Channel time correlation perception multi-user alternating training method

Abstract

The invention provides a channel time correlation perception multi-user alternating training method. The method comprises the steps of system modeling, antenna selection and alternating training, data transmission and conditional distribution matrix updating. In the system modeling stage, a system transmission model and a time correlation channel model are constructed; in the antenna selection and alternative training stage, an antenna training sequence is guided according to a conditional mean matrix and a conditional variance matrix before training, training and feedback are mutually linked, channel state information (CSI) of a trained antenna is fed back, and whether training is stopped is determined; in the data transmission stage, the base station performs downlink beam forming and data transmission; in the conditional distribution matrix updating stage, the base station updates the conditional mean value matrix and the conditional variance matrix based on the channel model according to the trained antenna set and the channel state information. Compared with the existing multi-user alternate training scheme, the alternate training scheme further reduces the system training overhead while not reducing the outage probability of the system.

Description

Technical Field

[0001] The invention belongs to the technical field of wireless communications, and in particular relates to a channel time-correlated perception multi-user alternating training method. Background Art

[0002] Massive MIMO systems can achieve high communication performance by leveraging the spatial degrees of freedom provided by massive antenna arrays. A fundamental issue is the acquisition of channel state information (CSI). The availability and quality of CSI can determine the performance of MIMO and massive MIMO systems. For massive MIMO downlink, CSI acquisition is particularly challenging. How to ensure that the system can obtain accurate CSI while reducing the huge training and feedback overhead caused by the significant improvement of base station antennas has become an urgent problem to be solved.

[0003] In some scenarios, the base station only needs to ensure the transmission rate required by the user, and the interruption probability becomes a key indicator. Therefore, the traditional fixed-length training scheme aims to obtain complete channel information, which may generate unnecessary training redundancy.

[0004] Unlike traditional training methods, the alternating training scheme achieves a close integration of training and feedback. In this scheme, the base station can obtain feedback from the user end in real time, and the training process will stop immediately once the trained channel is stable enough to avoid system interruption. By evaluating the interruption risk of the trained channel in real time, this scheme can adaptively adjust the training length, thereby reducing the average training time.

[0005] However, the existing alternating training scheme mainly focuses on independent and identically distributed (IID) channels, and the system will immediately terminate the training if the condition of no interruption occurs. But in practical applications, the channel coefficients at adjacent moments often show a certain correlation. If this time correlation is taken into account, the system's training overhead performance may be further optimized. Summary of the invention

[0006] The purpose of the present invention is to provide a channel time correlation perception multi-user alternating training method, which further explores the time correlation of the channel on the basis of the existing alternating training scheme, so as to solve the technical problem of the sharp increase in training overhead caused by the sharp increase in the channel dimension of the large-scale MIMO system;

[0007] In order to achieve the above object, the present invention adopts the following technical solution: a channel time correlation perception multi-user alternating training method, comprising:

[0008] Step 1: System modeling phase: construct the system transmission model and model the time-dependent channel, then define and mathematically represent the various parameters involved, and initialize the various parameters involved;

[0009] Step 2: Antenna selection and alternating training phase: The base station in the system transmission model guides the antenna training sequence at the current moment according to the conditional mean matrix and the conditional variance matrix, and sends a pilot signal from the transmitting antenna on the selected base station side for the user to receive. Then the user estimates the channel state information according to the pilot signal sent by the base station and feeds it back to the base station. At the same time, the base station determines whether the trained antenna meets the non-interruption condition and decides whether to interrupt the training process.

[0010] Step 3: Data transmission phase: The base station performs beamforming and data transmission based on the feedback channel state information;

[0011] Step 4: Conditional distribution matrix update phase: Update the conditional mean matrix and conditional variance matrix according to the trained antenna set and channel state information under the time-correlated channel model.

[0012] Furthermore, the step 1 specifically includes:

[0013] Step 11: Based on the downlink of a single-cell multi-user massive MIMO system, M transmitting antennas are configured at the base station, and K users are set. At the same time, a single receiving antenna is configured for each user to obtain the system transmission model, where the expression of the system transmission model is:

[0014]

[0015] In the formula, y is the user received signal; is the normalized beamforming vector matrix, ||W|| F =1; s is the transmission symbol, and the average energy is normalized; P is the transmission power, and the value range is [0,3,6,5,10,15]; n is the normalized noise, obeying is the complete channel matrix, represents the channel coefficient between the base station antenna and the kth user, h K,M represents the channel coefficient between the Mth transmitting antenna of the base station and the Kth user;

[0016] Step 12: Use a first-order autoregressive model to model the time-varying channel to obtain a time-dependent channel model. The expression of the time-dependent channel model is:

[0017]

[0018] In the formula, H t is the complete channel matrix at time t, is the uncorrelated complex Gaussian noise introduced at time t, and all elements of it obey ρ is the time correlation coefficient, ranging from [0.1, 0.3, 0.5, 0.7, 0.9], which is used to characterize the strength of the channel correlation at adjacent moments;

[0019] Step 13: Record the set of all trained antenna numbers as The channel state information obtained by the base station through user feedback is: Defining the Matrix Then we have:

[0020]

[0021] in, express The element in the kth row and ith column of ;

[0022] The base station then uses the channel state information Design the MRT precoding matrix, which is expressed as:

[0023]

[0024] Then the received signal to interference and noise ratio SINR of user k under the precoding matrix is ​​expressed as:

[0025]

[0026] in express The channel state information between the antenna and user k.

[0027] Furthermore, the step 2 specifically includes:

[0028] Step 21: The base station calculates the probability that all untrained antennas meet the SINR threshold of each user based on the conditional mean matrix and the conditional variance matrix, and selects antennas accordingly. The selected antenna number is recorded as n. ＊ ;

[0029] Step 22: The base station sends a pilot signal to the user to train the selected antenna. The user then estimates the channel state information of the currently trained antenna based on the pilot signal sent by the base station and feeds back the corresponding channel state information. To the base station;

[0030] Step 23: The base station updates the mth round of trained antenna set based on the antenna selected in step 21 And according to the corresponding channel state information in step 22 Calculate the received signal to interference and noise ratio SINR provided by the currently trained antenna and compare it with the target received signal to interference and noise ratio SINR. If the received signal to interference and noise ratio SINR of all users is greater than or equal to the target received signal to interference and noise ratio SINR, or all antennas have completed training, stop training. Otherwise, go to step 21.

[0031] Furthermore, in step 21, the next antenna is selected according to the probability that all untrained antennas meet the SINR threshold of each user, and the formula is:

[0032]

[0033] Where, Pr([SINR k ] m+1 ≥γ k )for:

[0034]

[0035] in,

[0036] μ k,n represents the mean of the channel distribution between user k and antenna n, σ k,n represents the standard deviation of the channel distribution between user k and antenna n, represents the variance of the channel distribution between user j and antenna n, is the receiving SINR threshold for user k.

[0037] Furthermore, the specific steps of step 3 are:

[0038] When the received signal to interference and noise ratios of all users are greater than or equal to the target received signal to interference and noise ratio SINR, the base station performs beamforming and data transmission according to the fed-back channel state information.

[0039] Furthermore, the specific steps of step 4 are:

[0040] Based on the time-correlated channel model, combined with the trained antenna set and channel state information For the conditional mean matrix M t and the conditional variance matrix Σ t Update, where the conditional mean matrix Μ t The update formula is:

[0041]

[0042] In the formula, h t,k,i represents the channel state information between user k and antenna i in time slot t, μ t,k,i represents the mean of the channel distribution between user k and antenna i in time slot t;

[0043] Conditional variance matrix Σ t The update formula is:

[0044]

[0045] In the formula, represents the variance of the channel distribution between user k and antenna i in time slot t.

[0046] Beneficial effect: The channel time-correlation-aware multi-user alternating training scheme proposed in the present invention further exploits the time correlation of the channel on the basis of the existing alternating training scheme, and uses the conditional mean matrix and the conditional variance matrix to guide the antenna training sequence, thereby further reducing the system training overhead without losing the system interruption performance. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 is a flow chart of an embodiment of the present invention;

[0048] Figure 2 Flowchart for antenna selection and alternating training phases;

[0049] Figure 3 When the number of users K = 10 and the transmission power P = [5, 10, 15] dB, the average training length comparison diagram of the existing alternating training scheme and the proposed channel time-correlated sensing alternating training scheme;

[0050] Figure 4 Comparison of the average training lengths of the existing alternating training scheme and the proposed channel time-correlation-aware alternating training scheme when the number of users K = 4 and the transmit power P = [0, 3, 6] dB. DETAILED DESCRIPTION

[0051] The present invention will be further explained below in conjunction with the accompanying drawings.

[0052] like Figure 1 As shown, the present invention provides a channel time correlation perception multi-user alternating training method, comprising:

[0053] Step 1: System modeling phase: Construct the system transmission model and model the time-dependent channel, then define and mathematically represent the various parameters involved, and initialize the various parameters involved.

[0054] Step 2: Antenna selection and alternating training phase: The base station in the system transmission model guides the antenna training sequence at the current moment according to the conditional mean matrix and the conditional variance matrix, and sends a pilot to the user to train the antenna. Then the user estimates the channel state information according to the pilot sent by the base station and feeds it back to the base station. At the same time, the base station determines whether the system meets the non-interruption conditions and decides whether to interrupt the training process.

[0055] Step 3: Data transmission phase: The base station performs beamforming and data transmission based on the fed-back channel state information.

[0056] Step 4: Conditional distribution matrix update phase: Update the conditional mean matrix and conditional variance matrix according to the trained antenna set and channel state information under the time-correlated channel model.

[0057] In step 1, the specific steps include:

[0058] Step 11: Based on the downlink of a single-cell multi-user massive MIMO system, M transmitting antennas are configured at the base station, and K users are set. At the same time, a single antenna is configured for each user to obtain the system transmission model, where the expression of the system transmission model is:

[0059]

[0060] In the formula, y is the user received signal; is the normalized beamforming vector matrix, ||W|| F =1; s is the transmission symbol, and the average energy is normalized; P is the transmission power, and the value range is [0,3,6,5,10,15]; n is the normalized noise, obeying is the complete channel matrix, represents the channel coefficient between the base station antenna and the kth user, h K,M Represents the channel coefficient between the Mth transmitting antenna of the base station and the Kth user.

[0061] Step 12: Use a first-order autoregressive model to model the time-varying channel to obtain a time-dependent channel model. The expression of the time-dependent channel model is:

[0062]

[0063] In the formula, H t is the complete channel matrix at time t, is the uncorrelated complex Gaussian noise introduced at time t, and all elements of it obey ρ is the time correlation coefficient, which ranges from [0.1, 0.3, 0.5, 0.7, 0.9] and is used to characterize the strength of channel correlation at adjacent moments.

[0064] Step 13: Record the set of all trained antenna numbers as The channel state information obtained by the base station through user feedback is: Defining the Matrix Then we have:

[0065]

[0066] in, express The element in the kth row and ith column of ;

[0067] The base station then uses the channel state information Design the MRT precoding matrix, which is expressed as:

[0068]

[0069] Then the received signal to interference and noise ratio SINR of user k under the precoding matrix is ​​expressed as:

[0070]

[0071] in express The channel state information between the antenna and user k.

[0072] like Figure 2 As shown, in step 2, the base station transmitter first calculates the probability that all untrained antennas meet the SINR threshold of each user according to the conditional mean matrix and the conditional variance matrix, and selects an antenna accordingly. The formula for selecting the next antenna according to the probability that all untrained antennas meet the SINR threshold of each user is:

[0073]

[0074] Where, Pr([SINR k ] m+1 ≥γ k )for:

[0075]

[0076] in, μ k,n represents the mean of the channel distribution between user k and antenna n, σ k,n represents the standard deviation of the channel distribution between user k and antenna n, represents the variance of the channel distribution between user j and antenna n, is the receiving SINR threshold for user k.

[0077] The base station then sends a pilot signal to the user to select antenna n ＊ The user estimates the channel state information of the currently trained antenna based on the pilot signal sent by the base station and feeds back the corresponding state information. to the base station.

[0078] Then, the base station feeds back the corresponding status information Update the trained antenna set And calculate the received signal to interference and noise ratio SINR provided by the currently trained antenna, and compare it with the target received signal to interference and noise ratio SINR. If the received signal to interference and noise ratio of all users is greater than or equal to the target received signal to interference and noise ratio SINR, or all antennas have completed training, stop training and complete antenna training. Otherwise, return to the antenna selection step and repeat the above operation.

[0079] In this embodiment, during the first training, since there is no historical information, the conditional mean matrix M is initialized. t =0 K×M , the conditional variance matrix Σ t =1 K×M .

[0080] In step 3, after completing antenna training, the base station performs downlink beamforming and data transmission.

[0081] In step 4, based on the time-dependent channel model, combined with the trained antenna set and channel state information For the conditional mean matrix M t and the conditional variance matrix Σ t Update, where the conditional mean matrix Μ t The update formula is:

[0082]

[0083] In the formula, h t,k,i represents the channel state information between user k and antenna i in time slot t, μ t,k,i represents the mean of the channel distribution between user k and antenna i in time slot t;

[0084] Conditional variance matrix Σ t The update formula is:

[0085]

[0086] In the formula, represents the variance of the channel distribution between user k and antenna i in time slot t.

[0087] like Figure 3 As shown, the number of base station antennas M = 64, the number of users K = 10, and the number of users K = 10. th =2bps / Hz, the proposed channel time correlation perception multi-user alternating training scheme is simulated and verified with the basic alternating training scheme, that is, the training scheme in sequence according to the antenna order. The value range of the time correlation coefficient ρ is: [0.1, 0.3, 0.5, 0.7, 0.9]. Compared with the basic scheme, the proposed scheme can effectively mine the time correlation of the channel, reduce the training overhead, and have a more obvious performance gain.

[0088] like Figure 4 As shown, the number of base station antennas M = 32, the number of users K = 4, and the number of users K = 4. th =2bps / Hz, the proposed channel time-correlation-aware multi-user alternating training scheme is simulated and verified with the basic alternating training scheme, that is, the training scheme is performed in sequence according to the antenna order. The value range of the time correlation coefficient ρ is: [0.1, 0.3, 0.5, 0.7, 0.9]. Compared with the basic scheme, when the number of users is small, the proposed scheme still has considerable performance gain, which shows that the approximation of Chebyshev's law of large numbers is very reasonable.

[0089] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A channel time-correlation perception multi-user alternating training method, characterized in that: include: Step 1: System modeling phase: construct the system transmission model and model the time-dependent channel, then define and mathematically represent the various parameters involved, and initialize the various parameters involved; Step 2: Antenna selection and alternating training phase: The base station in the system transmission model guides the antenna training sequence at the current moment according to the conditional mean matrix and the conditional variance matrix, and sends a pilot signal from the transmitting antenna on the selected base station side for the user to receive. Then the user estimates the channel state information according to the pilot signal sent by the base station and feeds it back to the base station. At the same time, the base station determines whether the trained antenna meets the non-interruption condition and decides whether to interrupt the training process. Step 3: Data transmission phase: The base station performs beamforming and data transmission based on the feedback channel state information; Step 4: Conditional distribution matrix update phase: Update the conditional mean matrix and conditional variance matrix according to the trained antenna set and channel state information under the time-correlated channel model.

2. The channel time correlation perception multi-user alternating training method according to claim 1, characterized in that: The step 1 specifically includes: Step 11: Based on the downlink of a single-cell multi-user massive MIMO system, M transmitting antennas are configured at the base station, and K users are set. At the same time, a single receiving antenna is configured for each user to obtain the system transmission model, where the expression of the system transmission model is: In the formula, y is the user received signal; is the normalized beamforming vector matrix, ||W|| F =1; s is the transmission symbol, and the average energy is normalized; P is the transmission power, and the value range is [0,3,6,5,10,15]; n is the normalized noise, obeying is the complete channel matrix, represents the channel coefficient between the base station antenna and the kth user, h K,M represents the channel coefficient between the Mth transmitting antenna of the base station and the Kth user; Step 12: Use a first-order autoregressive model to model the time-varying channel to obtain a time-dependent channel model. The expression of the time-dependent channel model is: In the formula, H t is the complete channel matrix at time t, is the uncorrelated complex Gaussian noise introduced at time t, and all elements of it obey ρ is the time correlation coefficient, ranging from [0.1, 0.3, 0.5, 0.7, 0.9], which is used to characterize the strength of the channel correlation at adjacent moments; Step 13: Record the set of all trained antenna numbers as The channel state information obtained by the base station through user feedback is: Defining the Matrix Then we have: in, express The element in the kth row and ith column of ; The base station then uses the channel state information Design the MRT precoding matrix, which is expressed as: Then the received signal to interference and noise ratio SINR of user k under the precoding matrix is ​​expressed as: in express The channel state information between the antenna and user k.

3. The channel time correlation perception multi-user alternating training method according to claim 1, characterized in that: The step 2 specifically includes: Step 21: The base station calculates the probability that all untrained antennas meet the SINR threshold of each user based on the conditional mean matrix and the conditional variance matrix, and selects antennas accordingly. The selected antenna number is recorded as n. ＊ ; Step 22: The base station sends a pilot signal to the user to train the selected antenna. Then the user estimates the channel state information of the currently trained antenna based on the pilot signal sent by the base station and feeds back the corresponding channel state information [H t ] 1:K,n＊ To the base station; Step 23: The base station updates the mth round of trained antenna set based on the antenna selected in step 21 And according to the corresponding channel state information [H t ] 1:K,n＊ Calculate the received signal to interference and noise ratio SINR provided by the currently trained antenna and compare it with the target received signal to interference and noise ratio SINR. If the received signal to interference and noise ratio SINR of all users is greater than or equal to the target received signal to interference and noise ratio SINR, or all antennas have completed training, stop training. Otherwise, go to step 21.

4. The channel time correlation perception multi-user alternating training method according to claim 3, characterized in that: In step 21, the next antenna is selected according to the probability that all untrained antennas meet the SINR threshold of each user, and the formula is: In the formula, Pr([SINR k ] m+1 ≥γ k )for: in, μ k,n represents the mean of the channel distribution between user k and antenna n, σ k,n represents the standard deviation of the channel distribution between user k and antenna n, represents the variance of the channel distribution between user j and antenna n, is the receiving SINR threshold for user k.

5. The channel time correlation perception multi-user alternating training method according to claim 3, characterized in that: The specific steps of step 3 are: When the received signal to interference and noise ratios of all users are greater than or equal to the target received signal to interference and noise ratio SINR, the base station performs beamforming and data transmission according to the fed-back channel state information.

6. The channel time correlation perception multi-user alternating training method according to claim 2, characterized in that: The specific steps of step 4 are: Based on the time-correlated channel model, combined with the trained antenna set and channel state information For the conditional mean matrix M t and the conditional variance matrix Σ t Update, where the conditional mean matrix Μ t The update formula is: In the formula, h t,k,i represents the channel state information between user k and antenna i in time slot t, μ t,k,i represents the mean of the channel distribution between user k and antenna i in time slot t; Conditional variance matrix Σ t The update formula is: In the formula, represents the variance of the channel distribution between user k and antenna i in time slot t.

Images