A wave number domain channel sensing method based on asymmetric architecture in super-massive MIMO

Abstract

The application relates to the technical field of wireless communication, in particular to a wave number domain channel sensing method based on an asymmetric architecture in a super large scale MIMO, which comprises the following steps: determining the uplink antenna position by using a coprime array to extract a virtual array signal; designing an alternating direction multiplier method to estimate the wave number domain channel variance and noise variance; according to the mapping relationship between the wave number direction and the angle, deducing the arrival angle information from the wave number direction in the support set; performing MMSE on the uplink receiving signal to estimate the complex path gain and obtain the estimation of the channel; constructing the coprime array according to the antenna index from small to large to obtain the first angle estimation information; constructing the coprime array according to the antenna index from large to small to obtain the second angle estimation information; calculating the distance from the base station to the scattering cluster to realize the wave number domain channel sensing of the super large scale MIMO. The application can accurately estimate the statistical CSI of the wave number domain channel and the instantaneous CSI of the current channel under the condition that the sparsity is unknown, and reduces the hardware cost and the calculation complexity of the base station.

Description

Technical Field

[0001] This invention relates to the field of wireless communication technology, specifically to a wavenumber domain channel sensing method based on an asymmetric architecture in ultra-large-scale MIMO, which can be used for channel sensing at the base station side in fifth-generation mobile communication systems (5G) and their evolved versions (6G). Background Technology

[0002] In future sixth-generation mobile communication systems, 3GPP believes there is a need to further apply and expand ultra-large-scale MIMO technology, demonstrating the superior performance of MIMO arrays in terms of rate, spectral efficiency, and reliability. However, hardware cost, data processing burden, and power consumption requirements limit the practical application of all-digital ultra-large-scale MIMO technology. Using fewer radio frequency chains can effectively reduce the base station's signal processing pressure and uplink power consumption. Therefore, researchers, based on the current situation where uplink demand is less than downlink demand, have proposed using an asymmetric architecture instead of an all-digital architecture. The asymmetric transceiver architecture, by configuring all antennas and radio frequency links for the downlink and only enabling some receiving antennas and radio frequency links for the uplink, significantly reduces the hardware cost and baseband processing pressure of uplink reception while achieving all-digital beamforming gain in the downlink. However, due to the asymmetry in the number of uplink and downlink radio frequency links, uplink channel state information can only be obtained through some antennas, and the full-dimensional downlink channel must be recovered from the reduced-dimensional uplink observations. Existing research has proposed a series of methods to address this channel recovery problem. For example, the paper “Asymmetrical Uplink and Downlink Transceivers in Massive MIMO Systems” (IEEE Transactions on Vehicular Technology, Vol. 70, No. 11, November 2021) proposes an uplink-to-downlink channel transfer algorithm based on discrete Fourier transform and modified Newton orthogonal matched pursuit. It uses a plane wave angle parameterization model to extract the angle of arrival and complex gain from the uplink estimate and then reconstructs the downlink channel.

[0003] However, the aforementioned literature and existing work all rely on the plane wavefront assumption to study asymmetric architectures. As array apertures continue to increase and enter the realm of ultra-large-scale MIMO, the array size far exceeds the Rayleigh distance, the electromagnetic wavefront curvature becomes non-negligible, and the channel needs to be accurately characterized using a spherical wavefront, rather than a plane wavefront. Under these near-field conditions, the angle parameterization model based on the plane wave assumption will suffer severe model mismatch, leading to a significant degradation in the performance of existing uplink channel estimation and downlink channel recovery algorithms, fundamentally limiting the usability of asymmetric architectures in ultra-large-scale MIMO. Although some researchers have keenly noticed this problem and attempted improvements, these works either make local modifications to existing far-field algorithms or approximate compensation of the plane wave model within a finite distance, failing to truly break through the constraints of the plane wave assumption. It is noteworthy that the paper "Fourier Plane-Wave Series Expansion for Holographic MIMO Communications" (IEEE Transactions on Wireless Communications, Vol. 21, No. 9, September 2022) establishes a wavenumber domain channel model based on Fourier plane wave series expansion, starting from the fundamental principles of electromagnetic wave propagation. This model can uniformly describe the propagation characteristics of the channel in both the far and near fields, possessing clear physical meaning and providing a fundamental theoretical tool for analyzing ultra-large-scale MIMO channels. However, this wavenumber domain analysis method has not yet been introduced into the study of asymmetric transceiver architectures. Therefore, if the wavenumber domain channel model can be used to re-examine the spatial undersampling problem under asymmetric reception, it is hoped that the limitations of the plane wave assumption can be overcome, establishing a new characterization and reconstruction framework for full-dimensional downlink channel recovery in asymmetric transceiver architectures of ultra-large-scale MIMO systems.

[0004] Furthermore, within the current 3GPP standard framework (including 5G-Advanced and the 6G technology under research), the emerging Integrated Sensing and Communication (ISAC) research project (see the consensus of RAN1#124 meeting) defines the sensing measurement parameters that base stations can report, including delay, Doppler, and angle of arrival, but it does not explicitly specify the angular spread information of scattering clusters and distance information. These two pieces of information are crucial for accurately describing the spatial distribution of environmental scatterers and beam prediction. Therefore, how to estimate the angular spread and distance of scattering clusters and report this sensing information in a standard-compliant manner has become a pressing technical problem to be solved. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of existing technologies and the lack of reporting of scattering cluster angle spread and distance in current 3GPP standards. It provides a wavenumber domain channel sensing method based on an asymmetric architecture for ultra-large-scale MIMO. This method can accurately estimate the statistical CSI of the wavenumber domain channel even with unknown sparsity, and simultaneously obtain the instantaneous CSI of the current channel. It can also effectively reduce the hardware cost and computational complexity of the base station. Furthermore, it reports angle spread and distance information according to the 3GPP ISAC framework, helping the base station better perceive its surrounding environment.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0007] A wavenumber domain channel sensing method based on an asymmetric architecture in ultra-large-scale MIMO includes the following steps:

[0008] S100, constructing an ultra-large-scale MIMO system consisting of one base station and one single-antenna user; wherein, the base station is equipped with a... A uniform linear array composed of antennas, and simultaneously within the uniform linear array... All antennas are connected to the downlink RF chain, and in Selecting from root antennas The root antenna is connected to the uplink RF chain;

[0009] S200. Based on the differential co-array completeness criterion, the uplink antenna position is determined using a coprime array, and the virtual array signal is extracted from the autocorrelation matrix of the uplink received signal under the antenna configuration.

[0010] S300. Based on the extracted virtual array signal, a method combining the alternating direction multiplier method improved by Fused Lasso is designed to estimate the channel variance and noise variance in the wavenumber domain.

[0011] S400. Based on the estimated channel variance in the wavenumber domain, the support set is extracted. Then, based on the mapping relationship between wavenumber direction and angle, the arrival angle information, including mode direction and angle spread, is inferred from the wavenumber direction in the support set.

[0012] S500: Based on the estimated variance and wavenumber domain codebook, perform MMSE on the uplink received signal to estimate the complex path gain and further obtain an estimate of the channel.

[0013] S600. Based on the idea of ​​solving triangles, design a two-stage distance estimation scheme:

[0014] In the first stage, the first reference point is set as the first antenna, and a coprime array is constructed according to the antenna index from small to large. The first angle estimation information is obtained by estimating the variance.

[0015] In the second stage, the second reference point is set as the Mth antenna, and a coprime array is constructed from large to small according to the antenna lines. The second angle estimation information is obtained by estimating the variance. Based on the first angle estimation information and the second angle estimation information, as well as the distance between the first reference point and the second reference point, the distance from the base station to the scattering cluster is calculated. Finally, wavenumber domain channel sensing for ultra-large-scale MIMO is realized.

[0016] Preferably, step S200 specifically includes:

[0017] S201, to Perform prime factorization to obtain the first prime number. and the second prime number You can choose any of them. and The combination only requires... ,and ;

[0018] S202. Based on the selected first and second prime numbers, construct a coprime array in ascending order of antenna indices: select antennas from the first to the second... The first antenna is selected as the first subarray, and the antenna indices in the first subarray are consecutive; the first antenna is selected as the second antenna. root antenna to the first The first antenna is used as the second subarray, and the index difference between adjacent antennas in the second subarray is 1. The first and second subarrays are coprime arrays, and the total number of uplink antennas satisfies the following condition: ;

[0019] S203, Based on the uplink received signal obtained in the coprime array ,pass The autocorrelation matrix is ​​obtained from the second snapshot. The expression is: ;

[0020] S204. Due to the use of coprime arrays, the autocorrelation matrix... Contains from arrive The continuous linear delay terms; extract and process the elements at the positions of the non-negative delay terms in the matrix:

[0021] If multiple elements in the matrix correspond to the same non-negative delay term, the average value is taken; if only one element in the matrix corresponds to a certain non-negative delay term, no processing is required; for the processed elements, they are arranged in ascending order of delay term index to obtain the virtual array signal. .

[0022] Preferably, step S300 specifically includes:

[0023] S301, Obtain virtual array signal dictionary matrix and preset regularization parameters , and penalty factor The dictionary matrix A has L+1 columns of vectors, with the last column being a vector. ,forward Listed as wavenumber field codebook , Wavenumber direction index set Total number of elements in the middle; wavenumber direction index set , will set The elements in the array are arranged in ascending order; among them, For wavelength, The aperture length of the ULA;

[0024] S302, Based on the virtual array signal The quadratic matrix is ​​calculated from the dictionary matrix A. and gradient vector Construct a first-order difference matrix Its elements satisfy , The rest are zero; thus generating the projection matrix. and ;in, yes Unit matrix of order, yes A zero-dimensional vector;

[0025] S303, Calculate the matrix And find the inverse;

[0026] in Indicates the penalty parameter;

[0027] S304. Initialization: Set the auxiliary variables... , , and dual variables , , Set the zero vector, and set the maximum number of iterations and convergence tolerance;

[0028] S305, Enter the iteration, for the... Next iteration:

[0029] Construct vectors:

[0030] And update the variance estimate vector. The iterative expression is obtained as follows: ;

[0031] Where d represents the gradient vector. The parameters represent the penalty term, and P and Q represent the projection matrices. , , Describes the auxiliary variable for the k-th iteration. , , Let U represent the dual variable of the k-th iteration, and let U represent the matrix calculated in step S303.

[0032] Update the first auxiliary variable The iterative expression is: ;

[0033] Update the second auxiliary variable The iterative expression is: ;

[0034] in, This represents the preset regularization parameters;

[0035] Update the third auxiliary variable The iterative expression is:

[0036] ;

[0037] in, This represents the preset regularization parameters;

[0038] Update the first auxiliary variable The iterative expression is: ;

[0039] Update the second auxiliary variable The iterative expression is: ;

[0040] Update the third auxiliary variable The iterative expression is: ;

[0041] Calculate the original residuals and dual residuals If all values ​​are less than the convergence tolerance or the maximum number of iterations is reached, the iteration is terminated; otherwise, the iteration continues.

[0042] The expression for calculating the original residual is as follows:

[0043] ;

[0044] The expression for calculating the dual residual is: ;

[0045] S306. After the iteration is completed, the variance estimation vector is... The final output, after applying a non-negativity truncation, is expressed as: .

[0046] Preferably, step S400 specifically includes:

[0047] S401, Extracting the variance estimation vector The former Use elements to construct a temporary vector. ; in descending order, process the temporary vectors Arrange all elements in order to obtain a sorted vector. and its corresponding set of original index mappings ;

[0048] S402, Calculate the sorting vector forward The sum of the terms ,in ;

[0049] S403, Obtain the threshold Calculate the absolute energy threshold Then find the smallest integer. , making ;

[0050] S404, Extract Set The front of the middle Each element serves as the support set. The estimated wavenumber domain channel variance is obtained. The elements satisfy: ;

[0051] S405, For sets any index in Calculate the corresponding wavenumber direction For the support set any index in The corresponding wavenumber direction is obtained. ;

[0052] S406, Based on Free Space Wavenumber , will support the collection The corresponding wavenumber direction is converted to an angle. Generate angle estimation vector The elements satisfy: ;

[0053] S407, Estimation of angular extension and modal orientation:

[0054] In a physical propagation environment, since any scattering cluster corresponds to multiple paths arriving from the same set of scatterers, and these paths exhibit a statistical distribution within a finite angular range around a certain principal direction in the angular domain, the macroscopic position of the cluster is determined by the modal direction. The degree of energy diffusion in space is determined by the angle. Quantization, therefore the angle estimation vector Non-zero values ​​in the vector will form consecutive blocks. Identify the angle estimation vector. For any consecutive index segment with non-zero values, Extract its starting index and end index Calculate continuous index segments The corresponding angle extension is expressed as: ;

[0055] Calculate contiguous index segments The corresponding modal direction is expressed as:

[0056] ;

[0057] The continuous index segment The range of values ​​for depends on the number of scattering clusters in the environment; assuming there are a total of scattering clusters in the environment. If there are scattering clusters, then The total number of consecutive index segments is also ,have .

[0058] Preferably, step S500 specifically includes:

[0059] S501, Extracting the variance estimation vector The last element yields the noise estimate. ;

[0060] S502. Constructing a matrix of known information , where the matrix ;

[0061] S503. Perform MMSE on the uplink received signal to obtain the complex path gain estimate. The specific expression is: ;

[0062] S504. Calculate downlink channel estimation The expression is: .

[0063] Preferably, step S600 specifically includes:

[0064] S601, First-stage angle estimation:

[0065] Treat the current coprime array as the first array and estimate the current angle. As the result of the first stage angle estimation The first array is characterized in that the first antenna is set as the first reference point, and a coprime array is constructed in ascending order of antenna index.

[0066] Identify the first-stage angle estimation results For any consecutive index segment with non-zero values, Extract its starting index and end index ;

[0067] Calculate contiguous index segments Corresponding angle extension: ,

[0068] And modal orientation: ;

[0069] S602, Second Stage Angle Estimation:

[0070] Reconstruct the coprime array as the second array; the second array will... The first antenna is set as the second reference point, and a coprime array is constructed according to the antenna index from largest to smallest;

[0071] Repeat steps S200 and S300 to obtain the second-stage angle estimation results. ;

[0072] Identify the second-stage angle estimation results For any consecutive index segment with non-zero values, Extract its starting index and end index Calculate continuous index segments Corresponding angle extension and modal direction ;

[0073] S603, Estimated Distance Information:

[0074] Calculate the distance between the first reference point and the second reference point. ;

[0075] Considering the generalized stationarity of the physical propagation environment, the two-stage estimation results and The total number of consecutive index segments is ,have At the same time, when At that time, these two consecutive index segments describe the same scattering cluster; based on the angle estimation results of the two stages. and and spacing Calculate the distance from the base station to the... Distance of each scattering cluster The expression is:

[0076] ;

[0077] At this point, the base station achieved wavenumber domain channel awareness for ultra-large-scale MIMO.

[0078] Preferably, in step S400, after calculating the angular spread, the base station reports the angular spread of the scattering cluster according to the format of the 3GPP Integrated Sensing and Communication (ISAC) measurement level; wherein the reported content is a newly added optional field of the measurement level.

[0079] Preferably, in step S600, after the base station calculates the distance to the scattering cluster, it reports the distance to the scattering cluster in accordance with the 3GPP ISAC measurement level format; wherein the reported content is a newly added optional field of the measurement level.

[0080] Compared with the prior art, the beneficial effects achieved by the present invention are:

[0081] This invention fully utilizes the wavenumber domain channel model's ability to uniformly characterize far-field and near-field propagation characteristics during downlink channel recovery in ultra-large-scale MIMO asymmetric transceiver architectures. It achieves joint acquisition of wavenumber domain statistical channel state information and instantaneous channel state information even with unknown channel sparsity. Compared to existing angle-parameterized channel recovery methods based on the planar wavefront assumption, this invention overcomes the model mismatch constraint caused by the near-field spherical wavefront curvature, fundamentally solving the full-dimensional downlink channel recovery problem in asymmetric reception under ultra-large-scale arrays, while simultaneously reducing the requirements for prior channel knowledge.

[0082] This invention proposes an alternating direction multiplier method improved by Fused Lasso, which effectively controls computational complexity during channel recovery by leveraging the low-rank characteristics and clear physical interpretation of the wavenumber domain channel representation. Compared with existing near-field channel recovery methods based on angle grid search or iterative optimization, this invention avoids fine-grained searches in high-dimensional angle spaces and complex decoupling operations of spherical wave parameters, requiring only computations matching the effective dimension of the wavenumber domain. This significantly reduces the computational complexity of channel recovery, making it more suitable for the deployment and application of ultra-large-scale arrays in practical systems.

[0083] This invention is the first to introduce the angular extension of the scattering cluster and the distance from the base station to the scattering cluster into the 3GPP ISAC measurement framework, which supplements the shortcomings of the existing measurement levels and provides richer spatial information for environmental perception and beam prediction. Attached Figure Description

[0084] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0085] Figure 1 This is a flowchart of an embodiment of the present invention;

[0086] Figure 2 This is a simulation diagram of an embodiment of the present invention. Detailed Implementation

[0087] 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, and 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.

[0088] Please see Figures 1-2 The present invention provides the following technical solution:

[0089] Example 1: This embodiment of the invention proposes a wavenumber domain channel sensing method based on an asymmetric architecture in ultra-large-scale MIMO. It can accurately estimate the statistical CSI of the wavenumber domain channel even with unknown sparsity, and simultaneously obtain the instantaneous CSI of the current channel. It can also effectively reduce the hardware cost and computational complexity of the base station. Furthermore, it reports angle extension and distance information according to the 3GPP ISAC framework, helping the base station better perceive its surrounding environment. Figure 1 As shown, this embodiment specifically includes the following steps:

[0090] S100, constructing an ultra-large-scale MIMO system consisting of one base station and one single-antenna user; wherein, the base station is equipped with a... A uniform linear array composed of antennas, and simultaneously within the uniform linear array... All antennas are connected to the downlink RF chain, and in Selecting from root antennas The root antenna is connected to the uplink RF chain; where The The antennas are numbered sequentially as follows: ;

[0091] S200. Based on the differential co-array completeness criterion, the uplink antenna position is determined using a coprime array, and the virtual array signal is extracted from the autocorrelation matrix of the uplink received signal under the antenna configuration.

[0092] Preferably, step S200 specifically includes:

[0093] S201, to Perform prime factorization to obtain the first prime number. and the second prime number You can choose any of them. and The combination only requires... ,and ;

[0094] S202. Based on the selected first and second prime numbers, construct a coprime array in ascending order of antenna indices: select antennas from the first to the second... The first antenna is selected as the first subarray, and the antenna indices in the first subarray are consecutive; the first antenna is selected as the second antenna. root antenna to the first The first antenna is used as the second subarray, and the index difference between adjacent antennas in the second subarray is 1. The first and second subarrays are coprime arrays, and the total number of uplink antennas satisfies the following condition: ;

[0095] S203, Based on the uplink received signal obtained in the coprime array ,pass The autocorrelation matrix is ​​obtained from the second snapshot. The expression is: ;

[0096] S204. Due to the use of coprime arrays, the autocorrelation matrix... Contains from arrive The continuous linear delay terms; extract and process the elements at the positions of the non-negative delay terms in the matrix:

[0097] If multiple elements in the matrix correspond to the same non-negative delay term, the average value is taken; if only one element in the matrix corresponds to a certain non-negative delay term, no processing is required; for the processed elements, they are arranged in ascending order of delay term index to obtain the virtual array signal. .

[0098] S300. Based on the extracted virtual array signal, an alternating direction multiplier method combined with Fused Lasso is designed to estimate the channel variance and noise variance in the wavenumber domain.

[0099] Preferably, step S300 specifically includes:

[0100] S301, Obtain virtual array signal dictionary matrix and preset regularization parameters , and penalty factor The dictionary matrix A has L+1 columns of vectors, with the last column being a vector. ,forward Listed as wavenumber field codebook , Wavenumber direction index set Total number of elements in the middle; wavenumber direction index set , will set The elements in the array are arranged in ascending order; among them, For wavelength, The aperture length of the ULA;

[0101] S302, Based on the virtual array signal The quadratic matrix is ​​calculated from the dictionary matrix A. and gradient vector Construct a first-order difference matrix Its elements satisfy , The rest are zero; thus generating the projection matrix. and ;in, yes Unit matrix of order, yes A zero-dimensional vector;

[0102] S303, Calculate the matrix And find the inverse;

[0103] in Indicates the penalty parameter;

[0104] S304. Initialization: Set the auxiliary variables... , , and dual variables , , Set the zero vector, and set the maximum number of iterations and convergence tolerance;

[0105] S305, Enter the iteration, for the... Next iteration:

[0106] Construct vectors:

[0107] And update the variance estimate vector. The iterative expression is obtained as follows: ;

[0108] Where d represents the gradient vector. The parameters represent the penalty term, and P and Q represent the projection matrices. , , Describes the auxiliary variable for the k-th iteration. , , Let U represent the dual variable of the k-th iteration, and let U represent the matrix calculated in step S303.

[0109] Update the first auxiliary variable The iterative expression is: ;

[0110] Update the second auxiliary variable The iterative expression is: ;

[0111] in, This represents the preset regularization parameters;

[0112] Update the third auxiliary variable The iterative expression is:

[0113] ;

[0114] in, This represents the preset regularization parameters;

[0115] Update the first auxiliary variable The iterative expression is: ;

[0116] Update the second auxiliary variable The iterative expression is: ;

[0117] Update the third auxiliary variable The iterative expression is: ;

[0118] Calculate the original residuals and dual residuals If all values ​​are less than the convergence tolerance or the maximum number of iterations is reached, the iteration is terminated; otherwise, the iteration continues.

[0119] The expression for calculating the original residual is as follows:

[0120] ;

[0121] The expression for calculating the dual residual is: ;

[0122] S306. After the iteration is completed, the variance estimation vector is... The final output, after applying a non-negativity truncation, is expressed as: .

[0123] S400: Based on the estimated wavenumber domain channel variance, extract the support set, and then, according to the mapping relationship between wavenumber direction and angle, infer the angle of arrival information from the wavenumber direction in the support set, including mode direction and angle spread.

[0124] Preferably, step S400 specifically includes:

[0125] S401, Extracting the variance estimation vector The former Use elements to construct a temporary vector. ; in descending order, process the temporary vectors Arrange all elements in order to obtain a sorted vector. and its corresponding set of original index mappings ;

[0126] S402, Calculate the sorting vector forward The sum of the terms ,in ;

[0127] S403, Obtain the threshold Calculate the absolute energy threshold Then find the smallest integer. , making ;

[0128] S404, Extract Set The front of the middle Each element serves as the support set. The estimated wavenumber domain channel variance is obtained. The elements satisfy: ;

[0129] S405, For sets any index in Calculate the corresponding wavenumber direction For the support set any index in The corresponding wavenumber direction is obtained. ;

[0130] S406, Based on Free Space Wavenumber , will support the collection The corresponding wavenumber direction is converted to an angle. Generate angle estimation vector The elements satisfy: ;

[0131] S407, Estimation of angular extension and modal orientation:

[0132] In a physical propagation environment, since any scattering cluster corresponds to multiple paths arriving from the same set of scatterers, and these paths exhibit a statistical distribution within a finite angular range around a certain principal direction in the angular domain, the macroscopic position of the cluster is determined by the modal direction. The degree of energy diffusion in space is determined by the angle. Quantization, therefore the angle estimation vector Non-zero values ​​in the vector will form consecutive blocks. Identify the angle estimation vector. For any consecutive index segment with non-zero values, Extract its starting index and end index Calculate continuous index segments The corresponding angle extension is expressed as: ;

[0133] Calculate contiguous index segments The corresponding modal direction is expressed as:

[0134] ;

[0135] The continuous index segment The range of values ​​for depends on the number of scattering clusters in the environment; assuming there are a total of scattering clusters in the environment. If there are scattering clusters, then The total number of consecutive index segments is also ,have .

[0136] Preferably, in step S400, after calculating the angular spread, the base station reports the angular spread of the scattering cluster according to the format of the 3GPP Integrated Sensing and Communication (ISAC) measurement level; wherein the reported content is a newly added optional field of the measurement level.

[0137] S500 performs MMSE on the uplink received signal based on the estimated variance and wavenumber domain codebook to estimate the complex path gain and further obtain an estimate of the channel.

[0138] Preferably, step S500 specifically includes:

[0139] S501, Extracting the variance estimation vector The last element yields the noise estimate. ;

[0140] S502. Constructing a matrix of known information , where the matrix ;

[0141] S503. Perform MMSE on the uplink received signal to obtain the complex path gain estimate. The specific expression is: ;

[0142] S504. Calculate downlink channel estimation The expression is: .

[0143] S600. Based on the idea of ​​solving triangles, design a two-stage distance estimation scheme:

[0144] In the first stage, the first reference point is set as the first antenna, and a coprime array is constructed according to the antenna index from small to large. The first angle estimation information is obtained by estimating the variance.

[0145] In the second stage, the second reference point is set as the Mth antenna, and a coprime array is constructed from large to small according to the antenna lines. The second angle estimation information is obtained by estimating the variance. Based on the first angle estimation information and the second angle estimation information, as well as the distance between the first reference point and the second reference point, the distance from the base station to the scattering cluster is calculated. Finally, wavenumber domain channel sensing for ultra-large-scale MIMO is realized.

[0146] Preferably, step S600 specifically includes:

[0147] S601, First-stage angle estimation:

[0148] Treat the current coprime array as the first array and estimate the current angle. As the result of the first stage angle estimation The first array is characterized in that the first antenna is set as the first reference point, and a coprime array is constructed in ascending order of antenna index.

[0149] Identify the first-stage angle estimation results For any consecutive index segment with non-zero values, Extract its starting index and end index ;

[0150] Calculate contiguous index segments Corresponding angle extension: ,

[0151] And modal orientation: ;

[0152] S602, Second Stage Angle Estimation:

[0153] Reconstruct the coprime array as the second array; the second array will... The first antenna is set as the second reference point, and a coprime array is constructed according to the antenna index from largest to smallest;

[0154] Repeat steps S200 and S300 to obtain the second-stage angle estimation results. ;

[0155] Identify the second-stage angle estimation results For any consecutive index segment with non-zero values, Extract its starting index and end index Calculate continuous index segments Corresponding angle extension and modal direction ;

[0156] S603, Estimated Distance Information:

[0157] Calculate the distance between the first reference point and the second reference point. ;

[0158] Considering the generalized stationarity of the physical propagation environment, the two-stage estimation results and The total number of consecutive index segments is ,have At the same time, when At that time, these two consecutive index segments describe the same scattering cluster; based on the angle estimation results of the two stages. and and spacing Calculate the distance from the base station to the... Distance of each scattering cluster The expression is:

[0159] ;

[0160] At this point, the base station achieved wavenumber domain channel awareness for ultra-large-scale MIMO.

[0161] Preferably, in step S600, after the base station calculates the distance to the scattering cluster, it reports the distance to the scattering cluster in accordance with the 3GPP ISAC measurement level format; wherein the reported content is a newly added optional field of the measurement level, which is used for subsequent reporting of sensing information.

[0162] Figure 2 The results shown are obtained from the simulation verification of this invention, with the following parameter settings: the base station has a total of 331 antennas, but only 166 antenna users form a coprime array to receive uplink signals. Figure 2 The horizontal axis represents the signal-to-noise ratio (SNR) in decibels (dB), and the vertical axis represents the normalized mean square error of the full-dimensional channel estimation in decibels (dB).

[0163] The four curves in the figure represent the effects of the proposed scheme. The Orthogonal Matching Pursuit (MMSE) is a greedy iterative algorithm that selects the atom that best matches the current residual and updates the residual using orthogonal projection, thus gradually approximating the sparse representation of the variance vector and finally obtaining the channel estimate through MMSE. The Graph Cut method estimates the wavenumber domain channel variance based on graph theory, but this method cannot estimate the noise variance. For fair comparison, all compared schemes use the actual noise variance in MMSE estimation. The lower bound is the channel estimate obtained after performing MMSE on the channel with known wavenumber domain channel variance and noise variance. It is clear from the figure that the proposed scheme outperforms both the Orthogonal Matching Pursuit and Graph Cut methods, and the performance of the proposed scheme is extremely close to the lower bound. Furthermore, since frequent inversion is not required, the computational complexity of the proposed scheme is much lower than that of the Orthogonal Matching Pursuit.

[0164] Finally, it should be noted that the above descriptions are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A wavenumber domain channel sensing method based on an asymmetric architecture in ultra-large-scale MIMO, characterized in that, Includes the following steps: S100, constructing an ultra-large-scale MIMO system consisting of one base station and one single-antenna user; wherein, the base station is equipped with a... A uniform linear array composed of antennas, and simultaneously within the uniform linear array... All antennas are connected to the downlink RF chain, and in Selecting from root antennas The root antenna is connected to the uplink RF chain; S200. Based on the differential co-array completeness criterion, the uplink antenna position is determined using a coprime array, and the virtual array signal is extracted from the autocorrelation matrix of the uplink received signal under the antenna configuration. S300. Based on the extracted virtual array signal, a method combining the alternating direction multiplier method improved by Fused Lasso is designed to estimate the channel variance and noise variance in the wavenumber domain. S400. Based on the estimated channel variance in the wavenumber domain, the support set is extracted. Then, based on the mapping relationship between wavenumber direction and angle, the arrival angle information, including mode direction and angle spread, is inferred from the wavenumber direction in the support set. S500: Based on the estimated variance and wavenumber domain codebook, perform MMSE on the uplink received signal to estimate the complex path gain and further obtain an estimate of the channel. S600. Based on the idea of ​​solving triangles, design a two-stage distance estimation scheme: In the first stage, the first reference point is set as the first antenna, and a coprime array is constructed according to the antenna index from small to large. The first angle estimation information is obtained by estimating the variance. In the second stage, the second reference point is set as the Mth antenna, and a coprime array is constructed from large to small according to the antenna lines. The second angle estimation information is obtained by estimating the variance. Based on the first angle estimation information and the second angle estimation information, as well as the distance between the first reference point and the second reference point, the distance from the base station to the scattering cluster is calculated. Finally, wavenumber domain channel sensing for ultra-large-scale MIMO is realized.

2. The wavenumber domain channel sensing method based on an asymmetric architecture in ultra-large-scale MIMO as described in claim 1, characterized in that, Step S200 specifically includes: S201, to Perform prime factorization to obtain the first prime number. and the second prime number You can choose any of them. and The combination only requires... ,and ; S202. Based on the selected first and second prime numbers, construct a coprime array in ascending order of antenna indices: select antennas from the first to the second... The first antenna is selected as the first subarray, and the antenna indices in the first subarray are consecutive; the first antenna is selected as the second antenna. root antenna to the first The first antenna is used as the second subarray, and the index difference between adjacent antennas in the second subarray is 1. The first and second subarrays are coprime arrays, and the total number of uplink antennas satisfies the following condition: ; S203, Based on the uplink received signal obtained in the coprime array ,pass The autocorrelation matrix is ​​obtained from the second snapshot. The expression is: ; S204. Extract and process the elements at the positions of the non-negative delay terms in the matrix: If multiple elements in the matrix correspond to the same non-negative delay term, the average value is taken; if only one element in the matrix corresponds to a certain non-negative delay term, no processing is required; for the processed elements, they are arranged in ascending order of delay term index to obtain the virtual array signal. .

3. The wavenumber domain channel sensing method based on an asymmetric architecture in ultra-large-scale MIMO as described in claim 1, characterized in that, Step S300 specifically includes: S301, Obtain virtual array signal dictionary matrix and preset regularization parameters , and penalty factor The dictionary matrix A has L+1 columns of vectors, with the last column being a vector. ,forward Listed as wavenumber field codebook , Wavenumber direction index set Total number of elements in the middle; wavenumber direction index set , will set The elements in the array are arranged in ascending order; among them, For wavelength, The aperture length of the ULA; S302, Based on the virtual array signal The quadratic matrix is ​​calculated from the dictionary matrix A. and gradient vector Construct a first-order difference matrix Its elements satisfy , The rest are zero; thus generating the projection matrix. and ;in, yes Unit matrix of order, yes A zero-dimensional vector; S303, Calculate the matrix And find the inverse; in Indicates the penalty parameter; S304. Initialization: Set the auxiliary variables... , , and dual variables , , Set the zero vector, and set the maximum number of iterations and convergence tolerance; S305, Entering the iteration, for the... Next iteration: Construct vectors: And update the variance estimate vector. The iterative expression is obtained as follows: ; Where d represents the gradient vector. The parameters represent the penalty term, and P and Q represent the projection matrices. , , Describes the auxiliary variable for the k-th iteration. , , Let U represent the dual variable of the k-th iteration, and let U represent the matrix calculated in step S303. Update the first auxiliary variable The iterative expression is: ; Update the second auxiliary variable The iterative expression is: ; in, This represents the preset regularization parameters; Update the third auxiliary variable The iterative expression is: ； in, This represents the preset regularization parameters; Update the first auxiliary variable The iterative expression is: ; Update the second auxiliary variable The iterative expression is: ; Update the third auxiliary variable The iterative expression is: ; Calculate the original residuals and dual residuals If all values ​​are less than the convergence tolerance or the maximum number of iterations is reached, the iteration is terminated; otherwise, the iteration continues. The expression for calculating the original residual is as follows: ； The expression for calculating the dual residual is: ; S306. After the iteration is completed, the variance estimation vector is... The final output, after applying nonnegativity truncation, is expressed as: .

4. The wavenumber domain channel sensing method based on an asymmetric architecture in ultra-large-scale MIMO as described in claim 1, characterized in that, Step S400 specifically includes: S401, Extracting the variance estimation vector The former Use elements to construct a temporary vector. ; in descending order, process the temporary vectors Arrange all elements in order to obtain a sorted vector. and its corresponding set of original index mappings ; S402, Calculate the sorting vector forward The sum of the terms ,in ; S403, Obtain the threshold Calculate the absolute energy threshold Then find the smallest integer. , making ; S404, Extract Set The front of the middle Each element serves as the support set. The estimated wavenumber domain channel variance is obtained. The elements satisfy: ; S405, For sets any index in Calculate the corresponding wavenumber direction For the support set any index in The corresponding wavenumber direction is obtained. ; S406, Based on Free Space Wavenumber , will support the collection The corresponding wavenumber direction is converted into an angle. Generate angle estimation vector The elements satisfy: ; S407, Estimation of angular extension and modal orientation: Identify angle estimation vector For any consecutive index segment with non-zero values, Extract its starting index and end index Calculate continuous index segments The corresponding angle extension is expressed as: ; Calculate contiguous index segments The corresponding modal direction is expressed as: ； The continuous index segment The range of values ​​for depends on the number of scattering clusters in the environment; assuming there are a total of scattering clusters in the environment. If there are scattering clusters, then The total number of consecutive index segments is also ,have .

5. The wavenumber domain channel sensing method based on an asymmetric architecture in ultra-large-scale MIMO as described in claim 1, characterized in that, Step S500 specifically includes: S501, Extracting the variance estimation vector The last element yields the noise estimate. ; S502. Constructing a matrix of known information , where the matrix ; S503. Perform MMSE on the uplink received signal to obtain the complex path gain estimate. The specific expression is: ; S504. Calculate downlink channel estimation The expression is: .

6. The wavenumber domain channel sensing method based on an asymmetric architecture in ultra-large-scale MIMO as described in claim 1, characterized in that, Step S600 specifically includes: S601, First-stage angle estimation: Treat the current coprime array as the first array and estimate the current angle. As the result of the first stage angle estimation The first array sets the first antenna as the first reference point and constructs a coprime array according to the antenna index from smallest to largest. Identify the first-stage angle estimation results For any consecutive index segment with non-zero values, Extract its starting index and end index ; And calculate continuous index segments Corresponding angle extension: And modal orientation: ; S602, Second Stage Angle Estimation: Reconstruct the coprime array as the second array; the second array will... The first antenna is set as the second reference point, and a coprime array is constructed according to the antenna index from largest to smallest; Repeat steps S200 and S300 to obtain the second-stage angle estimation results. ; Identify the second-stage angle estimation results For any consecutive index segment with non-zero values, Extract its starting index and end index Calculate continuous index segments Corresponding angle extension and modal direction ; S603, Estimated Distance Information: Calculate the distance between the first reference point and the second reference point. ; Considering the generalized stationarity of the physical propagation environment, the two-stage estimation results and The total number of consecutive index segments is ,have At the same time, when At that time, these two consecutive index segments describe the same scattering cluster; based on the angle estimation results of the two stages. and and spacing Calculate the distance from the base station to the... Distance of each scattering cluster The expression is: ； At this point, the base station achieved wavenumber domain channel awareness for ultra-large-scale MIMO.

7. The wavenumber domain channel sensing method based on an asymmetric architecture in ultra-large-scale MIMO as described in claim 1, characterized in that, In step S400, after calculating the angular spread, the base station reports the angular spread of the scattering cluster in the format of the 3GPP Integrated Sensing and Communication (ISAC) measurement level. The reported content includes a newly added optional field for the measurement level.

8. The wavenumber domain channel sensing method based on an asymmetric architecture in ultra-large-scale MIMO as described in claim 7, characterized in that, In step S600, after calculating the distance between itself and the scattering cluster, the base station reports the distance between itself and the scattering cluster in the format of 3GPP ISAC measurement level. The reported content includes a newly added optional field for the measurement level.

Images