CN-122027057-A - 5G-R channel model optimization method for high-speed rail large-scale station

CN122027057ACN 122027057 ACN122027057 ACN 122027057ACN-122027057-A

Abstract

The invention discloses a 5G-R channel model optimization method for a high-speed rail large-scale station, which comprises the steps of collecting channel data, wherein collected reference signals comprise amplitude and phase changes. In the data receiving process, the characteristics of gain, phase, time delay and the like of a channel are estimated in real time through a reference signal, and CSI data is generated. The method and the device can fully adapt to the characteristics of complex structure, serious multipath, strong scene dynamic property and the like of a high-speed railway large-scale station under 2100MHz frequency band and 2X 10MHz bandwidth, obviously improve the channel capacity of a 5G-R system, and maximize the throughput and data transmission efficiency of the communication system.

Inventors

Liu Yanjian
WANG WEIGANG
ZHANG HAITAO
YU HANBIN
WANG FANG
LI YI
Guan Zebin
LUO YUHAO
DU YUMENG

Assignees

中国国家铁路集团有限公司
南京邮电大学
中国铁道科学研究院集团有限公司

Dates

Publication Date: 20260512
Application Date: 20251231
Priority Date: 20250604

Claims (9)

1. The 5G-R channel model optimization method for the high-speed rail large-scale station is characterized by comprising the following steps of: step 1, channel data acquisition; Acquiring signals in a downlink synchronous signal block SSB and a channel state information reference signal CSI-RS in a 5G-R system under 2100MHz frequency band and 2X 10MHz bandwidth configuration through an adaptive sensor and equipment, wherein the signals are used for channel estimation to help extract channel characteristics from a 5G base station gNB to intelligent super-surface RIS, RIS to user equipment UE and gNB directly to users; step 2, channel estimation; when receiving a reference signal, the system carries out channel estimation in real time, and estimates the characteristics of gain, phase, time delay and the like of a channel, so as to obtain Channel State Information (CSI) for subsequent feature extraction and optimization; Step 3, convolutional neural network CNN feature extraction; The CNN is designed to conduct automatic feature extraction and compression on the channel matrix data, and the high-dimensional CSI data is reduced to a low-dimensional feature vector with obvious representative features so as to represent time sequence and nonlinear features in the CSI data; Step 4, manifold learning dimension reduction; based on the CNN extracted features, the local linear embedded LLE algorithm is adopted to further reduce the data dimension, and the lower-dimension feature representation is obtained through neighborhood structure maintenance, so that the local structure of the original data is effectively maintained, and the calculation complexity is reduced; Step 5, RIS reflection matrix optimization; Taking low-dimensional CSI features extracted by CNN and LLE as inputs, constructing an optimization model taking channel capacity as an objective function, performing global optimization on an RIS reflection matrix of an intelligent super surface by adopting a differential evolution algorithm, calculating an equivalent channel matrix of a communication system after obtaining the optimized reflection matrix, and evaluating the channel capacity of the system after optimizing by utilizing a Shannon capacity formula, thereby quantitatively measuring the change of communication performance before and after optimizing.
2. The method for optimizing the 5G-R channel model of the high-speed rail-oriented large-scale station according to claim 1, wherein the channel data acquisition is used for acquiring signals in downlink SSB and CSI-RS in a 5G-R system through adaptive sensors and equipment under 2100MHz frequency band and 2 × MHz bandwidth configuration, wherein the signals are used for channel estimation and help to extract channel characteristics from gNB to RIS, RIS to UE and gNB directly to users; in 2100MHz frequency band, 5G-R system has stronger signal penetrability and lower propagation loss, and is especially suitable for high-speed mobile environment, and the free space path loss can be expressed as: L=20log10(f)+20log10(d)-147.55(dB) The method comprises the following steps of (a) setting a frequency band, wherein f is the frequency, d is the propagation distance, a signal can well penetrate through an obstacle in the frequency band, meanwhile, lower attenuation is kept, communication reliability in a high-speed scene is guaranteed, and as the multipath propagation effect in a high-speed moving environment is obvious, the signal forms a plurality of paths after being reflected, scattered and diffracted by the ground, a building and a car body, so that the difference between channel gains h i and phase offsets theta i of different paths is obvious, and the expression is: Because the signal reflection paths are more, the gain and the phase difference on different paths are particularly obvious in the CSI data, and the frequency selective fading characteristic is shown; Meanwhile, under the configuration of 2×10MHz bandwidth, the uplink and downlink respectively occupy 10MHz bandwidth, and the bandwidth allocation is helpful to more finely describe the fading characteristics among subcarriers, and the complex gain of the kth subcarrier can be expressed as: wherein h i,k represents the gain of the ith path on the kth subcarrier, θ i,k is phase offset, and N is the total number of paths, and under the bandwidth configuration, the refined allocation of spectrum resources can capture the frequency selective fading characteristics among the subcarriers, especially in the scene of significant multipath effect, the channel gain and phase variation difference of different subcarriers are larger.
3. The method for optimizing a 5G-R channel model for a large-sized station for high-speed rail according to claim 1, wherein the channel estimation is performed based on a received reference signal, and the gain, phase and delay characteristics of the channel are estimated, and the calculation formula of the estimation process is as follows: where H is the estimated CSI, r i is the i-th signal of the reference signals, G i is the corresponding channel gain, N is the number of reference signals, and epsilon is the estimation error.
4. The method for optimizing the 5G-R channel model of the large-scale station for the high-speed rail according to claim 1, wherein the CNN feature extraction adopts a multilayer one-dimensional convolutional neural network structure, takes CSI data under 2100MHz frequency band and 2X 10MHz bandwidth configuration as input, the CSI data not only comprises signal gain, phase and time delay in a channel, but also comprises physical features such as frequency selective fading, path loss, multipath effect and the like in a 5G-R system, and the calculation formula of the CNN feature extraction is as follows: Wherein, the The j-th output characteristic diagram is used for capturing channel gain, phase offset and path delay characteristics; For convolution kernels, CSI features on different paths are extracted by sliding window operations, including primary path gain and interference path attenuation, For bias term, f (·) is a nonlinear activation function for enhancing the variability between features, and M j is the set of input channels connected to the feature map j.
5. The optimization method of the 5G-R channel model for the high-speed rail large-scale station according to claim 1, wherein the manifold learning dimension reduction adopts an LLE method for carrying out dimension reduction processing on high-dimension channel characteristics extracted by CNN, and the calculation formula is as follows: Wherein, X i represents a high-dimensional data point, Y i represents a data point after dimension reduction, W ij is a reconstruction weight, and N (i) is a neighbor set.
6. The optimization method of the 5G-R channel model for the high-speed rail large-scale station according to claim 1, wherein the optimization of the RIS reflection matrix adopts a differential evolution algorithm, and the mutation and crossover operations are carried out according to the following strategies: Wherein, the The variation vector of the ith individual at the g+1st generation is determined by three individuals randomly selected from the population Differential operation generation is carried out, F is a variation factor, CR is a crossover probability, The jth dimension value at the g generation for the ith individual, For the test individuals of the ith individual at the g+1st generation, random j (0, 1 is a random number uniformly distributed) was generated by mutation and crossover operation.
7. The method for optimizing the 5G-R channel model of the large-sized station for the high-speed rail according to claim 1, wherein a variation factor F and a crossover probability CR in the differential evolution algorithm adopt an adaptive updating mechanism, and an updating formula is as follows: wherein G is the current iteration number, G max is the maximum iteration number, and F min ,F max ,CR min ,CR max is a preset parameter threshold.
8. The method for optimizing a 5G-R channel model for a large-sized station for high-speed rail according to claim 1, wherein the optimized RIS reflection matrix Φ opt is used for constructing an optimized equivalent channel matrix of a communication system The calculation formula is as follows: wherein, H d is a direct link channel matrix from gNB to user, H r is a channel matrix from RIS to user, H t is a channel matrix from gNB to RIS, and phi opt is an optimized RIS reflection coefficient matrix; The reflection coefficient in the RIS refers to the reflection intensity and phase adjustment capability of each unit of the super surface to an incident electromagnetic signal, which is usually expressed in a complex form and comprises two dimensions of amplitude (amplitude) and phase, by adjusting and controlling the electromagnetic characteristics (such as capacitance, inductance or other material parameters) of the super surface unit, the reflection coefficient of each unit can be accurately controlled, and the adjusted reflection coefficient can be used for realizing the directional reflection or refraction of the signal, effectively enhancing the expected direction signal and inhibiting the interference direction signal, thereby optimizing the communication link.
9. The method for optimizing the 5G-R channel model of the large-scale station for the high-speed rail according to claim 1, wherein the system evaluates the communication rate by adopting the following shannon channel capacity formula under the configuration of the optimized RIS reflection matrix so as to measure the theoretical maximum channel capacity: Wherein C represents the channel capacity (unit is bps/Hz) of the system, ρ is the signal-to-noise ratio (SNR), N t is the number of antennas at the transmitting end, I is the identity matrix, Represents the conjugate transpose of H eff ; By calculating channel capacity values C before and C after before and after optimization, respectively, and taking the difference: ΔC=C after -C before 。

Description

5G-R channel model optimization method for high-speed rail large-scale station Technical Field The invention relates to a 5G-R channel model optimization method for a high-speed rail large-scale station, and belongs to the technical field of wireless communication. Background With the rapid development of 5G and future 6G communication technologies, high-mobility and complex-structure scenes such as typical high-speed rail large stations are more highly required for communication systems. In such environments, 5G-R systems are assuming high capacity, low latency, high reliability communication tasks as a wireless communication solution for rail traffic. However, with the continuous increase of communication demands, the data volume is exponentially increased, and the conventional wireless system has difficulty in adapting to complex channel environments with limited space, dense reflection paths and prominent multipath effect of a large-scale station of a high-speed railway by increasing the transmitting power of a base station or densely deploying infrastructure to improve coverage and capacity, and also has problems of high construction cost, shortage of spectrum resources and the like. RIS is used as an emerging passive communication auxiliary technology, can realize intelligent reflection and directional control of incident signals by regulating and controlling electromagnetic characteristics of surface units, has the advantages of low power consumption, low cost, no need of radio frequency links and the like, and is particularly suitable for complex communication environments with wide space, serious shielding and remarkable multipath effect such as high-speed railway large-scale stations. The introduction of the method in the 5G-R system provides an effective means for improving the channel capacity and the communication reliability in the station scene. However, the RIS faces the problems of real-time performance and computational complexity of reflection coefficient optimization in practical application, and the existing method is dependent on high-cost iterative computation, so that the method is difficult to adapt to a rapidly-changing channel environment. In this context, deep learning techniques, especially CNNs, exhibit good performance in processing complex high-dimensional channel data, enabling automatic extraction of key features in CSI. However, the feature dimension of CNN extraction is high, and direct use for optimization will increase the computational burden, and it is difficult to meet the real-time requirement. Therefore, features need to be compressed further in combination with dimension reduction techniques. Although the traditional method such as principal component analysis has certain maintenance reduction capability, local structural characteristics of data are difficult to maintain, and the effect is limited when CSI data with manifold characteristics in a large-scale station of a high-speed rail is processed. Therefore, it is necessary to develop a novel optimization method, effectively combine the RIS technology with CNN feature extraction, manifold learning dimension reduction and efficient optimization algorithm, and perform modeling optimization on the channel characteristics of the 5G-R system of the high-speed rail large-scale station so as to improve the capacity, instantaneity and robustness of the system in a high-dynamic and multipath complex environment. Disclosure of Invention Aiming at overcoming the defects and shortcomings of the prior art, the invention provides a 5G-R channel model optimization method for a high-speed rail large-scale station, which can integrate intelligent super-surface, deep learning feature extraction, manifold dimension reduction and evolutionary algorithm optimization and aims at remarkably improving the channel capacity, the frequency spectrum efficiency and the real-time response capability of a 5G-R communication system. The technical scheme adopted for solving the technical problems is that the 5G-R channel model optimization method for the high-speed rail large-scale station comprises the following steps: step 1, channel data acquisition; signals in downlink SSB and CSI-RS in a 5G-R communication system in 2100MHz band and 2 x 10MHz bandwidth configuration are collected by an adapted sensor and device and used for channel estimation to help extract channel characteristics from the gNB to the RIS, the RIS to the UE, and the gNB directly to the user. Under 2100MHz frequency band, the 5G-R system has stronger signal penetrability and lower propagation loss, and is particularly suitable for high-speed mobile environments. Its free space path loss can be expressed as: L=20log10(f)+20log10(d)-147.55(dB) Where f is the frequency and d is the propagation distance. Under the frequency band, the signal can penetrate through the barrier better, meanwhile, the attenuation is kept low, and the communication reliability in a high-speed scene i