IET Communications最新文献

Ultra-Wideband (UWB) is a wireless communication technology that uses Radio Frequency (RF) to transmit and receive signals between devices. Beamforming in UWB is a technique that uses multiple antennas simultaneously to focus on specific directions. In beamforming, Deep Learning (DL) techniques are applied to enhance signal processing and optimise beam pattern generation by utilising neural networks for efficient and accurate spatial filtering. However, existing DL techniques suffer from catastrophic forgetting, in which the testing data forgets previously learnt data due to the lack of knowledge distillation in other layers. Therefore, this research proposes a Regularised Hyperparameter Bilevel Optimisation with Continual Learning-based Deep Neural Network (RHBO-CLDNN) for beamforming in UWB systems. RHBO optimises hyperparameter efficiency at both the upper and lower levels, thereby enabling the DNN to accurately capture UWB channel characteristics, which improves channel estimation and enhances the Signal-to-Noise Ratio (SNR). CL is applied to dynamically adapt to changing environmental conditions without requiring complete retraining, making it suitable for real-time applications. Elastic Weight Consolidation (EWC) regularisation is also applied, which mitigates catastrophic forgetting by preserving weights from learnt tasks and enables the model to adapt to channel conditions without losing previous knowledge. Experiments on the DeepMIMO dataset show that RHBO-CLDNN enhances the sum-rate by up to 18% and achieves an inference time of 0.025 s over Convolutional Neural Network (CNN), thereby demonstrating its suitability for real-time beamforming.

The rapid growth of 6G Internet of Things (IoT) networks demands scalable and secure learning systems that can support massive device connectivity with minimal coordination overhead. Federated learning (FL) over grant-free non-orthogonal multiple access (GF-NOMA) offers a promising approach by enabling distributed model training with asynchronous uplink access and low signalling cost. However, this setup introduces coupled vulnerabilities: The uncoordinated nature of GF-NOMA leads to random collisions and residual interference, while the decentralised nature of FL exposes the system to poisoning, Sybil and jamming attacks. These cross-layer threats jointly degrade model convergence and communication reliability. To address this, we propose Security-Aware Proximal Policy Optimisation (SA-PPO), a reinforcement learning framework that co-designs communication security for FL over GF-NOMA. SA-PPO jointly embeds physical-layer features (e.g., SINR and interference) and learning-layer signals (e.g., anomaly scores and trust values) into its state, action and reward spaces. This enables the base station to optimise admission control, resource allocation and trust-weighted aggregation in a unified loop. Unlike prior methods that treat communication and security independently, SA-PPO learns coordinated strategies that attenuate adversarial impact while preserving update diversity. Simulation results show that SA-PPO achieves over 90% anomaly detection accuracy, sustains secure participation above 80% and reduces collision-induced decoding errors by 25% under scenarios with up to 40% compromised devices, while incurring only modest increases in energy and latency. These results demonstrate SA-PPO's effectiveness for secure, scalable and resilient edge intelligence in future 6G IoT environments.

This paper explores the practical application of source polar codes to entropy coding tasks in modern transform coding pipelines. Transform coding remains the predominant and rapidly evolving framework for compressing complex real-world data. Despite the strong theoretical guarantees of polar codes, conventional polarization-based compression techniques follow a “construct-then-use” paradigm, which proves inefficient and inaccurate when applied to transform coding scenarios characterized by highly dynamic entropy models. To overcome this limitation, we propose a construction-free, plug-and-play polar compression scheme. Rather than relying on precomputed polarized entropies, our method selects output symbols based on probability vectors generated by a conditional entropy model. These vectors can be computed with low complexity and exact numerical precision, enabling efficient adaptation across diverse entropy coding tasks. The proposed approach offers greater flexibility than classical methods and achieves superior performance in the finite-length regime.

Generative adversarial networks (GANs) with their strong generative capabilities have shown significant promise in visual privacy protection. However, when applied to image full-body visual privacy protection in open-world scenarios, where abnormal visual privacy data may exist in the training data, issues such as mode collapse and instability in GANs can be severely exacerbated. This leads to a significant reduction in both image quality and utility preservation. In this paper, we propose an end-to-end, contrastive GANs-based framework, FBPPGAN, for image full-body visual privacy protection, specifically designed to address these challenges. First, we introduce the architecture of FBPPGAN, which is tailored for full-body visual privacy protection. Second, we propose a novel adversarial loss function aimed at mitigating mode collapse and instability, particularly in the presence of abnormal images in open-world environments. We also design a content mapping network and a content loss function based on contrastive learning to address the issue of insufficient color gamut in generated images. Furthermore, a stylized loss function is introduced to more accurately measure the difference between the generated and target domains. Experimental results across four public datasets demonstrate that FBPPGAN effectively overcomes mode collapse and instability, delivering superior image quality and utility preservation. Compared to the existing state-of-the-art methods, FBPPGAN outperforms in terms of convergence, stability, computational complexity, processing speed, and effectiveness. To the best of our knowledge, this is the first GAN-based framework for image full-body visual privacy protection in open-world scenarios.

Since the opening of the 6 GHz bands for unlicensed radio access technologies (RATs), new coexistence mechanisms leveraging the currently uninhabited 6 GHz bands have been investigated, aiming for fair coexistence of Wi-Fi 6 and 5G new radio unlicensed (NR-U). However, our study shows that the utilization of the highly attractive 6 GHz bands can be significantly enhanced by exploring additional spectrum access opportunities, which remain unrealized in the existing channel contention mechanisms. We propose a novel two-stage channel contention mechanism for the coexistence of Wi-Fi 6 and 5G NR-U in the 6 GHz bands to explore these missed spectrum access opportunities. We formulate the probability of interference to ongoing transmissions and utilize this probability to enhance the utilization of radio resources by allowing simultaneous transmissions on a channel. We incorporate cross-technology communication (CTC) to compute this probability and formulate an optimization problem to derive the optimal CTC information required for the computation. Extensive simulation results show that the proposed framework significantly outperforms legacy channel contention mechanisms in terms of spectrum utilization while ensuring the ongoing transmissions unharmed.

Cognitive Spectrum Sensing (CSS) stands as a foundational component in 5G and emerging 6G wireless communication systems, enabling intelligent identification and dynamic utilisation of underutilised spectrum bands. However, the implementation of CSS in dense and heterogeneous 5G/6G environments presents significant challenges, including high spectral dynamics, multi-protocol interference, and the requirement for real-time decision-making across diverse frequency bands. Existing methods such as deep belief networks, CNN-PSO hybrids, and DQN-based models suffer from limited adaptability, insufficient spatial-temporal learning, and poor generalisation in real-world RF environments. The proposed model includes a dual-stream deep learning architecture which has a 1D convolutional neural network (CNN)-based spectral encoder and a graph convolutional network (GCN)-based spatial encoder for extracting the frequency-domain and node-topology features. Experimental analysis of proposed model is performed using the Real-World Wireless Communication Dataset containing Wi-Fi, LTE, and 5G RF signals. The dataset is pre-processed using Fast Fourier Transformation (FFT) transformation and labelled through a signal-power-based thresholding mechanism. Results indicate the Spectral-Spatial Dual Encoder with Bio-Inspired Swarm Adaptation (SSDE-BSA) achieves an accuracy of 96.1%, an F1-score of 96.1%, and a spectral efficiency of 91. These results confirm the model's superiority in adapting to real-world spectrum dynamics, offering a robust and scalable solution for cognitive spectrum sensing in next-generation wireless networks.

We employ convolutional neural networks (CNNs) with distance feature and satellite image for path loss (PL) estimation at sub-6 GHz and millimetre wave (mmWave) frequencies. In order to avoid complex preprocessing of embedding distance feature into the image, we append this feature at the earliest, after the convolutional blocks of a CNN-based VGG-16 architecture. This is intuitive since the following fully-connected (FC) layer performs feature aggregation, thus, it combines the injected distance feature with the extracted features from the image. We propose three VGG-16 structures which vary in how the distance information is included. Performance is then evaluated in terms of training and prediction times, root mean square error (RMSE) and correlation coefficient, while performance without appended distance serves as benchmark. We observe that the inclusion of distance parameter gives more accurate estimation in terms of RMSE and a very strong correlation between the predicted and estimated PL values. Moreover, the proposed structures typically converge more quickly. Among the proposed structures, the one aided by a logarithm-of-distance model, is the most computationally efficient, leading to $��50�\%�$ and $��27�\%�$ reduction of training time and prediction time, respectively. Additionally, the VGG-16-based PL predictors yield lower RMSE by up to 2.4 dB and $��21�\%�$ higher correlation compared to the 3GPP 38.901 urban macro cell (UMa) empirical model.

Non-orthogonal multiple access (NOMA) combined with integrated sensing and communication (ISAC) presents a promising solution to 6G spectrum scarcity. However, non-line-of-sight channels caused by obstacles significantly degrade performance. Unmanned aerial vehicles (UAVs) can establish line-of-sight links, thereby improving communication rates. Existing research primarily focuses on stationary targets, while studies on moving targets rely on single-UAV sensing, leading to substantial estimation errors in dynamic environments. To address this, we propose a UAV-enabled NOMA-ISAC network that employs an extended Kalman filter for accurate target tracking. We aim to jointly optimize the UAV trajectory and beamforming vector using a block coordinate descent framework, where the subproblems are efficiently solved via successive convex approximation and semidefinite programming. Numerical results demonstrate that the proposed scheme achieves an 18.47% higher sum rate compared to benchmark schemes.

This study presents the design and optimization of a microstrip monopole antenna for 5G sub-6 GHz applications, employing a deep learning-based surrogate model combined with honeybee mating optimization (HBMO). The studied antenna structure employs air via arrays, intended to enhance antenna performance, including improved impedance matching and increased bandwidth. It is important to note that, unlike conventional antennas, the proposed design does not include a fully enclosed metallic cavity similar to a substrate integrated waveguide (SIW) antenna designs. A sensitivity analysis was conducted to assess the impact of these parameters, emphasizing the need for optimal tuning. To generate training and test datasets efficiently, Latin hypercube sampling (LHS) was used. A convolutional neural network (CNN) surrogate model was trained, outperforming other machine learning (ML) algorithms in predictive accuracy and generalization. The proposed CNN-HBMO framework reduced computational costs by minimizing the need for expensive electromagnetic (EM) simulations, enabling rapid design space exploration. The optimized antenna was fabricated and validated through experimental measurements, achieving 2–3 dBi gain and 𝑆₁₁ < −10 dB across the 2.7–5.2 GHz band. Compared to existing designs, the proposed antenna offers a compact size (34 × 34 mm) with competitive performance, making it suitable for multi-band 5G applications.

Reconfigurable intelligent surface (RIS)-assisted received spatial modulation (RIS-RSM) has emerged as a promising technique to enhance spectral and energy efficiency in next-generation wireless systems. However, performing accurate signal detection without channel state information (CSI) remains a critical challenge, particularly in blind detection scenarios. In this paper, we propose a novel clustering-based blind detector named energy-tiered structure initialization (ETSI). The proposed method exploits the amplitude heterogeneity of modulation symbols—originating from the unequal energy levels of QAM or hybrid constellations—by partitioning the signal space into multiple energy tiers. Each receive antenna is associated with a structure prototype, representing a normalized statistical channel pattern, which is initialized using the distribution of the underlying fading model (e.g., Rayleigh) rather than instantaneous CSI. During clustering, these prototypes are iteratively refined through tier-wise averaging of normalized signal samples, thereby enforcing structural consistency and mitigating amplitude-induced bias. After convergence, the constellation-aligned cluster centres are reconstructed by combining the updated prototypes with their corresponding modulation amplitudes, inherently enabling the joint detection of both the modulation symbol and the RIS-assisted spatial index. Simulation results show that ETSI achieves around 0.5–1 dB SNR gain over amplitude–phase aware clustering under 8PSK, and about 1–1.5 dB improvement under 16QAM, while outperforming the CSI-based greedy detector (GD) across both modulations. Moreover, ETSI achieves BER performance close to that of the perfect-CSI maximum likelihood detector, confirming its accuracy, scalability and practical feasibility for blind RIS-RSM detection.