A low-orbit satellite channel prediction method and system based on Gaussian process regression

Abstract

A method and system for predicting a low-orbit satellite channel based on Gaussian process regression, the prediction method comprising the following steps: obtaining the geographical location of a user terminal within a set time range and the corresponding channel state information to construct a training sample; utilizing the prior knowledge that the low-orbit satellite channel is affected by line-of-sight paths, non-line-of-sight paths, and noise to design the kernel function used in the Gaussian process, and using the designed kernel function as a measure of the correlation between the training samples to train the Gaussian process; and predicting the channel state information of the channel between the user terminal and the low-orbit satellite through the trained Gaussian process according to the geographical location of the user terminal. The present invention also proposes a low-orbit satellite channel prediction system, electronic device, and computer-readable storage medium based on Gaussian process regression. In the case where a large amount of sufficient training data cannot be obtained, the prediction method of the present invention can achieve a better low-orbit satellite channel prediction effect with relatively less training data.

Description

Technical Field

[0001] The present invention belongs to the technical field of satellite communications, and in particular relates to a low-orbit satellite channel prediction method and system based on Gaussian process regression. Background Art

[0002] With the development of communication technology, the Internet of Things is widely affecting people's production and living quality. The Internet of Everything will be an inevitable development direction in the present and future. With the gradual application and deployment of the fifth-generation mobile communication system, the sixth-generation mobile communication system is being discussed and studied. The sixth-generation mobile communication system has put forward higher requirements for the Internet of Everything. The sixth-generation mobile communication system hopes to build a global network with wide-area coverage and support for larger-scale device connections. Due to the influence of realistic conditions and costs, most of the existing communication networks are deployed in densely populated areas. For sparsely populated areas such as mountainous areas and forests, it is difficult for existing communication networks to provide services. Therefore, the sixth-generation mobile communication system hopes to use satellite communications to achieve extensive connections around the world. Low-orbit satellites with relatively low latency, path loss and launch costs are more favored by future industrial communications.

[0003] With the help of effective channel state information, low-orbit satellites can achieve better transmission performance by selecting appropriate transmission power, coding methods, etc. However, due to the high mobility of low-orbit satellites and the influence of transmission delay, the channel state information obtained by low-orbit satellites faces the problem of being outdated. Although some solutions have been made to improve the impact of outdated channel state information, these solutions either passively compensate for the loss of system performance at the cost of scarce wireless resources, or only achieve a part of the optimal performance under outdated channel state information. Summary of the invention

[0004] The purpose of the present invention is to provide a low-orbit satellite channel prediction method and system based on Gaussian process regression in order to solve the problems in the above-mentioned prior art. The kernel function of the Gaussian process is designed by using the prior knowledge of the low-orbit satellite channel so that the kernel function can better express the characteristics of the low-orbit satellite channel. At the same time, Gaussian process regression is used for channel prediction, thereby achieving a higher low-orbit satellite channel prediction performance with a smaller number of training samples.

[0005] In order to achieve the above object, the present invention has the following technical solutions:

[0006] A low-orbit satellite channel prediction method based on Gaussian process regression comprises the following steps:

[0007] Obtaining the geographical location of the user terminal within a set time range and the corresponding channel state information to construct a training sample;

[0008] The kernel function used in the Gaussian process is designed by using the prior knowledge that the low-orbit satellite channel is affected by the line-of-sight path, non-line-of-sight path and noise. The designed kernel function is used as a measure of the correlation between training samples to train the Gaussian process.

[0009] According to the geographical location of the user terminal, the channel state information of the channel between the user terminal and the low-orbit satellite is predicted through the trained Gaussian process.

[0010] As a preferred solution, when obtaining the geographical location of the user terminal within a set time range and the corresponding channel state information to construct the training sample, the geographical location data of the user terminal within the set time range is expressed as:

[0011] X=(x1,x2,…,x N ) T

[0012] Where N is the size of the training set; x n =(x n,1 ,x n,2 ) T Indicates the two-dimensional geographic coordinates of the user terminal, x n,1 ,x n,2 are the first and second dimension geographic coordinates respectively;

[0013] The channel state information expression corresponding to the geographical location of the user terminal is:

[0014]

[0015] In the formula, Indicates the channel state information of the nth sample of the user terminal on the mth subcarrier;

[0016] Then, the constructed training sample set is expressed as (X,H A ).

[0017] As a preferred solution, the step of designing a kernel function used for the Gaussian process by using the prior knowledge that the low-orbit satellite channel is affected by the line-of-sight path, the non-line-of-sight path and the noise includes:

[0018] A square exponential kernel is used to describe the influence of the line-of-sight path. The square exponential kernel is expressed as:

[0019]

[0020]

[0021] A rational quadratic kernel is used to describe the impact of non-line-of-sight paths. The rational quadratic kernel is expressed as:

[0022]

[0023]

[0024] A white noise kernel is used to describe the effect of noise. The white noise kernel is expressed as:

[0025]

[0026]

[0027] Then, the designed kernel function is expressed as:

[0028] k DK (x i ,x j ) = k SE (x i ,x j )+k RQ (x i ,x j )+k WK (x i ,x j )

[0029] The hyperparameters to be trained in the kernel function are:

[0030] Θ={σ SE ,l SE ,σ rq ,l rq ,α,σ wk}

[0031] In the formula, x i 、x j are any two samples in the training set, where i, j = 1, 2, …, N; σ SE is the variance parameter of the squared exponential kernel, l SE is the length scale parameter of the squared exponential kernel, σ rq is the variance parameter of the rational quadratic kernel, l rq is the length scale parameter of the rational quadratic kernel, α is the mixing scale parameter of the rational quadratic kernel, σ wk is the noise variance parameter.

[0032] As a preferred solution, the step of using the designed kernel function as a measure of the correlation between training samples and training the Gaussian process includes:

[0033] The log-likelihood function logp(H A |X):

[0034]

[0035] Where K is the abbreviation of the covariance matrix K(X,X), and the calculation expression of K(X,X) is:

[0036]

[0037] By gradient descent method, we maximize the log-likelihood function logp(H A |X), determine the hyperparameter Θ={σ SE ,l SE ,σ rq ,l rq ,α,σ wk}, complete the Gaussian process training.

[0038] As a preferred solution, when predicting the channel state information of the channel between the user terminal and the low-orbit satellite through the trained Gaussian process according to the geographical location of the user terminal, the channel state information at the next moment is predicted according to the geographical location of the k user terminal:

[0039] Enter the two-dimensional geographic coordinates x of the kth user terminal at the next moment * ;

[0040] The channel state information of the channel between the k user terminal and the low-orbit satellite at the next moment is predicted as follows:

[0041]

[0042] In the formula, k * K(x * ,X), K(x * ,X) is calculated as follows:

[0043] K(x * ,X)=(k DK (x * ,x1),k DK (x * ,x2),…,k DK (x * ,x N )) T .

[0044] As a preferred solution, when predicting the channel state information of the channel between the user terminal and the low-orbit satellite through the trained Gaussian process according to the geographical location of the user terminal, the channel state information of the user terminal at time k+1 is predicted according to the geographical location of the user terminal at time k and the corresponding channel state information: the training set is updated to X according to the geographical location of the user terminal at time k and the corresponding channel state information new =(x2,…,x N ,x * ) T ,

[0045] The updated training set is used to complete the prediction of the user terminal channel state information at time k+1;

[0046] The above process is repeated continuously to complete continuous low-orbit satellite channel prediction.

[0047] A low-orbit satellite channel prediction system based on Gaussian process regression, comprising:

[0048] A training sample construction module is used to obtain the geographical location of the user terminal within a set time range and the corresponding channel state information to construct a training sample;

[0049] The Gaussian process training module is used to design the kernel function used in the Gaussian process by using the prior knowledge that the low-orbit satellite channel is affected by the line-of-sight path, non-line-of-sight path and noise, and use the designed kernel function as a measure of the correlation between training samples to train the Gaussian process;

[0050] The prediction module is used to predict the channel state information of the channel between the user terminal and the low-orbit satellite through a trained Gaussian process according to the geographical location of the user terminal.

[0051] An electronic device, comprising:

[0052] a memory storing at least one instruction; and

[0053] The processor executes the instructions stored in the memory to implement the low-orbit satellite channel prediction method based on Gaussian process regression.

[0054] A computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method for predicting a low-orbit satellite channel based on Gaussian process regression is implemented.

[0055] Compared with the prior art, the present invention has at least the following beneficial effects:

[0056] As an active prediction scheme, channel prediction can effectively compensate for the system performance loss caused by outdated channel state information without sacrificing other wireless resources. The present invention uses the prior knowledge that the low-orbit satellite channel is affected by the main line-of-sight path, the secondary non-line-of-sight path and a small amount of noise to design the kernel function used for the Gaussian process, and uses this kernel function as a measure of the correlation between input samples to train the Gaussian process. In the channel prediction stage, according to the geographical location of the user terminal at this time, the channel state information of the channel between the user terminal and the low-orbit satellite is predicted through the Gaussian process, thus completing the low-orbit satellite channel prediction at this moment. In the case where a large amount of sufficient training data cannot be obtained, the low-orbit satellite channel prediction method of the present invention can achieve better low-orbit satellite channel prediction effects with relatively less training data. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1 A schematic diagram of a low-orbit satellite communication system model according to an embodiment of the present invention;

[0058] Figure 2 This is a flow chart of a method for predicting a low-orbit satellite channel based on Gaussian process regression according to an embodiment of the present invention;

[0059] Figure 3 This is a curve chart showing the change in training loss value based on the long short-term memory network solution;

[0060] Figure 4 This is a comparison chart of the low-orbit satellite channel prediction performance of the method of the present invention and existing deep learning solutions. DETAILED DESCRIPTION

[0061] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, ordinary technicians in this field can also obtain other embodiments without making creative work.

[0062] Channel prediction uses historical channel state information to achieve reasonable predictions of channel state information at future moments, thereby tracking channel changes. With the development of machine learning and deep learning, there are currently some studies using deep learning algorithms for low-orbit satellite channel prediction. For example, there have been studies using long short-term memory networks to learn the correlation between historical channel state information to predict low-orbit satellite channels; there have been studies using convolutional neural networks to extract spatial features of channel information, and then using long short-term memory networks to learn the temporal correlation of channel information to predict low-orbit satellite channels. However, the amount of training data required for channel prediction using deep neural networks is usually very large. The resources carried by low-orbit satellites are usually very limited, and in practice they cannot support the collection and training process of such a large amount of data. When the number of training samples is insufficient, the performance of low-orbit satellite channel prediction methods based on deep neural networks will drop significantly.

[0063] See also Figure 1 , low-orbit satellites establish communication services with user terminals on the ground. With the help of effective channel state information, low-orbit satellites can achieve better communication performance by adjusting transmission parameters. However, due to the high mobility and high transmission delay of low-orbit satellites, channel state information faces the problem of being outdated. Channel prediction is to use historical channel state information to predict the channel at future times, so as to reduce the impact of outdated channel state information on the performance of the communication system. Figure 2 The low-orbit satellite channel prediction method based on Gaussian process regression in the embodiment of the present invention follows the following steps:

[0064] 1) Before the prediction begins, the low-orbit satellite obtains the geographical location of the user terminal within a certain period of time and its corresponding channel state information as training samples:

[0065] Specifically, the geographical location data of the user terminal within a period of time is:

[0066] X=(x1,x2,…,x N ) T

[0067] Among them, N represents the size of the training set, x n =(x n,1 ,x n,2 ) T Indicates the two-dimensional geographic coordinates of the user terminal, x n,1 ,x n,2 are the first and second dimension geographic coordinates respectively.

[0068] Considering the OFDM system, the channel state information corresponding to the geographical location of the user terminal is:

[0069]

[0070] in, Represents the channel state information of the nth sample of the user terminal on the mth subcarrier.

[0071] The constructed training sample set is (X,H A ).

[0072] 2) Using the prior knowledge of low-orbit satellite channels to design the kernel function used in the Gaussian process:

[0073] The LEO satellite channel is affected by a primary line-of-sight path, a secondary non-line-of-sight path, and a small amount of noise.

[0074] Therefore, a square exponential kernel is used to describe the influence of the main line-of-sight path. The square exponential kernel is expressed as:

[0075]

[0076]

[0077] The influence of the secondary non-line-of-sight paths is described by a rational quadratic kernel, which is expressed as:

[0078]

[0079]

[0080] A white noise kernel is used to describe the effect of noise. The white noise kernel is expressed as:

[0081]

[0082]

[0083] The designed kernel function is:

[0084] k DK (x i ,x j ) = k SE (x i ,x j )+k RQ (x i ,x j )+k wK (x i ,x j )

[0085] Where Θ={σ SE ,l SE ,σ rq ,l rq ,α,σ wk} is the hyperparameter to be trained.

[0086] 3) The designed kernel function is used as a measure of the correlation between the Gaussian process training samples to train the Gaussian process. The training process includes the following steps:

[0087] Step 1: Calculate the log-likelihood function logp(H A |X).

[0088]

[0089] Here, K is the abbreviation of the covariance matrix K(X,X), which is:

[0090]

[0091] Step 2: Maximize the log-likelihood function logp(H by gradient descent A |X), optimize and determine the hyperparameter Θ={σ SE ,l SE ,σ rq ,l rq ,α,σ wk}, completing the training process.

[0092] 4) After the training is completed, in the prediction stage, the channel state information at the next moment is predicted by the Gaussian process according to the geographical location of the k user terminal. The prediction process is as follows:

[0093] a) Input the two-dimensional geographic coordinates x of the user terminal k at the next moment * .

[0094] b) At this time, the channel state information of the channel between the user terminal and the low-orbit satellite is predicted as:

[0095]

[0096] Among them, k * K(x * ,X), K(x * ,X) is:

[0097] K(x * ,X)=(k DK (x * ,x1),k DK (x * ,x2),…,k DK (x * ,x N )) T

[0098] This completes the prediction of the k low-orbit satellite channels at the next moment.

[0099] 5) In the process of continuous channel prediction, if the prediction is not terminated, the training set needs to be updated when predicting the channel state information of the user terminal at time k+1. The training set can be updated to X according to the geographical coordinates and channel state information of the user terminal at time k. new =(x2,…,x N ,x * ) T , The prediction of the user terminal channel state information at time k+1 can be completed by referring to the previous training and prediction process. This process can be repeated continuously to complete the continuous low-orbit satellite channel prediction until the prediction is completed.

[0100] Simulation results show that this method can achieve better low-orbit satellite channel prediction effect with relatively fewer training samples.

[0101] In order to verify the performance of the low-orbit satellite channel prediction method proposed in the present invention, the following simulation is performed:

[0102] Simulation conditions

[0103] The simulation experiment of the present invention is carried out on an Intel(R) Core(TM) i5-10400F CPU@2.9GHz, 64-bit Windows operating system, and the simulation software uses MATLAB. This simulation experiment uses Quasi DeterministicRadio Channel Generator (QuaDRiGa) to generate a low-orbit satellite channel that is closer to the actual scene. In this simulation experiment, the center frequency is set to 2.2GHz, the baseband bandwidth is 20MHz, the number of subcarriers is 128, and the scene uses the 3GPP_38.881_Urban_LOS satellite scene. The satellite is 1400km away from the earth's surface, and the elevation angle of the satellite relative to the user is 29.7° and the azimuth angle is 172.7°. The satellite antenna uses a parabolic antenna with a diameter of 3m, a gain of 44dBi, and a polarization mode of LHCP polarization. The user terminal antenna uses an LHCP polarized patch antenna pointing to the sky. The user terminal moves at a uniform speed of 1m / s in the northeast direction on the ground. The channel is sampled every 0.01s, and the first 2000 samples are used to construct the training set.

[0104] Simulation content and result analysis

[0105] In the above environment, the present invention (DKGP) is compared with a Gaussian process prediction scheme (GP) based on a square exponential kernel function, a prediction scheme based on a gated recurrent unit (GRU), and a prediction scheme based on a long short-term memory network (LSTM) in the prior art. Figure 3 The figure shows the change of loss value during the training process based on the long short-term memory network scheme when the number of training samples is 2000. Figure 3 It can be seen that when the number of training samples is 2000, the loss value of the LSTM network can only drop to about 0.017 during training, while the loss value of the network when fully trained generally drops to 10 -4 , 10 -5 At this time, the insufficient number of training samples leads to the inability to fully train the network, which limits the prediction performance of the long short-term memory network. Figure 4 As shown in Figure 2. Among them, the prediction error is reflected by three indicators: mean square error (MSE), root mean square error (RMSE), and mean absolute error (MAE). Figure 4 It can be seen that when the number of training samples is not sufficient, the performance of the low-orbit satellite channel prediction scheme based on the deep neural network will be limited. Whether it is a prediction scheme based on a gated recurrent unit or a prediction scheme based on a long short-term memory network, its prediction performance is lower than that of a scheme based on a Gaussian process. At the same time, since the kernel function of the Gaussian process is designed using the prior knowledge of the low-orbit satellite channel in the present invention so that it can better reflect the channel characteristics, the prediction effect of the low-orbit satellite channel in the present invention is also better than the general Gaussian process prediction scheme based on the square exponential kernel function. The low-orbit satellite channel prediction method based on Gaussian process regression in the embodiment of the present invention uses a Gaussian process for prediction. The number of training samples required for prediction is less than that of the deep learning scheme, and since the kernel function of the Gaussian process is designed using the prior knowledge of the low-orbit satellite channel, the prediction performance of the low-orbit satellite channel is further improved.

[0106] Therefore, it can be seen from the above that the low-orbit satellite channel prediction method based on Gaussian process regression proposed in the present invention can achieve better low-orbit satellite channel prediction effect when the number of training samples is not sufficient.

[0107] Another embodiment of the present invention further provides a low-orbit satellite channel prediction system based on Gaussian process regression, comprising:

[0108] A training sample construction module is used to obtain the geographical location of the user terminal within a set time range and the corresponding channel state information to construct a training sample;

[0109] The Gaussian process training module is used to design the kernel function used in the Gaussian process by using the prior knowledge that the low-orbit satellite channel is affected by the line-of-sight path, non-line-of-sight path and noise, and use the designed kernel function as a measure of the correlation between training samples to train the Gaussian process;

[0110] The prediction module is used to predict the channel state information of the channel between the user terminal and the low-orbit satellite through a trained Gaussian process according to the geographical location of the user terminal.

[0111] Another embodiment of the present invention further provides an electronic device, including:

[0112] a memory storing at least one instruction; and

[0113] The processor executes the instructions stored in the memory to implement the low-orbit satellite channel prediction method based on Gaussian process regression.

[0114] Another embodiment of the present invention further proposes a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the low-orbit satellite channel prediction method based on Gaussian process regression is implemented.

[0115] Exemplarily, the instructions stored in the memory may be divided into one or more modules / units, which are stored in a computer-readable storage medium and executed by the processor to complete the low-orbit satellite channel prediction method based on Gaussian process regression of the present invention. The one or more modules / units may be a series of computer-readable instruction segments capable of completing specific functions, which are used to describe the execution process of the computer program in the server.

[0116] The electronic device may be a computing device such as a smart phone, a notebook, a PDA, and a cloud server. The electronic device may include, but is not limited to, a processor and a memory. Those skilled in the art will appreciate that the electronic device may also include more or fewer components, or a combination of certain components, or different components, for example, the electronic device may also include an input / output device, a network access device, a bus, etc.

[0117] The processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc.

[0118] The memory may be an internal storage unit of the server, such as a hard disk or memory of the server. The memory may also be an external storage device of the server, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, etc. equipped on the server. Furthermore, the memory may include both an internal storage unit of the server and an external storage device. The memory is used to store the computer-readable instructions and other programs and data required by the server. The memory may also be used to temporarily store data that has been output or is to be output.

[0119] It should be noted that the information interaction, execution process and other contents between the above-mentioned module units are based on the same concept as the method embodiment. Their specific functions and technical effects can be found in the method embodiment part and will not be repeated here.

[0120] The technicians in the relevant field can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional units and modules is used as an example for illustration. In practical applications, the above-mentioned function allocation can be completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiment can be integrated in a processing unit, or each unit can exist physically separately, or two or more units can be integrated in one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of software functional units. In addition, the specific names of the functional units and modules are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of this application. The specific working process of the units and modules in the above-mentioned system can refer to the corresponding process in the aforementioned method embodiment, which will not be repeated here.

[0121] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present application implements all or part of the processes in the above-mentioned embodiment method, which can be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and the computer program can implement the steps of the above-mentioned various method embodiments when executed by the processor. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may at least include: any entity or device that can carry the computer program code to the camera device / terminal device, recording medium, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electrical carrier signal, telecommunication signal and software distribution medium. For example, a USB flash drive, a mobile hard disk, a disk or an optical disk.

[0122] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described or recorded in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0123] The embodiments described above are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, a person skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application, and should all be included in the protection scope of the present application.

Claims

1. A low-orbit satellite channel prediction method based on Gaussian process regression, characterized in that: The following steps are involved: Obtaining the geographical location of the user terminal within a set time range and the corresponding channel state information to construct a training sample; When obtaining the geographical location of the user terminal within the set time range and the corresponding channel state information to construct the training sample, the geographical location data expression of the user terminal within the set time range is: X=(x1,x2,…,x N ) T Where N is the size of the training set; x n =(x n,1 ,x n,2 ) T Indicates the two-dimensional geographic coordinates of the user terminal, x n,1 ,x n,2 are the first and second dimension geographic coordinates respectively; The channel state information expression corresponding to the geographical location of the user terminal is: In the formula, Indicates the channel state information of the nth sample of the user terminal on the mth subcarrier; Then, the constructed training sample set is expressed as (X,H A ); The kernel function used in the Gaussian process is designed by using the prior knowledge that the low-orbit satellite channel is affected by the line-of-sight path, non-line-of-sight path and noise. The designed kernel function is used as a measure of the correlation between training samples to train the Gaussian process. The step of designing the kernel function used for the Gaussian process by utilizing the prior knowledge that the low-orbit satellite channel is affected by the line-of-sight path, the non-line-of-sight path and noise comprises: A square exponential kernel is used to describe the influence of the line-of-sight path. The square exponential kernel is expressed as: A rational quadratic kernel is used to describe the impact of non-line-of-sight paths. The rational quadratic kernel is expressed as: A white noise kernel is used to describe the effect of noise. The white noise kernel is expressed as: Then, the designed kernel function is expressed as: k DK (x i ,x j )=k SE (x i ,x j )+k RQ (x i ,x j )+k WK (x i ,x j ) The hyperparameters to be trained in the kernel function are: Θ={σ SE ,l SE ,s rq ,l rq ,a,s wk } In the formula, x i 、x j are any two samples in the training set, where i, j = 1, 2, …, N; σ SE is the variance parameter of the squared exponential kernel, l SE is the length scale parameter of the squared exponential kernel, σ rq is the variance parameter of the rational quadratic kernel, l rq is the length scale parameter of the rational quadratic kernel, α is the mixing scale parameter of the rational quadratic kernel, σ wk is the noise variance parameter; According to the geographical location of the user terminal, the channel state information of the channel between the user terminal and the low-orbit satellite is predicted through the trained Gaussian process.

2. The low-orbit satellite channel prediction method based on Gaussian process regression according to claim 1, characterized in that: The steps of using the designed kernel function as a measure of the correlation between training samples and training the Gaussian process include: The log-likelihood function logp(H A |X): Where K is the abbreviation of the covariance matrix K(X,X), and the calculation expression of K(X,X) is: By gradient descent method, we maximize the log-likelihood function logp(H A |X), determine the hyperparameter Θ={σ SE ,l SE ,σ rq ,l rq ,α,σ wk }, complete the Gaussian process training.

3. The low-orbit satellite channel prediction method based on Gaussian process regression according to claim 2, characterized in that: When the channel state information of the channel between the user terminal and the low-orbit satellite is predicted by the trained Gaussian process according to the geographical location of the user terminal, the channel state information at the next moment is predicted according to the geographical location of the k user terminal: Enter the two-dimensional geographic coordinates x of the kth user terminal at the next moment * ; The channel state information of the channel between the k user terminal and the low-orbit satellite at the next moment is predicted as follows: In the formula, k * K(x * ,X), K(x * ,X) is calculated as follows: K(x * ,X)=(k DK (x * ,x1), k DK (x * ,x2),…,k DK (x * , x N )) T 。 4. The low-orbit satellite channel prediction method based on Gaussian process regression according to claim 3 is characterized in that: When the channel state information of the channel between the user terminal and the low-orbit satellite is predicted by the trained Gaussian process according to the geographical location of the user terminal, the channel state information of the user terminal at time k+1 is predicted according to the geographical location of the user terminal at time k and the corresponding channel state information: the training set is updated to X according to the geographical location of the user terminal at time k and the corresponding channel state information new =(x2,…,x N ,x * ) T , The updated training set is used to complete the prediction of the user terminal channel state information at time k+1; The above process is repeated continuously to complete continuous low-orbit satellite channel prediction.

5. A low-orbit satellite channel prediction system based on Gaussian process regression, characterized in that: include: A training sample construction module is used to obtain the geographical location of the user terminal within a set time range and the corresponding channel state information to construct a training sample; When obtaining the geographical location of the user terminal within the set time range and the corresponding channel state information to construct the training sample, the geographical location data expression of the user terminal within the set time range is: X=(x1,x2,…,x N ) T Where N is the size of the training set; x n =(x n,1 ,x n,2 ) T Indicates the two-dimensional geographic coordinates of the user terminal, x n,1 ,x n,2 are the first and second dimension geographic coordinates respectively; The channel state information expression corresponding to the geographical location of the user terminal is: In the formula, Indicates the channel state information of the nth sample of the user terminal on the mth subcarrier; Then, the constructed training sample set is expressed as (X,H A ); The Gaussian process training module is used to design the kernel function used in the Gaussian process by using the prior knowledge that the low-orbit satellite channel is affected by the line-of-sight path, non-line-of-sight path and noise, and use the designed kernel function as a measure of the correlation between training samples to train the Gaussian process; The step of designing the kernel function used for the Gaussian process by utilizing the prior knowledge that the low-orbit satellite channel is affected by the line-of-sight path, the non-line-of-sight path and noise comprises: A square exponential kernel is used to describe the influence of the line-of-sight path. The square exponential kernel is expressed as: A rational quadratic kernel is used to describe the impact of non-line-of-sight paths. The rational quadratic kernel is expressed as: A white noise kernel is used to describe the effect of noise. The white noise kernel is expressed as: Then, the designed kernel function is expressed as: k DK (x i ,x j )=k SE (x i ,x j )+k RQ (x i ,x j )+k WK (x i ,x j ) The hyperparameters to be trained in the kernel function are: Θ={σ SE ,l SE ,s rq ,l rq ,a,s wk } In the formula, x i 、x j are any two samples in the training set, where i, j = 1, 2, …, N; σ SE is the variance parameter of the squared exponential kernel, l SE is the length scale parameter of the squared exponential kernel, σ rq is the variance parameter of the rational quadratic kernel, l rq is the length scale parameter of the rational quadratic kernel, α is the mixing scale parameter of the rational quadratic kernel, σ wk is the noise variance parameter; The prediction module is used to predict the channel state information of the channel between the user terminal and the low-orbit satellite through a trained Gaussian process according to the geographical location of the user terminal.

6. An electronic device, characterized in that: include: A memory storing at least one instruction; and A processor executes instructions stored in the memory to implement the low-orbit satellite channel prediction method based on Gaussian process regression as described in any one of claims 1 to 4.

7. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the low-orbit satellite channel prediction method based on Gaussian process regression as described in any one of claims 1 to 4 is implemented.

Images