Super-large-scale MIMO near-field channel estimation method based on planar array

By using the near-field spherical wave channel estimation method of planar arrays in super-large-scale MIMO systems, decoupling azimuth angle and pitch angle, combined with grid processing and algorithms, the accuracy problem of near-field channel estimation under base station plane array is solved. Efficient channel parameter estimation is achieved.

CN118945016AActive Publication Date: 2024-11-12SUN YAT SEN UNIV
View PDF 3 Cites 0 Cited by

Patent Information

Application Number
CN202410313716.0
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-03-19
Publication Date
2024-11-12
Estimated Expiration
2044-03-19

AI Technical Summary

Technical Problem

In super-large-scale MIMO systems, when the base station deploys a planar array and the user is in the near-field area, it is difficult to obtain accurate channel state information (CSI). The existing algorithms are mainly based on uniform linear arrays, and there is a lack of near-field channel estimation method for planar arrays .

Method used

A near-field spherical wave channel estimation method based on planar array is proposed. By defining a new plane coordinate system decoupling azimuth and pitch angle, using OMP algorithm and LS algorithm, combining angle and distance grid processing, the elevation angle is gradually estimated , distance and azimuth angle reduce calculation complexity and improve estimation accuracy.

Benefits of technology

High-precision channel estimation in super-large-scale MIMO systems is realized, which reduces the computational complexity and ensures the accuracy of channel estimation. It is suitable for the near-field channel model of super-large-scale MIMO systems.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118945016A_ABST
    Figure CN118945016A_ABST
Patent Text Reader

Abstract

The invention discloses a super-large-scale MIMO near-field channel estimation method based on a planar array. According to the method, the near-field spherical wave super-large-scale MIMO channel model under the planar array is established, and the azimuth angle and the pitch angle are decoupled by defining a new planar coordinate system, so that the model is simplified. According to the method, elevation angle and distance parameters are independent by means of the structural characteristics of received data, coarse estimation values of the elevation angle and distance parameters are obtained by dividing angle and distance grids and utilizing an OMP algorithm, and final estimation results of the elevation angle and the distance are obtained by further respectively and finely dividing the angle grids and the distance grids. Moreover, the estimated values of the elevation angle and the distance are used as known parameters to estimate an azimuth angle, and an LS algorithm is used to estimate a gain coefficient. According to the method, angle and distance parameters can be estimated by selecting a small amount of receiving data by using the simplified planar array near-field model, so that the calculation complexity is reduced, and the estimation precision is also ensured.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of communication technology, and more specifically, to a channel estimation method based on a planar array in a very large-scale MIMO system in a near-field scenario. Background Art

[0002] Massive Multiple Input Multiple Output (Massive MIMO) is an enabling technology for the current fifth generation (5G) mobile communication system, and plays a key role in improving spectrum efficiency. With the development of the sixth generation (6G) system, the future intelligent information society will increase the demand for high spectrum efficiency and high energy efficiency, which will require the number of antennas at the base station (BS) to increase further. Compared with massive MIMO, the number of antennas equipped at the BS of the ultra-large-scale MIMO (XL-MIMO) system can reach thousands of units, and the achievable spectrum efficiency can be increased several times. Therefore, XL-MIMO technology has broad development and application potential in the future 6G era.

[0003] In very large-scale MIMO systems, beamforming technology can bring huge spatial multiplexing gains, thereby improving spectrum efficiency, but it is heavily dependent on channel state information (CSI). In addition, to meet the large-scale connection of IoT devices, the system needs to perform a large number of channel accesses, which leads to huge consumption of channel estimation. Therefore, it is of great significance to study channel estimation in very large-scale MIMO systems.

[0004] Considering the huge number of base station antennas, uniform planar array (UPA) is a feasible compact antenna array deployment scheme in very large-scale MIMO systems. In addition, UPA can not only avoid beam squint (i.e., the negative spatial broadband effect that often occurs in uniform linear array (ULA) due to the huge array size), but also realize three-dimensional beamforming. Therefore, channel estimation under UPA configuration is of great value for very large-scale MIMO systems. At the same time, the number of antennas in very large-scale MIMO systems is huge, which leads to fundamental changes in the electromagnetic field structure. The radiation field of electromagnetic waves can be divided into two regions, namely the far field region and the near field region. The boundary dividing the near field and the far field is determined by the Rayleigh distance. When the array aperture increases to a large extent (tens to hundreds of wavelengths), the near field region in XL-MIMO systems can no longer be ignored. At this time, a spherical wave model that is closer to the real propagation mechanism should be used in channel estimation to avoid model mismatch. When the receiver is located in the near field region, the wavefront needs to be accurately simulated under the spherical wavefront assumption, which depends not only on the angle but also on the distance between the BS and the scattering point.

[0005] In recent years, research on ultra-large-scale MIMO channel estimation technology based on the near-field spherical wave model has just begun. In addition, existing algorithms, such as the near-field OMP algorithm, the reduced-rank MUSIC algorithm, and the deep learning algorithm, are all based on the premise that the base station is a uniform linear array, while there are very few studies on ultra-large-scale MIMO near-field channel estimation methods based on planar arrays. Summary of the invention

[0006] Aiming at the problem that it is difficult for an ultra-large-scale MIMO system to obtain accurate CSI when a base station deploys a planar array and considers that the user is in the near-field area, the present invention proposes a high-precision and feasible channel estimation method based on near-field spherical wave channel model parameter estimation.

[0007] The present invention establishes a near-field spherical wave ultra-large-scale MIMO channel model under a planar array, in which not only the angle parameters in the azimuth and elevation directions are considered, but also the distance parameters increased due to the near-field effect of the ultra-large-scale MIMO system are considered. The present invention decouples the azimuth and elevation angles by defining a new plane coordinate system, simplifies the planar array channel model, and based on this, proposes a near-field channel estimation method. The method uses the structural characteristics of the received data, and obtains the rough estimation values ​​of the elevation angle and distance parameters by dividing the angle and distance grids using the OMP algorithm, and uses the distance parameter and the gain coefficient as redundant parameters to refine the angle grid, and further obtains the final estimation result of the elevation angle, thereby obtaining the final estimation result of the distance. In addition, the estimated values ​​of the elevation angle and the distance are used as known parameters to estimate the azimuth angle, and the gain coefficient is estimated using the LS algorithm. The method uses the simplified near-field model of the planar array, selects a small amount of received data, and can estimate the angle and distance parameters, while reducing the calculation complexity and ensuring the estimation accuracy.

[0008] To achieve the above objectives, the technical solution of the present invention is as follows:

[0009] A method for near-field channel estimation of ultra-large-scale MIMO based on a planar array, characterized by comprising the following steps:

[0010] S1: Receive signals using a near-field uniform planar array;

[0011] S2: Take the first column of the received signal matrix, separate the elevation and azimuth parameters, and use this small portion of data to estimate the elevation and distance parameters;

[0012] S3: Grid the range of elevation angle and distance, construct a sparse reconstruction problem, and use the orthogonal matching pursuit algorithm to estimate the rough estimate of the elevation angle and distance parameters;

[0013] S4: fixing the distance value to the rough estimated value of the distance obtained in step S3, and performing a grid search on the elevation angle again to obtain a precise estimated value of the elevation angle;

[0014] S5: fixing the elevation angle to the precise elevation angle estimate value obtained in step S4, and performing a grid search on the distance again to obtain a precise distance estimate value;

[0015] S6: for the estimated elevation angle and distance parameters, determine whether the corresponding remaining residuals meet the iteration stop condition;

[0016] S7: Estimate azimuth parameters according to the estimated elevation angle and distance parameters;

[0017] S8: reconstruct the channel response according to the path gain coefficient;

[0018] S9: Output channel estimation result.

[0019] Preferably, the near-field uniform planar array in step S1 is specifically:

[0020] S1.1: The near-field uniform planar array is a two-dimensional array with M*N array elements and the distances between adjacent antennas in the x-direction and the y-direction are d x , d y , where the antenna numbers in the x-direction of the near-field uniform planar array increase from 0 to M-1, and the antenna numbers in the y-direction increase from 0 to N-1. Figure 2 As shown in the figure, the angle between the projection of the propagation path on the YOZ plane and the propagation path is defined as the elevation angle θ; the angle between the projection of the propagation path on the XOZ plane and the propagation path is defined as the azimuth angle The distance between the scatterer and the reference array element is r.

[0021] Preferably, in step S1, the near-field uniform planar array receives the signal, specifically:

[0022] S1.2: Based on the newly defined positional relationship between elevation angle and azimuth angle, the channel model is constructed as follows:

[0023]

[0024] In the formula, α l is the gain coefficient of the lth path, θ l , r l are respectively the elevation angle, azimuth angle and distance between the scatterer and the reference array element of the lth path, L is the multipath number, and L<<M×N. is the steering matrix, and the element at the (p,q) position is:

[0025]

[0026] In the formula, In the channel model, The first-order term in the exponential part is only related to the angle, and decouples the elevation angle and the azimuth angle, thereby simplifying the channel model. The second-order term reflects the impact of the ultra-large-scale MIMO near-field effect - the distance parameter is taken into account in the model.

[0027] S1.3: The base station receives the pilot signal as follows:

[0028]

[0029] Where P is the transmit power and N is Gaussian white noise.

[0030] Preferably, in step S2, the first column of data of the received signal matrix is ​​taken to estimate the elevation angle and distance parameters, specifically:

[0031] S2.1: Take the first column of the received signal matrix Y and get y0∈C M×1 ,as follows:

[0032]

[0033] In the formula, h represents the corresponding channel vector and n is the noise vector. The expression of h is:

[0034]

[0035] In the formula, a(θ l ,r l )∈C M×1 is the corresponding steering vector,

[0036]

[0037] From the above expression, we can see that, in addition to the gain coefficient, y0 only contains elevation and distance parameters, but no azimuth parameters. Independently of other parameters, a small portion of data is first used to estimate the elevation angle and distance, thereby reducing the amount of calculation of the method of the present invention.

[0038] Preferably, in step S3, the range of values ​​of the elevation angle and the distance is gridded, and the rough estimated values ​​of the elevation angle and the distance parameters are estimated using the orthogonal matching pursuit algorithm, specifically:

[0039] S3.1: The sine value of the angle is sampled uniformly in the range [-1,1]; the distance is sampled exponentially in the near field range (0,Z] (Z is the boundary between the far field and the near field - the Rayleigh distance, and Where D is the array aperture size and λ is the wavelength). Then M 2 The sampling grid Ψ of angle and distance sampling values c It can be expressed as:

[0040]

[0041] S3.2: In traditional far-field massive MIMO systems, when the number of paths is not comparable to the number of antennas, the channel is sparse in the angular domain and the paths can be extracted. This sparsity is partly due to the directionality of the array pattern, and the array pattern of ultra-large-scale MIMO in near-field conditions also has directionality and channel sparsity. Therefore, sparse signal recovery algorithms, such as the orthogonal matching pursuit (OMP) algorithm, can also be used to estimate the channel and obtain parameter values ​​in the channel.

[0042] The received signal vector y0 obtained in step S2 is approximately sparsely represented as:

[0043]

[0044] In the formula, is the corresponding sparse channel vector, is the perception matrix, which can be expressed as:

[0045] B co =[a(θ1,r1),…,a(θ1,r M ),…,a(θ M ,r1),…,a(θ M ,r M )]

[0046] In the formula, (θ m ,r m )∈Ψ c .

[0047] Therefore, the elevation angle and distance parameter estimation problem can be transformed into:

[0048]

[0049] In the formula, η is a very small constant. c A rough estimate of the elevation angle and distance can be obtained by searching within

[0050] Preferably, the fixed distance value in step S4 is the rough estimated value of the distance obtained in step S3, and the elevation angle is grid searched again to obtain the precise estimated value of the elevation angle, specifically:

[0051] S4.1: Fix the distance value to the rough estimate obtained in step S3 A rough estimate of the elevation angle The perception matrix W of the refined estimated angle is obtained again by sinusoidal uniform quantization within the nearby grid range θ ∈C M×M :

[0052]

[0053] In the formula,

[0054] S4.2: The refined estimation result of the elevation angle can be obtained by solving the following problems through the OMP algorithm:

[0055]

[0056] In the formula, and η θ are the corresponding channel vector and constant respectively.

[0057] Preferably, in step S5, the fixed elevation angle is taken as the elevation angle precision estimation value obtained in step S4, and the distance is grid searched again to obtain the distance precision estimation value, specifically:

[0058] S5.1: Fix the elevation angle to the estimated result in step S4 The rough distance estimate obtained in step S3 The perception matrix W of the refined estimated distance is obtained again by exponential sampling in the vicinity of the nearby grid range r ∈C M×M :

[0059]

[0060] In the formula,

[0061] S5.2: The OMP algorithm can be used to solve the following problems to obtain a refined distance estimate:

[0062]

[0063] In the formula, and η r are the corresponding channel vector and constant respectively.

[0064] Preferably, in step S6, for the estimated elevation angle and distance parameters, it is determined whether the corresponding remaining residual satisfies the iteration stop condition, specifically:

[0065] S6.1: After the l-1th iteration of OMP, l-1 sets of parameters and paths are estimated, and the remaining residual is:

[0066]

[0067] In the formula, the path gain

[0068] S6.2: Considering that the number of propagation paths is unknown in the channel estimation problem, this method uses the method of making the remaining residual as small as possible (when only noise is included) to stop the iteration process. If the estimation result meets the following iteration stop condition, execute step S7 to estimate the azimuth; if it does not meet the following condition, continue to estimate the new elevation angle and distance parameter group.

[0069]

[0070] In the formula, is the Fourier transform, P fa is the false alarm probability.

[0071] Preferably, in step S7, the azimuth parameter is estimated according to the estimated elevation angle and distance parameter, specifically:

[0072] S7.1: Vectorize the received matrix Y row by row as follows:

[0073]

[0074] In the formula, is the vectorized noise, is the steering vector after row vectorization, which is:

[0075]

[0076]

[0077] S7.2: Further, using the elevation angle and distance estimates obtained in steps S4 and S5, As a known condition, estimate the azimuth Specifically:

[0078] Angle In range The uniform quantization sampling method is used, and the elevation angle and distance precision estimation values ​​of each path are substituted respectively. Construct the perception matrix D l :

[0079]

[0080] Where Q is the angle The number of quantized samples.

[0081] Therefore, the vectorized received signal can be The sparse representation is:

[0082]

[0083] In the formula, is the sparse representation of vec(H).

[0084] S7.3: Construct the following sparse reconstruction problem and use the OMP algorithm to perform a sparse dictionary search to obtain the azimuth estimate of the lth path:

[0085]

[0086] Preferably, in step S8, the channel response is reconstructed according to the path gain coefficient, specifically:

[0087] S8.1: Based on the estimated elevation angle, azimuth angle and range values, the LS algorithm can be used to obtain an estimated value of the gain factor:

[0088]

[0089] In the formula, is the estimated number of paths, It is composed of vectorized steering vectors:

[0090]

[0091] S8.2: Reconstruct the near-field channel as:

[0092] BRIEF DESCRIPTION OF THE DRAWINGS

[0093] Figure 1 It is a schematic diagram of the method flow of the present invention.

[0094] Figure 2 It is a schematic diagram of the near-field uniform planar array of the present invention.

[0095] Figure 3 Provided for the embodiment, a schematic diagram of the performance comparison between the method of the present invention and other channel estimation algorithms as the signal-to-noise ratio SNR increases. DETAILED DESCRIPTION

[0096] The drawings are for illustrative purposes only and should not be construed as limiting the present patent;

[0097] In order to better illustrate the present embodiment, some parts in the drawings may be omitted, enlarged or reduced, and do not represent the size of the actual product;

[0098] It is understandable to those skilled in the art that some well-known structures and their descriptions may be omitted in the drawings.

[0099] The technical solution of the present invention is further described below in conjunction with the accompanying drawings and embodiments.

[0100] Example 1

[0101] This example provides a planar array-based ultra-large-scale MIMO near-field channel estimation method, characterized in that it includes the following steps:

[0102] S1: Receive signals using a near-field uniform planar array;

[0103] S2: Take the first column of the received signal matrix, separate the elevation and azimuth parameters, and use this small portion of data to estimate the elevation and distance parameters;

[0104] S3: Grid the range of elevation angle and distance, construct a sparse reconstruction problem, and use the orthogonal matching pursuit algorithm to estimate the rough estimate of the elevation angle and distance parameters;

[0105] S4: fixing the distance value to the rough estimated value of the distance obtained in step S3, and performing a grid search on the elevation angle again to obtain a precise estimated value of the elevation angle;

[0106] S5: fixing the elevation angle to the precise elevation angle estimate value obtained in step S4, and performing a grid search on the distance again to obtain a precise distance estimate value;

[0107] S6: for the estimated elevation angle and distance parameters, determine whether the corresponding remaining residuals meet the iteration stop condition;

[0108] S7: Estimate azimuth parameters according to the estimated elevation angle and distance parameters;

[0109] S8: reconstruct the channel response according to the path gain coefficient;

[0110] S9: Output channel estimation result.

[0111] Example 2

[0112] This embodiment further discloses the following contents based on the embodiment 1:

[0113] The near-field uniform planar array in step S1 is specifically:

[0114] S1.1: The near-field uniform planar array is a two-dimensional array with M*N array elements and the distances between adjacent antennas in the x-direction and the y-direction are d x , d y , where the antenna numbers in the x direction of the near-field uniform planar array increase from 0 to M-1, and the antenna numbers in the y direction increase from 0 to N-1. The angle between the projection of the propagation path on the YOZ plane and the propagation path is defined as the elevation angle θ; the angle between the projection of the propagation path on the XOZ plane and the propagation path is defined as the azimuth angle The distance between the scatterer and the reference array element is r.

[0115] In step S1, the near-field uniform planar array receives the signal, specifically:

[0116] S1.2: Based on the newly defined positional relationship between elevation angle and azimuth angle, the channel model is constructed as follows:

[0117]

[0118] In the formula, α l is the gain coefficient of the lth path, θ l , r l are respectively the elevation angle, azimuth angle and distance between the scatterer and the reference array element of the lth path, L is the multipath number, and L<<M×N. is the steering matrix, and the element at the (p,q) position is:

[0119]

[0120] In the formula, In the channel model, The first-order term in the exponential part is only related to the angle, and decouples the elevation angle and the azimuth angle, thereby simplifying the channel model. The second-order term reflects the impact of the ultra-large-scale MIMO near-field effect - the distance parameter is taken into account in the model.

[0121] S1.3: The base station receives the pilot signal as follows:

[0122]

[0123] Where P is the transmit power and N is Gaussian white noise.

[0124] In step S2, the first column of the received signal matrix is ​​taken to estimate the elevation angle and distance parameters, specifically:

[0125] S2.1: Take the first column of the received signal matrix Y and get y0∈C M×1 ,as follows:

[0126]

[0127] In the formula, h represents the corresponding channel vector and n is the noise vector. The expression of h is:

[0128]

[0129] In the formula, a(θ l ,r l )∈C M×1 is the corresponding steering vector,

[0130]

[0131] From the above expression, we can see that, in addition to the gain coefficient, y0 only contains elevation and distance parameters, but no azimuth parameters. Independently of other parameters, a small portion of data is first used to estimate the elevation angle and distance, thereby reducing the amount of calculation of the method of the present invention.

[0132] In step S3, the range of elevation angle and distance is gridded, and the rough estimated values ​​of elevation angle and distance parameters are estimated using the orthogonal matching pursuit algorithm, specifically:

[0133] S3.1: The sine value of the angle is sampled uniformly in the range [-1,1]; the distance is sampled exponentially in the near field range (0,Z] (Z is the boundary between the far field and the near field - the Rayleigh distance, and Where D is the array aperture size and λ is the wavelength). Then M 2 The sampling grid Ψ of angle and distance sampling values c It can be expressed as:

[0134]

[0135] S3.2: In traditional far-field massive MIMO systems, when the number of paths is not comparable to the number of antennas, the channel is sparse in the angular domain and the paths can be extracted. This sparsity is partly due to the directionality of the array pattern, and the array pattern of ultra-large-scale MIMO in near-field conditions also has directionality and channel sparsity. Therefore, sparse signal recovery algorithms, such as the orthogonal matching pursuit (OMP) algorithm, can also be used to estimate the channel and obtain parameter values ​​in the channel.

[0136] The received signal vector y0 obtained in step S2 is approximately sparsely represented as:

[0137]

[0138] In the formula, is the corresponding sparse channel vector, is the perception matrix, which can be expressed as:

[0139] B co =[a(θ1,r1),…,a(θ1,r M ),…,a(θ M ,r1),…,a(θ M ,r M )]

[0140] In the formula, (θ m ,r m )∈Ψ c.

[0141] Therefore, the elevation angle and distance parameter estimation problem can be transformed into:

[0142]

[0143] In the formula, η is a very small constant. c A rough estimate of the elevation angle and distance can be obtained by searching within

[0144] In step S4, the fixed distance value is taken as the rough estimated value of the distance obtained in step S3, and the elevation angle is grid searched again to obtain the precise estimated value of the elevation angle, which is specifically:

[0145] S4.1: Fix the distance value to the rough estimate obtained in step S3 A rough estimate of the elevation angle The perception matrix W of the refined estimated angle is obtained again by sinusoidal uniform quantization within the nearby grid range θ ∈C M×M :

[0146]

[0147] In the formula,

[0148] S4.2: The refined estimation result of the elevation angle can be obtained by solving the following problems through the OMP algorithm:

[0149]

[0150] In the formula, and η θ are the corresponding channel vector and constant respectively.

[0151] In step S5, the fixed elevation angle is taken as the elevation angle precision estimated value obtained in step S4, and the distance is grid searched again to obtain the distance precision estimated value, which is specifically:

[0152] S5.1: Fix the elevation angle to the estimated result in step S4 The rough distance estimate obtained in step S3 The perception matrix W of the refined estimated distance is obtained again by exponential sampling in the vicinity of the nearby grid range r ∈C M×M :

[0153]

[0154] In the formula,

[0155] S5.2: The OMP algorithm can be used to solve the following problems to obtain a refined distance estimate:

[0156]

[0157] In the formula, and η r are the corresponding channel vector and constant respectively.

[0158] In step S6, for the estimated elevation angle and distance parameters, it is determined whether the corresponding residuals meet the iteration stop condition, specifically:

[0159] S6.1: After the l-1th iteration of OMP, l-1 sets of parameters and paths are estimated, and the remaining residual is:

[0160]

[0161] In the formula, the path gain

[0162] S6.2: Considering that the number of propagation paths is unknown in the channel estimation problem, this method uses the method of making the remaining residual as small as possible (when only noise is included) to stop the iteration process. If the estimation result meets the following iteration stop condition, execute step S7 to estimate the azimuth; if it does not meet the following condition, continue to estimate the new elevation angle and distance parameter group.

[0163]

[0164] In the formula, is the Fourier transform, P fa is the false alarm probability.

[0165] In step S7, the azimuth parameter is estimated according to the estimated elevation angle and distance parameters, specifically:

[0166] S7.1: Vectorize the received matrix Y row by row as follows:

[0167]

[0168] In the formula, is the vectorized noise, is the steering vector after row vectorization, which is:

[0169]

[0170]

[0171] S7.2: Further, using the elevation angle and distance estimates obtained in steps S4 and S5, As a known condition, estimate the azimuth Specifically:

[0172] Angle In range The uniform quantization sampling method is used, and the elevation angle and distance precision estimation values ​​of each path are substituted respectively. Construct the perception matrix D l :

[0173]

[0174] Where Q is the angle The number of quantized samples.

[0175] Therefore, the vectorized received signal can be The sparse representation is:

[0176]

[0177] In the formula, is the sparse representation of vec(H).

[0178] S7.3: Construct the following sparse reconstruction problem and use the OMP algorithm to perform a sparse dictionary search to obtain the azimuth estimate of the lth path:

[0179]

[0180] In step S8, the channel response is reconstructed according to the path gain coefficient, specifically:

[0181] S8.1: Based on the estimated elevation angle, azimuth angle and range values, the LS algorithm can be used to obtain an estimated value of the gain factor:

[0182]

[0183] In the formula, is the estimated number of paths, It is composed of vectorized steering vectors:

[0184]

[0185] S8.2: Reconstruct the near-field channel as:

[0186]

[0187] Example 3

[0188] This embodiment provides the following specific embodiments based on Embodiment 1 and Embodiment 2:

[0189] In the specific implementation process, the simulation parameters are set as follows: the number of array elements M*N=35*35, the array element spacing in the x direction and the y direction is half a wavelength, and the false alarm probability P fa = 0.01, the number of near-field paths is L = 2, and the normalized mean square error of the estimated result is calculated

[0190] The performance of the method proposed in the present invention is compared with that of the LS algorithm and the OMP algorithm, as follows: Figure 3 In the figure, the SI-OMP curve represents the new method proposed in the present invention; the Nopt-OMP curve represents the algorithm that only considers one layer of coarse estimation grid search when estimating the elevation angle and distance; the NFOMP curve represents the near-field OMP algorithm under the uniform linear array; and the LS curve represents the least squares method. As can be seen from the figure, the near-field channel estimation scheme based on planar arrays (SI-OMP algorithm) proposed in the present invention is significantly better than the existing channel estimation algorithm. The reason is that, on the one hand, the algorithm proposed in the present invention solves the near-field channel estimation problem in the scenario where the base station is deployed in a planar rectangular array, while the NFOMP algorithm is only applicable to the channel estimation problem of linear arrays. On the other hand, after the SI-OMP algorithm obtains the elevation angle and distance based on only one column of data in the receiving matrix, it effectively optimizes and estimates the parameters and has higher channel estimation accuracy.

[0191] The same or similar reference numerals correspond to the same or similar components;

[0192] The positional relationships described in the drawings are only for illustrative purposes and should not be construed as limiting the present patent.

[0193] Obviously, the above examples of the present invention are only examples for clearly explaining the present invention, and are not intended to limit the embodiments of the present invention. For those skilled in the art, other different forms of changes or modifications can be made based on the above description. It is not necessary and impossible to list all the embodiments here. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the claims of the present invention.

Claims

1. A planar array-based ultra-large-scale MIMO near-field channel estimation method, characterized in that: The following steps are involved: S1: Receive signals using a near-field uniform planar array; S2: Take the first column of the received signal matrix, separate the elevation and azimuth parameters, and use this small portion of data to estimate the elevation and distance parameters; S3: Grid the range of elevation angle and distance, construct a sparse reconstruction problem, and use the orthogonal matching pursuit algorithm to estimate the rough estimate of the elevation angle and distance parameters; S4: fixing the distance value to the rough estimated value of the distance obtained in step S3, and performing a grid search on the elevation angle again to obtain a precise estimated value of the elevation angle; S5: fixing the elevation angle to the precise elevation angle estimate value obtained in step S4, and performing a grid search on the distance again to obtain a precise distance estimate value; S6: for the estimated elevation angle and distance parameters, determine whether the corresponding remaining residuals meet the iteration stop condition; S7: Estimate azimuth parameters according to the estimated elevation angle and distance parameters; S8: reconstruct the channel response according to the path gain coefficient; S9: Output the channel estimation result.

2. The planar array-based ultra-large-scale MIMO near-field channel estimation method according to claim 1, characterized in that: The near-field uniform planar array in step S1 is specifically: The near-field uniform planar array is a two-dimensional array, the number of array elements is M*N, and the distances between adjacent antennas in the x-direction and the y-direction are d x , d y , where the antenna numbers in the x direction of the near-field uniform planar array increase from 0 to M-1, and the antenna numbers in the y direction increase from 0 to N-1. As shown in Figure 2, the angle between the projection of the propagation path on the YOZ plane and the propagation path is defined as the elevation angle θ; the angle between the projection of the propagation path on the XOZ plane and the propagation path is defined as the azimuth angle The distance between the scatterer and the reference array element is r.

3. The planar array-based ultra-large-scale MIMO near-field channel estimation method according to claim 2, characterized in that: In step S1, the near-field uniform planar array receives the signal, specifically: Through the newly defined positional relationship between elevation angle and azimuth angle, the channel model is constructed as follows: In the formula, α l is the gain coefficient of the lth path, θ l , r l are respectively the elevation angle, azimuth angle and distance between the scatterer and the reference array element of the lth path, L is the multipath number, and L<<M×N. is the steering matrix, and the element at the (p,q) position is: In the formula, In the channel model, The first-order term in the exponential part is only related to the angle, and decouples the elevation angle and the azimuth angle, thereby simplifying the channel model. The second-order term reflects the impact of the ultra-large-scale MIMO near-field effect - the distance parameter is taken into account in the model. The base station receives the pilot signal as: Where P is the transmit power and N is Gaussian white noise.

4. The planar array-based ultra-large-scale MIMO near-field channel estimation method according to claim 3, characterized in that: In step S2, the first column of the received signal matrix is ​​taken to estimate the elevation angle and distance parameters, specifically: Take the first column of the received signal matrix Y and get y0∈C M×1 ,as follows: In the formula, h represents the corresponding channel vector and n is the noise vector. The expression of h is: In the formula, a(θ l ,r l )∈C M×1 is the corresponding steering vector, From the above expression, we can see that, in addition to the gain coefficient, y0 only contains elevation and distance parameters, but no azimuth parameters. Independently of other parameters, a small portion of data is first used to estimate the elevation angle and distance, thereby reducing the amount of calculation of the method of the present invention.

5. The planar array-based ultra-large-scale MIMO near-field channel estimation method according to claim 4, characterized in that: In step S3, the range of elevation angle and distance is gridded, and the rough estimated values ​​of elevation angle and distance parameters are estimated using the orthogonal matching pursuit algorithm, specifically: The sine value of the angle is sampled uniformly in the range of [-1,1]; the distance is sampled exponentially in the near field range (0,Z] (Z is the boundary between the far field and the near field - the Rayleigh distance, and Where D is the array aperture size and λ is the wavelength). Then M 2 The sampling grid Ψ of angle and distance sampling values c It can be expressed as: In traditional far-field massive MIMO systems, when the number of paths is not comparable to the number of antennas, the channel is sparse in the angular domain and the paths can be extracted. This sparsity is partly due to the directionality of the array pattern, and the array pattern of ultra-large-scale MIMO under near-field conditions also has directionality and channel sparsity. Therefore, sparse signal recovery algorithms, such as the orthogonal matching pursuit (OMP) algorithm, can also be used to estimate the channel and obtain parameter values ​​in the channel. The received signal vector y0 obtained in step S2 is approximately sparsely represented as: In the formula, is the corresponding sparse channel vector, is the perception matrix, which can be expressed as: B co =[a(θ1,r1),…,a(θ1,r M ),…,a(θ M ,r1),…,a(θ M ,r M )] In the formula, (θ m ,r m )∈Ψ c . Therefore, the elevation angle and distance parameter estimation problem can be transformed into: In the formula, η is a very small constant. c A rough estimate of the elevation angle and distance can be obtained by searching within 6. The planar array-based ultra-large-scale MIMO near-field channel estimation method according to claim 5, characterized in that: In step S4, the fixed distance value is taken as the rough estimated value of the distance obtained in step S3, and the elevation angle is grid searched again to obtain the precise estimated value of the elevation angle, which is specifically: The fixed distance value is the rough estimate obtained in step S3 A rough estimate of the elevation angle The perception matrix W of the refined estimated angle is obtained again by sinusoidal uniform quantization within the nearby grid range θ ∈C M×M : In the formula, The OMP algorithm can be used to solve the following problems to obtain a refined estimate of the elevation angle: In the formula, and η θ are the corresponding channel vector and constant respectively.

7. The planar array-based ultra-large-scale MIMO near-field channel estimation method according to claim 6, characterized in that: In step S5, the fixed elevation angle is taken as the elevation angle precision estimated value obtained in step S4, and the distance is grid searched again to obtain the distance precision estimated value, which is specifically: The fixed elevation angle is the estimated result in step S4. The rough distance estimate obtained in step S3 The perception matrix W of the refined estimated distance is obtained again by exponential sampling in the vicinity of the nearby grid range r ∈C M×M : In the formula, By solving the following problems with the OMP algorithm, a refined distance estimation result can be obtained: In the formula, and η r are the corresponding channel vector and constant respectively.

8. The planar array-based ultra-large-scale MIMO near-field channel estimation method according to claim 7, characterized in that: In step S6, for the estimated elevation angle and distance parameters, it is determined whether the corresponding residuals meet the iteration stop condition, specifically: After OMP performs the l-1th iteration, l-1 sets of parameters and paths are estimated, and the remaining residual is: In the formula, the path gain Considering that the number of propagation paths is unknown in the channel estimation problem, this method uses the method of making the remaining residual as small as possible (when only noise is included) to stop the iteration process. If the estimation result meets the following iteration stop condition, step S7 is executed to estimate the azimuth; if the following condition is not met, the new elevation angle and distance parameter group are continuously estimated. In the formula, is the Fourier transform, P fa is the false alarm probability.

9. The planar array-based ultra-large-scale MIMO near-field channel estimation method according to claim 8, characterized in that: In step S7, the azimuth parameter is estimated according to the estimated elevation angle and distance parameters, specifically: The receiving matrix Y is vectorized row by row as follows: In the formula, is the vectorized noise, is the steering vector after row vectorization, which is: Further, using the elevation angle and distance accurate estimates obtained in steps S4 and S5 As a known condition, estimate the azimuth Specifically: Angle In range The uniform quantization sampling method is used, and the elevation angle and distance precision estimation values ​​of each path are substituted respectively. Construct the perception matrix D l : Where Q is the angle The number of quantized samples is therefore the vectorized received signal The sparse representation is: In the formula, is the sparse representation of vec(H), The following sparse reconstruction problem is constructed, and the sparse dictionary search is performed using the OMP algorithm to obtain the azimuth angle estimate of the lth path: 。 10. The planar array-based ultra-large-scale MIMO near-field channel estimation method according to claim 9, characterized in that: In step S8, the channel response is reconstructed according to the path gain coefficient, specifically: Based on the estimated elevation angle, azimuth angle and distance values, the LS algorithm can be used to obtain the estimated value of the gain coefficient: In the formula, is the estimated number of paths, It is composed of vectorized steering vectors: The reconstructed near-field channel is: 。

Citation Information

Patent Citations

  • Super-large-scale MIMO channel estimation method based on Newton orthogonal matching pursuit

    CN112565122A

  • Orthogonal position design method in near-field super-large-scale planar array communication system

    CN115996392A

  • Near-field full-dimensional non-orthogonal codebook design method for super-large-scale MIMO system

    CN117544205A