A Beam Domain Channel Estimation Method for Spatial Non-Stationary Large-Scale MIMO

Abstract

This invention discloses a beam-domain channel estimation method for spatially non-stationary large-scale MIMO, comprising the following steps: First, a beam-domain channel model for spatially non-stationary large-scale MIMO is constructed using the visible region; then, based on the sparsity of the beam-domain channel and the influence of power leakage, the beam-domain channel estimation problem is transformed into a sparse channel reconstruction problem; next, based on the sparse structure of the cross-blocks of the beam-domain channel and the power ratio threshold, a sparsity-adaptive matched pursuit scheme based on the beam-domain structure is proposed; finally, simulation results verify that the proposed scheme has lower pilot overhead and higher accuracy and effectiveness compared with traditional schemes. The beam-domain channel estimation method for spatially non-stationary large-scale MIMO provided by this invention can be effectively applied to channel estimation with non-stationary characteristics, and has significant advantages in estimation accuracy and complexity.

Description

Technical Field

[0001] This invention relates to the field of channel estimation technology in wireless communication technology, and in particular to a beam domain channel estimation method for spatially non-stationary large-scale MIMO. Background Technology

[0002] To meet the ever-increasing demands for communication capacity, high speed, and full coverage from 5G and 6G mobile communication systems, massive MIMO plays a crucial role in improving spectrum and energy efficiency. Obtaining accurate channel state information is essential for effectively utilizing massive MIMO technology.

[0003] Channel measurement studies have shown that large-scale MIMO channels exhibit spatial non-stationarity and inherent sparsity. Therefore, channel estimation for large-scale MIMO systems can be viewed as a sparse signal reconstruction problem. To obtain channel state information for large-scale MIMO systems with low pilot overhead, traditional channel estimation schemes based on compressed sensing theory mostly treat large-scale MIMO channels as spatially stationary channels with a common sparse structure. However, in reality, spatially non-stationary channels do not exhibit a conventional sparse block structure. Therefore, simply using a common sparse structure to divide large-scale MIMO channels into conventional sparse block structures inevitably leads to poor estimation performance. On the other hand, when performing beam domain channel estimation, determining the beam dominance term forming the beam domain channel matrix solely based on power leakage, while ignoring the impact of channel spatial non-stationarity on power leakage, causes the channel estimation to deviate from the actual channel model, resulting in distorted estimation performance. To more accurately perform beam domain channel estimation for spatially non-stationary large-scale MIMO, a new channel estimation method needs to be proposed, and its correctness needs to be verified. Summary of the Invention

[0004] The technical problem to be solved by this invention is to propose a beam domain channel estimation method for spatially nonstationary large-scale MIMO that takes into account more realistic beam domain channels that consider spatial nonstationarity and power leakage, so as to further improve the channel estimation performance with lower algorithm complexity.

[0005] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows:

[0006] A beam domain channel estimation method for spatially nonstationary large-scale MIMO, characterized by the following steps:

[0007] S1. Construct a beam domain channel model for a spatially non-stationary large-scale MIMO system;

[0008] S2. Obtain the sparse beam structure based on the beam domain channel model, obtain the power ratio threshold based on the power leakage effect of the beam domain channel, and transform the beam domain channel estimation problem into a sparse channel reconstruction problem.

[0009] S3. Based on the beam sparse structure, obtain the dominant beam support, refine the dominant beam support according to the power ratio threshold, and use a sparsity adaptive matched pursuit scheme based on the beam domain structure to obtain the beam support set. Based on the beam support set, reconstruct the beam domain channel vector of a single user in sequence to obtain the estimated channel matrix.

[0010] Furthermore, the specific steps of step S1 are as follows:

[0011] Step S101: In the spatially non-stationary massive MIMO system, the base station is equipped with P=P h ×P v A uniform planar antenna array, P h and P v The horizontal and vertical dimensions of the uniform planar array are respectively. The base station serves U single-antenna users. All scattering clusters in the massive MIMO channel are divided into fully visible scattering clusters and partially visible scattering clusters. Each scattering cluster has a corresponding visible area. The visible area of ​​the fully visible scattering cluster is the entire uniform planar antenna array, and the visible area of ​​the partially visible scattering cluster is part of the antenna array. The ratio of the partially visible scattering clusters to the total scattering clusters is ρ.

[0012] Step S102: Construct a geometric random channel model: The antenna domain channel matrix of the u-th user is represented as follows:

[0013]

[0014] Where N1 and N2 are fully visible scattering clusters and partially visible scattering clusters, respectively, and M... n It is the total number of rays in the scattering cluster, f represents the carrier frequency, and β n,m τ n,m and Φ n,m It represents the initial path gain, time delay, and phase of the m-th ray in the n-th scattering cluster. and The antenna array steering matrices for the visible areas of fully visible scattering clusters and partially visible scattering clusters, respectively, satisfy the following conditions: ⊙ represents the Hadamard product, ξ n,m It is a matrix consisting only of 0s and 1s;

[0015] Step S103: Construct the beam domain channel model for a spatial non-stationary large-scale MIMO system: Convert the antenna domain channel matrix into a beam domain channel matrix through two-dimensional DFT processing.

[0016]

[0017] in{·} * It is a complex conjugate operation, {·} T It is the transpose operation, F el and F az These are the elevation beamforming matrix and the azimuth beamforming matrix, respectively.

[0018] Furthermore, the beam sparse structure is a beam crossblock structure, and the impact of power leakage includes two cases:

[0019] When the ratio ρ of the partially visible scattering clusters to the total scattering clusters is 0, imperfect beam sampling causes power leakage.

[0020] When the ratio ρ of the partially visible scattering clusters to the total scattering clusters is not 0, the spatial resolution is low due to the partial visibility of the non-stationary channel's visible area, inevitably resulting in power leakage.

[0021] Furthermore, the sparse channel reconstruction problem is described as follows:

[0022] In a spatially nonstationary massive MIMO system, the base station repeatedly sends orthogonal pilot sequences to U users Q times. The pilot matrix is ​​obtained from these orthogonal pilot sequences sent by the base station to the users. The channel experiences the same fading during the K = U × Q time slot. The base station employs an analog precoder. When transmitting the q-th pilot sequence, the received signal of the u-th user is represented as:

[0023]

[0024] in Represents the Kronecker product. It is an additive white Gaussian noise vector, σ 2 Representing the noise power, after Q repetitions of the pilot sequence, the received signal matrix of the u-th user is:

[0025]

[0026] in For the measurement matrix, This is the noise matrix.

[0027] Furthermore, the sparse adaptive matched pursuit scheme based on beam domain structure is as follows:

[0028] Input: Received signal y u Measurement matrix Φ, power ratio threshold μ, step size s;

[0029] Output: Estimated beam domain channel matrix:

[0030] (a) Initialization: Initial residual vector r0 = y u Beam support set The number of iterations k = 1, and the step size s = 1;

[0031] (b) Based on the residual vector r of the (k-1)th iteration k-1 And the p-th column of the measurement matrix Φ p Find the column most relevant to the residual vector. To achieve initial dominant beam support;

[0032] (c) Lock S according to the beam crossblock structure k The dominant beam supports at the top, bottom, left, and right are analyzed, and the power ratio of each dominant beam support to all dominant beam supports is calculated sequentially.

[0033] (d) Compare the power ratios The dominant beam support set is refined and updated based on the power ratio threshold μ.

[0034] (e) Let C k =Ω s ∪S k By merging Ω s and S k Obtain the beam index set;

[0035] (f) Let pass Obtain the spatially non-stationary massive MIMO channel H B,u The least squares estimate, where F is the final set of beam indices for a single iteration, card(C) k ) represents C k The number of elements, Φ F This represents the corresponding column of the measurement matrix obtained from F;

[0036] (g) Update residuals:

[0037] (h) If the residual satisfies r F r k-1 If the update step s = s + 1, return to step (b) and continue iterating; if the residual satisfies Where SNR is the signal-to-noise ratio, let Ω s =F,r k =r F If neither of the above conditions is met, then Ω terminates the iteration and proceeds to step (i); if neither of the above conditions is met, then Ω s =F,r k =rF , k = k + 1; when k ∈ K, stop the iteration and proceed to step (i);

[0038] (i) via Obtain an estimate of the beam domain channel.

[0039] Furthermore, the measurement matrix Φ is a Bernoulli random matrix, and the elements of Φ are from the set Selected randomly with equal probability, satisfying the mutual interference of columns. Smaller requirements, Φ i and Φ j To measure different columns of matrix Φ.

[0040] The beneficial effect of this invention is that, considering the effects of spatial nonstationarity and power leakage, it provides a beam domain channel estimation method for spatially nonstationary large-scale MIMO, which is aimed at more realistic beam domain channels, so as to perform more effective channel estimation for large-scale MIMO systems with lower algorithm complexity. Attached Figure Description

[0041] Figure 1 Flowchart for constructing a spatial nonstationary large-scale MIMO beam domain channel model for this invention;

[0042] Figure 2 This is a structural diagram of a large-scale MIMO system based on a uniform planar antenna array.

[0043] Figure 3 This is a schematic diagram of the sparse structure of the large-scale MIMO beam domain channel of the present invention;

[0044] Figure 4 This is a comparison of the NMSE performance of different channel estimation schemes for a spatially nonstationary large-scale MIMO system in an embodiment of the present invention.

[0045] Figure 5 This is a comparison of the NMSE performance of different channel estimation schemes for a spatially stationary large-scale MIMO system in an embodiment of the present invention.

[0046] Figure 6 The figure shows the NMSE performance comparison results of the sparse adaptive matched tracking scheme based on beam domain structure under different numbers of antennas in the embodiments of the present invention. Detailed Implementation

[0047] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0048] Example 1:

[0049] This embodiment provides a beam domain channel estimation method for spatially non-stationary large-scale MIMO, specifically including the following steps:

[0050] S1. Construct a beam domain channel model for a spatially non-stationary large-scale MIMO system;

[0051] S2. Obtain the sparse beam structure based on the beam domain channel model, obtain the power ratio threshold based on the power leakage effect of the beam domain channel, and transform the beam domain channel estimation problem into a sparse channel reconstruction problem.

[0052] S3. Based on the beam sparse structure, obtain the dominant beam support, refine the dominant beam support according to the power ratio threshold, and use a sparsity adaptive matched pursuit scheme based on the beam domain structure to obtain the beam support set. Based on the beam support set, reconstruct the beam domain channel vector of a single user in sequence to obtain the estimated channel matrix.

[0053] Specifically, in this embodiment, reference Figure 1 The specific steps of step S1 in this invention are as follows:

[0054] Step S101: In the spatially non-stationary massive MIMO system, the base station is equipped with P=P h ×P v A uniform planar antenna array, P h and P v The horizontal and vertical dimensions of the uniform planar array are respectively. The base station serves U single-antenna users. All scattering clusters in the massive MIMO channel are divided into fully visible scattering clusters and partially visible scattering clusters. Each scattering cluster has a corresponding visible area. The visible area of ​​the fully visible scattering cluster is the entire uniform planar antenna array, and the visible area of ​​the partially visible scattering cluster is part of the antenna array. The ratio of the partially visible scattering clusters to the total scattering clusters is ρ.

[0055] Step S102: Construct a geometric random channel model: The antenna domain channel matrix of the u-th user is represented as follows:

[0056]

[0057] Where N1 and N2 are fully visible scattering clusters and partially visible scattering clusters, respectively, and M... n It is the total number of rays in the scattering cluster, f represents the carrier frequency, and β n,m τ n,m and Φ n,m It represents the initial path gain, time delay, and phase of the m-th ray in the n-th scattering cluster. and The antenna array steering matrices for the visible areas of fully visible scattering clusters and partially visible scattering clusters, respectively, satisfy the following conditions: ⊙ represents the Hadamard product, ξ n,m It is a matrix consisting only of 0s and 1s;

[0058] Step S103: Construct the beam domain channel model for a spatial non-stationary large-scale MIMO system: Convert the antenna domain channel matrix into a beam domain channel matrix through two-dimensional DFT processing.

[0059]

[0060] in{·} * It is a complex conjugate operation, {·} T It is the transpose operation, F el and F az These are the elevation beamforming matrix and the azimuth beamforming matrix, respectively.

[0061] In this embodiment, as Figure 3 As shown, the beam sparse structure is a beam crossblock structure, and the power leakage effect includes two cases:

[0062] (1) When the ratio ρ of the partially visible scattering clusters to the total scattering clusters is 0, imperfect beam sampling causes power leakage.

[0063] (2) When the ratio ρ of the partially visible scattering clusters to the total scattering clusters is not 0, the spatial resolution is low due to the partial visibility of the non-stationary channel's visible area, inevitably resulting in power leakage.

[0064] Specifically, in this embodiment, the sparse channel reconstruction problem is described as follows:

[0065] In a spatially nonstationary massive MIMO system, the base station repeatedly sends orthogonal pilot sequences to U users Q times. The pilot matrix is ​​obtained from these orthogonal pilot sequences sent by the base station to the users. The channel experiences the same fading during the K = U × Q time slot. The base station employs an analog precoder. When transmitting the q-th pilot sequence, the received signal of the u-th user is represented as:

[0066]

[0067] in Represents the Kronecker product. It is an additive white Gaussian noise vector, σ 2 Representing the noise power, after Q repetitions of the pilot sequence, the received signal matrix of the u-th user is:

[0068]

[0069] in For the measurement matrix, This is the noise matrix.

[0070] Specifically, the sparse adaptive matched pursuit scheme based on beam domain structure is as follows:

[0071] Input: Received signal y u Measurement matrix Φ, power ratio threshold μ, step size s;

[0072] Output: Estimated beam domain channel matrix:

[0073] (a) Initialization: Initial residual vector r0 = y u Beam support set The number of iterations k = 1, and the step size s = 1;

[0074] (b) Based on the residual vector r of the (k-1)th iteration k-1 And the p-th column of the measurement matrix Φ p Find the column most relevant to the residual vector. To achieve initial dominant beam support;

[0075] (c) Lock S according to the beam crossblock structure k The dominant beam supports at the top, bottom, left, and right are analyzed, and the power ratio of each dominant beam support to all dominant beam supports is calculated sequentially.

[0076] (d) Compare the power ratios The dominant beam support set is refined and updated based on the power ratio threshold μ.

[0077] (e) Let C k =Ω s ∪S k By merging Ω s and S k Obtain the beam index set;

[0078] (f) Let pass Obtain the spatially non-stationary massive MIMO channel HB,u The least squares estimate, where F is the final set of beam indices for a single iteration, card(C) k ) represents C k The number of elements, Φ F This represents the corresponding column of the measurement matrix obtained from F;

[0079] (g) Update residuals:

[0080] (h) If the residual satisfies r F r k-1 If the update step s = s + 1, return to step (b) and continue iterating; if the residual satisfies Where SNR is the signal-to-noise ratio, let Ω s =F,r k =r F If neither of the above conditions is met, then Ω terminates the iteration and proceeds to step (i); if neither of the above conditions is met, then Ω s =F,r k =r F , k = k + 1; when k ∈ K, stop the iteration and proceed to step (i);

[0081] (i) via Obtain an estimate of the beam domain channel.

[0082] Specifically, the measurement matrix Φ is a Bernoulli random matrix, and the elements of Φ are from the set Selected randomly with equal probability, satisfying the mutual interference of columns. Smaller requirements, Φ i and Φ j To measure different columns of matrix Φ.

[0083] To verify the beam domain channel estimation method for spatially nonstationary large-scale MIMO provided in this embodiment, and especially to verify the performance of the sparse adaptive matched pursuit scheme based on beam domain structure, normalized mean square error is used. To measure the estimation accuracy of the estimation scheme, when constructing the spatial non-stationary large-scale MIMO beam domain channel model, the carrier frequency is set to 11 GHz, and other channel parameters are set with reference to the urban microcell communication scenario in 3GPP.

[0084] Figure 4 This indicates that under different signal-to-noise ratio conditions, when P h =32, P vThe NMSE performance of different channel estimation schemes for spatially non-stationary large-scale MIMO systems is compared when K=32, K=256, and ρ=0.45. The Oracle LS-based algorithm, block orthogonal matched pursuit algorithm, and adaptive support detection algorithm are used for comparison. For the Oracle LS-based algorithm, it assumes that the user knows the total sparsity of the channel and can directly select the dominant beam support based on the sparsity; this is an ideal estimation and can be considered as the upper bound of the estimation performance. It can be seen that compared with the block orthogonal matched pursuit algorithm and the adaptive support detection algorithm, the sparsity-based adaptive matched pursuit scheme based on beam domain structure proposed in this invention can significantly improve the NMSE estimation accuracy, especially in the high signal-to-noise ratio region. Furthermore, since the number of pilots K is much smaller than the dimension P of the beam domain channel, this scheme has lower pilot overhead.

[0085] Figure 5 This indicates that under different signal-to-noise ratio conditions, when P h =32, P v The performance comparison of NMSE of different channel estimation schemes for spatially stationary large-scale MIMO systems at K=32, K=256, and ρ=0 is used to test the beam domain channel estimation performance of the proposed sparse adaptive matched pursuit scheme based on beam domain structure in spatially stationary large-scale MIMO. The estimation performance of spatially stationary channels is generally better than that of spatially non-stationary channels, but this comes at the cost of sacrificing channel accuracy. In this case, the proposed sparse adaptive matched pursuit scheme based on beam domain structure outperforms the block orthogonal matched pursuit algorithm and the adaptive support detection algorithm in NMSE estimation accuracy, especially in the high SNR region. Furthermore, the performance of the proposed sparse adaptive matched pursuit scheme based on beam domain structure is similar to that of the adaptive support detection algorithm in the low SNR region.

[0086] Figure 6 The results show the NMSE performance comparison of the sparse adaptive matched pursuit scheme based on beam domain structure under different antenna numbers, where the antenna numbers P are 32×32, 48×48, and 64×6, and the number of pilots K = P / 4. The results indicate that in both low and high signal-to-noise ratio (SNR) regions, the NMSE performance improves with increasing antenna number. Therefore, the sparse adaptive matched pursuit scheme based on beam domain structure proposed in this invention can accurately estimate the channel of spatially non-stationary large-scale MIMO.

[0087] This invention proposes a beam-domain channel estimation method for spatially non-stationary large-scale MIMO. Compared with existing channel estimation methods, the proposed method considers the non-stationarity and power leakage of large-scale MIMO channels. The results are compared with those of different estimation schemes, verifying the accuracy and effectiveness of the proposed sparse adaptive matched pursuit scheme based on beam-domain structure compared to traditional estimation schemes. The beam-domain channel estimation method for spatially non-stationary large-scale MIMO provided by this invention can be effectively applied to channel estimation with non-stationary characteristics, exhibiting significant advantages in estimation accuracy and complexity.

[0088] Any aspects of this invention not described in detail are well-known to those skilled in the art.

[0089] The preferred embodiments of the present invention have been described in detail above. It should be understood that those skilled in the art can make numerous modifications and variations based on the concept of the present invention without creative effort. Therefore, all technical solutions that can be obtained by those skilled in the art based on the concept of the present invention through logical analysis, reasoning, or limited experimentation on the basis of existing technology should be within the scope of protection defined by the claims.

Claims

1. A beam domain channel estimation method for spatially nonstationary large-scale MIMO, characterized in that, Includes the following steps: S1. Construct a beam domain channel model for a spatially nonstationary large-scale MIMO system; S2. Obtain the sparse beam structure based on the beam domain channel model, obtain the power ratio threshold based on the power leakage effect of the beam domain channel, and transform the beam domain channel estimation problem into a sparse channel reconstruction problem. S3. Obtain the dominant beam support based on the beam sparse structure, refine the dominant beam support according to the power ratio threshold, obtain the beam support set using the sparsity adaptive matched pursuit method based on the beam domain structure, and reconstruct the individual user beam domain channel vectors sequentially according to the beam support set to obtain the estimated channel matrix. The specific steps of step S1 are as follows: Step S101: Construct a spatial non-stationary massive MIMO system, wherein each base station in the spatial non-stationary massive MIMO system is equipped with A uniform planar antenna array, and The horizontal and vertical dimensions of the uniform planar array are respectively the number of the base station serving U single-antenna users. In the spatially non-stationary massive MIMO system, all scattering clusters in the channel are divided into fully visible scattering clusters and partially visible scattering clusters. Each scattering cluster has a corresponding visible area. The visible area of ​​the fully visible scattering cluster is the entire uniform planar antenna array, and the visible area of ​​the partially visible scattering cluster is a portion of the antenna array. The ratio of the partially visible scattering clusters to the total scattering clusters is [ratio missing]. ; Step S102: Construct a geometric random channel model: the antenna domain channel matrix of the u-th user. Represented as in and These are fully visible scattering clusters and partially visible scattering clusters, respectively. It is the total number of rays in the scattering cluster. Indicates the carrier frequency. , and It represents the initial path gain, time delay, and phase of the m-th ray in the n-th scattering cluster. and The antenna array steering matrices for the visible areas of fully visible scattering clusters and partially visible scattering clusters, respectively, satisfy the following conditions: , This represents the Hadamard product. It is a matrix consisting only of 0s and 1s; Step S103: Construct the beam domain channel model of the spatial non-stationary large-scale MIMO system: This involves converting the antenna domain channel matrix... Converted into a beam-domain channel matrix through two-dimensional DFT processing: in It is a complex conjugate operation. It's a transpose operation. and These are the elevation beamforming matrix and the azimuth beamforming matrix, respectively.

2. The beam domain channel estimation method for spatially non-stationary large-scale MIMO according to claim 1, characterized in that, The beam sparse structure is a beam crossblock structure, and the impact of power leakage includes two cases: The ratio of visible scattering clusters to total scattering clusters in the aforementioned portion. When the value is 0, imperfect beam sampling causes power leakage; The ratio of visible scattering clusters to total scattering clusters in the aforementioned portion. When the value is not 0, the spatial resolution is low due to the partial visibility of the non-stationary channel's visible area, inevitably resulting in power leakage.

3. The beam domain channel estimation method for spatially non-stationary large-scale MIMO according to claim 1, characterized in that, The sparse channel reconstruction problem is described as follows: In a spatially nonstationary massive MIMO system, the base station repeatedly sends orthogonal pilot sequences to U users Q times. The pilot matrix is ​​obtained based on these orthogonal pilot sequences sent by the base station to the users. The channel... Experiencing the same fading during the time slot, the base station uses an analog precoder. When transmitting the q-th pilot sequence, the received signal of the u-th user... Represented as in , Represents the Kronecker product. It is an additive white Gaussian noise vector, and after Q repetitions of the pilot sequence, it is the received signal matrix of the u-th user. for in , For the measurement matrix, This is the noise matrix.

4. The beam domain channel estimation method for spatially non-stationary large-scale MIMO according to claim 3, characterized in that, The sparse adaptive matched pursuit method based on beam domain structure specifically includes the following steps: Input: Received signal Measurement matrix Power ratio threshold Step size s; Output: Estimated beam domain channel matrix: ; (a): Initial residual vector Beam support set Number of iterations Step length ; (b): According to the first The residual vector of the next iteration and measurement matrix column p Find the column most relevant to the residual vector. To obtain initial dominant beam support; (c): Locking based on beam crossblock structure The dominant beam supports at the top, bottom, left, and right are analyzed, and the power ratio of each dominant beam support to all dominant beam supports is calculated sequentially. ; (d): Compare the power ratios With the power ratio threshold To refine and update the dominant beam support set; (e): Let Through merger and Obtain beam index set ; (f): Let ,pass Obtaining spatially non-stationary massive MIMO channels The least squares estimate, where For the final set of beam indices from a single iteration, express The number of elements, Indicates according to The corresponding columns of the obtained measurement matrix; (g): Update residuals: ; (h): If the residual satisfies Then update step Return to step (b) and continue iterating; if the residual satisfies Where SNR is the signal-to-noise ratio, then let , If neither of the above conditions is met, then terminate the iteration and proceed to step (i); if neither of the above conditions is met, then... , , ;when Stop the iteration and proceed to step (i); (i): Through Obtain an estimate of the beam domain channel.

5. The beam domain channel estimation method for spatially non-stationary large-scale MIMO according to claim 4, characterized in that, The measurement matrix For Bernoulli random matrices, The elements from the set Selected randomly with equal probability, satisfying the mutual interference of columns. Smaller requirements and For measurement matrix Different columns.

Images