Method, apparatus, and device for selecting parameters of channel state information, and storage medium

Abstract

The present application discloses a method, apparatus, and device for selecting parameters of channel state information, and a storage medium. The method comprises: acquiring sampling data of a target location at different moments, the target location being the location of parameters to be selected; on the basis of the sampling data, determining the number of target data streams of the target location; on the basis of the number of target data streams and a preset neural network model, determining a target precoding matrix of the target location; and, on the basis of the target precoding matrix, determining a channel quality indicator of the target location, the number of data streams, the precoding matrix, and the channel quality indicator all belonging to parameters of channel state information.

Description

Channel state information parameter selection method, device, equipment and storage medium

[0001] Related applications

[0002] This application claims priority to Chinese patent application No. 202410042195.X filed on January 10, 2024, the entire contents of which are incorporated by reference into this application. Technical Field

[0003] The present application relates to the field of wireless communication technology, and in particular to a method, apparatus, device, and storage medium for selecting parameters of channel state information. Background Art

[0004] In order to improve signal quality, MIMO (Multiple-Input Multiple-Output) wireless communication technology is currently used to implement signal transmission between a transmitter and a receiver.

[0005] Common MIMO requires multiple antennas at both the transmitter and receiver ends to send and receive signals. To achieve high-speed downlink transmission, the base station sends a pilot sequence to the user. The user uses this pilot sequence to estimate the transmission channel and calculate channel state information based on the channel. This channel state information is then fed back to the base station to guide MIMO transmission. However, as the number of antennas increases, the pilot signal and channel state feedback information also increase, resulting in a decrease in the actual effective data transmission rate.

[0006] Summary of the Invention

[0007] The main purpose of this application is to provide a method, device, equipment and storage medium for selecting parameters of channel state information, aiming to solve the technical problem in the prior art that as the number of antennas increases, the pilot and channel state feedback information also increases, resulting in a decrease in the actual effective data transmission rate.

[0008] To achieve the above objectives, the present application provides a method for selecting parameters of channel state information, the method comprising:

[0009] Acquire sampling data at target location points at different times, where the target location points are location points of the parameters to be selected, and the target location points are location points that need to be sampled among the geographical locations within the coverage area of ​​the base station;

[0010] Determine the target data stream number of the target location point based on the sampling data;

[0011] Determining a target precoding matrix for the target location point based on the target number of data streams and a preset neural network model;

[0012] A channel quality indicator of the target location point is determined based on the target precoding matrix, wherein the number of data streams, the precoding matrix, and the channel quality indicator are all parameters of channel state information.

[0013] In one embodiment, the preset neural network model includes a VAE, and the step of determining a target precoding matrix for the target location point based on the target number of data streams and the preset neural network model includes:

[0014] If the target location point is a known location point, determining an optimal precoding matrix set corresponding to the target number of data streams;

[0015] Based on the VAE, the optimal precoding matrix set is encoded and reconstructed to obtain the target precoding matrix of the target position point. The VAE takes the optimal precoding matrix set of each data stream number as input and the precoding matrix of the position point corresponding to each data stream number as output, and trains the first neural network model to be trained.

[0016] In one embodiment, the VAE includes an encoder and a decoder, and the step of encoding and reconstructing the optimal precoding matrix set based on the VAE to obtain the target precoding matrix of the target location point includes:

[0017] Performing dimension conversion on each optimal preset coding matrix in the optimal preset coding matrix set based on the encoder to obtain a corresponding Gaussian variable;

[0018] Performing an average operation on the means and variances of all the Gaussian variables based on the VAE to determine the average mean and average variance of the Gaussian variables;

[0019] Based on the VAE, searching for a target Gaussian variable having the smallest difference from the average mean and the average variance from each of the Gaussian variables;

[0020] A precoding matrix is ​​reconstructed based on the decoder and the mean and variance of the target Gaussian variable to obtain a target precoding matrix for the target location point.

[0021] In one embodiment, the step of determining the channel quality indicator of the target location point based on the target precoding matrix includes:

[0022] Determining a target SINR for each of the data streams based on the target precoding matrix;

[0023] Performing an average operation on each of the target SINRs to obtain an equivalent SINR value;

[0024] Search a preset mapping table for a channel quality indicator corresponding to the equivalent SINR value.

[0025] In one embodiment, the step of determining the target data stream number of the target location point based on the sampled data includes:

[0026] If the target location point is a known location point, determining the mutual information of the channels on each subcarrier at the target location point based on the sampled data;

[0027] Performing a sum operation on the mutual information quantities to obtain a mutual information sum;

[0028] Determine the number of data streams at each moment based on the maximum value of the mutual information sum, and obtain a data stream number set of the target location point;

[0029] Determining the mode of the data stream set, and judging whether the mode satisfies a preset proportion;

[0030] If the majority satisfies the preset proportion, determining the majority as the target data flow number of the target location point;

[0031] Or if the majority does not satisfy the preset proportion, the majority minus one is determined to be the target data stream number of the target location point.

[0032] In one embodiment, the preset neural network model includes a GPR, and the step of determining a target precoding matrix for the target location point based on the target number of data streams and the preset neural network model further includes:

[0033] If the target location point is an unknown location point, query the target known location point closest to the target location point;

[0034] Determining the reference flow number of the data flow at the target known position point;

[0035] Selecting a corresponding target GPR based on the reference stream number, wherein the GPR is obtained by iteratively training a second neural network model to be trained, taking a sample position point as input and outputting a mean and a variance of a Gaussian variable corresponding to a precoding matrix of the sample position point;

[0036] Generate a reference mean and a reference variance of the Gaussian variable of the precoding matrix corresponding to the target known position point based on the target GPR;

[0037] The precoding matrix is ​​reconstructed based on the encoder in the VAE, the reference mean and the reference variance to obtain a target precoding matrix for the target position point.

[0038] In one embodiment, the step of determining the target data stream number of the target location point based on the sampled data further includes:

[0039] If the target location point is an unknown location point, the reference stream number is used as the target data stream number of the target location point.

[0040] In addition, to achieve the above-mentioned purpose, the present application also provides a channel state information parameter selection device, which includes:

[0041] An acquisition module is used to acquire sampling data at a target location point at different times, wherein the target location point is a location point of a parameter to be selected, and the target location point is a location point that needs to be sampled among geographical locations within the coverage area of ​​the base station;

[0042] A flow number determination module, configured to determine a target data flow number of the target location point based on the sampled data;

[0043] a matrix determination module, configured to determine a target precoding matrix for the target location point based on the target number of data streams and a preset neural network model;

[0044] The quality determination module is configured to determine a channel quality indicator of the target location point based on the target precoding matrix, wherein the number of data streams, the precoding matrix, and the channel quality indicator are all parameters of the channel state information.

[0045] In addition, to achieve the above-mentioned purpose, the present application also proposes a parameter selection device for channel state information, which includes: a memory, a processor, and a parameter selection program for channel state information stored on the memory and runnable on the processor, wherein the parameter selection program for channel state information is configured to implement the steps of the parameter selection method for channel state information as described above.

[0046] In addition, to achieve the above-mentioned purpose, the present application also proposes a storage medium, on which a parameter selection program for channel state information is stored. When the parameter selection program for channel state information is executed by a processor, the steps of the parameter selection method for channel state information as described above are implemented.

[0047] The present application provides a parameter selection method, device, equipment and storage medium for channel state information. Compared with the prior art in which the pilot and channel state feedback information increases with the increase in the number of antennas, resulting in a decrease in the actual effective data transmission rate, in the present application, sampling data at a target location point at different times is obtained, the target location point is the location point of the parameter to be selected, and the target location point is the location point that needs to be sampled among the geographical points within the base station coverage area; the target number of data streams at the target location point is determined based on the sampling data; the target precoding matrix of the target location point is determined based on the target number of data streams and a preset neural network model; the channel quality indicator of the target location point is determined based on the target precoding matrix, wherein the number of data streams, the precoding matrix and the channel quality indicator are all parameters of the channel state information. That is, in the present application, the target number of data streams at the target location point is determined by sampling data at different times, and then the precoding matrix and channel quality indicator of any target location point are inferred based on the target number of data streams and the preset neural network model, thereby avoiding the feedback process and avoiding the increase of channel state feedback information when the number of antennas increases, thereby improving the actual effective data transmission rate. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0049] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0050] FIG1 is a schematic diagram of a device mechanism for selecting parameters of channel state information in a hardware operating environment according to an embodiment of the present application;

[0051] FIG2 is a flow chart of a first embodiment of a method for selecting parameters for channel state information of the present application;

[0052] FIG3 is a flow chart of a second embodiment of a method for selecting parameters of channel state information of the present application;

[0053] FIG4 is a schematic diagram of a specific implementation flow of selecting parameters for known position points in the parameter selection method for channel state information of the present application;

[0054] FIG5 is a flow chart of a third embodiment of a method for selecting parameters for channel state information of the present application;

[0055] FIG6 is a schematic diagram of a specific implementation flow of selecting parameters for unknown position points in the parameter selection method for channel state information of the present application;

[0056] FIG7 is a schematic diagram of the structural configuration of the channel state information parameter selection device of the present application.

[0057] The realization of the objectives, functional features and advantages of this application will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0058] It should be understood that the specific embodiments described herein are only used to explain the present application and are not intended to limit the present application.

[0059] Refer to Figure 1, which is a schematic diagram of the structure of a parameter selection device for channel state information in a hardware operating environment involved in an embodiment of the present application.

[0060] As shown in FIG1 , the channel state information parameter selection device may include: a processor 1001, such as a central processing unit (CPU), a communication bus 1002, a user interface 1003, a network interface 1004, and a memory 1005. The communication bus 1002 is used to implement communication between these components. The user interface 1003 may include a display screen and an input unit such as a keyboard. The user interface 1003 may also include a standard wired interface or a wireless interface. The network interface 1004 may include a standard wired interface or a wireless interface (such as a wireless fidelity (Wi-Fi) interface). The memory 1005 may be a high-speed random access memory (RAM) or a stable non-volatile memory (NVM), such as a disk storage device. The memory 1005 may also be a storage device independent of the aforementioned processor 1001.

[0061] Those skilled in the art will appreciate that the structure shown in FIG1 does not limit the device for selecting channel state information parameters, and may include more or fewer components than shown in the figure, or combine certain components, or arrange the components differently.

[0062] As shown in FIG. 1 , the memory 1005 as a storage medium may include an operating system, a network communication module, a user interface module, and a parameter selection program for channel state information.

[0063] In the channel state information parameter selection device shown in Figure 1, the network interface 1004 is mainly used for data communication with other devices; the user interface 1003 is mainly used for data interaction with the user; the processor 1001 and the memory 1005 in the channel state information parameter selection device of the present application can be set in the channel state information parameter selection device, and the channel state information parameter selection device calls the channel state information parameter selection program stored in the memory 1005 through the processor 1001, and executes the channel state information parameter selection method provided in the embodiment of the present application.

[0064] An embodiment of the present application provides a method for selecting parameters of channel state information. Referring to FIG. 2 , FIG. 2 is a flow chart of a first embodiment of a method for selecting parameters of channel state information of the present application.

[0065] It should be noted that the executor of this embodiment may be a parameter selection device for the channel state information, and the parameter selection device for the channel state information may be an electronic device such as a personal computer, a smart phone, a tablet computer, or other other devices that can achieve the same or similar functions. This embodiment does not limit this. In this embodiment and the following embodiments, the parameter selection method for the channel state information of this application is described using the parameter selection device for the channel state information as an example.

[0066] In this embodiment, the method for selecting parameters of the channel state information includes:

[0067] Step S10: acquiring sampling data at target location points at different times, wherein the target location points are location points of parameters to be selected, and the target location points are location points that need to be sampled among geographical locations within the coverage area of ​​the base station.

[0068] Among them, the location point can be understood as a geographical location point within the coverage area of ​​the base station. The location point can be a known location point or an unknown location point. A known location point can be understood as a location point where communication transmission data has been measured, and an unknown location point can be understood as a location point where communication transmission data has not been measured. The target location point can be a location point where parameters need to be configured, and the location point can be any location point on the transmission channel.

[0069] Sampling can be understood as sampling of the channel matrix of a location point in the time domain and on different subcarriers.

[0070] In a specific implementation, the target location point requiring configuration parameters is first determined, and then SVD (Singular Value Decomposition) is performed on the channel matrix of the target location point at different times to obtain a precoding matrix of the number of streams of the right unitary matrix. Based on the precoding matrix, the equivalent channel data of the target location point, the noise power during communication, and the zero-forcing equalizer at the receiving end are obtained, so as to determine the number of data streams at the target location point based on these data. The acquisition time can be customized by the user according to needs or determined based on historical experience, and there is no specific limitation. The channel matrix is ​​a matrix with a dimension of the number of transmitting and receiving antennas on different subcarriers.

[0071] Step S20: determining the target data flow number of the target location point based on the sampling data.

[0072] The target data flow number can be understood as the number of data flows passing through the target location point at a certain moment.

[0073] It should be noted that the sampling data can be used to first determine the ratio of the signal to interference plus noise that can be reached by each data stream at the target location point, and then the ratio can be used to determine the mutual information that can be reached by each subcarrier passing through the target location point. Since the mutual information can be understood as the difference between the probability of the transmitter sending a signal and the probability of the receiver receiving the signal, it can accurately reflect the strength of the correlation between the transmitter and the receiver. Therefore, determining the target number of data streams at the target location point through the mutual information can ensure the accuracy of determining the target number of data streams.

[0074] In a specific implementation, the ratio of the signal to interference plus noise that can be achieved for each data stream can be determined based on the zero-forcing equalizer data in the sampled data, the number of data streams, and the noise power during communication. The mutual information that can be achieved for each subcarrier at the target location can then be determined based on this ratio. The target number of streams at the target location can also be determined based on the mutual information. The achievable mutual information can be understood as the maximum achievable mutual information, and the achievable signal can be understood as the maximum signal that can be passed through the channel at the target location at the sampling moment.

[0075] Step S30: determining a target precoding matrix for the target location point based on the target number of data streams and a preset neural network model.

[0076] Among them, the preset neural network model is obtained by iteratively training the training model by taking the optimal precoding matrix of each data stream as input and the precoding matrix of the target position point as output.

[0077] It should be noted that by inferring the target precoding matrix in the parameters of the channel state information of any position point through a preset neural network model, the channel feedback process is avoided, and the calculation process of the feedback data is avoided, thereby improving the rate of determining the target precoding matrix of the target position point.

[0078] It should be noted that, since the neural network model is updated iteratively in real time, determining the target precoding matrix by presetting the neural network model can improve the accuracy of obtaining the target precoding matrix.

[0079] The preset neural networks may be multiple or a composite neural network.

[0080] In a specific implementation, the preset neural network may include VAE (Variational Auto-Encoder) and GPR (Gaussian Process Regression). If the target position point is a known position point, the optimal precoding matrix set corresponding to the number of streams of the target position point can be determined according to the target data stream number, and the optimal precoding matrix set is input into the VAE, so that the optimal precoding matrix set is converted into a low-dimensional Gaussian variable through the VAE, and the precoding matrix is ​​reconstructed through the Gaussian variable to obtain the target coding matrix of the target position point; if the target position point is an unknown position point, the neighboring point interpolation method can be used to use the corresponding data of the target precoding matrix of the nearest known position point, GPR and VAE to reconstruct the precoding matrix to obtain the target precoding matrix of the unknown position point.

[0081] In a specific implementation, since the target precoding matrix is ​​obtained based on the optimal precoding matrix of each data stream, and the Gaussian variable corresponding to the target precoding matrix is ​​a set of Gaussian variables corresponding to the optimal precoding matrix set, but since new parameters may be added during the precoding matrix reconstruction process, the target precoding matrix may not be the same as any optimal precoding matrix in the optimal precoding matrix set.

[0082] Step S40: determining a channel quality indicator of the target location point based on the target precoding matrix, wherein the number of data streams, the precoding matrix, and the channel quality indicator are all parameters of channel state information.

[0083] It should be noted that Channel Quality Indicator (CQI) is a process for establishing and managing continuous improvement within the supply chain, emphasizing defect prevention and reducing variation and waste. Because the target precoding matrix is ​​not a known optimal precoding matrix, using the target precoding matrix to determine the CQI requires re-determining the signal-to-interference-plus-noise ratio for each data stream to accurately obtain the CQI at the target location. This CQI does not require receiver feedback.

[0084] In a specific implementation, if the target location point is a known location point, the ratio of the signal to interference plus noise reachable by the target precoding matrix corresponding to the data stream can be recalculated, and then the channel quality indication of the target location point can be determined from the preset mapping table based on the ratio.

[0085] It should be noted that since the number of data streams, precoding matrix, and channel quality indicator are all parameters of the channel state information, and no feedback information from the receiving end is required when determining these parameters, instead, methods such as VAE are used to select representative CSI (Channel State Information) parameters at known locations. The CSI parameters of any location point are inferred through methods such as VAE and GPR, thereby improving the actual effective data transmission rate.

[0086] This embodiment provides a parameter selection method for channel state information. Compared with the prior art in which the pilot and channel state feedback information increases with the increase in the number of antennas, resulting in a decrease in the actual effective data transmission rate, in this application, sampling data at a target location point at different times is obtained, the target location point is the location point of the parameter to be selected, and the target location point is the location point that needs to be sampled among the geographical points within the base station coverage area; the target number of data streams at the target location point is determined based on the sampling data; the target precoding matrix of the target location point is determined based on the target number of data streams and a preset neural network model; the channel quality indicator of the target location point is determined based on the target precoding matrix, wherein the number of data streams, the precoding matrix, and the channel quality indicator are all parameters of the channel state information. That is, in this application, the target number of data streams at the target location point is determined by sampling data at different times, and then the precoding matrix and channel quality indicator of any target location point are inferred based on the target number of data streams and the preset neural network model, thereby avoiding the feedback process and avoiding the increase of channel state feedback information when the number of antennas increases, thereby improving the actual effective data transmission rate.

[0087] Refer to FIG3 , which is a flow chart of a second embodiment of a method for selecting parameters of channel state information of the present application.

[0088] Based on the above embodiment, in this embodiment, when the target location point is a known location point, the preset neural network model includes VAE, and the step of determining the target precoding matrix of the target location point based on the target number of data streams and the preset neural network model includes:

[0089] Step S3a1: if the target location point is a known location point, determining an optimal precoding matrix set corresponding to the target number of data streams;

[0090] Step S3a2, based on the VAE, the optimal precoding matrix set is encoded and reconstructed to obtain the target precoding matrix of the target position point. The VAE takes the optimal precoding matrix set of each data stream number as input and the precoding matrix of the position point corresponding to each data stream number as output, and trains the first neural network model to be trained.

[0091] The optimal precoding matrix is ​​the precoding matrix used when the sum of the mutual information amount achievable on each subcarrier at the target location point is maximized, and each number of data streams corresponds to an optimal precoding matrix.

[0092] It should be noted that, since VAE takes the optimal precoding matrix set of the number of data streams as input and the precoding matrix of each position point corresponding to the number of data streams as output to train the model to be trained, and the model to be trained is a neural network model, the use of VAE can increase the accuracy and efficiency of determining the target precoding matrix at the target position point and reduce the amount of calculation, thereby shortening the time required for the target position point to select the target precoding matrix when transmitting data, and speeding up the transmission of effective data.

[0093] It should be noted that VAE needs to be created before step S3a2;

[0094] In the specific implementation, referring to Figure 4, step S201: historical channel data is a data set that is sampled in the time domain for the channel matrix of a finite position point and saved. For a known position point, at a certain time domain sampling point, the channel H on each subcarrier is first decomposed by SVD: H = U*S*V, and the first N_s columns of the right unitary matrix V are the precoding matrix with N_s streams, where U is the left matrix and S is the diagonal matrix; after decomposition, the H matrix is ​​decomposed into the U matrix, the S matrix and the V matrix. Here, N_s is taken from the beginning, and the maximum value of the number of data streams is the smaller of the number of transmitting antennas and the number of receiving antennas. Then, fix N_s from one to the maximum, use the first N_s columns of the V matrix as the precoding matrix W, and calculate the signal to interference plus noise ratio (SINR) that can be reached for each data stream:

[0095] Among them, E is the zero-forcing equalizer at the receiving end, G is the equivalent channel, and F represents the power. is the noise power. The numerator represents the useful signal power, the first term in the denominator is the interference power between the data streams, and the second term is the enhanced noise power. Next, calculate the mutual information achievable on each subcarrier, and find the precoding matrix used when the sum of the mutual information is maximized. The precoding matrix is ​​the optimal precoding matrix for the time domain sampling point with the number of streams being N_s. Each position point has a different time domain sampling channel. Through this processing, the optimal precoding matrix set with the number of streams from one to the maximum can be obtained; the calculation formula for the mutual information is I=log2(1+SINR). Step S202: Train the VAE with the number of streams from one to the maximum. For different numbers of data transmission streams, use the optimal precoding matrix after SVD decomposition of the corresponding number of streams as the input of the VAE. The encoder reduces the input into a low-dimensional Gaussian variable, and the decoder reconstructs the original input. During the training process, the learning rate is set to 10 -3 , a total of 100 rounds of training, in which 128 input and output samples are extracted in each round, and the ADAM algorithm is used to update the gradient of the VAE neural network parameters.

[0096] In one embodiment, the VAE includes an encoder and a decoder, and the step of encoding and reconstructing the optimal precoding matrix set based on the VAE to obtain the target precoding matrix of the target location point includes:

[0097] Step S3a21, performing dimension conversion on each optimal preset coding matrix in the optimal preset coding matrix set based on the encoder to obtain a corresponding Gaussian variable;

[0098] Step S3a22, performing an average operation on the means and variances of all the Gaussian variables based on the VAE to determine the average mean and average variance of the Gaussian variables;

[0099] Step S3a23, searching for a target Gaussian variable having the smallest difference from the average mean and the average variance from each of the Gaussian variables based on the VAE;

[0100] Step S3a24: reconstructing a precoding matrix based on the decoder and the mean and variance of the target Gaussian variable to obtain a target precoding matrix for the target location point.

[0101] It should be noted that by converting the optimal preset coding matrix into a Gaussian variable, the noise interference to the quantized signal during channel propagation can be simulated through the Gaussian variable, so as to deeply understand and analyze the transmission characteristics of the message in a noisy and interfered channel, and average the mean and variance of the Gaussian variable, and use the average mean and average variance to accurately determine the central variable of the Gaussian variable, so as to determine the target precoding matrix that best represents the target position point through the central variable.

[0102] In the specific implementation, refer to Figure 4, step S207: according to the determined number of data streams RI_fixed, the optimal precoding matrix set corresponding to the number of streams at the position point is input into the VAE of the corresponding number of streams. Then, the encoder will output low-dimensional latent space Gaussian variables, which are characterized by their respective means and variances. Next, the means and variances of all Gaussian variables are averaged, and these two values ​​are used as standards to measure the gap between each Gaussian variable and the statistical average. Finally, the Gaussian variable with the smallest gap is selected as the representative variable, and its mean and variance are input into the decoder to obtain the reconstructed precoding matrix, which is the representative precoding matrix.

[0103] Furthermore, the step of determining the channel quality indicator of the target location point based on the target precoding matrix includes:

[0104] Step S4a1, determining a target SINR for each of the data streams based on the target precoding matrix;

[0105] Step S4a2, performing an average operation on each of the target SINRs to obtain an equivalent SINR value;

[0106] Step S4a3: searching a preset mapping table for a channel quality indicator corresponding to the equivalent SINR value.

[0107] The preset mapping table may represent a mapping relationship between the SINR value and the channel quality indicator.

[0108] It should be noted that since new parameters may be introduced in the process of reconstructing the precoding matrix, the target precoding matrix finally obtained is different from the best precoding matrix in the best precoding matrix set, and the channel index needs to be obtained by using a preset mapping table to search according to the SINR value corresponding to the target precoding matrix. Therefore, in order to ensure the accuracy of the determined channel quality index, it is necessary to recalculate the SINR value according to the target precoding matrix, determine the target SINR value of each data stream number, and average the target SINR values ​​of all data stream numbers to obtain an equivalent SINR value that can characterize all data stream numbers.

[0109] In the specific implementation, referring to Figure 4, step S208: for the channel matrix sampled in a certain time domain at the location point, use the precoding matrix fixed in step S207 to calculate the SINR achievable on each subcarrier and each data stream. Then, all SINR values ​​are averaged to obtain an equivalent SINR value. Through the preset correspondence table between SINR and CQI (Channel Quality Indicator), the equivalent SINR is mapped to the CQI value. In this way, the historical CQI data of the location point in the time domain can be obtained. Finally, the historical CQI data is averaged and rounded down. This integer value is the representative value of the CQI at this location point.

[0110] Furthermore, the step of determining the target data stream number of the target location point based on the sampled data includes:

[0111] Step S2a1: if the target location point is a known location point, determining the mutual information of the channels on each subcarrier at the target location point based on the sampled data;

[0112] Step S2a2, performing a summation operation on each mutual information to obtain a mutual information sum;

[0113] Step S2a3, determining the number of data streams at each moment based on the maximum value of the mutual information sum, and obtaining a data stream number set of the target location point;

[0114] Step S2a4, determining the mode of the data stream set, and judging whether the mode satisfies a preset proportion;

[0115] Step S2a5: if the majority satisfies the preset proportion, determining the majority as the target data flow number of the target location point;

[0116] Step S2a6, or if the majority does not satisfy the preset proportion, determining the majority minus one as the target data stream number of the target location point.

[0117] It should be noted that since the mode is the number that appears the most times in a set, the mode can accurately represent the set. Therefore, the mode of the data stream number set of the target location point at different times is used as the target data stream number of the target location point, that is, the data stream number with the most repetitions is screened out from the data stream number set at different times as the target data stream number.

[0118] In a specific implementation, referring to Figure 4, step S203: For a channel sampled at a certain time domain at a certain location, calculate the mutual information of the channel on each subcarrier when using the optimal precoding matrix with a maximum number of streams from one. Then, select the number of streams used when the mutual information is maximized as the number of data streams used at that sampling moment. This way, the historical number of data streams at that location in the time domain is obtained. Finally, the mode appearing in the historical number of data streams is counted and recorded as RI_mode. Step S204: For that location, determine whether the proportion of RI_mode in the historical number of data streams reaches N. If so, execute step S205; if not, execute step S206. Step S205: Fix the number of data streams used at that location, RI_fixed, to RI_mode, and proceed to step S207. Step S206: Fix the number of data streams used at that location, RI_fixed, to RI_mode-1, and proceed to step S207.

[0119] Refer to FIG5 , which is a flowchart of a third embodiment of a method for selecting parameters of channel state information of the present application.

[0120] Based on the above embodiment, in this embodiment, when the target location point is an unknown location point, the preset neural network model includes a GPR, and the step of determining a target precoding matrix for the target location point based on the target number of data streams and the preset neural network model further includes:

[0121] Step S3b1: if the target location point is an unknown location point, query the target known location point closest to the target location point;

[0122] Step S3b2, determining the reference flow number of the data flow at the target known location point;

[0123] Step S3b3: selecting a corresponding target GPR based on the reference stream number, wherein the GPR is obtained by iteratively training a second neural network model to be trained, using the sample position point as input and the mean and variance of the Gaussian variable corresponding to the precoding matrix of the sample position point as output;

[0124] Step S3b4, generating a reference mean and a reference variance of the Gaussian variable of the precoding matrix corresponding to the target known position point based on the target GPR;

[0125] Step S3b5: reconstruct the precoding matrix based on the encoder in the VAE, the reference mean and the reference variance to obtain the target precoding matrix of the target position point.

[0126] It should be noted that since the number of reference streams of the data stream at the target known position point is known, the corresponding target GPR can be selected according to the number of reference streams, so as to generate the reference mean and reference variance of the Gaussian variable of the corresponding precoding matrix of the target known position point through the target GPR. The encoder in the VAE reconstructs the precoding matrix according to the reference mean and reference variance to obtain the target precoding matrix of the unknown target position point. That is, the target precoding matrix of any position point can be inferred by using GPR and VAE without the need for a feedback process.

[0127] In the specific implementation, referring to Figure 6, step S304: for the precoding matrix of the unknown position point, according to the number of streams obtained in step S303, the GPR of the corresponding number of streams is selected to generate the mean and variance of the Gaussian variable in the latent space corresponding to the precoding matrix, and input it into the decoder of the VAE of the corresponding number of streams to generate the precoding matrix; step S305: for the CQI of the unknown position point, based on the CQI scatter value fixed at the known position point, the CQI value of the unknown position point is inferred using natural neighbor interpolation.

[0128] It should be noted that the GPR needs to be created before step S3b3.

[0129] In the specific implementation, referring to Figure 6, step S301: the number of streams ranges from one to the maximum, and for the representative precoding matrix selected by VAE, the mean and variance of its latent space Gaussian variables are saved to obtain a data set of latent space Gaussian variables corresponding to the representative precoding matrix; step S302: training GPR with the number of streams ranging from one to the maximum. During training, the input of GPR is a known position point, and the output is the mean and variance of the representative precoding matrix in the VAE latent space at the corresponding position point. Setting the kernel function to a Gaussian function and giving the corresponding input and output to the existing GPR fitter can automatically adjust the kernel function parameters to fit the input-output relationship.

[0130] In one embodiment, the step of determining the target data stream number of the target location point based on the sampled data further includes:

[0131] Step S2b1: If the target location point is an unknown location point, the reference stream number is used as the target data stream number of the target location point.

[0132] In the specific implementation, referring to Figure 6, step S303: For an unknown point, the number of data streams must first be determined. By using the nearest neighbor interpolation method, the nearest known point to the unknown point is found, and the fixed number of streams at this known point is assigned to the unknown point.

[0133] The present application also provides a device for selecting parameters of channel state information. Referring to FIG7 , the device for selecting parameters of channel state information includes:

[0134] An acquisition module 701 is configured to acquire sampling data at a target location point at different times, where the target location point is a location point of a parameter to be selected, and the target location point is a location point that needs to be sampled among geographical locations within a base station coverage area;

[0135] A flow number determination module 702 is configured to determine a target data flow number of the target location point based on the sampled data;

[0136] A matrix determination module 703 is configured to determine a target precoding matrix for the target location based on the target number of data streams and a preset neural network model;

[0137] The quality determination module 704 is configured to determine a channel quality indicator of the target location point based on the target precoding matrix, wherein the number of data streams, the precoding matrix, and the channel quality indicator are all parameters of the channel state information.

[0138] In one embodiment, the preset neural network model includes VAE;

[0139] The matrix determination module 703 is also used to determine the optimal precoding matrix set corresponding to the target number of data streams if the target position point is a known position point; the optimal precoding matrix set is encoded and reconstructed based on the VAE to obtain the target precoding matrix of the target position point, and the VAE is obtained by training the first neural network model to be trained by taking the optimal precoding matrix set of each data stream number as input and the precoding matrix of each position point corresponding to the data stream number as output.

[0140] In one embodiment, the VAE includes an encoder and a decoder;

[0141] The matrix determination module 703 is also used to perform dimension conversion on each optimal preset coding matrix in the optimal preset coding matrix set based on the encoder to obtain corresponding Gaussian variables; perform averaging operation on the means and variances of all the Gaussian variables based on the VAE to determine the average mean and average variance of the Gaussian variables; query the target Gaussian variable with the smallest difference from the average mean and the average variance from each Gaussian variable based on the VAE; reconstruct the precoding matrix based on the decoder and the mean and variance of the target Gaussian variable to obtain the target precoding matrix of the target position point.

[0142] In one embodiment, the quality determination module 704 is further configured to determine a target SINR for each of the data streams based on the target precoding matrix; perform an average operation on each of the target SINRs to obtain an equivalent SINR value; and search a preset mapping table for a channel quality indicator corresponding to the equivalent SINR value.

[0143] In one embodiment, the stream number determination module 702 is further used to, if the target location point is a known location point, determine the mutual information of the channels on each subcarrier of the target location point based on the sampled data; perform a sum operation on each of the mutual information amounts to obtain a mutual information sum; determine the number of data streams at each moment based on the maximum value of the mutual information sum to obtain a data stream number set for the target location point; determine the mode of the data stream number set, and determine whether the mode satisfies a preset proportion; if the mode satisfies the preset proportion, determine the mode as the target data stream number for the target location point; or if the mode does not satisfy the preset proportion, determine the mode minus one as the target data stream number for the target location point.

[0144] In one embodiment, the preset neural network model includes GPR;

[0145] The matrix determination module 703 is also used to query the target known position point closest to the target position point if the target position point is an unknown position point; determine the reference stream number of the data stream at the target known position point; select the corresponding target GPR based on the reference stream number, the GPR is obtained by iteratively training the second neural network model to be trained by taking the sample position point as input and the mean and variance of the Gaussian variables corresponding to the precoding matrix of the sample position point as output; generate the reference mean and reference variance of the Gaussian variables of the precoding matrix corresponding to the target known position point based on the target GPR; reconstruct the precoding matrix based on the encoder in the VAE, the reference mean and the reference variance to obtain the target precoding matrix of the target position point.

[0146] In one embodiment, the stream number determination module 702 is further configured to use the reference stream number as the target data stream number of the target location point if the target location point is an unknown location point.

[0147] The specific implementation of the channel state information parameter selection device of the present application is basically the same as the various embodiments of the channel state information parameter selection method described above, and will not be repeated here.

[0148] An embodiment of the present application provides a storage medium, and the storage medium stores one or more programs, and the one or more programs can also be executed by one or more processors to implement the steps of any of the above-mentioned channel state information parameter selection methods.

[0149] The specific implementation of the storage medium of the present application is basically the same as the embodiments of the parameter selection method of the channel state information described above, and will not be repeated here.

[0150] It should be noted that, in this document, the terms "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, method, article, or apparatus comprising a series of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus. In the absence of further limitations, an element defined by the phrase "comprising an etc." does not exclude the presence of other identical elements in the process, method, article, or apparatus comprising the element.

[0151] The serial numbers of the above embodiments of the present application are for description only and do not represent the advantages or disadvantages of the embodiments.

[0152] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus the necessary general hardware platform, and of course can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes a number of instructions for enabling a terminal device (which can be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods of each embodiment of the present application.

[0153] The above are merely optional embodiments of the present application and do not limit the patent scope of the present application. Any equivalent structure or equivalent process transformation made using the contents of the present application specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present application.

Claims

1. A method for parameter selection of channel state information, wherein, The method for parameter selection of the channel state information includes: Obtaining sampling data at the target position points at different times, where the target position points are the position points of the parameters to be selected, and the target position points are the position points that need to be sampled among the geographical position points within the base station coverage area; Determining the target data stream number of the target position points based on the sampling data; Determining the target precoding matrix of the target position points based on the target data stream number and the preset neural network model; Determining the channel quality indication of the target position points based on the target precoding matrix, where the data stream number, precoding matrix, and the channel quality indication all belong to the parameters of the channel state information.

2. The method for parameter selection of channel state information according to claim 1, wherein, The preset neural network model includes VAE. The step of determining the target precoding matrix of the target position points based on the target data stream number and the preset neural network model includes: If the target position point is a known position point, determining the optimal precoding matrix set corresponding to the target data stream number; Encoding and reconstructing the optimal precoding matrix set based on the VAE to obtain the target precoding matrix of the target position point. The VAE takes the optimal precoding matrix set of each data stream number as the input and takes the precoding matrix of the corresponding position points of each data stream number as the output, and is obtained by training the first neural network model to be trained.

3. The method for parameter selection of channel state information according to claim 2, wherein, The VAE includes an encoder and a decoder. The step of encoding and reconstructing the optimal precoding matrix set based on the VAE to obtain the target precoding matrix of the target position point includes: Converting the dimensions of each optimal preset coding matrix in the optimal preset coding matrix set based on the encoder to obtain the corresponding Gaussian variables; Performing an averaging operation on the means and variances of all the Gaussian variables based on the VAE to determine the average mean and average variance of the Gaussian variables; Querying the target Gaussian variable with the smallest difference from the average mean and the average variance from each of the Gaussian variables based on the VAE; Reconstructing the precoding matrix based on the decoder and the mean and variance of the target Gaussian variable to obtain the target precoding matrix of the target position point.

4. The method for parameter selection of channel state information according to claim 1, wherein, The step of determining the channel quality indication of the target position points based on the target precoding matrix includes: Determining the target SINR of each data stream number based on the target precoding matrix; Performing an averaging operation on each of the target SINRs to obtain an equivalent SINR value; Searching for the channel quality indication corresponding to the equivalent SINR value from the preset mapping table.

5. The method for parameter selection of channel state information according to claim 1, wherein, The step of determining the target data stream number of the target position points based on the sampling data includes: If the target position point is a known position point, determining the mutual information amount of the channels on each subcarrier of the target position point based on the sampling data; Performing a summation operation on each of the mutual information amounts to obtain a mutual information sum; Determining the data stream number at each moment based on the maximum value of the mutual information sum to obtain the data stream number set of the target position points; Determining the mode of the data stream number set and judging whether the mode meets the preset proportion; If the mode satisfies the preset ratio, determine that the mode is the target data stream number of the target position point; Or if the mode does not satisfy the preset ratio, determine that the mode minus one is the target data stream number of the target position point.

6. The method for parameter selection of channel state information according to any one of claims 1 to 5, wherein, The preset neural network model includes GPR. The step of determining the target precoding matrix of the target position point based on the target data stream number and the preset neural network model further includes: If the target position point is an unknown position point, query the target known position point closest to the target position point; Determine the reference stream number of the data stream at the target known position point; Select a corresponding target GPR based on the reference stream number. The GPR takes the sample position point as the input and the mean and variance of the Gaussian variable corresponding to the precoding matrix of the sample position point as the output, and is obtained by iteratively training the second neural network model to be trained; Generate the reference mean and reference variance of the Gaussian variable corresponding to the precoding matrix of the target known position point based on the target GPR; Reconstruct the precoding matrix based on the encoder in the VAE, the reference mean, and the reference variance to obtain the target precoding matrix of the target position point.

7. The method for parameter selection of channel state information according to claim 6, wherein, The step of determining the target data stream number of the target position point based on the sampling data further includes: If the target position point is an unknown position point, use the reference stream number as the target data stream number of the target position point.

8. A parameter selection device for channel state information, wherein, The parameter selection device for the channel state information includes: An acquisition module, configured to acquire sampling data at the target position point at different times. The target position point is the position point of the parameter to be selected, and the target position point is the position point that needs to be sampled among the geographical position points within the base station coverage area; A stream number determination module, configured to determine the target data stream number of the target position point based on the sampling data; A matrix determination module, configured to determine the target precoding matrix of the target position point based on the target data stream number and the preset neural network model; A quality determination module, configured to determine the channel quality indication of the target position point based on the target precoding matrix, where the data stream number, the target precoding matrix, and the channel quality indication all belong to the parameters of the channel state information.

9. A parameter selection device for channel state information, wherein, The parameter selection device for the channel state information includes: a memory, a processor, and a parameter selection program for the channel state information stored on the memory and executable on the processor. The parameter selection program for the channel state information is configured to implement the steps of the parameter selection method for the channel state information as described in any one of claims 1 to 7.

10. A storage medium, wherein, A program for implementing the parameter selection method for the channel state information is stored on the storage medium. The program for implementing the parameter selection method for the channel state information is executed by the processor to implement the steps of the parameter selection method for the channel state information as described in any one of claims 1 to 7.

Images