In anti-jamming wireless communications, single-channel mix source separation (SCMSS) is an effective way to combat full-band jamming. Conventional SCMSS methods typically depend on prior knowledge of jamming, and existing deep learning-based SCMSS methods require extensive training samples, limiting their applications in practical scenarios. Alternatively, we exploit large language models (LLMs) to design a novel SCMSS method, including a new LLM-based deep neural network (DNN), and a new fine-tuning algorithm. By harnessing the extremely powerful feature extraction and cross-domain knowledge transfer capabilities of LLMs, our method can effectively separate target signals from full-band jamming after fine-tuning with a small number of labeled data samples. Experimental results demonstrate superior bit error ratio (BER) and generalization performance over traditional and existing deep learning-based anti-jamming methods with significantly reduced training samples.

Accurate behavioral modeling of power amplifiers (PAs) is crucial for digital pre-distortion and energy-efficient vehicular wireless transceivers, yet remains challenging due to bandwidth-dependent nonlinearities and fast-varying drive conditions in practical RF front-ends. Conventional PA models struggle to capture coupled temporal memory effects and spatial signal dependencies, limiting their effectiveness in broadband vehicular transmitters. To address this issue, we propose a Kolmogorov–Arnold Convolutional Liquid Neural Network (KACLN) that jointly leverages Kolmogorov–Arnold representation learning and liquid neuron adaptive dynamics to model nonlinear distortion with strong temporal generalization. The proposed KACLN integrates one-dimensional and two-dimensional convolutional operator kernels to simultaneously characterize long- and short-term PA memory, frequency-selective distortion patterns, and cross-sample bandwidth interactions. Experiments over diverse vehicular RF operating regimes demonstrate that KACLN improves modeling precision and robustness, achieving 1.789 dB normalized mean square error (NMSE) reduction and 2.639 dB adjacent channel error power ratio (ACEPR) suppression gains compared with the best performed MCLDNN. The proposed KACLN further mitigates efficiency degradation associated with static power back-off, enhancing effective output utilization under fast-varying channel and wideband signal envelopes. These results validate KACLN as a practical, real-time capable nonlinear surrogate suitable for next-generation vehicular PA linearization, DPD-aided waveform shaping, and low-latency wireless transmitter design.

This paper presents a resilient multi-cell distributed antenna system (DAS) capable of reconfiguring user-clusters and cell structures after antenna failure. A constrained K-means algorithm is employed to jointly optimize user-clustering, ensuring balanced load and effective antenna sharing under various antenna failure conditions. Two types of cell-restructuring are considered: a fixed-cell-number (FCN) approach that keeps the number of cells constant, and a variable-cell-number (VCN) approach that reduces the total number of cells compared with the normal state. Simulation results demonstrate that, even with a 50% antenna failure ratio, the proposed multi-cell DAS recovers over 40% of the original link capacity for the worst 1% users by cell-restructuring.

We investigate covert communications in an intelligent reflecting surface (IRS)-assisted symbiotic radio (SR) system under the parasitic SR (PSR) and the commensal SR (CSR) cases, where an IRS is exploited to create a double reflection link for legitimate users and degrade the detection performance of the warden (W). Specifically, we derive an analytical expression for the average detection error probability of W and design an optimal strategy to determine the transmit power and backscatter reflection coefficient. To further enhance the covert performance based on the derived results, the joint optimization of the source transmit power, backscatter device reflection coefficient, and IRS phase-shifter is formulated as an exponential-based quadratic-fractional problem. By reformulating the original problem into a quartic polynomial problem, we further develop the phase alignment pursuit and the power leakage minimization algorithms for the PSR and the CSR cases, respectively. Numerical results confirm the accuracy of the derived results and the superiority of our proposed strategy in terms of covertness.

A fluid antenna system (FAS) can improve both energy harvesting and data transmission in wireless powered communication (WPC) by exploiting spatial diversity with a single RF chain. In this paper, we study the outage probability of a FAS-assisted WPC network under the port selection criterion that maximizes the product of the channel gains of the two links. Adopting the block-correlation (BC) decomposition fitted to the Jakes’ autocorrelation function, we derive a semi-analytical outage expression via nested Gauss–Chebyshev quadrature with singularity-removing variable substitutions. Numerical results under the exact Jakes’ correlation confirm the analytical predictions and show that exploiting both links in port selection yields much lower outage probabilities than single-link selection. Moreover, the advantage becomes more pronounced as the aperture increases, and for sufficiently large apertures it further increases with the number of ports. For the considered settings, significant diversity gains from increasing the number of ports are observed at W = 4λ.

Massive multiple-input multiple-output (MIMO) array systems are a cornerstone technology for beyond fifth-generation (B5G) and sixth-generation (6G) wireless communications. This paper proposes a novel algorithm for the joint estimation of direction-of-departure (DOD) and direction-of-arrival (DOA) in massive MIMO systems under unknown gain-phase errors. The proposed method first exploits a normalized rotational invariance property to extract the relative amplitude-phase difference vectors between adjacent antenna elements. By incorporating the prior knowledge that the transmitter and receiver phase errors follow a zero-mean distribution, we formulate two decoupled cost functions to enable joint DOD and DOA estimation. We then obtain that the corresponding angular parameters efficiently through low-complexity spectral searches. Notably, the proposed method requires only one well-calibrated transmitter and one well-calibrated receiver, thereby substantially reducing the calibration effort compared with existing approaches. The gain errors are directly estimated from the amplitude-phase difference vectors, while the phase error vectors are reconstructed using the estimated DOD and DOA values. The proposed framework accommodates general conformal transceiver array geometries and effectively mitigates error accumulation in gain-phase calibration. Simulation results verify that the proposed algorithm achieves superior angular estimation accuracy and calibration precision compared with state-of-the-art techniques.

By leveraging the position flexibility of the emerging fluid antenna (FA) technology, this correspondence investigates the joint optimization of beamforming and antenna position in FA-assisted dual-functional radar-communication systems. To characterize the performance trade-off between radar and communication functionalities, we propose to maximize the rate region of the system under both transmit power and antenna position constraints. To solve the formulated non-convex problem, an alternating optimization based method is proposed, where a semidefinite relaxation and an exponential gradient descent-based method are adopted to solve the beamforming and the antenna position optimization subproblems, respectively. Simulation results demonstrate that the proposed scheme effectively reduces grating lobes and enhances the overall system performance in terms of both radar sensing and communication throughput.

Synthetic Aperture Radar (SAR) target recognition faces great challenges in open environments due to unknown targets, which introduces significant epistemic uncertainty. Traditional closed-set methods often fail to reject outliers and suffer from overconfidence, leading to high false alarm rates. To address this, we propose a novel Evidential Complex-Valued Network (ECVNet) for explicit epistemic uncertainty quantification in open-set SAR recognition, synergizing Composite Pseudo-Labeling (CPL), Complex-Valued Neural Networks (CVNNs), and Generalized Quantum Evidence Theory (GQET). Inspired by GQET, CPL builds an evidential label space via unsupervised clustering, explicitly modeling singleton class probabilities, inter-class confusion, and unknown prototypes—key components of epistemic uncertainty in OSR. CVNNs process raw complex SAR data to leverage amplitude-phase coupling, extracting robust features compatible with GQET’s Hilbert space representation. By optimizing the network to regress toward evidence-based soft labels, probability mass is reserved for the open space, and epistemic uncertainty is rigorously quantified. Experiments on MSTAR, EuroSAT, and FUSAR-Ship show that ECVNet outperforms baselines in closed-set accuracy and open-set unknown rejection, with marked FPR95 reductions, thanks to reliable epistemic uncertainty quantification.

The synergy of fluid antenna systems (FAS) and reconfigurable intelligent surfaces (RIS) promises robust vehicle-to-everything (V2X) links, yet most analyses invoke the central limit theorem (CLT) and thus fail to capture practical, finite-size deployments. This paper develops a realistic and tractable framework for FAS-RIS systems with a finite number of elements. We approximate the cascaded end-to-end gain via a Gamma distribution using moment matching, model spatial dependence across FAS ports with a block-correlation structure, and derive an accurate, closed-form approximation for the outage probability using Gauss-Chebyshev quadrature. Extensive simulations show that the proposed Gamma-based analysis markedly outperforms CLT-based baselines-especially for small numbers of RIS elements and ports-while converging to them as the array grows. The results provide actionable guidance for V2X design and dimensioning under practical constraints on RIS size, FAS aperture, and channel correlation.

Artificial noise (AN) has been recognized as an effective physical-layer security scheme impairing the eavesdropper (Eve). Recently, artificial noise elimination (ANE) has emerged as a promising strategy to mitigate the impact of AN at Eves. However, conventional ANE schemes rely on prior knowledge, such as legitimate channel state information (CSI) or classification information, which may limit their practical applicability. To address these practical challenges, we propose an ANE scheme beyond prior knowledge (BPK) by leveraging machine learning algorithms. Firstly, a coarse projection is applied to partially eliminate the impact of AN using maximum likelihood estimation on the equivalent AN matrix. Secondly, a density clustering algorithm is introduced to obtain classification information based on the coarsely-projected observed vectors. Thirdly, a generalized principal component analysis (PCA)-based ANE algorithm is developed to effectively mitigate the residual AN using the obtained classification information. Furthermore, the artificial-noise-to-signal ratio (ANSR) and computational complexity are analyzed for performance revaluation, and a redefinition of several AN design principles is provided for scenarios involving a powerful Eve equipped with the BPK-ANE scheme by deriving the validity boundary. Finally, numerical results reveal key insights into four principles of AN: 1) Allocating less power to AN; 2) Reducing the randomness of AN; 3) Increasing the number of transmit antennas; and 4) Increasing the modulation order.

Hybrid analog-digital structures (HADS) have emerged as an efficient solution for mitigating transmission loss and reducing power consumption in multiple-input multiple-output (MIMO) systems. However, the limited number of radio frequency (RF) chains and the presence of array mutual coupling (MC) pose significant challenges to achieving high-performance direction-of-arrival (DOA) estimation, thereby hindering effective downlink beamforming. To address these challenges, an efficient DOA estimation method specifically designed for HADS is proposed, effectively mitigating the impact of MC by employing a two-stage framework. The first stage reconstructs the spatial covariance matrix (SCM) by adjusting switch states and leveraging a middle subarray, combined with the real-valued subspace technique for initial DOA estimation. Using these initial estimates, MC is modeled and compensated by adjusting amplifiers and phase shifters. In the second stage, enhanced DOA estimation is achieved by fully exploiting the data from the entire array. Simulation results validate the effectiveness of the proposed method, demonstrating its capability to mitigate the MC effect and deliver accurate DOA estimation under practical conditions.

This paper presents a WiFi Channel State Information (CSI)-based human activity recognition framework that fuses class-aware textual prompts with CSI patch6 tokens through a Prompt-Query CrossModal Attention front end. A lightweight “patch reprogramming” layer aligns CSI tokens to the languageembedding space, after which a frozen GPT-2 performs contextual modeling; only small task-specific heads are trained. To improve representation learning without altering inference, three optional training-only auxiliaries, supervised contrastive learning, masked reconstruction, and temporal order prediction-act as complementary regularizers. The overall objective is optimized as a simple weighted sum without uncertainty weighting. Experiments on two public CSI datasets show consistent gains over representative Convolutional Neural Network (CNN), Recurrent Neural Network (RNN), Residual Network (ResNet), and Vision Transformer (ViT) baselines, faster and smoother convergence, robustness under Signal-to-Noise Ratio (SNR) degradation (where SNR refers to the quality of the received WiFi signal in typical indoor multipath environments), and cleaner confusion matrices on hard class pairs. These results suggest that injecting textual priors and aligning CSI tokens before using a frozen LLM is an effective way to create stronger, noise-resilient CSI representations.

In surface micro-vibration-based water–air cross-medium communication technology, existing research primarily focuses on idealized experimental settings involving single underwater sound sources and noise-free conditions. However, real-world ocean environments are often accompanied by multi-source interference, complex background noise, and nonlinear medium propagation, which pose significant challenges. To address these issues, this paper proposes an underdetermined blind source separation framework that combines a dual-feature-optimized generative adversarial network with subspace-based sparse feature extraction for surface micro-vibration signal detection. Specifically, a dual-feature-optimized U-Net GAN is designed to fuse temporal and time-frequency features, with blind source separation accuracy introduced as a feedback mechanism to guide adversarial training and improve the reconstruction of non-linear mixed signals. Subsequently, a subspace covariance-based mixing matrix estimation method is employed, coupled with a sparse-domain inverse transformation algorithm, to achieve accurate separation and reconstruction of source signals under underdetermined conditions. Experimental results demonstrate that the reconstructed source signals achieve 4.33% improvement in waveform similarity, 1.81 increase in source signal energy proportion, and 1.28 reduction in spectral distance compared to baseline algorithms.

Spectrum prediction plays a pivotal role in the cognitive radio process, facilitating spectrum management and enhancing spectrum utilization. Due to the complexity and spatiotemporal correlation of spectrum data, spectrum prediction remains a challenging issue today. Building upon previous work, this paper addresses the insufficiency of spectrum spatial correlation in conventional prediction methods and proposes a novel spectrum prediction approach based on Closed-form Continuous-time Neural Network With Graph Attention (GA-CfC), which achieves precise prediction of multidimensional spectrum data. This method employs a Graph Attention Network (GAT) to extract spectral spatial correlations from data that has been preliminarily processed by convolutional layers across different frequency bands. The extracted features are then merged with the original spectrum data and fed into a CfC for iterative learning in the temporal dimension. Finally, a prediction is made based on temporal and spectral correlations using an attention mechanism-based hidden layer integration module. We conducted experiments on real-world data to validate the excellent performance of our method and compared it with other spectrum prediction methods. The experimental results demonstrate that our proposed method exhibits superior predictive performance.

In this paper, we analyze the role of fluid antenna systems (FAS) in multi-user systems with hardware impairments (HIs). Specifically, we investigate a scenario where a base station (BS) equipped with multiple fluid antennas communicates with multiple communication users (CUs), each equipped with a single fluid antenna. Our objective is to maximize the minimum communication rate among all users by jointly optimizing the BS's transmit beamforming, the positions of its transmit fluid antennas, and the positions of the CUs' receive fluid antennas. To address this non-convex problem, we propose a block coordinate descent (BCD) algorithm integrating semidefinite relaxation (SDR), rank-one constraint relaxation (SRCR), successive convex approximation (SCA), and majorization-minimization (MM). Simulation results demonstrate that FAS significantly enhances system performance and robustness, with notable gains when both the BS and CUs are equipped with fluid antennas. Even under low transmit power conditions, deploying FAS at the BS alone yields substantial performance gains. However, the effectiveness of FAS depends on the availability of sufficient movement space, as space constraints may limit its benefits compared to fixed antenna strategies. Our findings highlight the potential of FAS to mitigate HIs and enhance multi-user system performance, while emphasizing the need for practical deployment considerations.

The rapid development of wireless communication and Internet of Things (IoT) devices has exacerbated the challenge of spectrum scarcity. Spectrum prediction plays a critical role in enhancing resource utilization and facilitating dynamic spectrum access (DSA). However, existing methods often face difficulties with missing data arising from sensor failures or environmental changes, while prioritizing prediction accuracy over computational efficiency, hindering deployment. To overcome these challenges, this paper introduces the generative augmented cascade broad learning network (GA-CBLN) for spectrum prediction. This approach combines a deep learning (DL)-based pre-trained model for data imputation with an online broad learning (BL) process to ensure efficient prediction. Initially, a DL-based regressor imputes missing values by learning spatiotemporal patterns, generating a complete dataset. Subsequently, a generative adversarial network (GAN) produces synthetic samples to augment the dataset, enhancing the model’s adaptability and generalization. Finally, the cascade broad learning network (CBLN) conducts feature extraction and prediction using both original and augmented data, exploiting the efficient learning capabilities of BL. Simulations on real spectrum datasets from four widely used frequency bands show that GA-CBLN surpasses traditional methods in computational efficiency, maintaining robust performance even under high rates of missing data.

UAV RF surveillance is becoming a critical layer of low-altitude airspace security, yet most fingerprinting pipelines still assume that every signal belongs to a known class. In open-world deployments, where new platforms, protocol updates, and changing propagation conditions are routine, that assumption makes conventional classifiers brittle and often overconfident on unseen emitters. Foundation models offer a compelling alternative because their pretrained priors can support richer sequence understanding beyond task-specific training. We present SkyLLM, a foundation-model-driven framework that aligns RF time-series with the token space of a frozen LLM through patch-based embedding, cross-channel fusion, and prompt conditioning, so that waveform fragments can be interpreted in a representation space better suited to novelty-aware decision making. Experiments on real UAV RF data show that SkyLLM preserves strong discrimination while delivering markedly more reliable unknown rejection than lightweight baselines across increasingly open settings. More broadly, the study suggests that future trusted spectrum intelligence will depend not only on better classifiers, but on representation layers that remain stable as the airspace evolves. SkyLLM therefore offers a practical path toward 6G-era UAV monitoring systems that must combine recognition, uncertainty awareness, and operational trust.

Fluid antenna systems (FAS) are emerging as a transformative enabler for sixth-generation (6G) wireless communications, providing unprecedented spatial diversity through dynamic reconfiguration of antenna ports. However, the inherent spatial correlation among ports poses significant challenges for accurate analysis. Conventional models such as Jakes are analytically intractable, while oversimplified constant-correlation models fail to capture the true behavior. in this work, we address these challenges by applying the variable block-correlation model (VBCM)-originally proposed by Ramirez-Espinosa et al. in 2024-to FAS security analysis, and by developing comprehensive optimization methods to enhance analytical accuracy. We derive new closed-form expressions for average secrecy capacity (ASC) and secrecy outage probability (SOP), demonstrating that the VBCM framework achieves simulation-aligned accuracy, with relative errors consistently below 5% (compared to 10-15% for constant-correlation models). To maximize ASC, we further design two algorithms: a grid search (GS) method and a gradient descent (GD) method. Numerical results reveal that the VBCM-based approach not only provides reliable insights into FAS security performance, but also yields substantial gains-ASC improvements exceeding 120% in high-threat scenarios and 18-19% performance enhancements for compact antenna configurations. These findings underscore the practical value of integrating VBCM into FAS security analysis and optimization, establishing it as a powerful tool for advancing 6G communication systems.

The openness of the space environment renders satellite Internet vulnerable to external malicious attacks. As a typical approach of physical-layer authentication, radio frequency fingerprint (RFF) identification can be exploited to identify wireless user devices by extracting their unique hardware characteristics, thereby becoming a promising solution to strengthen the authentication security of satellite Internet. However, since satellite Internet is expected to provide globally seamless services to massive users, it is not only overhead-costly but also privacy-risky to construct a centralized RFF identification model by sharing wireless signals of all users in a huge dataset for RFF extraction. Fortunately, by collaborative model training among satellites, federated learning (FL) offers a viable solution for building up an on-orbit distributed RFF identification model that does not require sharing the received signals among satellites. Inspired by this fact, this article investigates the multi-satellite FL framework design for on-orbit RFF identification. In specific, we first briefly introduce the fundamentals of RFF identification. Then, following the idea of pioneering work [1], a vanilla multi-satellite FL framework is presented for on-orbit RFF identification, followed by technical challenges of its practical application, including heterogeneous local datasets, non-identical satellite resources, and dynamic link topology. To overcome these challenges, we further propose a hierarchically collaborative multi-satellite FL framework, under which three key techniques, say scheduling management, model aggregation, and model personalization are also investigated in-depth. Finally, we outline the future research directions and challenges.