A Massive MIMO Channel Estimation Optimization Method

Abstract

The present invention discloses a large-scale MIMO channel estimation optimization method, comprising the following steps: (1) constructing an uplink unified system model for a large-scale MIMO communication system including a base station configured with M antennas and K users configured with single antennas; (2) constructing a channel model according to the large-scale fading and small-scale fading characteristics of the communication channel between the base station and each user; (3) using the least squares method to initialize and estimate the channel matrix to obtain an initial estimation value of the channel matrix; (4) using a cyclic iteration method to perform channel estimation and evaluate performance; using the obtained initial estimation value as the initialization of the loop, and cyclically iterating and calculating the channel matrix estimation value until convergence; the present invention improves the accuracy of channel estimation.

Description

Technical Field

[0001] The present invention relates to the technical field of communication engineering, and in particular to a large-scale MIMO channel estimation optimization method. Background Art

[0002] With the rapid development of 5G communication technology, the information society has an increasingly urgent demand for high-speed, low-latency and high-reliability wireless communications. As one of the core technologies of 5G and future wireless communication networks, Massive MIMO (Massive Multiple Input Multiple Output) has the potential to significantly improve the system's spectrum efficiency and energy efficiency. Massive MIMO deploys a large number of antennas at base stations, enabling the system to provide high-data-rate transmission services to multiple user devices at the same time. However, with the increase in the number of antennas and user devices, the channel estimation problem becomes extremely complex. How to efficiently and accurately perform channel estimation in complex channel environments has become a key challenge for massive MIMO technology.

[0003] In MIMO systems, the wireless channel between the base station and the user equipment is affected by complex environmental factors such as multipath propagation. In order to perform reliable transmission, the base station must accurately estimate the channel state information (CSI). The accuracy of CSI directly determines the performance of the MIMO system, including rate, coverage, energy efficiency, etc. In traditional small-scale MIMO systems, the number of pilot signals and the dimension of the channel matrix are relatively small, and the classic linear estimation methods such as the least squares (LS) and minimum mean square error (MMSE) can meet the channel estimation requirements to a certain extent. However, with the rapid growth in the number of antennas and user equipment, traditional channel estimation methods face huge challenges, especially in large-scale MIMO scenarios, where the exponential growth of channel dimensions leads to a significant increase in computational complexity, and pilot resources become very tight. In short, with the widespread application of large-scale MIMO technology, channel estimation optimization methods face a series of challenges such as computational complexity, pilot resource limitations, and multi-user interference. Summary of the invention

[0004] Purpose of the invention: The purpose of the present invention is to provide a large-scale MIMO channel estimation optimization method, which solves the problems existing in the background technology by improving the algorithm, adopting sparse signal processing technology, and iterative recursive method.

[0005] Technical solution: A large-scale MIMO channel estimation optimization method described in the present invention comprises the following steps:

[0006] (1) For a massive MIMO communication system consisting of a base station with M antennas and K users with single antennas, a unified uplink system model is constructed.

[0007] (2) Construct a channel model based on the large-scale fading and small-scale fading characteristics of the communication channel between the base station and each user;

[0008] (3) Use the least squares method to initialize and estimate the channel matrix to obtain the initial estimated value of the channel matrix;

[0009] (4) A loop iterative method is used to estimate the channel and evaluate the performance; the obtained initial estimate is used as the initialization of the loop, and the channel matrix estimate is calculated by loop iterative calculation until convergence.

[0010] Furthermore, in step (1), the uplink unified system model formula is as follows:

[0011] ;

[0012] in, is the signal matrix received by the base station, is the channel matrix, , represents the pilot sequence sent by the i-th user. The length of the pilot is τ, and the pilot sequence is designed to obey the Bernoulli distribution. is the additive white Gaussian noise matrix.

[0013] Furthermore, in step (2), the channel modeling formula between the kth user and the base station is as follows:

[0014] ;

[0015] in, is the large-scale fading coefficient, is the small-scale fading coefficient, and .

[0016] Furthermore, step (4) includes the following steps:

[0017] (41) Through linear transformation Perform decorrelation operation with X to remove the correlation between the pilot matrix and the estimated channel matrix and generate an intermediate estimate ;

[0018] (42) Yes Perform nonlinear processing and use divergence-free functions right Mapping, we get , combined with the channel The prior distribution is used to calculate the posterior probability distribution p, and then the expectation of p is taken as the output of this round of iteration ;

[0019] (43) recursively process the error terms estimated in step (41) and step (42), analyze the evolution of the algorithm, and estimate the minimum mean square error (MSE) to evaluate the performance of the algorithm;

[0020] (44) Convergence conditions Is it satisfied? If so, As the final output, and exit the loop; if not satisfied, then , and enter the next round of iterative calculation; among them, is an arbitrarily small positive number.

[0021] Furthermore, step (41) is specifically as follows:

[0022] ;

[0023] in, is the channel matrix value to be estimated, X is the pilot signal, is the output of the linear estimation, is the decorrelation matrix, that is, it needs to satisfy

[0024] ;

[0025] The specific construction method is as follows:

[0026] ;

[0027] in, .

[0028] Furthermore, in step (42), if there is no divergence function, then: Expressing hope; ;

[0029] The specific construction method is:

[0030] ;

[0031] in, is a soft threshold function, that is , is a non-negative threshold parameter, here taken as the noise variance ;

[0032] The final output of the algorithm in this round of iteration is:

[0033] .

[0034] Furthermore, in step (42), the posterior probability distribution is calculated as follows:

[0035] Set the channel of the active device Belongs to the modulation constellation set , that is, take a value in this set; let the prior probability distribution of the kth device be:

[0036] ;

[0037] in is the length of the pilot sequence, is the sparsity, The function is an impulse function;

[0038] According to the observation model:

[0039] ;

[0040] And the prior distribution, we get The approximate posterior distribution of is:

[0041] ;

[0042] in,

[0043] ;

[0044] ;

[0045] parameter is the posterior sparsity ratio, representing Based on the linear estimation of the observed When the sparsity , the user is considered active and has sent a pilot sequence.

[0046] Furthermore, in step (43), the error terms of the linear and nonlinear estimation modules are recursively processed as follows: Based on the original error term definition:

[0047] ;

[0048] ;

[0049] The recursive formula of the error term of the algorithm obtained by mathematical derivation is as follows:

[0050] ;

[0051] ;

[0052] in,

[0053] ;

[0054] Where I is the identity matrix and the superscript t represents the tth iteration.

[0055] An electronic device described in the present invention includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the computer program is loaded into the processor, any one of the large-scale MIMO channel estimation optimization methods described is implemented.

[0056] Beneficial effects: Compared with the prior art, the present invention has the following significant advantages: it improves the existing algorithm, utilizes the sparsity of the signal and orthogonal processing, and realizes efficient and accurate channel estimation in large-scale MIMO systems. In a complex environment with multiple antennas and multiple users, this method can effectively reduce the correlation between the pilot matrix and the channel matrix, improve the accuracy of channel estimation, and significantly improve the performance in scenarios with low signal-to-noise ratio and high sparsity. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1 It is a system model diagram of the present invention;

[0058] Figure 2 It is the algorithm flow chart of the present invention;

[0059] Figure 3 State evolution and MSE effect of algorithm simulation of the present invention. DETAILED DESCRIPTION

[0060] The technical solution of the present invention is further described below in conjunction with the accompanying drawings.

[0061] like Figure 1 As shown, an embodiment of the present invention provides a large-scale MIMO channel estimation optimization method, comprising the following steps:

[0062] S1 is aimed at Figure 1 The massive MIMO communication system shown in the figure includes a base station with M antennas and K users with single antennas, and constructs an uplink unified system model:

[0063] ;

[0064] in, is the signal matrix received by the base station, is the channel matrix, , represents the pilot sequence sent by the i-th user. The length of the pilot is τ, and the pilot sequence is designed to obey the Bernoulli distribution. is the additive white Gaussian noise matrix;

[0065] S2 then constructs specific channel models based on the large-scale fading and small-scale fading characteristics of the communication channel between the base station and each user. The channel model between the kth user and the base station is:

[0066] ;

[0067] in, is the large-scale fading coefficient, is the small-scale fading coefficient, and .

[0068] S3 is based on the above model, such as Figure 2 As shown, we first perform an initial estimation and use the least squares method to To make an initial estimate, the formula is as follows:

[0069] ;

[0070] And set the initial iteration parameters according to the estimated values ​​and noise characteristics.

[0071] S4 then enters the linear estimation module to perform decorrelation operation on the input signal and generate an intermediate estimate value, the formula is as follows:

[0072] ;

[0073] in, is the channel matrix value to be estimated, X is the pilot signal, is the output of the linear estimation, is the decorrelation matrix, that is, it needs to satisfy

[0074] ;

[0075] The specific construction method is as follows:

[0076] ;

[0077] in: .

[0078] The non-divergence function in the nonlinear estimation module (i.e., it needs to satisfy: Expressing expectation. . )

[0079] The specific construction method is:

[0080] ;

[0081] in is a soft threshold function, that is , is a non-negative threshold parameter, here taken as the noise variance .

[0082] The final output of the algorithm in this round of iteration is:

[0083] ;

[0084] The expected value of the posterior probability distribution is calculated under known prior estimates.

[0085] The posterior probability distribution is calculated as follows:

[0086] Assume the channel of the active device Belongs to the modulation constellation set , that is, take a value from this set.

[0087] Based on this, the prior probability distribution of the kth device is considered to be:

[0088] ;

[0089] in is the length of the pilot sequence, is the sparsity, The function is an impulse function.

[0090] Combined with observation model:

[0091] ;

[0092] And the prior distribution, we can get The approximate posterior distribution of is:

[0093] ;

[0094] in, ;

[0095] ;

[0096] parameter is the posterior sparsity ratio, which represents Based on the linear estimation of the observed When the sparsity , the user is considered active and has sent a pilot sequence. Each user has a certain probability of Send a pilot sequence, and another 1- The probability of not sending is that it is inactive.

[0097] The error term recursive processing method of the linear and nonlinear estimation modules is as follows:

[0098] According to the original error term definition

[0099] ;

[0100] ;

[0101] Through appropriate mathematical derivation, the recursive formula of the algorithm's error term can be obtained as follows:

[0102] ;

[0103] ;

[0104] in:

[0105] ;

[0106] I is the identity matrix, and the superscript t represents the tth iteration.

[0107] The minimum mean square error (MSE) of the algorithm is defined as follows:

[0108] ;

[0109] The performance of MSE is analyzed by coding simulation comparison.

[0110] Processing of inactive users:

[0111] For inactive users, the channel matrix is ​​directly set to zero, thereby reducing the computational complexity and unnecessary computational burden.

[0112] Finally, the state evolution process of algorithm simulation and the MSE effect are as follows Figure 3 As shown. The base station is equipped with 64 antennas, the number of antennas for each user is 1, there are 50 active users, each user sends a pilot signal of Bernoulli distribution, and the algorithm is applied to perform channel estimation based on the pilot signal and the signal received by the base station. The MSE of the optimization method proposed in the present invention is compared with that of the existing method (here, the MSE is processed as follows in order to intuitively compare the performance of the algorithm: , that is, the MSE unit in the simulation diagram is decibel), it is found that the MSE of the optimization method proposed in the present invention is much smaller than that of the existing method, indicating that the proposed optimization algorithm has higher accuracy in channel estimation and better performance.

[0113] An embodiment of the present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the computer program is loaded into the processor, any one of the large-scale MIMO channel estimation optimization methods described above is implemented.

Claims

1. A large-scale MIMO channel estimation optimization method, characterized in that: The following steps are involved: (1) For a massive MIMO communication system consisting of a base station with M antennas and K users with single antennas, a unified uplink system model is constructed. (2) Construct a channel model based on the large-scale fading and small-scale fading characteristics of the communication channel between the base station and each user; (3) Use the least squares method to initialize and estimate the channel matrix to obtain the initial estimated value of the channel matrix; (4) Using a loop iteration method to perform channel estimation and performance evaluation; using the obtained initial estimation value as the initialization of the loop, and looping and iterating to calculate the channel matrix estimation value until convergence; comprising the following steps: (41) Through linear transformation Perform decorrelation operation with X to remove the correlation between the pilot matrix and the estimated channel matrix and generate an intermediate estimate ; The details are as follows: ； in, is the channel matrix value to be estimated, X is the pilot, is the output of the linear estimation, is the decorrelation matrix, that is, it needs to satisfy ； The specific construction method is as follows: ； in, ; (42) Yes Perform nonlinear processing and use divergence-free functions right Mapping, we get , combined with the channel The prior distribution is used to calculate the posterior probability distribution p, and then the expectation of p is taken as the output of this round of iteration ; No divergence function satisfies: Expressing hope; ; The specific construction method is: ； in, is a soft threshold function, that is , is a non-negative threshold parameter, here taken as the noise variance ; The final output of the algorithm in this round of iteration is: ； The posterior probability distribution is calculated as follows: Set the channel of the active device Belongs to the modulation constellation set , that is, take a value in this set; let the prior probability distribution of the kth device be: ； in is the length of the pilot sequence, is the sparsity, The function is an impulse function; According to the observation model: ； And the prior distribution, we get The approximate posterior distribution of is: ； in, ； ； parameter is the posterior sparsity ratio, representing Based on the linear estimation of the observed When the sparsity When , the user is considered to be active and has sent a pilot sequence; (43) The error terms estimated in step (41) and step (42) are recursively processed, and the evolution process of the algorithm is analyzed to estimate the minimum mean square error (MSE) to evaluate the performance of the algorithm; the error term recursive processing method of the linear and nonlinear estimation modules is as follows: According to the original error term definition: ； ； The recursive formula of the error term of the algorithm obtained by mathematical derivation is as follows: ； ； in, ； Where I is the identity matrix, and the superscript t represents the tth iteration; (44) Convergence conditions Is it satisfied? If so, As the final output, and exit the loop; if not satisfied, then , and enter the next round of iterative calculation; among them, is an arbitrarily small positive number.

2. A method for optimizing large-scale MIMO channel estimation according to claim 1, characterized in that: In step (1), the uplink unified system model formula is as follows: ； in, is the signal matrix received by the base station, is the channel matrix, , represents the pilot sequence sent by the i-th user. The length of the pilot is τ, and the pilot sequence is designed to obey the Bernoulli distribution. is the additive white Gaussian noise matrix.

3. A method for optimizing large-scale MIMO channel estimation according to claim 2, characterized in that: In step (2), the channel modeling formula between the kth user and the base station is as follows: ； in, is the large-scale fading coefficient, is the small-scale fading coefficient, and .

4. The method for optimizing large-scale MIMO channel estimation according to claim 1, wherein: In step (43), the minimum mean square error MSE is defined as follows: ； in is the output matrix of the t-th iteration, is the actual channel matrix.

5. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the computer program is loaded into a processor, a large-scale MIMO channel estimation optimization method according to any one of claims 1 to 4 is implemented.

Images