing and Communication systems (ISAC). We conduct a comparative analysis of radar and communication performance for waveforms optimized using classical PAPR reduction techniques, primarily Tone Reservation (TR) and Selected Mapping (SLM), as well as their state-of-the-art sensing-aware variants, across radar- and communication-centric scenarios. Additionally, we investigate the pilot-aided case with pseudorandom pilot tones. The results reveal a trade-off between PAPR reduction, radar sensing capability, and computational cost, offering practical insights for efficient OFDM-based ISAC system design.

Understanding the spatial and temporal patterns of environmental exposure to radio-frequency electromagnetic fields (RF - EMF) is essential for conducting risk assessments. These assessments aim to explore potential connections between RF-EMF exposure and its effects on human health, as well as on wildlife and plant life. Existing research has used different machine learning tools for EMF exposure estimation; however, a comparative analysis of these techniques is required to better understand their performance for real-world datasets. In this work, we present both finite and infinite-width convolutional network-based methods to estimate and assess EMF exposure levels from 70 real-world sensors in Lille, France. A comparative analysis has been conducted to analyze the performance of the methods' execution time and estimation accuracy. To improve estimation accuracy for higher-resolution grids, we utilized a preconditioned gradient descent method for kernel estimation. Root Mean Square Error (RMSE) is used as the evaluation criterion for comparing the performance of these deep learning models.

Understanding the spatial and temporal patterns of environmental exposure to radio-frequency electromagnetic fields (RF-EMF) is essential for conducting risk assessments. These assessments aim to explore potential connections between RF-EMF exposure and its effects on human health, as well as on wildlife and plant life. Existing research has used different machine learning tools for EMF exposure estimation; however, a comparative analysis of these techniques is required to better understand their performance for real-world datasets. In this work, we present both finite and infinite-width convolutional network-based methods to estimate and assess EMF exposure levels from 70 real-world sensors in Lille, France. A comparative analysis has been conducted to analyze the performance of the methods' execution time and estimation accuracy. To improve estimation accuracy for higher-resolution grids, we utilized a preconditioned gradient descent method for kernel estimation. Root Mean Square Error (RMSE) is used as the evaluation criterion for comparing the performance of these deep learning models.

The support of artificial intelligence (AI) based decision-making is a key element in future 6G networks. Moreover, AI is widely employed in critical applications such as autonomous driving and medical diagnosis. In such applications, using AI as black-box models is risky and challenging. Hence, it is crucial to understand and trust the decisions taken by these models. Tackling this issue can be achieved by developing explainable AI (XAI) schemes that aim to explain the logic behind the black-box model behavior, and thus, ensure its efficient and safe deployment. Highlighting the relevant inputs the black-box model uses to accomplish the desired prediction is essential towards ensuring its interpretability. Recently, we proposed a novel perturbation-based feature selection framework called XAI-CHEST and oriented toward channel estimation in wireless communications. This manuscript provides the detailed theoretical foundations of the XAI-CHEST framework. In particular, we derive the analytical expressions of the XAI-CHEST loss functions and the noise threshold fine-tuning optimization problem. Hence the designed XAI-CHEST delivers a smart low-complex one-shot input feature selection methodology for high-dimensional model input that can further improve the overall performance while optimizing the architecture of the employed model. Simulation results show that the XAI-CHEST framework outperforms the classical feature selection XAI schemes such as local interpretable model-agnostic explanations (LIME) and shapley additive explanations (SHAP), mainly in terms of interpretability resolution as well as providing better performance-complexity trade-off.

The rapid growth of 5G and 6G networks, with their dense deployments, millimeter-wave communications, and dynamic beamforming, necessitates scalable simulation tools for performance evaluation, network design, and electromagnetic field (EMF) exposure assessment. EMF simulations model exposure across frequency, space, and time, typically based on base stations, user devices, and deterministic and empirical models. These simulations are essential for network operators and researchers for network planning, assessing human exposure to RF-EMF, and ensuring compliance with safety regulations considering factors like frequency, power levels, antenna configurations, and the environment, but they often require extensive computational resources and large simulation time. To address this, we propose an infinitely wide convolutional neural network approach for fast and accurate EMF exposure estimation. Remarkably, taking the width of a neural network to infinity allows for improved computational performance. We compute a convolutional neural tangent kernel from the infinite-width network to perform matrix imputation for exposure estimation. Proposed method estimates exposure fields using less than 7% of simulation points and outperforms other machine learning models, predicting exposure levels which fall below the safety limit set by ICNIRP in under 3,66x10(-3) seconds.

Integrated Sensing and Communications (ISAC) has garnered significant attention as a promising technology for next-generation wireless and vehicular communications. Among candidate waveforms, Orthogonal Frequency Division Multiplexing (OFDM) has been extensively investigated over the past decade for its robustness against frequency-selective fading and its favorable ranging performance. However, the waveform's sensing and communication (S C) performance depends strongly on the modulation scheme; while variable-amplitude constellations such as quadrature amplitude (QAM) are more efficient for communication, constant-modulus modulations such as phase shift keying (PSK) are more suitable for sensing. Yet, it remains unclear whether these findings persist under power amplifier (PA) nonlinearity. Because OFDM signals exhibit a high peak-to-average power ratio (PAPR), they require highly linear PAs to avoid distortion, which conflicts with radar requirements, where high transmit power is always beneficial for sensing. In this work, we analyze the effect of PA-induced distortions on the sensing task for PSK and QAM constellations. By introducing the Signal-to-Distortion Ratio (SDR), we examine the extent of the distortion limitation on the ranging task. We complement simulation results with a theoretical characterization of the ambiguity function (AF), thereby explicitly demonstrating how distortion artifacts manifest in the zero-Doppler sidelobes (i.e, ranging sidelobes) and the zero-delay sidelobes. Simulations show that PA distortions impose a palpable performance ceiling for both constellations, reshape the AF, and reduce detection probability, diminishing the theoretical advantage of unimodular signaling and further compromising the OFDM sensing performance with non-uniform envelope signals.

Energy consumption remains a predominant constraint in numerous Internet of Things (IoT) applications, as microcontrollers typically exhibit excessively high power consumption. To address this issue, novel circuit designs have been introduced, and the utilization of spiking neurons and analog computing has emerged as a promising approach, enabling substantial reductions in power consumption. However, operating within the analog domain poses challenges in managing the sequential processing of incoming signals, which is essential for processing wireless communication signals. In this study, we leverage a bio-inspired phenomenon known as Saturating Synapses to create a temporal filter. We present a model of neurons whose synapses respond specifically within a defined range of delays between two incoming spikes, but remain unresponsive when the interspike timing (IST) falls outside this range. Our investigation delves into the model's parameters to better understand their selection criteria and adaptation for the IST. Subsequently, we show the system's efficiency in recognizing specific sequences defined by these ISTs. This article's novelty lies in the proposal of a new analog processing approach for managing temporal sequences, which permits a much lower energy consumption compared to classical approaches.

In this paper, we investigate the secrecy energy-efficiency (SEE) of a multi-user downlink non-orthogonal multiple access (NOMA) system assisted by multiple ambient backscatter communications (AmBC) in the presence of a passive eavesdropper. We analyze both the trade-off and the ratio between the achievable secrecy sum-rate and total power consumption. In the special case of two backscatter devices (BDs), we derive closed-form solutions for the optimal reflection coefficients and power allocation by exploiting the structure of the SEE objective and the Pareto boundary of the feasible set. When more than two BDs are present, the problem becomes analytically intractable. To address this, we propose two efficient optimization techniques: (i) an exhaustive grid- based benchmark method, and (ii) a scalable particle swarm optimization algorithm. Furthermore, we design a deep learning-based predictor using a feedforward neural network (FNN), which closely approximates the optimal solutions. Numerical results show that the inclusion of AmBC significantly improves SEE, with gains up to 615

In this paper, we investigate the secrecy energy-efficiency of a multi-user downlink non-orthogonal multiple access (NOMA) system with ambient backscatter communication in the presence of a passive eavesdropper. We formulate the optimization problem as a trade-off between secrecy sum-rate and total power consumption, which is inherently non-convex due to interference and variable coupling. A closed-form solution for the optimal reflection coefficient is derived, transforming the optimization problem into a convex power allocation that has been recently solved in a closed form. The solution can also be utilized to efficiently maximize the ratio between the secrecy sum-rate and power consumption, requiring only a line search. Simulation results show that the proposed approach outperforms conventional NOMA and orthogonal multiple access (OMA) schemes, with or without backscatter, in terms of secrecy energy-efficiency.

In an environment with strict constraints on maximum permissible delay and a high density of interconnected devices, achieving global system coordination becomes impractical. This lack of coordination results in increased noise due to interference. Under such conditions, interference levels can vary significantly from one packet to another, with only their statistical properties being estimable before transmission. Notably, the variance of the resulting noise remains unpredictable. To address this challenge, we propose an approach to optimize the modulation and coding schemes. Our method models interference using a mixture of exponential distributions. To accurately estimate its parameters, we employ a bootstrap method. We also propose a Quickest Change Detection approach to identify changes in the interference distribution. This approach allows us to determine transmission parameters that ensure a predefined success probability for transmitted packets, without requiring additional listening and estimation tasks on the end devices.

Deep learning (DL) algorithms have been widely integrated in various aspects of wireless communications research. In this paper, we investigate, in an Internet of Things context, the secrecy energy-efficiency (SEE) of a multi-user downlink non-orthogonal multiple access (NOMA) system in the presence of a passive eavesdropper. Hence, we formulate the convex optimization problem as maximizing the SEE defined as the trade-off between the secrecy sum-rate and the power budget. Notably, this optimization problem has been recently solved in a closed form. The solution can also be utilized to efficiently maximize the ratio between the secrecy sum-rate and power consumption, requiring only a line search. This approach is then used to generate training, validation, and test datasets. Our method relies on a deep neural network designed for resource allocation. The benefits of using a deep neural network include achieving optimal resource allocation results while minimizing complexity and latency. The results presented in this paper highlight the superiority and efficacy of DL optimization compared to traditional iterative search methods.

This paper presents a deep learning-based successive interference cancellation (SIC) scheme to improve the uplink in the LoRa Network. The proposed receiver can decode simultaneous transmissions from multiple users on the same frequency channel with the same spreading factor. The objective of the proposed approach is to reduce the error propagation issues of the classical SIC method. Each SIC step employs a specific convolutional neural network model to decode each user’s signal directly. Simulation results show that the proposed receiver significantly reduces the error propagation and increases the number of connected devices in the network compared to the classical SIC scheme.

Research into 6G networks has been initiated to support a variety of critical artificial intelligence (AI) assisted applications such as autonomous driving. In such applications, AI-based decisions should be performed in a real-time manner. These decisions include resource allocation, localization, channel estimation, etc. Considering the black-box nature of existing AI-based models, it is highly challenging to understand and trust the decision-making behavior of such models. Therefore, explaining the logic behind those models through explainable AI (XAI) techniques is essential for their employment in critical applications. This manuscript proposes a novel XAI-based channel estimation (XAI-CHEST) scheme that provides detailed reasonable interpretability of the deep learning (DL) models that are employed in doubly-selective channel estimation. The aim of the proposed XAI-CHEST scheme is to identify the relevant model inputs by inducing high noise on the irrelevant ones. As a result, the behavior of the studied DL-based channel estimators can be further analyzed and evaluated based on the generated interpretations. Simulation results show that the proposed XAI-CHEST scheme provides valid interpretations of the DL-based channel estimators for different scenarios.

IoT devices, constrained by limited resources, must balance energy consumption and performance, with the communication module often being the largest energy consumer. Among various communication technologies, Bluetooth Low Energy (BLE) is notable for its efficient design and performance. However, despite recent BLE optimizations, the Neighbor Discovery Process (NDP) remains energy-intensive. We propose a new approach that uses a modern microcontroller's multicore architecture to dynamically adjust the scanning duration based on the number of devices to be discovered. Our simulations show that avoiding a fixed scanning duration by concluding the NDP once the required devices are found can significantly reduce energy consumption. Validation through simulations and real-world experiments demonstrates up to 50% energy savings compared to a fixed 1-second scan using a single core during the NDP.

Electromagnetic field exposure (EMF) has grown to be a critical concern as a consequence of the ongoing installation of fifth-generation cellular networks (5G). The lack of measurements makes it difficult to accurately assess the EMF in a specific urban area, as Spectrum cartography (SC) relies on a set of measurements recorded by spatially distributed sensors for the generation of exposure maps. However, when the spatial sampling rate is limited, significant estimation errors occur. To overcome this issue, the exposure map estimation is addressed as a missing data imputation task. We compute a convolutional neural tangent kernel (CNTK) for an infinitely wide convolutional neural network whose training dynamics can be completely described by a closed-form formula. This CNTK is employed to impute the target matrix and estimate EMF exposure from few sensors sparsely located in an urban environment. Experimental results show that the kernel, even when only sparse sensor data are available, can produce accurate estimates. It is a promising solution for exposure map reconstruction that does not require large training sets. The proposed method is compared with other deep learning approaches and Gaussian Process regression.

Energy constraints are still a significant challenge in numerous IoT applications, particularly due to the excessive power consumption of microcontrollers. To overcome this limitation, novel circuit designs have been introduced, with the integration of spiking neurons and analog computing emerging as a promising solution, facilitating substantial reductions in power consumption. However, the operation within the analog domain introduces complexities in managing the sequential processing of incoming signals, a critical requirement for diverse applications. This study employs the Saturating Synapses Leaky Integrate and Fire (SLIF) model, a bio-inspired neuron model, to develop a signature recognition system based on a Spiking Neural Network, without the need of non-biological techniques such as synaptic delays. SLIF neurons exhibit spiking behavior exclusively in response to two consecutive spikes with an Inter Spike Timing (IST) within a specific range, remaining unresponsive to other ISTs. We present the joint design of IST-based signatures and the corresponding network. Subsequently, we evaluate the system's efficiency in recognizing its specific sequence and discriminating against alternative sequences. The novelty of this paper lies in the proposition of a new type of temporal sequence recognition networks based on ISTs, offering significantly lower energy consumption compared to conventional approaches.

In Spectrum cartography (SC), the generation of exposure maps for radio frequency electromagnetic fields (RF-EMF) spans dimensions of frequency, space, and time, which relies on a sparse collection of sensor data, posing a challenging ill-posed inverse problem. Cartography methods based on models integrate designed priors, such as sparsity and low-rank structures, to refine the solution of this inverse problem. In our previous work, EMF exposure map reconstruction was achieved by Generative Adversarial Networks (GANs) where physical laws or structural constraints were employed as a prior, but they require a large amount of labeled data or simulated full maps for training to produce efficient results. In this paper, we present a method to reconstruct EMF exposure maps using only the generator network in GANs which does not require explicit training, thus overcoming the limitations of GANs, such as using reference full exposure maps. This approach uses a prior from sensor data as Local Image Prior (LIP) captured by deep convolutional generative networks independent of learning the network parameters from images in an urban environment. Experimental results show that, even when only sparse sensor data are available, our method can produce accurate estimates.