Radio maps (RMs) serve as a critical foundation for enabling environment-aware wireless communication, as they provide the spatial distribution of wireless channel characteristics. Despite recent progress in RM construction using data-driven approaches, most existing methods focus solely on pathloss prediction in a fixed 2D plane, neglecting key parameters such as direction of arrival (DoA), time of arrival (ToA), and vertical spatial variations. Such a limitation is primarily due to the reliance on static learning paradigms, which hinder generalization beyond the training data distribution. To address these challenges, we propose UrbanRadio3D, a large-scale, high-resolution 3D RM dataset constructed via ray tracing in realistic urban environments. UrbanRadio3D is over 37x larger than previous datasets across a 3D space with 3 metrics as pathloss, DoA, and ToA, forming a novel 3Dx 3D dataset with 7x more height layers than prior state-of-the-art (SOTA) dataset. To benchmark 3D RM construction, a UNet with 3D convolutional operators is proposed. Moreover, we further introduce RadioDiff-3D, a diffusion-model-based generative framework utilizing the 3D convolutional architecture. RadioDiff-3D supports both radiation-aware scenarios with known transmitter locations and radiation-unaware settings based on sparse spatial observations. Extensive evaluations on UrbanRadio3D validate that RadioDiff-3D achieves superior performance in constructing rich, high-dimensional radio maps under diverse environmental dynamics. This work provides a foundational dataset and benchmark for future research in 3D environment-aware communication.

Integrating large language models (LLMs) into wireless communication optimization is a promising yet challenging direction. Existing approaches either use LLMs as black-box solvers or code generators, tightly coupling them with numerical computation. However, LLMs lack the precision required for physical-layer optimization, and the scarcity of wireless training data makes domain-specific fine-tuning impractical. We propose BeamAgent, an LLM-aided MIMO beamforming framework that explicitly decouples semantic intent parsing from numerical optimization. The LLM serves solely as a semantic translator that converts natural language descriptions into structured spatial constraints. A dedicated gradient-based optimizer then jointly solves the discrete base station site selection and continuous precoding design through an alternating optimization algorithm. A scene-aware prompt enables grounded spatial reasoning without fine-tuning, and a multi-round interaction mechanism with dual-layer intent classification ensures robust constraint verification. A penalty-based loss function enforces dark-zone power constraints while releasing optimization degrees of freedom for bright-zone gain maximization. Experiments on a ray-tracing-based urban MIMO scenario show that BeamAgent achieves a bright-zone power of 84.0 dB, outperforming exhaustive zero-forcing by 7.1 dB under the same dark-zone constraint. The end-to-end system reaches within 3.3 dB of the expert upper bound, with the full optimization completing in under 2 s on a laptop.

This paper investigates a multi-Unmanned Aerial Vehicle (UAV) joint base station-assisted Internet of Vehicles (IoV) task offloading system in dense urban environments. To minimize system delay and energy consumption under strict coupling constraints, the complex non-convex optimization problem is decoupled into a hierarchical execution framework. First, a sequential distributed optimization algorithm based on Second-Order Cone Programming (SOCP) is proposed to optimize the 3D flight trajectory of each UAV, ensuring adaptive network coverage. Second, a novel hybrid resource scheduling paradigm synergizing Deep Reinforcement Learning (DRL) and Large Language Models (LLMs) is developed. Within this framework, the DRL agent dictates the initial resource allocation, while the LLM acts as a semantic macro-scheduler to rectify long-tail allocation imbalances for failed and surplus tasks. Crucially, a reward decoupling mechanism is introduced to isolate DRL training from external LLM interventions, thereby ensuring policy convergence. Finally, the task offloading ratios are precisely determined via Linear Programming (LP) within an alternating optimization loop. Simulation results demonstrate that the proposed method significantly outperforms traditional multi-agent reinforcement learning baselines in terms of task success rate and system efficiency.

The paradigm of semantic communication prioritizes the recovery of semantically meaningful information over pixel-level fidelity, which is crucial for bandwidth-constrained and extreme environments such as deep-space exploration and disaster response. However, existing deep learning-based approaches often rely on task-specific training and struggle with the compound distortion of aggressive compression and severe channel noise. In this paper, we formulate semantic-aware image transmission as a signal recovery problem from incomplete and noisy measurements, guided by semantic importance. We propose SemAID, a novel framework that integrates an adaptive, semantic-driven non-uniform sampler at the transmitter and a generative diffusion-based Bayesian estimator at the receiver. The transmitter dynamically masks pixels based on saliency and edge information, effectively implementing content-aware compression. The receiver, leveraging a pre-trained diffusion model as a strong generative prior, solves the corresponding inverse problem via posterior sampling without any retraining. Extensive experiments demonstrate that SemAID achieves superior performance in both perceptual quality and reconstruction fidelity under low-SNR conditions, showcasing remarkable generalization across varying channel states and unseen data distributions. This work validates the potent efficacy of generative priors for semantic signal recovery under the compound challenge of non-uniform sampling and channel noise, offering a practical zero-shot solution for extreme communication environments.

Angle power spectrum (APS) characterizes the directional distribution of received signal power and is directly relevant to beam management and MIMO processing. While environment-aware learning has been widely studied for radio maps and path loss, direct map-to-APS prediction still lacks a standardized large-scale benchmark. This paper presents Map2APS, a physically grounded benchmark constructed from intelligent ray-tracing (IRT) path-level propagation records. Map2APS covers 51 equal-height urban maps and approximately 2.55 million Tx–Rx samples, with a strict cross-map split for evaluating generalization to unseen urban layouts. We benchmark representative model families and introduce MS-AReg as a strong reference baseline. On the full held-out test set of 249,993 samples, MS-AReg achieves a cosine similarity of 0.948, a peak location error of 1.20^∘, and an inference latency of 0.101 ms/sample. We further report dominant-direction metrics, including Top-1 dominant peak hit rate and dominant peak recall, to evaluate whether predicted spectra preserve decision-relevant arrival directions. The benchmark, code, and evaluation scripts are released at https://github.com/UNIC-Lab/aps-data.

Driven by the evolution toward 6G and AI-native edge intelligence, network operations increasingly require predictive and risk-aware adaptation under stringent computation and latency constraints. Network Traffic Matrix (TM), which characterizes flow volumes between nodes, is a fundamental signal for proactive traffic engineering. However, accurate TM forecasting remains challenging due to the stochastic, non-linear, and bursty nature of network dynamics. Existing discriminative models often suffer from over-smoothing and provide limited uncertainty awareness, leading to poor fidelity under extreme bursts. To address these limitations, we propose LEAD, a Large Language Model (LLM)-Enhanced Adapter-based conditional Diffusion model. First, LEAD adopts a "Traffic-to-Image" paradigm to transform traffic matrices into RGB images, enabling global dependency modeling via vision backbones. Then, we design a "Frozen LLM with Trainable Adapter" model, which efficiently captures temporal semantics with limited computational cost. Moreover, we propose a Dual-Conditioning Strategy to precisely guide a diffusion model to generate complex, dynamic network traffic matrices. Experiments on the Abilene and GEANT datasets demonstrate that LEAD outperforms all baselines. On the Abilene dataset, LEAD attains a remarkable 45.2

The rapid development of the low-altitude economy is shaping industries such as transportation, logistics, and agriculture. The profound integration of artificial intelligence and low-altitude economic networks, particularly in unmanned aerial vehicles (UAVs), has created an expansive market for low-altitude applications, including logistics and aerial services. Deep learning technologies, including image classification, target recognition, and segmentation, have markedly improved UAVs’ intelligent perception and operational capabilities. However, the deployment of deep learning models confronts substantial problems due to computing resource limits created by the storage space and maximum payload of UAVs. Filter pruning is an efficient model compression technique to reduce the dimensions of deep learning models and enhance inference performance. This paper introduces a soft filter pruning technique with adaptive layer pruning ratios, named AdaSP, which obviates the necessity for pre-training and fine-tuning, executing all computations through a "filter pruning-filter regrowing" process within a single training from scratch. Furthermore, we propose an algorithm based on updated weights for determining the real-time pruning ratio of each network layer. This algorithm does not require complete model training and does not rely on any additional dataset information. The optimal real-time pruning ratio for each layer can be computed during each training iteration. We compress the ResNet model for image classification tasks using ImageNet and remote sensing datasets (AID and NWPU45). AdaSP possesses a marginal advantage over alternative pruning techniques in enhancing top-1 accuracy and decreasing model FLOPs.

In Automatic Modulation Classification (AMC), deep learning methods have shown remarkable performance, offering significant advantages over traditional approaches and demonstrating their vast potential. Nevertheless, notable drawbacks, particularly in their high demands for storage, computational resources, and large-scale labeled data, limit their practical application in real-world scenarios. To tackle this issue, this paper proposes an automatic modulation classification model based on the Adaptive Lightweight Wavelet Neural Network (ALWNN) and the few-shot framework (MALWNN). The ALWNN model, by integrating the adaptive wavelet neural network and depth separable convolution, reduces the number of model parameters and computational complexity. The MALWNN framework, using ALWNN as an encoder and incorporating prototype network technology, decreases the model's dependence on the quantity of samples. Simulation results indicate that this model performs remarkably well on mainstream datasets. Moreover, in terms of floating point operations per second (FLOPS) and normalized multiply-accumulate complexity (NMACC), ALWNN significantly reduces computational complexity compared to existing methods. This is further validated by real-world system tests on Universal Software Radio Peripheral (USRP) and Raspberry Pi platforms. Experiments show MALWNN's superior performance in few-shot learning scenarios compared to other algorithms.

Radio maps (RMs) serve as environment-aware electromagnetic (EM) representations that connect scenario geometry and material properties to the spatial distribution of signal strength, enabling localization without costly in-situ measurements. However, constructing high-fidelity indoor RMs remains challenging due to the prohibitive latency of EM solvers and the limitations of learning-based methods, which often rely on sparse measurements or assumptions of homogeneous material, which are misaligned with the heterogeneous and multipath-rich nature of indoor environments. To overcome these challenges, we propose iRadioDiff, a sampling-free diffusion-based framework for indoor RM construction. iRadioDiff is conditioned on access point (AP) positions, and physics-informed prompt encoded by material reflection and transmission coefficients. It further incorporates multipath-critical priors, including diffraction points, strong transmission boundaries, and line-of-sight (LoS) contours, to guide the generative process via conditional channels and boundary-weighted objectives. This design enables accurate modeling of nonstationary field discontinuities and efficient construction of physically consistent RMs. Experiments demonstrate that iRadioDiff achieves state-of-the-art performance in indoor RM construction and received signal strength based indoor localization, which offers effective generalization across layouts and material configurations. Code is available at https://github.com/UNIC-Lab/iRadioDiff.

This paper investigates a dual radio map-aware inspection system that employs a cellular-connected unmanned aerial vehicle (UAV) in rural environments. The UAV is tasked with executing a series of mission-critical inspection operations while transmitting the collected data in real time under the guidance of pre-constructed radio maps. To enhance inspection efficiency and ensure low-latency data feedback, a joint optimization problem is formulated to simultaneously determine the inspection sequence and flight trajectory, with the objective of minimizing both mission completion time and data transmission delay. To solve this problem, we propose a novel framework that integrates an improved genetic algorithm (IGA) with deep reinforcement learning (DRL). The IGA first derives a near-optimal inspection sequence by jointly considering flight distance, uplink communication throughput, and downlink outage duration based on ground-to-air and air-to-ground radio maps. Subsequently, the DRL module refines the continuous flight trajectory. The simulation results demonstrate that the proposed DRL-IGA outperforms benchmark schemes, achieving up to a 20% reduction in both total task completion time and transmission delay.

Autonomous aerial vehicles (AAVs) enable data collection for sixth-generation Internet-of-Things networks, but their trajectories couple nonlinear wireless rates with long-horizon service progress. This paper views the evolution of AAV kinematics, channel state, and user backlog as a structured differentiable world model and develops Learn for Variation (L4V) to exploit that model efficiently. L4V replaces a discontinuous completion-time objective with a cumulative-backlog surrogate, unrolls the mission dynamics, and propagates pathwise sensitivities to a neural policy through the discrete adjoint recursion. The resulting derivative is exact conditional on a fixed exogenous-noise realization; stochastic expected-objective optimization still requires sampling. We show that the structured adjoint grows at most polynomially with the horizon and establish a stationary-point rate for fixed-step full-gradient descent under standard smoothness assumptions. The framework also learns shared OFDMA allocation under reparameterized shadowing and Rician fading, while distributional pretraining amortizes model-based optimization into forward-only deployment on unseen layouts. Paired stress tests cover channel-generator mismatch, noisy partial observations, a fixed-resource two-AAV extension, and a circular no-fly region. Code and configurations are available at https://github.com/UNIC-Lab/L4V-AAV. Against genetic-algorithm, DQN, A2C, DDPG, and differentiable model-predictive-control implementations, L4V reduces mission time by up to 65%, executes a default mission in 53 ms, and completes all 60 frozen-policy tests after pretraining on 1,600 layouts.

Radio map (RM) estimation is a key enabler for environment-aware optimization in 6G wireless networks. In practice, RM construction increasingly relies on crowdsourced received signal strength (RSS) feedback that is inherently sparse and noisy. A further and often overlooked challenge is location drift, whereby privacy constraints and user mobility cause reported sampling coordinates to deviate from the true measurement locations. Unlike additive measurement noise, location drift perturbs the sensing operator itself, since each RSS sample effectively queries the underlying RM at an incorrect spatial coordinate. This operator uncertainty, compounded with sparse noisy sensing, renders the inverse problem severely ill-posed and limits conventional estimators that rely on analytically specified priors. This paper proposes RadioDiff-Inv2, a differentiable diffusion inversion framework that estimates RMs from sparse noisy measurements under location drift. A Gaussian resampling scheme is introduced to construct a differentiable, drift-aware measurement operator on grid-based maps, and the probability-flow ordinary differential equation (ODE) is exploited to cast the diffusion sampler as a deterministic, differentiable mapping from an initial noise code to the estimated RM. By optimizing the noise code via backpropagation against a drift-marginalized data-fidelity objective, RadioDiff-Inv2 produces reconstructions that are both prior-plausible and measurement-consistent without costly posterior sampling. Extensive experiments show that RadioDiff-Inv2 outperforms the best competing baseline by 4 to 14 dB in PSNR across varying sparsity and drift levels. The advantage is most pronounced in low-SNR regimes, where the learned diffusion prior maintains near-constant reconstruction fidelity while conventional methods degrade severely.

In this paper, we investigate the power control and spectrum allocation challenge in vehicle-to-everything (V2X) networks with uncertain channels, where two types of services coexist, i.e., the large-capacity services supported by vehicle-to-infrastructure (V2I) links and the ultra-reliable services supported by vehicle-to-vehicle (V2V) links. Our goal is to maximize the transmission rate of V2I whilst ensuring the V2V outage probability constraint. To deal with the channel uncertainty, we first mine the channel correlation by mapping historical channel samples into high-dimensional space via a designed deep neural network (DNN). Then, a support vector clustering (SVC) based uncertainty set is derived as a union of convex subsets according to the output of piecewise activation DNN. Afterwards, a power control algorithm is proposed for the transformed non-convex problem. Specifically, a bisection-based method is developed to tackle the fractional signal-to-interference-plus-noise ratio (SINR) expression, and an exploration-comparison-approximation method is proposed to relax the union of convex subsets. Additionally, we design a set of virtual V2V links to enforce the spectrum allocation problem into a feasible bipartite graph matching problem. Simulation results demonstrate the proposed method overcomes the conservatism of previous uncertainty set construction methods, and outperforms other methods on V2I transmission rate.

The evolution toward sixth-generation (6G) wireless networks is driving larger antenna arrays and highly directional multi-beam transmission, making accurate knowledge of beam-dependent spatial coverage important for beam management and environment-aware network operation. Radio maps (RMs) provide such a representation, yet conventional RM prediction assumes omnidirectional or transmitter-level radiation. In beamformed multiple-input multiple-output (MIMO) systems, one propagation scene instead gives rise to many configuration-dependent beam radio maps (BeamRMs), creating challenges in beam representation and generalization. Existing methods either condition prediction on beam descriptors or use beam maps as auxiliary inputs to generic architectures. We propose BeamRMX, which, to the best of our knowledge, is the first dedicated framework to treat the spatial radiation pattern as the primary BeamRM query and learn how scene geometry transforms it into the received power field. XBase learns multiscale interactions between the radiation query and scene geometry, while an optional Evidence Adapter uses a few cross-configuration BeamRMs from the same scene. Matched-domain and zero-shot experiments show consistent gains over deterministic and diffusion baselines, including mean absolute error reductions of 26.1% on unseen scenes and 47.8% on an unseen configuration. Cross-configuration evidence further improves reconstruction and intra-sector beam refinement.

This paper investigates backdoor attacks in image-oriented semantic communications. The threat of backdoor attacks on symbol reconstruction in semantic communication (SemCom) systems has received limited attention. Previous research on backdoor attacks targeting SemCom symbol reconstruction primarily focuses on input-level triggers, which are impractical in scenarios with strict input constraints. In this paper, we propose a novel channel-triggered backdoor attack (CT-BA) framework that exploits inherent wireless channel characteristics as activation triggers. Our key innovation involves utilizing fundamental channel statistics parameters, specifically channel gain with different fading distributions or channel noise with different power, as potential triggers. This approach enhances stealth by eliminating explicit input manipulation, provides flexibility through trigger selection from diverse channel conditions, and enables automatic activation via natural channel variations without adversary intervention. We extensively evaluate CT-BA across four joint source-channel coding (JSCC) communication system architectures and three benchmark datasets. Simulation results demonstrate that our attack achieves near-perfect attack success rate (ASR) while maintaining effective stealth. Finally, we discuss potential defense mechanisms against such attacks.

Hyperparameter optimization (HPO) is crucial for federated learning (FL) performance. Given the inherent data heterogeneity across clients, recent research has focused on providing personalized hyperparameters for individual clients. However, such personalized approaches introduce exponential search complexity as the number of clients increases, significantly reducing the efficiency of existing HPO methods. To address this challenge, we propose pFedDHPO, a novel personalized HPO framework that efficiently optimizes hyperparameters in a differentiable manner. Specifically, pFedDHPO formulates personalized HPO as an optimization problem targeting joint distribution parameters within the clients’ search space and leverages gradient information from differentiable validation loss to substantially enhance the efficiency of the HPO process. Experimental results demonstrate that pFedDHPO achieves state-of-the-art performance compared to baseline methods, improving accuracy by up to 18.35% under extreme Non-IID data distributions. Additionally, the framework reduces communication overhead by 41.2% compared to conventional HPO methods, making it highly scalable for resource-constrained FL deployments.

Interference recognition serves as the preprocessing technology for interference suppression. Researchers typically employ the method based on the time-frequency graph - deep convolutional neural network (CNN) in pursuit of recognition accuracy. Nevertheless, certain time-frequency analysis processing approaches or network models with excessive parameters make it challenging to apply this technology to devices with limited resources. In this paper, a weight pruning and parameter quantization assisted multi-Scale lightweight network named WPMNet is proposed, which can directly take the raw time series data as input. The network model comprises three multi-scale convolutional modules, an attention mechanism, and fully connected layers. The multi-scale convolutional modules are composed of convolutional kernels of different sizes in parallel, aiming to extract deep-level features and features of various granularities simultaneously. To compress the model, we introduce an adaptive search algorithm for determining the optimal pruning threshold for weight pruning, ensuring that the accuracy decline is within an acceptable range. And the model is further compressed through parameter quantization. The experimental outcomes demonstrate that prior to the lightweighting of the model, the average recognition accuracy was 97.6%. After the lightweight processing, the compression ratio of the network attained 84.5%, while the average recognition accuracy merely declined slightly to 95.08%. This substantiates that the proposed approach is applicable to scenarios that simultaneously demand both high precision and model deployability, such as embedded systems or mobile devices.

Integrated sensing and communication (ISAC) has emerged as a promising paradigm for next-generation mobile wireless communication systems. In this paper, we propose a novel framework for full-duplex ISAC systems that jointly incorporates dynamic tracking and robust secure beamforming. Specifically, a dual-functional radar-communication base station employs an extended Kalman filter to dynamically estimate the trajectories of mobile downlink (DL) users. To mitigate the impact of imperfect channel state information from multiple eavesdroppers, a robust beamforming strategy is devised by quantifying angular uncertainty via the Cram & eacute;r-Rao bound (CRB). A total transmit power minimization problem is formulated under secrecy rate constraints for both DL and uplink (UL) communications, while simultaneously ensuring sensing accuracy through CRB and beam tracking mean squared error metrics. The optimization jointly considers the beamforming matrices, artificial noise covariance matrix, and UL power allocation strategy. To address the formulated non-convex problem, the S-procedure is employed to transform semi-infinite constraints into linear matrix inequalities. Successive convex approximation technique is then iteratively applied to obtain high-quality solutions. Extensive simulations verify the effectiveness of the proposed framework, demonstrating superiority in secrecy rate and power efficiency compared to benchmark schemes. Moreover, the results highlight the inherent trade-offs between secure communications performance and sensing accuracy, thereby offering insights into the design of future full-duplex ISAC systems.

With the significant advance of wireless communication technology, more networked control systems are looped via wireless networks. However, the dynamics and uncertainties of wireless channels as well as the limitation of radio resources make it challenging to close all loops at each control step. As the open-loop control will lead to performance deterioration, it is essential to jointly optimize the uplink and downlink transmissions for the full-loop control. In this paper, we analyze the impact of transmission delay in uplink and downlink on the full-loop control performance. We then propose a novel multicast transmission scheme for the latency-critical full-loop control. Accordingly, the uplink-downlink transmission and the full-loop control are jointly considered to minimize the control and communication cost. To effectively solve this mixed integer non-linear programming problem, the original problem is decomposed into the uplink transmission problem and the downlink transmission problem. Alternate resource optimization algorithm and multicast resource allocation algorithm are designed. Simulation results show that the proposed scheme has advantage on reducing both the communication and control cost.

This paper investigates over-the-air (OTA) computation federated learning (FL) in low-earth orbit (LEO) satellite networks, where we propose a novel joint design of adaptive beam hopping and power control to maximize the long-term total amount of training data. This problem is challenging due to the non-convex coupling between beam hopping patterns, power control, and global mean squared error (MSE) constraints, as well as the dynamic satellite coverage. To address these difficulties, we develop a proximal policy optimization (PPO)-based deep reinforcement learning framework that learns efficient scheduling policies. In particular, the proposed method jointly optimizes beam hopping patterns and transmission power, while incorporating MSE-aware reward shaping to balance data utilization and aggregation accuracy. Simulation results demonstrate that the proposed PPO approach achieves faster convergence, higher reward, and superior FL performance in terms of test accuracy and training loss, compared with soft actor-critic (SAC), deep deterministic policy gradient (DDPG), and a greedy baseline.