Low Earth Orbit (LEO) satellite communication constellations provide crucial support for the future of ubiquitous connectivity. However, the high dynamic nature of LEO satellites’ relative positions affects the stability and reliability of inter-satellite communication links while imposing stricter latency requirements. Fast adaptive beamforming technology offers a viable solution to address these challenges. To achieve fast adaptive beamforming, this paper proposes a two-step beamforming solution based on a Transformer neural network model and explores the feasibility of applying meta-heuristic algorithms for hyperparameter optimization in neural network (NN) models. Experimental results demonstrate that the angle-of-arrival predictor optimized using the Polar Lights Optimizer (PLO), a meta-heuristic algorithm inspired by the aurora phenomenon, significantly outperforms the unoptimized model, with its error consistently remaining within the half-power beamwidth of the proposed planar antenna array. Meanwhile, the beamforming accuracy improves by 15.2% compared to other NN-based models, and in all test scenarios, our algorithm reduced the average response time by approximately 82.1% compared to the null-steering beamforming algorithms. This study provides a novel solution for achieving low-latency, high-reliability 3D beamforming in LEO inter-satellite communication, integrating both speed and accuracy, thereby contributing to the advancement of 6G and future ubiquitous connectivity.

Satellite communication systems operate across diverse orbital configurations and terrestrial environments, where channel characteristics vary significantly due to differences in Doppler shifts, shadow fading, and propagation conditions. This heterogeneity causes severe domain shift, degrading the performance of interference recognition models trained on one channel environment when deployed in another. To address this challenge, this paper proposes a Cross-Domain Adversarial Disentanglement Network (DADN) that explicitly separates interference-invariant features from channel-dependent features through a dual-stream architecture. Unlike conventional domain adaptation methods that perform global feature alignment, DADN employs a parallel dual-stream architecture where the interference feature extractor and channel feature extractor operate independently with separate parameters, enabling explicit separation of transferable semantic information from environment-specific propagation effects.The interference feature extractor learns intrinsic signal characteristics shared across environments, while the channel feature extractor captures environment-specific propagation attributes. A gradient reversal mechanism adversarially removes residual channel information from interference features, and KL-divergence regularization prevents channel features from encoding interference category information, jointly ensuring feature orthogonality. Experiments across four representative satellite scenarios (LEO-Urban, LEO-Open, MEO-Suburban, GEO-Rain) demonstrate that DADN achieves 98.49% average recognition accuracy, outperforming CNN baseline by 2.08%, CLDNN by 1.24%, and DANN by 1.57%. Furthermore, DADN maintains over 97% accuracy with only 5% labeled target-domain data, confirming its strong cross-domain adaptability and data efficiency for practical satellite interference recognition applications.

3-D radio maps provide significantly richer information than their 2-D counterparts in spectrum cartography. However, reconstruction methods for 3-D radio maps remain underexplored. Classic methods, such as spatial interpolation and matrix completion, can be adapted for 3-D scenarios, but they are often computationally intensive and require a substantial number of sampling measurements. To address these challenges, we propose a novel and efficient 3-D radio map reconstruction paradigm inspired by cross tensor approximation (CTA). During the measurement phase, only piecewise straight-line paths are required, making this method well-suited for the flying trajectories of unmanned aerial vehicles (UAVs). In the reconstruction phase, we utilize the inherent tensor structure of 3-D radio maps by employing the fiber sampling tensor decomposition (FSTD) algorithm, which ensures high-quality and computationally efficient reconstruction from the measurements. To further reduce the number of measurements, we introduce a new algorithm called interpolation-integrated FSTD (II-FSTD). This algorithm builds on the original FSTD yet exploits the smoothness of radio maps to incorporate interpolation, thereby reducing the required measurement frequency. Extensive experimental results demonstrate that the proposed paradigm can reconstruct 3-D radio maps with high quality, using fewer measurements and less computational time compared to state-of-the-art techniques. Notably, in scenarios with lower measurement frequency, II-FSTD outperforms other methods, achieving superior reconstruction performance.

Spectrum prediction can further improve the performance of cognitive radio system, save spectrum sensing time and improve the communication quality of secondary users. To improve the accuracy of spectrum prediction and reduce the interference to the primiary users from the perspective of prediction, a hybrid spectrum prediction model based on deep learning is proposed in this paper. Specifically, the hybrid spectrum prediction model is composed of spectrum state regularity prediction (SSRP) model and enhanced available duration prediction (EADP) model, which respectively act on spectrum sensing and spectrum decision in cognitive radio. On the one hand, to improve the prediction precision of multi-channel joint spectrum state, SSRP model is proposed to analyze the regularity of joint spectrum state. SSRP model is used to extract frequency variation characteristics and spectrum occupancy characteristics by designing double networks, so as to excavate spectrum correlation in depth and achieve better prediction performance. On the other hand, to make full use of the joint spectrum state information without causing information redundancy, the EADP model of enhanced attention is proposed by integrating the attention mechanism with the spectrum sensing results. The simulation results show that SSRP model can obtain higher prediction accuracy and more stable distribution error to a certain extent, comparing with traditional spectrum state prediction methods. In addition, our innovatively proposed EADP model can reduce the available duration prediction error on the basis of integrating the attention mechanism.

In energy harvesting Ad-hoc networks, a multi-hop D2D link is made available for long-distance communication, which will cause interference with adjacent D2D nodes. The paper proposes a relay selection algorithm based on machine learning to maintain the supply of communication energy and reduce interference with other D2D nodes. First, the closed-form solution of the outage probability is analyzed. Then, the distance ratio factor (DRF) which affects the outage performance is derived. Based on the factor, the mapping DRF matrix is obtained and a support vector machine (SVM) is utilized to decrease the outage probability. Finally, the relay nodes are selected by the SVM - Dijkstra algorithm. Simulation experiments verify the performance of the proposed algorithm. The proposed algorithm outperforms the shortest path algorithm, the direct transmission method, and the DRF-Dijkstra algorithm.

Summary In this paper, a multi‐carrier cognitive decision engine based on a binary particle swarm optimization with a non‐linear decreasing inertia‐weight (NDI‐BPSO) is presented. Our main goal is to solve the optimization problem of transmitter parameters in different wireless communication modes for cognitive radio systems (CRSs), especially for the transmitter in communication systems based on the environment sensing. In the new algorithm, the multi‐carrier cognitive decision engine based on an NDI‐BPSO algorithm can mitigate the local extreme points effectively and reduce the oscillation phenomenon in the process of optimization. We apply the NDI‐BPSO to the cognitive orthogonal frequency division multiplexing (OFDM) system to determine the best parameters to obtain good performances in different communication modes. The simulation results show that the proposed multi‐objective cognitive decision engine, which has a high fitness value and strong robustness for different communication modes, is better than the existing engines. The novel NDI‐BPSO algorithm achieves the objective of parameter optimization effectively.

With the emergence of the information age, mobile data has become more random, heterogeneous and massive. Thanks to its many advantages, deep learning is increasingly applied in communication fields such as modulation recognition. However, recent studies show that the deep neural networks (DNN) is vulnerable to adversarial examples, where subtle perturbations deliberately designed by an attacker can fool a classifier model into making mistakes. From the perspective of an attacker, this study adds elaborate adversarial examples to the modulation signal, and explores the threats and impacts of adversarial attacks on the DNN-based modulation recognition in different environments. The results show that, regardless of a white-box or a black-box model, the adversarial attack can reduce the accuracy of the target model. Among them, the performance of the iterative attack is superior to the one-step attack in most scenarios. In order to ensure the invisibility of the attack (the waveform being consistent before and after the perturbations), an appropriate perturbation level is found without losing the attack effect. Finally, it is attested that the signal confidence level is inversely proportional to the attack success rate, and several groups of signals with high robustness are obtained.

The V2V (Vehicle to Vehicle) communication system in the Internet of vehicular networking is an important part of the future intelligent vehicle network, and it is extremely important to study the quality of the V2V communication channel. The double-Rayleigh fading model can better reflect the small-scale fading characteristics of the V2V channel. Therefore, this paper conducts experimental verification in this channel environment. First, using the gain matrix constructed by CSI, the images of time, frequency and related domain are obtained through the transformation of contour line, waterfall map and related diagram. Then, The image features transformed by the channel state information are extracted based on the improved multi-texton histogram. Finally, the V2V channel quality under slow fading conditions is discriminated under the SVM. The results show that the method not only simplifies the difficulty of channel state feature extraction, but also can effectively and reliably predict the instantaneous channel state.

In the era of 5G and Internet of Things, the number of connected devices has increased dramatically, which has placed a heavy burden on the network. So it is worthwhile to study the intelligent and accurate identification and control of devices. Radio Frequency (RF) fingerprinting technology has been widely used in wireless device identification. RF fingerprint contains rich nonlinear characteristics that reflect the uniqueness of wireless devices. However, redundant or irrelevant information in the features will result in poor recognition performance. In response to the problem, a novel integration feature selection method is proposed in this paper. The principle is to improve the identification performance through extracting the best-performing feature subset from the initial features. Signals from seven power amplifiers are collected as the dataset. The covariance feature is extracted as RF fingerprint and K-Nearest Neighbor (KNN) classifier is used for classification. The stability of the proposed method is evaluated by the Spearman correlation coefficient. The robustness is evaluated under the varying Signal-to-Noise Ratio (SNR). The identification results demonstrate the excellent performance of the proposal.

Network Intrusion Detection Systems (NIDS) have become a strong tool to alarm attacks in computer and communication systems. Machine learning, especially deep learning, has made huge success in fields of industry and academic. Network intrusion activity can be a time series event. In this paper, we adopt a time-related deep learning approach to detect network intrusions. A stacked sparse autoencoder (SSAE) is first built to extract the features with the greedy layer-wise strategy. And then, we propose a time- related intrusion detection system based on the variants of Recurrent Neural Network (RNN). We study the performance of proposed approach on the binary classification with a benchmark dataset UNSW- NB15. Based on the study of parameter time steps, it is proved that our time- related model is effective for intrusion detection. The experiment results show that the accuracy of the proposed approach reaches over 98% and the false alarm rate is as low as 1.8%. The performance of our model is superior to that of the standard RNN- based approach and approaches based on Deep Neural Network and shallow machine learning.

It has been recognized that fixed spectrum and channel allocation will lead to waste of spectrum resources when multiple agents communicate at the same time. Dynamic allocation of channels is proposed to maximize the utilization of spectrum resources. In the environment of multiple unmanned aerial vehicles (UAVs), it is necessary to ensure that each UAV can communicate successfully without interfering with other UAVs. Dynamic allocation of channels plays an important role in such systems. In this paper, we propose a dynamic channel allocation scheme based on deep reinforcement learning for multi-UAV systems. A slotted time system is used by all the UAVs. Di2642erent from the traditional method, the occupancy of each channel is scanned first in each time slot. Then a channel will be selected for data transmission, with feedback from the environment when the transmission is over. The proposed channel allocation scheme incorporates a long short-term memory (LSTM) into the deep reinforcement learning framework, to better learn from the past experience and better adapt to the the highly dynamic environment in a multi-UAV system. The experimental results show that compared with the traditional reinforcement learning method (Q- learning and Deep Q Network (DQN)), the proposed method achieves faster convergence and better performance with respect to average collision rate, average reward, and average successful communication rate.

Fault diagnosis is a typical multisensor information fusion problem. The information obtained from different sensors, such as sound, pressure, vibration, and temperature, can be considered as a piece of evidence. From the viewpoint of the evidence theory, the problem of multisensor fault diagnosis can be viewed as the problem of evidence fusion and decision. However, the information obtained from different sensors may be inaccurate, uncertain, fuzzy, or even conflict, so how to set up the fault diagnosis architecture of a distributed multisensor system and combine the conflict evidence should be taken into consideration. In this paper, the classical Dempster-Shafer evidence theory is described and the disadvantage of a classical Dempster's combination rule is discussed. In order to solve the counter-intuitive result when using the classical Dempster's combination rule, the Euclidean distance is proposed to characterize the differences between different pieces of evidence, and then the support degree of each evidence is generated and the weighted pieces of evidence can be combined directly using the classical Dempster's combination rule. Numerical simulation examples indicate that the proposed method has a better performance of analyzing the conflict between different pieces of evidence, especially for high conflict evidence. Therefore, compared with the existing methods, it has better applicability. According to the requirement of the Dempster-Shafer evidence theory, the fault diagnosis architecture of a distributed multisensor system is analyzed in detail, and a fault case of a rotating machine is used to illustrate that the proposed model is effective and superior, which can be used in practice.