Intelligent channel prediction plays a key role in artificial intelligence (AI)-optimized or AI-native communication networks for smart high-speed railways (HSRs). This paper investigates the spatial-temporal prediction of channel state information (CSI) and channel statistical characteristics (CSCs) based on deep-learning (DL) for the future smart HSR communication network. A propagation-graph simulation method is used to generate datasets of CSI and CSCs for massive multiple-input multiple-output (mMIMO) channels in a HSR cutting scenario, and realistic channel measurements are used to validate the datasets. Then, single-step ahead and multi-step ahead prediction problems are formulated with the consideration of both spatial and temporal information hidden in the datasets. By exploiting the temporal and spatial correlations of the HSR mMIMO channel, a novel spatial-temporal channel prediction model that combines the convolutional neural network (CNN) and convolutional long short-term memory (CLSTM) is proposed and called as Conv-CLSTM. Moreover, the hyper-parameters of the Conv-CLSTM model are determined by autocorrelation and similarity analysis and cross-validation. Finally, the performance of the Conv-CLSTM model is evaluated in terms of prediction accuracy and space and time computational complexity, and is compared with classical DL models. The evaluation results show that the proposed model has high prediction accuracy but acceptable computational complexity.

Spatial characteristics of the propagation channel have a vital impact on the application of multi-antenna techniques. This paper analyzes angular characteristics and the spatial correlation for high-speed railway (HSR) channels, based on a novel moving virtual antenna array (MVAA) measurement scheme. The principle of the MVAA scheme is deeply investigated and is further verified by a theoretical geometry-based stochastic model. Using the MVAA scheme, virtual single-input multiple-output (SIMO) channel impulse response data are derived from single-antenna measurements in typical HSR scenarios, involving viaduct, cutting, and station. Based on the SIMO channel data, angle of arrival is extracted according to the unitary estimation of signal parameters by the rotational invariance techniques algorithm, and is compared with the theoretical result. Moreover, power angular spectrum and root mean square (rms) angular spread (AS) are provided, and the rms AS results are statistically modeled and comprehensively compared. In addition, spatial correlation is calculated and analyzed, and a rms AS-dependent spatial correlation model is newly proposed to describe the relationship between the angular dispersion and the spatial correlation. The presented results could be used in multi-antenna channel modeling and will facilitate the assessment of multi-antenna technologies for future HSR mobile communication systems.

With the development of artificial intelligence (AI), wireless channel prediction based on deep learning (DL) has become a hot research issue. Channel prediction plays an important role in channel state information (CSI) feedback enhancement in AI-native air interface. To better predict the CSI, this paper investigates the Transformer network based channel prediction. Firstly, real channel data are obtained in Beijing-Tianjin railway line, and the channel prediction datasets are constructed through preprocessing. After formulating the channel prediction problem, a channel prediction model based on the Transformer network is newly proposed. The unique multi-head attention mechanism and position encoding of the Transformer network enable the proposed model to have more powerful parallel computation capability and better global information capture capability. Then, the hyper-parameters of the model are determined by autocorrelation analysis and cross-validation. Finally, the performance of the proposed model is evaluated in terms of prediction accuracy and space and time computational complexity using several evaluation metrics, and is compared with classical DL models. It is shown that the proposed model possesses higher prediction performance in the appropriate range of computational complexity.

Wireless channel scenarios identification is of pivotal significance for dedicated wireless communication design, especially for the heterogeneous network covering rich propagation environments. In this paper, the identification problem is investigated by machine learning approaches. To enhance the identification performance, some preprocessing methods, mainly referring to the data normalization and dimension reduction, are adopted. Then, both supervised and unsupervised learning algorithms, including k-nearest neighbor (k-NN), support vector machine (SVM), k-means, and Gaussian mixture model (GMM) are used to realize the scenarios identification, respectively. Finally, the identification performance of these four approaches are validated both on the actual measured HSR wireless channel data sets and the QuaDRiGa channel emulation platform with the ability of multiple scenarios emulation. Most of the results indicate that k-NN and SVM approaches can achieve an accuracy over 90%. As for those two unsupervised learning approaches, the GMM proves to be a promising approach by presenting a performance close to the former two approaches without training process, whereas the k-means yields an accuracy about 80%.

Channel measurements are the prerequisite condition for the design of the next-generation high-speed railway (HSR) communication system. Because of the measurement restriction and measurement efficiency issues of applying conventional channel sounders in HSR scenarios, railway-network-based channel measurements are becoming quite attractive. In this paper, a method that employs the cell-specific reference signal in long-term-evolution (LTE) railway networks as the excitation signal to perform HSR channel measurements is presented. This method is implemented by a novel channel measurement system composed of existing LTE railway networks and a dedicated LTE receiver, which can efficiently collect the time-frequency-space channel data and is able to measure different communication links simultaneously. Based on this system, field measurements that consider both direct coverage and relay coverage schemes are conducted on the Beijing-toTianjin HSR in China. Statistical results, involving path loss, Rician K-factor, root-mean-square delay spread, single-link spatial correlation (SC), and multilink SC, are derived, which not only confirm the viability of the proposed method but also provide realistic HSR channel characteristics for the research on future HSR communications.

The widespread popularity of high-speed railways (HSRs) urges a critical demand on high-data-rate railway communication services for both train operation and passenger experience. To satisfy the ever-increasing requirements, future HSR communication systems, such as long-term evolution for railway (LTE-R), fifth generation (5G) on HSR, and 5G for railway (5G-R), and corresponding transmission technologies, e.g., mobile relay, coordinated multipoint, massive multiple-input multiple-output (MIMO), and millimeter-wave (mmWave), have recently attracted much attention. Radio channel modeling is the foundation of design and evaluation of wireless systems and transmission technologies. This paper focuses on a survey of channel modeling for the future HSR communication systems. The significant requirements of future HSR channel models are highlighted, and recent advances in the HSR channel modeling are reviewed. Finally, potential research directions for future HSR channel modeling are outlined.