Geographic position-based high-speed rail mobile communication 6g fully-decoupled network uplink transmission method and system

Abstract

Disclosed in the present invention are a geographic position-based high-speed rail mobile communication 6G fully-decoupled network uplink transmission method and system. The method comprises: according to geographic positions of a train and each uplink base station, performing line-of-sight channel mapping, so as to serve as channel estimation to carry out a secondary combining joint reception processing of the uplink signals; designing a deep learning neural network, taking the mapped line-of-sight channel as an input, and outputting a multi-user joint precoding that meets constraint conditions; performing offline training on a network by using historical channel data of each position; in an actual deployment stage, inputting a line-of-sight channel corresponding to the train position, the network outputting a joint precoding at the corresponding position, so as to achieve transmission without feedback. Compared with a traditional precoding design solution, the present invention solves the problem in a high-speed rail scenario of untimely channel feedback and large resource overhead due to high-speed movement, improves the frequency effect of wireless communication in a high-speed rail scenario, and provides an effective uplink transmission method for future high-speed rail mobile communications.

Description

A method and system for uplink transmission of 6G fully decoupled network for high-speed rail mobile communication based on geographical location Technical Field

[0001] The present invention belongs to the field of wireless mobile communication technology, and relates to deep learning and multi-user multi-input multi-output transmission in high-speed rail scenarios. Specifically, it relates to a high-speed rail mobile communication 6G fully decoupled network uplink transmission method and system based on geographic location. Background Art

[0002] With the booming development of railway systems, high-speed train communications have attracted widespread attention. In fifth-generation mobile networks (5G), massive multiple-input, multiple-output (mMIMO) with joint precoding design has achieved significant performance improvements, fully leveraging diversity and multipath gain. However, high mobility degrades the channel in many ways, such as Doppler frequency offset (DFO), shorter coherence time, and phase noise, resulting in a significant performance degradation of high-speed communication systems. Furthermore, because 5G utilizes higher-frequency spectrum, handoffs between base stations will become more frequent in high-speed mobility scenarios.

[0003] Traditional solutions for Long Term Evolution for Railways (LTE-R) plan to use more frequent pilot transmissions and channel state information (CSI) feedback. This results in significant consumption of time and frequency resources. Furthermore, because feedback delay is exacerbated in high-speed mobility scenarios, even with more frequent pilot transmissions for channel estimation, accurate CSI is difficult to obtain.

[0004] Sixth-generation (6G) mobile communication networks are expected to provide more flexible resource allocation and enhance user experience. In their paper "A Fully-Decoupled RAN Architecture for 6G Inspired by Neurotransmission," Academician Yu Quan et al. proposed a 6G fully decoupled radio access network (FD-RAN) architecture, decoupling traditional base stations into uplink base stations (UBSs), downlink base stations (DBSs), and control base stations. FD-RAN offers significant advantages in high-speed rail communication systems. Firstly, all control plane signals are transmitted by the control base station. Since the control information is relatively small, the control base station can use a lower frequency band to cover a larger area, significantly reducing handover overhead. Secondly, FD-RAN employs a geolocation-based, feedback-free transmission scheme. This scheme not only mitigates the negative impacts of channel variations and resource overhead, but also avoids feedback delays through location prediction. Since train trajectories are fixed, their locations are easily known in advance. However, determining appropriate precoding based solely on location information remains challenging. Furthermore, considering the multiple mobile relays (MRs) on the train, a multi-user MIMO (MU-MIMO) scenario needs to be utilized, which necessitates joint precoding design. Furthermore, in order to conserve more pilot resources, signal reception at the base station also struggles to utilize the real-time channel estimated by the pilots.

[0005] In summary, the problems with existing technologies are: (1) Traditional pilot-based channel estimation results in greater resource overhead in high-speed rail scenarios, and due to feedback delay, it is difficult for the transmitter to obtain real-time channel information. (2) In 6G fully decoupled networks, the high feedback delay caused by the hardware decoupling and isolation architecture has a significant performance loss for transmission methods that rely on existing feedback mechanisms.

[0006] Summary of the Invention

[0007] Purpose of the invention: The purpose of the present invention is to provide a method and system for uplink transmission of a 6G fully decoupled network for high-speed rail mobile communications based on geographic location, which uses deep learning to fit the mapping relationship between the train's geographic location and precoding, and can directly select precoding based on the train's geographic location to replace the traditional channel estimation and channel feedback process.

[0008] Technical solution: To achieve the above-mentioned invention purpose, the technical solution adopted by the present invention is: a high-speed rail mobile communication 6G fully decoupled network uplink transmission method based on geographical location, comprising the following steps:

[0009] Step 1: Perform pure line-of-sight channel mapping based on the geographical location of the train and each uplink base station, using this as channel estimation to perform two-level combined joint reception processing on the uplink signal;

[0010] Step 2: Design a deep neural network that takes the pure line-of-sight channel information mapped between the train and the uplink base station as input and multi-user joint precoding information that satisfies the constant modulus constraint as output. Design the loss function of the deep neural network based on the system optimization goal, and use the precoding output of the deep neural network for data transmission.

[0011] Step 3: The deep neural network uses the historical channel data of each position sample to perform offline training and update parameters until convergence;

[0012] Step 4: In the actual deployment phase, the line-of-sight channel corresponding to the train position is input, and the deep neural network outputs the joint precoding at the corresponding position to achieve feedback-free transmission.

[0013] Furthermore, suppose there are L base stations serving K train mobile relays in the scenario, and the real channel between the mobile relay on the top of the kth train and the lth uplink base station is H kl , the pure line-of-sight channel mapping relationship is: in

[0014] represents the phase offset between antennas, θ kl represents the angle between the mobile relay on the top of the kth train and the lth uplink base station, f represents the carrier frequency, λ represents the carrier wavelength, v represents the train speed, c represents the speed of light, d B and d V Represents the antenna spacing of the base station and relay, N B and N V Represent the number of antennas of the base station and relay, β kl is the large-scale fading coefficient related to distance.

[0015] Furthermore, the precoding vector of the kth relay at the top of the train is Meeting power constraints as well as where p max The maximum power limit.

[0016] Furthermore, the joint reception processing of performing two-level combining on the uplink signal includes:

[0017] The base station first performs a distributed processing on the received signal. represents the combined vector of the kth relay at the lth base station, y l represents the received signal at the lth base station, and the signal estimation for the kth relay can be expressed as Based on pure line-of-sight channel mapping The local minimum mean square error (MMSE) method is used to calculate

[0018] The edge cloud performs a centralized second-level merging of the signals processed by the base station at the first level. The signal of the k-th relay after merging can be expressed as: Represents all secondary merging coefficients.

[0019] Furthermore, the corresponding signal to noise ratio when it is maximum can be obtained by calculation: in N k represents the interference term between relays, σ 2 represents the noise power,

[0020] Further, substituting into the calculation, the maximum frequency efficiency that can be achieved by the k-th relay is:

[0021] Among them, τ p and τ c denote the length of the pilot and coherence time respectively, and |·| denotes the determinant of the matrix.

[0022] Furthermore, using P = [P1,…,P K ] represents the precoding vector of all users. The system optimization goal takes relay frequency efficiency and maximization into consideration. The optimization problem of the system can be modeled as:

[0023] st|[P k ] i | 2 =1, for i=1,…,N V

[0024] The constraint condition is the constant modulus constraint of the coded phase shifter, which needs to ensure that the amplitude of each pre-coded element is 1. Under this optimization goal, the loss function of the deep neural network is defined as the negative of the system frequency effect sum, N represents the number of samples.

[0025] Furthermore, the designed deep learning neural network takes the channels between all relays and base stations as input and transforms the output of the neural network α k According to P k =cos(2πα k )+j·sin(2πα k ) is converted into a precoding vector that satisfies the constant modulus constraint, that is, the output of the neural network can be regarded as the precoding phase.

[0026] In specific implementation, the deep learning neural network can adopt a fully connected deep learning neural network or a residual network.

[0027] The present invention also provides a computer system, including a memory, a processor, and a computer program stored in the memory and executable on the processor. The computer program is loaded into the processor to implement the steps of the method for uplink transmission of a high-speed rail mobile communication 6G fully decoupled network based on geographic location.

[0028] Beneficial effects: Compared with the existing technology, the present invention has the following advantages: 1. The present invention provides a high-speed rail uplink communication system under the FD-RAN architecture, which avoids the problems of feedback delay and resource overhead in the high-speed rail scenario. 2. The present invention realizes multi-user multi-input multi-output precoding selection based on the train's geographic location, saving pilot estimation and feedback overhead, and can further improve the system frequency efficiency. 3. The present invention uses deep learning to fit the mapping relationship between the train's geographic location information and precoding, and adopts pure line-of-sight channel mapping before input to effectively alleviate the problem of low latitude of geographic location information. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] FIG1 is a schematic diagram of a scenario to which an embodiment of the present invention is applicable.

[0030] FIG2 is a schematic diagram of a neural network parameter updating process according to an embodiment of the present invention.

[0031] FIG3 is a schematic diagram of a neural network training convergence process according to an embodiment of the present invention.

[0032] FIG4 is a frequency-efficiency comparison curve diagram of an embodiment of the present invention under different architectures. DETAILED DESCRIPTION

[0033] In order to make the objectives, technical solutions, and advantages of the present invention more clearly understood, the following detailed description of an embodiment of the present invention is given in conjunction with the accompanying drawings. This embodiment is implemented based on the technical solutions of the present invention, and provides a detailed implementation method and specific operation process. It should be understood that the specific examples described herein are only used to illustrate the present invention, and the scope of protection of the present invention is not limited to the following embodiments.

[0034] An embodiment of the present invention discloses a method for uplink transmission of a high-speed rail mobile communication 6G fully decoupled network based on geographic location, the main steps of which are as follows:

[0035] Step 1: Perform pure line-of-sight channel mapping based on the geographical location of the train and each uplink base station, using this as channel estimation to perform two-level combined joint reception processing on the uplink signal;

[0036] Step 2: Design a deep neural network that takes the pure line-of-sight channel information mapped between the train and the uplink base station as input and outputs multi-user joint precoding information that meets the constant modulus constraint. Design the loss function of the deep neural network based on the system optimization goal, and use the precoding output of the deep neural network for data transmission.

[0037] Step 3: The deep neural network uses the historical channel data of each position sample to perform offline training and update parameters until convergence;

[0038] Step 4: In the actual deployment phase, the pure line-of-sight channel corresponding to the train position is input, and the deep neural network outputs the joint precoding at the corresponding position to achieve feedback-free transmission.

[0039] The specific steps of the embodiment of the present invention are described below with reference to the high-speed rail communication scenario shown in Figure 1. Assume that the uplink base station and the mobile relay are located in the same plane, there are L base stations serving K train mobile relays, and the real channel between the mobile relay on the top of the kth train and the lth uplink base station is H kl , the pure line-of-sight channel mapping relationship is: in

[0040] represents the phase offset between antennas, θ kl represents the angle between the mobile relay on the top of the kth train and the lth uplink base station, f represents the carrier frequency, λ represents the carrier wavelength, v represents the train speed, c represents the speed of light, d B and d V Represents the antenna spacing of the base station and relay, N B and N V Represent the number of antennas of the base station and relay, β kl is the large-scale fading coefficient related to distance.

[0041] The following is the specific process of the base station and edge cloud's two-level joint reception of signals.

[0042] The precoding vector of the kth relay is Meeting power constraints And the constant modulus constraint of the coded phase shifter satisfies the precoding amplitude of each antenna element to be 1, that is, The received signal at the lth base station is where x k Represents the signal of the kth relay, satisfying n l Represents the noise of the receiver.

[0043] The base station first performs a distributed processing on the received signal. represents the combined vector of the k-th relay at the l-th base station, and the signal estimation of the k-th relay can be expressed as Based on pure line-of-sight channel mapping The local minimum mean square error (MMSE) method is used to calculate in Represents the error between the mapped pure line-of-sight channel and the real channel, represents the error term due to estimation,

[0044] The edge cloud performs a centralized second-level merging of the signals processed by the base station at the first level, using a kl represents the secondary combining coefficient, then the signal of the kth relay after combining can be expressed as:

[0045] in represents the equivalent channel at the edge cloud, Denotes the equivalent channel of the k-th relay. Considering the interference between users, the signal-to-interference-noise ratio of the received signal is calculated using the UatF lower bound, and Represents all the secondary combining coefficients, then the signal to interference noise ratio of the k-th relay is in represents the interference term between relays, σ 2 Represents the noise power. By calculation, we can get the corresponding signal to noise ratio when it is the largest. Substituting into the calculation, the maximum frequency efficiency that can be achieved by the k-th relay is:

[0046] where τ p and τ c denote the length of the pilot and coherence time respectively, and |·| denotes the determinant of the matrix.

[0047] Use P=[P1,…,P K ] represents the precoding vector of all users. Considering the maximization of relay frequency efficiency and system frequency efficiency, the optimization problem of maximization can be modeled as:

[0048] st|[P k ] i | 2 =1, for i=1,...,N V

[0049] The constraint condition is the constant modulus constraint of the coded phase shifter, which ensures that the amplitude of each precoding element is 1. In addition to maximizing the frequency efficiency as in this embodiment, the system optimization goal can also be maximizing system energy efficiency or throughput.

[0050] The following describes the deep learning neural network structure used. In the embodiment of the present invention, the deep neural network can adopt a fully connected deep learning neural network to avoid the labeling overhead of supervised learning; it can also adopt a residual network (ResNet) to improve the training convergence speed of the deep neural network. The channels between all relays and base stations are taken as input. For example, after being converted into vectors by the flattening layer, the number of neurons is 1024, 512, 256 and K*N in turn. V , batch normalization is used after each layer to prevent overfitting. Lambda processing layer is designed before output to convert the output α of the neural network k According to P k =exp(j·2πα k )=cos(2πα k )+j·sin(2πα k ) is converted into a precoding vector that satisfies the constant modulus constraint, that is, the output of the neural network can be regarded as the precoding phase. For the above example of maximizing the system frequency effect sum, the loss function of the neural network is defined as the negative of the system frequency effect sum, that is N represents the number of samples. Each sample is obtained by sampling the historical channel data at that location.

[0051] The network training process is shown in Figure 2. As the loss function converges to its minimum, the network seeks the precoding solution that maximizes the system's frequency efficiency—the optimal solution to the optimization problem. This enables geolocation-based precoding design without channel feedback. The convergence curve during network training is shown in Figure 3, demonstrating that the network converges within 20 iterations, requiring relatively low computational resources.

[0052] To make this embodiment more intuitive and compare the advantages and disadvantages of the FD-RAN architecture with other architectures, Figure 4 shows a comparative diagram of high-speed rail relay frequency efficiency simulation results under different architectures. The results show that small cells perform worse than FD-RAN, as independent signal reception cannot eliminate significant inter-relay interference. Compared to cellular networks, higher frequency efficiency is achieved when the train is close to the cellular base station. However, frequency efficiency is lower at the cell edge. This demonstrates that the FD-RAN architecture in this embodiment can provide stable high frequency efficiency in high-speed rail communications.

[0053] A computer system disclosed in an embodiment of the present invention includes a memory, a processor, and a computer program stored in the memory and executable on the processor. The computer program is loaded into the processor to implement the steps of the aforementioned embodiment of the method for uplink transmission of a high-speed rail mobile communication 6G fully decoupled network based on geographic location.

[0054] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for uplink transmission of 6G fully decoupled network of high-speed railway mobile communication based on geographical location, characterized in that: The following steps are involved: Step 1: Perform pure line-of-sight channel mapping based on the geographical location of the train and each uplink base station, and use this as channel estimation to perform two-level combined joint reception processing on the uplink signal; Step 2: Design a deep neural network that takes the pure line-of-sight channel information mapped between the train and the uplink base station as input and the multi-user joint precoding information that satisfies the constant modulus constraint as output, and design the loss function of the deep neural network according to the system optimization goal, and use the precoding output by the deep neural network for data transmission; Step 3: The deep neural network uses the historical channel data of each position sample for offline training to update parameters until convergence; Step 4: In the actual deployment phase, the pure line-of-sight channel corresponding to the train position is input, and the deep neural network outputs the joint precoding at the corresponding position to achieve feedback-free transmission.

2. According to the geographical location-based high-speed railway mobile communication 6G fully decoupled network uplink transmission method according to claim 1, it is characterized by: The pure line-of-sight channel mapping relationship between the mobile relay on the top of the kth train and the lth uplink base station is: in, represents the phase offset between antennas, θ kl represents the angle between the mobile relay on the top of the kth train and the lth uplink base station, f represents the carrier frequency, λ represents the carrier wavelength, v represents the train speed, c represents the speed of light, and d B and d V Respectively represent the antenna spacing of the base station and the relay, N B and N V Respectively represent the number of antennas of the base station and relay, β kl is the attenuation coefficient.

3. According to the geographical location-based high-speed railway mobile communication 6G fully decoupled network uplink transmission method according to claim 1, it is characterized by: The precoding vector of the kth relay at the top of the train is Meeting power constraints and |[P k ] i | 2 =1,i=1,...,N V ; where N V represents the number of relay antennas, p max is the maximum power limit.

4. According to the geographical location-based high-speed railway mobile communication 6G fully decoupled network uplink transmission method according to claim 1, it is characterized by: The joint receiving process of performing two-level merging of uplink signals includes: the base station first performs a first-level distributed processing on the received signal, using represents the combined vector of the kth relay at the lth base station, y l represents the received signal at the lth base station, and the signal estimation for the kth relay is expressed as Based on pure line-of-sight channel mapping The local minimum mean square error merging method is used to calculate The edge cloud performs a centralized second-level merger of the signals processed by the base station at the first level. The signal of the k-th relay after the merger is expressed as: represents all secondary combining coefficients, and L represents the number of base stations.

5. According to the geographical location-based high-speed railway mobile communication 6G fully decoupled network uplink transmission method of claim 4, it is characterized by: By calculating the maximum signal-to-interference-noise ratio, we can get the in H nl represents the real channel between the mobile relay on the top of the nth train and the lth uplink base station, P n represents the precoding vector of the nth relay, P k represents the precoding vector of the kth relay, N k represents the interference term between relays, σ 2 represents the noise power, H kl represents the real channel between the mobile relay on the top of the kth train and the lth uplink base station, and K represents the number of train relays.

6. According to the geographical location-based high-speed railway mobile communication 6G fully decoupled network uplink transmission method of claim 5, it is characterized by: The maximum frequency efficiency that the k-th relay can achieve is: Among them, τ p and τ c denote the length of the pilot and coherence time respectively, |·| denotes the determinant of the matrix, 7. According to the geographical location-based high-speed railway mobile communication 6G fully decoupled network uplink transmission method according to claim 1, it is characterized by: Use P = [P1,…,P K ] represents the precoding vector of all users. The system optimization goal is to maximize the relay frequency efficiency. The optimization problem of the system is modeled as: Among them, SE k represents the maximum frequency efficiency that the k-th relay can achieve, N V represents the number of relay antennas, K represents the number of train relays; the loss function of the deep neural network is defined as the negative of the system frequency effect sum, N represents the number of samples.

8. The method for uplink transmission of a 6G fully decoupled network for high-speed railway mobile communication based on geographical location according to claim 1, characterized in that: The designed deep learning neural network uses the pure line-of-sight channel between all relays and base stations as input and transforms the output of the neural network α before output. k According to P k =cos(2πα k )+j·sin(2πα k ) is converted into a precoding vector that satisfies the constant modulus constraint.

9. The method for uplink transmission of a 6G fully decoupled network for high-speed railway mobile communication based on geographical location according to claim 1, characterized in that: The deep learning neural network adopts a fully connected deep learning neural network or a residual network.

10. A computer system comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: The computer program is loaded into the processor to implement the steps of the high-speed rail mobile communication 6G fully decoupled network uplink transmission method based on geographic location according to any one of claims 1-9.

Images