Channel estimation and blocking diagnosis method for RIS-assisted MIMO-OFDM system

Abstract

The invention provides a channel estimation and blocking diagnosis method for an RIS (Radio Information System)-assisted MIMO-OFDM (Multiple Input Multiple Output-Orthogonal Frequency Division Multiplexing) system, which comprises the following steps of: constructing an RIS-assisted millimeter wave MIMO-OFDM communication system, and blocking a part of RIS elements in the system; designing a frame-based structured training protocol, and modeling a received signal into a low-rank fourth-order tensor; analyzing uniqueness of tensor decomposition, and analyzing conditions which need to be met by a time frame number and a pilot frequency number; decomposing the received signal by using a non-iterative algorithm so as to obtain a factor matrix which is not influenced by a blocking vector and a coupling factor matrix containing residual channel parameters and the blocking vector; and channel parameters and RIS blocking vectors related to all factor matrixes are jointly estimated by using the sparse characteristic of the millimeter wave channel. According to the method, the problems of channel estimation and blocking diagnosis when part of RIS is blocked are solved, RIS blocking and channel parameters can be estimated, and then a cascade channel is recovered through the channel parameters.

Description

Technical Field

[0001] The present invention relates to the field of communication technology, and in particular to a channel estimation and blocking diagnosis method for a RIS-assisted MIMO-OFDM system. Background Art

[0002] Millimeter-wave (mmWave) communications are considered a key technology for next-generation cellular networks, leveraging abundant spectrum resources to deliver gigabit data rates. However, their practical deployment faces significant challenges, primarily due to severe path loss and susceptibility to environmental obstructions. Reconfigurable Intelligent Surfaces (RIS) are a promising solution. They consist of programmable passive components that dynamically alter electromagnetic wave propagation, potentially improving the coverage and energy efficiency of mmWave networks. However, RIS-assisted communications rely on accurate channel state information (CSI), which is crucial for effective beamforming design. Unlike active relays, passive RIS have limited signal processing capabilities and therefore rely on base stations (BSs) to estimate CSI. Furthermore, outdoor RIS are susceptible to interference from physical obstacles and weather factors such as precipitation, resulting in unpredictable phase and amplitude distortion, which compromises beamforming performance. Therefore, it is essential to jointly estimate CSI and diagnose the obstruction status of the RIS.

[0003] Due to the sparse nature of millimeter waves, the CSI estimation problem can be transformed into a finite set of channel parameter estimation problems. By estimating these channel parameters, the channel can be completely reconstructed. As for the diagnosis of RIS blocking, since the number of RIS blocking is generally small, the blocking diagnosis problem can be transformed into a sparse vector recovery problem. In the problem of jointly estimating CSI and diagnosing RIS blocking status, the focus of existing research is mainly on narrowband systems. In contrast, millimeter wave networks usually operate on wideband channels with frequency selectivity. For example, in Orthogonal Frequency Division Multiplexing (OFDM) systems, delay spread can cause a single RIS blocking to affect multiple subcarriers simultaneously. Therefore, how to take advantage of the key point that multiple carriers are affected by the same RIS blocking in OFDM systems to jointly improve the accuracy of channel estimation and blocking diagnosis is an urgent problem to be solved. Summary of the Invention

[0004] To address the shortcomings of the existing technology, the present invention explores the problems of cascaded channel estimation and blockage diagnosis in a passive RIS-assisted millimeter-wave multiple-input multiple-output (MIMO)-OFDM system when some RIS elements are blocked. The present invention can simultaneously estimate RIS blockage and channel parameters, and then restore the cascaded channel based on the channel parameters.

[0005] In order to achieve the above object, the technical solution of the present invention is achieved as follows:

[0006] A RIS-assisted channel estimation and blocking diagnosis method for a MIMO-OFDM system, comprising the following steps:

[0007] S1: Construct a RIS-assisted mmWave MIMO-OFDM communication system in which some RIS components are blocked;

[0008] S2: Design a frame-based structured training protocol to model the received signal as a low-rank fourth-order tensor and fully exploit the structural and coupling information of the low-rank fourth-order tensor;

[0009] S3: Before tensor decomposition, the uniqueness of the tensor decomposition is first analyzed, and the conditions that the number of time frames and the number of pilots must meet are analyzed;

[0010] S4: Decompose the received signal using a non-iterative algorithm to obtain a factor matrix that is not affected by the blocking vector and a coupling factor matrix that includes the remaining channel parameters and the blocking vector;

[0011] S5: For the factor matrix that is not affected by the blocking vector, the sparse characteristics of the millimeter wave channel are used to estimate some channel parameters related to the factor matrix;

[0012] S6: For the coupling factor matrix containing the remaining channel parameters and the blocking vector, the remaining channel parameters and the RIS blocking vector are jointly estimated by utilizing the sparsity of millimeter waves and blocking.

[0013] Preferably, the method for constructing the RIS-assisted millimeter wave MIMO-OFDM communication system is: establishing a passive RIS-assisted millimeter wave MIMO-OFDM communication system in an uplink scenario, in which the BS is equipped with M BS Antenna and N BS RF chain, and UE is equipped with M UE Antenna and N UE RF chain; The antenna structure of BS and UE both adopts uniform linear array (ULA), and RIS is composed of uniform planar array (UPA) of N passive elements; RIS has N in the horizontal direction x There are N elements in the vertical direction ycomponents, so the total number of components is N = N x ×N y ; The system uses a total of P0 OFDM subcarriers, of which P OFDM subcarriers are used for training.

[0014] Preferably, in the passive RIS-assisted millimeter wave MIMO-OFDM communication system, the Saleh-Valenzuela channel model is used to describe the UE-RIS channel and the BS-RIS channel;

[0015] Specifically, the definition of the UE-RIS channel in the delay domain is as follows:

[0016]

[0017] Where δ(τ) represents the Dirac-delta function, L1 represents the number of paths between UE and RIS, represents the path gain between UE and RIS, represents the azimuth in the RIS angle of arrival (AoA), represents the elevation angle in the RIS arrival angle, represents the angle of departure (AoD) at the UE, ι m represents the propagation delay;

[0018] Similarly, the RIS-BS channel is defined in the delay domain as follows:

[0019]

[0020] Where L2 represents the number of paths between RIS and BS, represents the path gain between RIS and BS, Indicates the azimuth in the RIS departure angle, Indicates the elevation angle in the RIS departure angle, represents the AoA of BS, κ n represents the propagation delay; the array response vector for ULA and UPA configurations is given by the following formula: for The transpose of , where X∈{UE,BS}; the array response at RIS is expressed as: in

[0021] After obtaining the delay domain channel model, the UE-RIS and RIS-BS channels under the p-th subcarrier in the frequency domain are expressed as follows:

[0022]

[0023] Among them, fs Indicates the sampling rate;

[0024] Due to the passive nature of RIS components, R p and G p is not feasible, therefore, it is necessary to estimate the cascade channel The cascaded channel is represented as:

[0025]

[0026] in, represents the Kronecker product, ⊙ represents the Khatri-Rao product, and the mapping process (a) is:

[0027]

[0028] in, represents the upper limit function, l=1,…,L1L2, From formula (4), it can be seen that in the case of multiple elements, H p Is a high-dimensional matrix, by estimating the parameters Completely reconstruct channel H p .

[0029] Preferably, the specific implementation method of step S2 is:

[0030] With a frame-based uplink training protocol, the channel estimation process for each subcarrier is performed in Q consecutive time frames, where each frame consists of T time slots. In this setup, the UE uses T different beamforming vectors in T time slots, while the RIS applies a specific phase shift vector s in the qth time slot. q to reflect the incident signal;

[0031] The transmission signal of the p-th subcarrier in the t-th time slot is expressed as:

[0032]

[0033] in, is the pilot symbol vector of the p-th subcarrier, is the baseband precoding matrix associated with the p-th subcarrier, is the RF precoder shared by all subcarriers; for all subcarriers, in the tth time slot, assuming F BB,t,p =F BB,t , m t,p =m t , in this case, the transmitted signal associated with the p-th subcarrier in the t-th time slot is: represents the hybrid combiner used by BS, where It is a common RF combiner for all subcarriers. Denote the baseband combiner associated with the p-th subcarrier, assuming the hybrid combiner remains unchanged:

[0034] Since RIS components are easily affected by environmental obstacles such as rain and snow, a RIS blocking vector e is introduced, where the nth element e n The characteristics are as follows:

[0035]

[0036] in, μ n and ω n They represent the amplitude attenuation and phase distortion caused by the blocking of the nth RIS element, respectively. Due to the frequency-flatness of RIS blocking, the blocking vectors on all subcarriers remain consistent. Therefore, during the qth time frame, the tth time slot, and the pth subcarrier, the signal received by the BS can be expressed as:

[0037]

[0038] Substituting (4) into (8), the received signal model is:

[0039]

[0040] As shown in (9), the channel parameters exhibit different blocking dependencies: Affected by the distortion caused by blocking, It has the characteristic of not being affected by blockage;

[0041] A two-stage estimation framework is designed to separate these parameters that are not affected by blocking from the blocking vector; specifically, in the first stage, the focus is on the blocking-invariant parameters The coupled observation matrix is ​​then derived based on the accurate extraction of To restore the cascade channel;

[0042] Considering the two key factors of the frequency flatness of the blocking vector in OFDM subcarriers and channel blocking decoupling, CPD is used as the main data processing tool; by taking advantage of the two inherent characteristics of the millimeter wave system: sparse multipath scattering and low-rank signal subspace structure, the received signal y is obtained by multi-dimensionally stacking Q time frames, T time slots and P subcarriers. q,t,p To construct a fourth-order tensor

[0043] By summing up the T time slot signals received at the BS, the composite received signal vector is defined Then the received signal at the p-th subcarrier in the q-th time frame is:

[0044]

[0045] in, By performing column union on the time slot observation data and converting the received signal y q,p Converted to N s ×T-dimensional matrix, the following signal model is obtained:

[0046]

[0047] Among them, N q,p Indicated by n q,p Transformed N s ×T matrix; the received signal from the BS is collected over Q time frames and a third-order tensor is constructed by stacking in This tensor representation is convenient for multi-dimensional signal processing. The composite received signal on the p-th subcarrier is expressed in tensor form as:

[0048]

[0049] in, Indicates stack N q,p The noise tensor formed, in is full column rank, define By summing the received signals from all training subcarriers (Equation (12)), the complete training signal can be represented as a fourth-order tensor in It follows the CPD format, which is:

[0050]

[0051] in, represents the noise tensor and The four factor matrices of are defined as follows:

[0052]

[0053] From formula (13) and formula (14), it can be seen that B, U and C are respectively determined by the channel parameters and Definition, where C is a Vandermonde structure; in addition, the factor matrix R contains the coupling parameters The corresponding parameters of each factor matrix can be estimated and then used to reconstruct the cascade channel matrix and estimate the blocking vector.

[0054] Preferably, the implementation method of step S3 is:

[0055] Regarding formula (13) The uniqueness analysis of CPD is generally determined by the Kruskal condition, that is:

[0056] k(B)+k(U)+k(R)+k(C)≥2L+3, (15);

[0057] where k(B) represents the k-rank of the matrix B. However, according to formulas (5) and (14), when L1 = 1, L2 = 1 and L1 ≥ 2, L2 ≥ 2, k(B) = 1 and k(U) = 1 do not satisfy formula (15). Considering the inherent Vandermonde structure in C, a more relaxed uniqueness criterion is established:

[0058] The tensor model in the noise-free case is The factor matrices are denoted as B, U, R, and C; if the following conditions hold:

[0059]

[0060] Then, the tensor The CPD of can be uniquely determined; r(R) represents the rank of the matrix R; Formula (16) is further expressed as r( C )=L and r(R)=L, and the number of subcarriers P and the number of time frames Q are greater than the total number of paths L, and and The parameters in these two sets are different.

[0061] Preferably, in step S4, the method for obtaining the factor matrix not affected by the blocking vector and the coupling factor matrix including the remaining channel parameters and the blocking vector is:

[0062] Using the inherent Vandermonde structure of the factor matrix C and assuming a noise-free case, the mode-2 expansion of the tensor N is considered as: right Do SVD decomposition:

[0063]

[0064] Ignoring the noise, we can find that there exists a non-singular matrix M such that U Y M=C⊙U⊙B; from this we can get:

[0065]

[0066] in, means removing the first row of matrix C, C It means removing the last row of matrix C. Using the Vandermonde structure of C, we can get:

[0067]

[0068] That is to say By performing eigenvalue decomposition, we can get the estimated values ​​of Z and M. and Then each column of the factor matrix C can be estimated as:

[0069]

[0070] in, is a matrix The lth element on the diagonal of byU Y M=C⊙U⊙B, each column of the estimated value of X is:

[0071]

[0072] After the factor matrices C and X are obtained, the estimated value of the factor matrix R is:

[0073]

[0074] The factor matrix estimated by non-iterative method is in and Does not contain blocking vectors, blocking vector e is only in the factor matrix The relationship between the true factor matrix and its estimated value can be expressed as:

[0075]

[0076] Where Ψ1 and Ψ2 are diagonal non-singular matrices satisfying Ψ1Ψ2=I, Ω represents the permutation matrix shared by all factor matrices, and {E1,E2,E3} represents the estimation error term.

[0077] Preferably, the implementation method of step S5 is:

[0078] As shown in formula (23), the estimated Compared with the true factor matrix C, there is no scale ambiguity, so it can be estimated directly; using the Vandermonde structure of C, it is estimated that The delay parameters are related to Therefore, the delay parameter can be estimated as:

[0079]

[0080] in, Indicates plural The actual parameter of the factor matrix X; each column of which is represented by the array response corresponding to the angle at the UE and the angle at the BS, so the angles at the BS and UE can be estimated by the correlation method, the purpose of which is to maximize the correlation between the estimated vector and the parameter vector; specifically, Rearrange to N s ×T matrix, where You can get:

[0081]

[0082] The rank-one matrix factorization algorithm can be used to estimate B and U; Performing truncated singular value decomposition, we can get:

[0083]

[0084] and The first column can be u x and v x To linearly represent, therefore, the angle at BS is estimated using normalized correlation maximization:

[0085]

[0086] in, Similarly, the angle at the UE can be estimated as:

[0087]

[0088] Therefore, the scale fuzziness of the factor matrix X is:

[0089]

[0090] in, The scale fuzzy of the factor matrix R can be obtained as:

[0091] Preferably, the implementation method of step S6 is:

[0092] For the factor matrix Each column is represented by the array response and blockage corresponding to the angle at RIS. The lth column of the factor matrix R can be expressed as:

[0093]

[0094] Based on formula (30), since the residual channel parameters are coupled with the blocking vector, a joint estimation framework is required to alternately optimize the angular domain and the blocking domain;

[0095] First, the RIS spatial angle is estimated by a normalized correlation maximization method:

[0096]

[0097] in, After obtaining the cascade angle at the RIS, reconstruct a RIS array response matrix: definition is represented as a sparse blocking vector, and can be expressed as Without noise influence So the path gain can be calculated as:

[0098]

[0099] use can be re-expressed as:

[0100]

[0101] Among them, N R Represents the estimation error, and the column transformation on both sides of Equation (33) is:

[0102]

[0103] Among them, n r =vec(N R ); in addition, define:

[0104]

[0105] Therefore, formula (34) can be transformed into:

[0106] r=Φk+n r . (36);

[0107] Therefore, the problem of solving the blocking vector is expressed as a sparse signal recovery problem; the sparse Bayesian algorithm is selected to solve it. In this case, assuming that the noise n r Obey complex Gaussian distribution β follows the gamma distribution Γ(β; c, d); the sparse vector k follows the complex Gaussian distribution Where Λ = diag(α), α follows the gamma distribution Γ(α; 1, ρ); the posterior probability distribution of the sparse vector k is:

[0108]

[0109] The mean and variance of the posterior probability distribution of the sparse vector k are:

[0110]

[0111] The update formula for hyperparameters α and β is:

[0112]

[0113] Preferably, a threshold-based detection rule is designed, which is expressed as follows:

[0114]

[0115] Among them, the predefined threshold δ determines the sparsity level.

[0116] Beneficial effects of the present invention:

[0117] 1) By designing a pilot frame structure and leveraging the high-dimensional nature of the received signal, the present invention stacks the received signal into a fourth-order tensor and constrains the blocking vector to a single factor matrix. Leveraging the Vandermonde structure of the partial factor matrix, the delay parameter, the transceiver angle parameter, and the blocking vector are decoupled, making these parameters unaffected by the blocking during the first partial channel estimation. Once the delay parameter and the transceiver angle parameter are obtained, the scale ambiguity introduced by the tensor's unique decomposition can be eliminated.

[0118] 2) For the estimation of the remaining channel parameters and the problem of blocking diagnosis, taking into account the strong coupling between these parameters, the present invention utilizes the inherent characteristics of the channel and the blocking and adopts an alternating iterative solution to solve it. Specifically, the structured characteristics of the channel are utilized to transform the parameter estimation problem into a problem of maximizing the normalized correlation, and the sparsity of the blocking is utilized to transform the blocking estimation problem into a sparse vector recovery problem. To this end, the present invention utilizes a hierarchical Bayesian model based on the Gaussian-inverseGamma prior distribution and applies sparse Bayesian learning technology to recover the blocking vector. This method can simultaneously recover the remaining channel parameters and the blocking vector. The present invention solves the channel estimation and blocking diagnosis problems when part of the RIS is blocked, and can make full use of the common characteristics of blocking in multiple subcarriers. While obtaining the CSI, the RIS blocking coefficient can be obtained at the same time. BRIEF DESCRIPTION OF THE DRAWINGS

[0119] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0120] Figure 1 Schematic diagram of the pilot training protocol based on time frame of the present invention.

[0121] Figure 2 Flowchart of the method of the present invention.

[0122] Figure 3 The number of time slots T is 6, the number of time frames Q is 16, the number of subcarriers P is 6, and the number of RIS blocking N is B Schematic diagram of the variation of NMSE with SNR for channel estimation and blocking diagnosis of the present invention and other algorithms when SNR is 6.

[0123] Figure 4 When the number of time slots T is 6, the number of time frames Q is 16, the number of subcarriers P is 6, and the SNR is 20, the NMSE of the blocking diagnosis and channel estimation decreases with the number of blockers (N B ) diagram of the changes.

[0124] Figure 5 The number of time slots T is 6, the number of subcarriers P is 6, and the number of RIS blocking is N. B Schematic diagram of the NMSE of blocking diagnosis and channel estimation changing with the number of time frames (Q) when the NMSE is 6 and the SNR is 20.

[0125] Figure 6 The number of time frames Q is 16, the number of subcarriers P is 6, and the number of RIS blocking is N. B Schematic diagram of the NMSE of blocking diagnosis and channel estimation changing with the number of slots (T) when the NMSE is 6 and the SNR is 20.

[0126] Figure 7 The number of time slots T is 6, the number of time frames Q is 16, and the number of RIS blocking is N. B Schematic diagram of the NMSE of blocking diagnosis and channel estimation changing with the number of subcarriers (P) when the NMSE is 6 and the SNR is 20.

[0127] Figure 8 The number of time slots T is 3, the number of time frames Q is 48, and the number of RIS blocking is N. B 6, RIS blocking number N B Schematic diagram of the variation of channel estimation and blocking diagnosis NMSE with SNR under different path numbers when the number of paths is 6. DETAILED DESCRIPTION

[0128] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without creative work are within the scope of protection of the present invention.

[0129] like Figure 2As shown, the embodiment of the present invention provides a RIS-assisted channel estimation and blocking diagnosis method for a MIMO-OFDM system, and the specific steps are as follows:

[0130] S1: Construct a RIS-assisted millimeter-wave MIMO-OFDM communication system in which some RIS elements are blocked. Both the base station and user equipment (UE) are equipped with hybrid precoding structures, and the antenna layout uses linear arrays, while the RIS elements are arranged in uniform planar arrays. Assume that the direct link between the UE and the base station is blocked.

[0131] In this embodiment, the specific process of step S1 is as follows:

[0132] A passive RIS-assisted millimeter-wave MIMO-OFDM communication system in the uplink scenario is established, in which the BS is equipped with M BS Antenna and N BS RF chain, and UE is equipped with M UE Antenna and N UE Radio Frequency (RF) chain. The antenna structure of BS and UE both adopts Uniform Linear Array (ULA). RIS is composed of N uniform planar arrays (UPA) of passive elements. RIS has N horizontal x There are N elements in the vertical direction y components, so the total number of components is N = N x ×N y The system uses a total of P0 OFDM subcarriers, of which P OFDM subcarriers are used for training.

[0133] The present invention uses the Saleh-Valenzuela channel model to describe the UE-RIS channel and the BS-RIS channel. Specifically, the UE-RIS channel is defined in the delay domain as follows:

[0134]

[0135] Where δ(τ) represents the Dirac-delta function, L1 represents the number of paths between UE and RIS, represents the path gain between UE and RIS, represents the azimuth angle in the RIS angle of arrival (Angles of Arrival, AoA), represents the elevation angle in the RIS arrival angle, represents the angle of departure (AoD) at the UE, mRepresents the propagation delay.

[0136] Similarly, the RIS-BS channel is defined in the delay domain as follows:

[0137]

[0138] Where L2 represents the number of paths between RIS and BS, represents the path gain between RIS and BS, Indicates the azimuth in the RIS departure angle, Indicates the elevation angle in the RIS departure angle, represents the AoA of BS, κ n represents the propagation delay. The array response vector for ULA and UPA configurations is given by the following formula: for The transpose of , where X∈{UE,BS}; the array response at RIS is expressed as: in

[0139] After obtaining the delay domain channel model, the UE-RIS and RIS-BS channels under the p-th subcarrier in the frequency domain are expressed as follows:

[0140]

[0141] Among them, f s Indicates the sampling rate; To design an accurate RIS phase shift matrix to maximize spectral efficiency, it is necessary to accurately understand the channel R p and G p However, due to the passive nature of RIS components, estimating R p and G p Therefore, it is necessary to estimate the cascade channel This is sufficient to optimize the RIS phase shift. The cascaded channel is expressed as:

[0142]

[0143] in, represents the Kronecker product, ⊙ represents the Khatri-Rao product, and the mapping process (a) is:

[0144]

[0145] in, represents the upper limit function, which gives the smallest integer greater than or equal to a, l=1,…,L1L2, From formula (4), it can be seen that in the case of multiple elements, Hp Is a high-dimensional matrix, directly estimate H p It will generate a lot of training overhead, but its inherent structural characteristics can be used to reduce this overhead. Analysis shows that the high-dimensional channel estimation problem can be simplified to estimating a limited set of parameters: By estimating these parameters, the channel H is fully reconstructed p .

[0146] S2: Design a frame-based structured training protocol to model the received signal as a low-rank fourth-order tensor and fully exploit the structural and coupling information of the low-rank fourth-order tensor;

[0147] With a frame-based uplink training protocol, the channel estimation process for each subcarrier is performed in Q consecutive time frames, where each frame consists of T time slots. In this setup, the UE uses T different beamforming vectors in T time slots, while the RIS applies a specific phase shift vector s in the qth time slot. q To reflect the incident signal. Figure 1 .

[0148] The transmission signal of the p-th subcarrier in the t-th time slot is expressed as:

[0149]

[0150] in, is the pilot symbol vector of the p-th subcarrier, is the baseband precoding matrix associated with the p-th subcarrier, is the RF precoder shared by all subcarriers; for all subcarriers, in the tth time slot, assuming F BB,t,p =F BB,t , m t,p =m t , in this case, the transmitted signal associated with the p-th subcarrier in the t-th time slot is: represents the hybrid combiner used by BS, where It is a common RF combiner for all subcarriers. Denote the baseband combiner associated with the p-th subcarrier, assuming the hybrid combiner remains unchanged:

[0151] It is worth noting that RIS elements are easily affected by environmental obstacles such as rain and snow, which can cause absorption and scattering effects on millimeter wave signals. For this reason, the present invention introduces the RIS blocking vector e, where the nth element e n The characteristics are as follows:

[0152]

[0153] in, μ n and ω n They represent the amplitude attenuation and phase distortion caused by the blocking of the nth RIS element. Due to the frequency flatness of RIS blocking, that is, the blocking vectors on all subcarriers remain consistent; therefore, during the qth time frame, tth time slot and pth subcarrier, the signal received by the BS can be expressed as:

[0154]

[0155] Substituting (4) into (8), the received signal model is:

[0156]

[0157] As shown in (9), the channel parameters exhibit different blocking dependencies: Affected by the distortion caused by blocking, This complex coupling poses a great challenge to estimating the effective channel parameters from the received training signal.

[0158] This paper will focus on designing a two-stage estimation framework to separate these parameters that are not affected by blocking from the blocking vector. Specifically: in the first stage, we will focus on the blocking-invariant parameters The coupled observation matrix is ​​then derived based on the accurate extraction of To restore the cascade channel.

[0159] Considering the two key factors of the frequency flatness of the blocking vector in OFDM subcarriers and the channel blocking decoupling, CPD is used as the main data processing tool. By taking advantage of the two inherent characteristics of the millimeter wave system: sparse multipath scattering and low-rank signal subspace structure, the received signal y is multi-dimensionally stacked with Q time frames, T time slots and P subcarriers. q,t,p To construct a fourth-order tensor

[0160] By summing up the T time slot signals received at the BS, the composite received signal vector is defined Then the received signal at the p-th subcarrier in the q-th time frame is:

[0161]

[0162] in, By performing column union on the time slot observation data and converting the received signal y q,p Converted to N s ×T-dimensional matrix, the following signal model is obtained:

[0163]

[0164] Among them, N q,p Indicated by n q,p Transformed N s ×T matrix; the received signal from the BS is collected over Q time frames and a third-order tensor is constructed by stacking in This tensor representation is convenient for multi-dimensional signal processing. The composite received signal on the p-th subcarrier is expressed in tensor form as:

[0165]

[0166] in, Indicates stack N q,p The noise tensor formed, in is full column rank, define By summing the received signals from all training subcarriers (Equation (12)), the complete training signal can be represented as a fourth-order tensor in It follows the CPD format, which is:

[0167]

[0168] in, represents the noise tensor and The four factor matrices of are defined as follows:

[0169]

[0170] From formula (13) and formula (14), it can be seen that B, U and C are respectively determined by the channel parameters and Definition, where C is a Vandermonde structure; in addition, the factor matrix R contains the coupling parameters Each factor matrix can have its corresponding parameters estimated and then used to reconstruct the concatenated channel matrix and estimate the blocking vector, but this simplification requires a comprehensive analysis of the unique conditions required for CPD.

[0171] S3: Before tensor decomposition, the uniqueness of the tensor decomposition is analyzed, including the conditions that must be met for the number of time frames and pilots. The accuracy of the tensor decomposition requires that the tensor decomposition be unique. The general condition for tensor CPD is to satisfy the Kruskal condition, but this condition cannot be met in some cases in this invention. Considering the Vandermonde structure of one of the factor matrices, a more relaxed uniqueness criterion can be established.

[0172] Regarding formula (13) The uniqueness analysis of CPD is generally determined by the Kruskal condition, that is:

[0173] k(B)+k(U)+k(R)+k(C)≥2L+3, (15);

[0174] where k(B) represents the k-rank of the matrix B. However, according to equations (5) and (14), when L1 = 1, L2 = 1 and L1 ≥ 2, L2 ≥ 2, k(B) = 1 and k(U) = 1 do not satisfy equation (15). Considering the inherent Vandermonde structure in C, a more relaxed uniqueness criterion is established.

[0175] The tensor model in the noise-free case is The factor matrices are denoted as B, U, R, and C. If the following conditions hold:

[0176]

[0177] Then, the tensor The CPD of can be uniquely determined, excluding the permutation and scaling ambiguities. r(R) represents the rank of the matrix R; Formula (16) is further expressed as r( C )=L and r(R)=L, which means that the number of subcarriers P and the number of time frames Q are greater than the total number of paths L, and and The parameters in the two sets are different. Since these parameters are randomly generated, the probability of this condition being true is 1. Compared with the classic Kruskal condition, this uniqueness criterion is less restrictive.

[0178] S4: A non-iterative algorithm is used to decompose the received signal to obtain a factor matrix unaffected by the blocking vector and a coupling factor matrix containing the remaining channel parameters and the blocking vector. Utilizing the inherent sparsity of the millimeter wave channel and the common sparsity of RIS blocking, the present invention models the received signal as a fourth-order tensor that obeys CPD. In this framework, the blocking vector is located only in a single factor matrix, and the channel parameters coupled to it are the cascade direction angle, elevation angle, and path gain at the RIS. The other three factor matrices, consisting of the delay term, the angle at the BS, and the angle at the UE, are not affected by blocking.

[0179] For general CPD problems, the Alternating Least Squares (ALS) method is usually used to estimate the factor matrix. However, the ALS algorithm also has some disadvantages: it requires multiple iterations, is prone to local convergence, is sensitive to random initialization, has a slow convergence speed, and has high computational complexity. A non-iterative decomposition algorithm is used to reduce the computational complexity.

[0180] Using the inherent Vandermonde structure of the factor matrix C and assuming a noise-free case, we consider the tensor The mode-2 is expanded to: right Do SVD decomposition:

[0181]

[0182] Ignoring the noise, we can find that there exists a non-singular matrix M such that U Y M=C⊙U⊙B; from this we can get:

[0183]

[0184] in, means removing the first row of matrix C, C It means removing the last row of matrix C. Using the Vandermonde structure of C, we can get:

[0185]

[0186] That is to say By performing eigenvalue decomposition, we can get the estimated values ​​of Z and M. and Then each column of the factor matrix C can be estimated as:

[0187]

[0188] in, is a matrix The lth element on the diagonal of byU Y M=C⊙U⊙B, each column of the estimated value of X is:

[0189]

[0190] After the factor matrices C and X are obtained, the estimated value of the factor matrix R is:

[0191]

[0192] The factor matrix estimated by non-iterative method is in and Does not contain blocking vectors, blocking vector e is only in the factor matrix The relationship between the true factor matrix and its estimated value can be expressed as:

[0193]

[0194] Here, Ψ1 and Ψ2 are diagonal nonsingular matrices satisfying Ψ1Ψ2 = I, Ω represents the permutation matrix shared by all factor matrices, and {E1, E2, E3} represents the estimation error term. The permutation ambiguity shared by all estimated factor matrices can be ignored in the subsequent analysis. However, scale ambiguity must be explicitly addressed during channel parameter estimation.

[0195] S5: For factor matrices unaffected by the blocking vector, the sparse nature of the millimeter-wave channel is exploited to estimate some channel parameters related to the factor matrix. This sparsity transforms the cascade channel matrix estimation problem into a channel parameter estimation problem. A correlation-based method is used to estimate the angular parameters of the base station and the user equipment (UE). This method employs the maximum likelihood criterion based on correlation, maximizing the inner product between the estimated array response and the theoretical direction vector. This is the first stage of channel estimation.

[0196] As shown in formula (23), the estimated Compared with the true factor matrix C, there is no scale ambiguity, so it can be estimated directly; using the Vandermonde structure of C, it is estimated that The delay parameters are related to Therefore, the delay parameter can be estimated as:

[0197]

[0198] in, Indicates plural For the factor matrix X, each column is represented by the array response corresponding to the angle at the UE and the angle at the BS, so the angles at the BS and UE can be estimated by the correlation method, the purpose of which is to maximize the correlation between the estimated vector and the parameter vector. Specifically, Rearrange to N s ×T matrix, where You can get:

[0199]

[0200] The rank-one matrix factorization algorithm can be used to estimate B and U; Performing truncated singular value decomposition (t-SVD), we can get:

[0201]

[0202] and The first column can be u xand v x To linearly represent, therefore, the angle at BS is estimated using normalized correlation maximization:

[0203]

[0204] in, Similarly, the angle at the UE can be estimated as:

[0205]

[0206] Therefore, the scale fuzziness of the factor matrix X is:

[0207]

[0208] in, The scale fuzzy of the factor matrix R can be obtained as:

[0209] The first stage of the estimation process is summarized in Table 1.

[0210] Table 1

[0211]

[0212] S6: For the coupling factor matrix containing the remaining channel parameters and the blocking vector, the sparsity of millimeter waves and blocking is utilized to jointly estimate the remaining channel parameters and the RIS blocking vector. In the second stage of channel estimation, the problem of obtaining the remaining channel parameters and the RIS blocking is that they are coupled to each other and therefore difficult to obtain separately. The present invention utilizes the intrinsic characteristics of the channel and blocking and adopts an alternating iterative strategy to solve this problem. Specifically, the structural characteristics of the channel are utilized to transform the parameter estimation task into a problem of maximizing the normalized correlation. At the same time, the sparsity of blocking is utilized to reformulate the blocking estimation as a sparse vector recovery problem. This method can simultaneously estimate the remaining channel parameters and the blocking vector.

[0213] For the factor matrix Each column is represented by the array response and blockage corresponding to the angle at RIS. The lth column of the factor matrix R can be expressed as:

[0214]

[0215] Based on formula (30), since the residual channel parameters are coupled with the blocking vector, a joint estimation framework is required to alternately optimize the angle domain and the blocking domain.

[0216] First, the RIS spatial angle is estimated by a normalized correlation maximization method:

[0217]

[0218] in, After obtaining the cascade angle at the RIS, reconstruct a RIS array response matrix: definition is represented as a sparse blocking vector, and can be expressed as Without noise influence So the path gain can be calculated as:

[0219]

[0220] use can be re-expressed as:

[0221]

[0222] Among them, N R Represents the estimation error, and the column transformation on both sides of Equation (33) is:

[0223]

[0224] Among them, n r =vec(N R ); in addition, define:

[0225]

[0226] Thus (34) can be transformed into:

[0227] r=Φk+n r . (36);

[0228] Therefore, the problem of solving the blocking vector is expressed as a sparse signal recovery problem; the sparse Bayesian algorithm is selected to solve it. Traditional compressed sensing algorithms are very sensitive to parameters. For example, the orthogonal matching pursuit algorithm (OMP) requires prior knowledge of the sparsity of the signal, and the least absolute shrinkage and selection operator (LASSO) requires manual parameter adjustment. Unlike traditional compressed sensing algorithms, the advantage of the Bayesian method is that it adaptively optimizes parameters through a hierarchical Bayesian inference framework, thus avoiding these problems. In this case, assuming that the noise n r Obey complex Gaussian distribution β follows the gamma distribution Γ(β; c, d); the sparse vector k follows the complex Gaussian distribution Where Λ = diag(α), α follows the gamma distribution Γ(α; 1, ρ); the posterior probability distribution of the sparse vector k is:

[0229]

[0230] The mean and variance of the posterior probability distribution of the sparse vector k are:

[0231]

[0232] The update formula for hyperparameters α and β is:

[0233]

[0234] This paper proposes a threshold-based detection mechanism that uses the inherent sparsity of blocking vectors to improve the accuracy of sparse signal recovery. The specific rules are as follows:

[0235]

[0236] Among them, the predefined threshold δ determines the sparsity level.

[0237] The second stage of the estimation process is summarized in Table 2.

[0238] Table 2

[0239]

[0240]

[0241] The computational complexity of Algorithm 1 is mainly determined by steps 2, 5, 7, and 8, and the corresponding complexities are and Where I represents the number of search points based on the correlation search process. For Algorithm 2, the computational complexity is mainly determined by formulas (31), (38) and (39), and the corresponding complexities are as follows: and Here, E represents the total number of iterations iter, and J represents the number of cycles of the inner loop in Algorithm 2.

[0242] The present invention provides numerical results to evaluate the performance of the proposed channel estimation and blocking diagnosis method. The specific parameters are set as follows: RIS is composed of N=N x ×N y =8×8 components, with N B Blocking unit. BS equipped with M BS = 32 antennas and N BS = 4 RF chains, UE equipped with M UE = 32 antennas and N UE = 1 RF chain. Complex gain and Obey complex Gaussian distribution All angles are uniformly distributed Delay Item and Uniform distribution The number of data streams is set to N s =4. The carrier frequency is f c =28GHz, the total number of subcarriers is P0=128, and the sampling rate is f s =0.32GHz. The signal-to-noise ratio (SNR) is defined as:

[0243]

[0244] The calculation formula of the channel normalized mean square error (NMSE) is:

[0245]

[0246] in, is the estimated channel matrix. The blocked NMSE is defined as

[0247] As far as we know, there is no known literature dealing with the problem of joint channel estimation and blocking diagnosis in millimeter wave MIMO-OFDM systems. To fill this gap, the present invention evaluates the performance of various sparse recovery methods in the second stage of the proposed algorithm. In order to determine the performance benchmark, the algorithm of the present invention is compared with two classical methods: OMP and LASSO. To ensure a fair comparison, all methods use the same first-stage decoupling procedure. In addition, in order to prove the effectiveness of the scheme proposed by the present invention, the present invention uses the estimation performance under ideal conditions as the lower limit of the comparison. Specifically, the curve labeled "k, ideal case" represents the NMSE lower limit of blocking estimation under perfect CSI, while the curve labeled "H, ideal case" represents the lower limit of channel estimation when the blocking vector is fully understood.

[0248] Figure 3A schematic diagram shows the variation of NMSE with SNR for channel estimation and blockage diagnosis using different algorithms. The proposed algorithm demonstrates consistent superiority over the CPD-OMP and CPD-LASSO baselines. This performance advantage stems from fundamental limitations of the compared methods: OMP requires prior knowledge of the sparsity of the blocking vector, while LASSO requires manual tuning of the regularization parameter. In contrast, the Bayesian framework can automatically adjust parameters without imposing sparsity constraints, thereby improving the estimation accuracy of the channel parameters and the blocking vector. For blockage diagnosis, a signal-to-noise ratio gap of approximately 8dB is found compared to the ideal k, indicating that the algorithm is more sensitive to changes in the channel matrix than to changes in the blocking vector. This difference in robustness highlights the algorithm's strong adaptability to dynamic blocking environments while maintaining stable channel estimation capabilities.

[0249] Figure 4 The NMSE of the blocking diagnosis and channel estimation is shown to vary with the number of blocking elements (N B ) changes. Compared with the CPD-OMP and CPD-LASSO methods, the framework proposed in this invention shows consistent advantages. In addition, as N B As N increases, the algorithm of the present invention shows relatively stable performance. B As NMSE increases, it also increases, which is mainly due to the larger N B This trend highlights the adverse impact of blockage on overall system performance.

[0250] Figure 5 A schematic diagram showing the variation of the NMSE of blockage diagnosis and channel estimation with the number of time frames (Q) is shown. The results show that the proposed scheme outperforms the CPD-OMP and CPD-LASSO methods. In addition, the NMSE decreases continuously with the increase of Q. This trend can be explained by (35), that is, the larger the Q, the larger the dimension of the observation vector r, which improves the performance of sparse recovery.

[0251] Figure 6 and Figure 7 Schematic diagrams show how the NMSE for blocking diagnosis and channel estimation varies with the number of slots (T) and subcarriers (P), respectively. In both cases, the proposed scheme outperforms the CPD-OMP and CPD-LASSO methods, demonstrating robust performance across both system parameters. Furthermore, the NMSE decreases with increasing T or P. This performance improvement is primarily attributed to two factors: an increase in T provides more training samples for more accurate parameter estimation. Increasing P provides better frequency domain observations, which is beneficial because blocking often exhibits common sparsity across subcarriers.

[0252] Figure 8 The relationship between NMSE and SNR is shown for different path numbers L. As can be observed, for both channel estimation and blockage diagnosis, NMSE increases with increasing L. This trend is mainly because the larger L, the more channel parameters need to be estimated, which ultimately leads to performance degradation with the same training overhead.

[0253] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A RIS-assisted channel estimation and blocking diagnosis method for MIMO-OFDM systems, characterized in that: The steps are as follows: S1: Construct a RIS-assisted mmWave MIMO-OFDM communication system in which some RIS components are blocked; S2: Design a frame-based structured training protocol to model the received signal as a low-rank fourth-order tensor and fully exploit the structural and coupling information of the low-rank fourth-order tensor; S3: Before tensor decomposition, the uniqueness of the tensor decomposition is first analyzed, and the conditions that the number of time frames and the number of pilots must meet are analyzed; S4: Decompose the received signal using a non-iterative algorithm to obtain a factor matrix that is not affected by the blocking vector and a coupling factor matrix that includes the remaining channel parameters and the blocking vector; S5: For the factor matrix that is not affected by the blocking vector, the sparse characteristics of the millimeter wave channel are used to estimate some channel parameters related to the factor matrix; S6: For the coupling factor matrix containing the remaining channel parameters and the blocking vector, the remaining channel parameters and the RIS blocking vector are jointly estimated by utilizing the sparsity of millimeter waves and blocking.

2. The RIS-assisted channel estimation and blocking diagnosis method for MIMO-OFDM system according to claim 1, characterized in that: The method for constructing the RIS-assisted millimeter wave MIMO-OFDM communication system is as follows: establishing a passive RIS-assisted millimeter wave MIMO-OFDM communication system in an uplink scenario, in which the BS is equipped with M BS Antenna and N BS RF chain, and UE is equipped with M UE Antenna and N UE RF chain; The antenna structure of BS and UE both adopts uniform linear array (ULA), and RIS is composed of uniform planar array (UPA) of N passive elements; RIS has N in the horizontal direction x There are N elements in the vertical direction y components, so the total number of components is N = N x ×N y ; The system uses a total of P0 OFDM subcarriers, of which P OFDM subcarriers are used for training.

3. The RIS-assisted channel estimation and blocking diagnosis method for MIMO-OFDM system according to claim 2, characterized in that: In the passive RIS-assisted millimeter-wave MIMO-OFDM communication system, the Saleh-Valenzuela channel model is used to describe the UE-RIS channel and the BS-RIS channel. Specifically, the definition of the UE-RIS channel in the delay domain is as follows: Where δ(τ) represents the Dirac-delta function, L1 represents the number of paths between UE and RIS, represents the path gain between UE and RIS, represents the azimuth in the RIS angle of arrival (AoA), represents the elevation angle in the RIS arrival angle, represents the angle of departure (AoD) at the UE, ι m represents the propagation delay; Similarly, the RIS-BS channel is defined in the delay domain as follows: Where L2 represents the number of paths between RIS and BS, represents the path gain between RIS and BS, Indicates the azimuth in the RIS departure angle, Indicates the elevation angle in the RIS departure angle, represents the AoA of BS, κ n represents the propagation delay; the array response vector for ULA and UPA configurations is given by the following formula: for The transpose of , where X∈{UE,BS}; the array response at RIS is expressed as: in After obtaining the delay domain channel model, the UE-RIS and RIS-BS channels under the p-th subcarrier in the frequency domain are expressed as follows: Among them, f s Indicates the sampling rate; Due to the passive nature of RIS components, R p and G p is not feasible, therefore, it is necessary to estimate the cascade channel The cascaded channel is represented as: in, represents the Kronecker product, ⊙ represents the Khatri-Rao product, and the mapping process (a) is: in, represents the upper limit function, l=1,…,L1L2, From formula (4), it can be seen that in the case of multiple elements, H p Is a high-dimensional matrix, by estimating the parameters Completely reconstruct channel H p .

4. The RIS-assisted channel estimation and blocking diagnosis method for MIMO-OFDM system according to claim 3, characterized in that: The specific implementation method of step S2 is: With a frame-based uplink training protocol, the channel estimation process for each subcarrier is performed in Q consecutive time frames, where each frame consists of T time slots. In this setup, the UE uses T different beamforming vectors in T time slots, while the RIS applies a specific phase shift vector s in the qth time slot. q to reflect the incident signal; The transmission signal of the p-th subcarrier in the t-th time slot is expressed as: in, is the pilot symbol vector of the p-th subcarrier, is the baseband precoding matrix associated with the p-th subcarrier, is the RF precoder shared by all subcarriers; for all subcarriers, in the tth time slot, assuming F BB,t,p =F BB,t , m t,p =m t , in this case, the transmitted signal associated with the p-th subcarrier in the t-th time slot is: represents the hybrid combiner used by BS, where It is a common RF combiner for all subcarriers. Denote the baseband combiner associated with the p-th subcarrier, assuming the hybrid combiner remains unchanged: Since RIS components are easily affected by environmental obstacles such as rain and snow, a RIS blocking vector e is introduced, where the nth element e n The characteristics are as follows: in, μ n and ω n They represent the amplitude attenuation and phase distortion caused by the blocking of the nth RIS element, respectively. Due to the frequency-flatness of RIS blocking, the blocking vectors on all subcarriers remain consistent. Therefore, during the qth time frame, the tth time slot, and the pth subcarrier, the signal received by the BS can be expressed as: Substituting (4) into (8), the received signal model is: As shown in (9), the channel parameters exhibit different blocking dependencies: Affected by the distortion caused by blocking, It has the characteristic of not being affected by blockage; A two-stage estimation framework is designed to separate these parameters that are not affected by blocking from the blocking vector; specifically, in the first stage, the focus is on the blocking-invariant parameters The coupled observation matrix is ​​then derived based on the accurate extraction of To restore the cascade channel; Considering the two key factors of the frequency flatness of the blocking vector in OFDM subcarriers and channel blocking decoupling, CPD is used as the main data processing tool; by taking advantage of the two inherent characteristics of the millimeter wave system: sparse multipath scattering and low-rank signal subspace structure, the received signal y is obtained by multi-dimensionally stacking Q time frames, T time slots and P subcarriers. q,t,p To construct a fourth-order tensor By summing up the T time slot signals received at the BS, the composite received signal vector is defined Then the received signal at the p-th subcarrier in the q-th time frame is: in, By performing column union on the time slot observation data and converting the received signal y q,p Converted to N s ×T-dimensional matrix, the following signal model is obtained: Among them, N q,p Indicated by n q,p Transformed N s ×T matrix; the received signal from the BS is collected over Q time frames and a third-order tensor is constructed by stacking in This tensor representation is convenient for multi-dimensional signal processing. The composite received signal on the p-th subcarrier is expressed in tensor form as: in, Indicates stack N q,p The noise tensor formed, in is full column rank, define By summing the received signals from all training subcarriers (Equation (12)), the complete training signal can be represented as a fourth-order tensor in It follows the CPD format, which is: in, represents the noise tensor and The four factor matrices of are defined as follows: From formula (13) and formula (14), it can be seen that B, U and C are respectively determined by the channel parameters and Definition, where C is a Vandermonde structure; in addition, the factor matrix R contains the coupling parameters The corresponding parameters of each factor matrix can be estimated and then used to reconstruct the cascade channel matrix and estimate the blocking vector.

5. The RIS-assisted channel estimation and blocking diagnosis method for MIMO-OFDM system according to claim 4, characterized in that: The implementation method of step S3 is: Regarding formula (13) The uniqueness analysis of CPD is generally determined by the Kruskal condition, that is: k(B)+k(U)+k(R)+k(C)≥2L+3, (15); where k(B) represents the k-rank of the matrix B. However, according to formulas (5) and (14), when L1 = 1, L2 = 1 and L1 ≥ 2, L2 ≥ 2, k(B) = 1 and k(U) = 1 do not satisfy formula (15). Considering the inherent Vandermonde structure in C, a more relaxed uniqueness criterion is established: The tensor model in the noise-free case is The factor matrices are denoted as B, U, R, and C; if the following conditions hold: Then, the tensor The CPD of can be uniquely determined; r(R) represents the rank of the matrix R; Formula (16) is further expressed as r( C )=L and r(R)=L, and the number of subcarriers P and the number of time frames Q are greater than the total number of paths L, and and The parameters in these two sets are different.

6. The RIS-assisted channel estimation and blocking diagnosis method for MIMO-OFDM system according to claim 5, characterized in that: The method for obtaining the factor matrix not affected by the blocking vector and the coupling factor matrix including the remaining channel parameters and the blocking vector in step S4 is as follows: Using the inherent Vandermonde structure of the factor matrix C and assuming a noise-free case, we consider the tensor The mode-2 is expanded to: right Do SVD decomposition: Ignoring the noise, we can find that there exists a non-singular matrix M such that U Y M=C⊙U⊙B; from this we can get: in, means removing the first row of matrix C, C It means removing the last row of matrix C. Using the Vandermonde structure of C, we can get: That is to say By performing eigenvalue decomposition, we can get the estimated values ​​of Z and M. and Then each column of the factor matrix C can be estimated as: in, is a matrix The lth element on the diagonal of byU Y M=C⊙U⊙B, each column of the estimated value of X is: After the factor matrices C and X are obtained, the estimated value of the factor matrix R is: The factor matrix estimated by non-iterative method is in and Does not contain blocking vectors, blocking vector e is only in the factor matrix The relationship between the true factor matrix and its estimated value can be expressed as: Where Ψ1 and Ψ2 are diagonal non-singular matrices satisfying Ψ1Ψ2=I, Ω represents the permutation matrix shared by all factor matrices, and {E1,E2,E3} represents the estimation error term.

7. The RIS-assisted channel estimation and blocking diagnosis method for MIMO-OFDM system according to claim 6, characterized in that: The implementation method of step S5 is: As shown in formula (23), the estimated Compared with the true factor matrix C, there is no scale ambiguity, so it can be estimated directly; using the Vandermonde structure of C, it is estimated that The delay parameters are related to Therefore, the delay parameter can be estimated as: in, Indicates plural The actual parameter of the factor matrix X; each column of which is represented by the array response corresponding to the angle at the UE and the angle at the BS, so the angles at the BS and UE can be estimated by the correlation method, the purpose of which is to maximize the correlation between the estimated vector and the parameter vector; specifically, Rearrange to N s ×T matrix, where You can get: The rank-one matrix factorization algorithm can be used to estimate B and U; Performing truncated singular value decomposition, we can get: and The first column can be u x and v x To linearly represent, therefore, the angle at BS is estimated using normalized correlation maximization: in, Similarly, the angle at the UE can be estimated as: Therefore, the scale fuzziness of the factor matrix X is: in, The scale fuzzy of the factor matrix R can be obtained as:

8. The RIS-assisted channel estimation and blocking diagnosis method for MIMO-OFDM system according to claim 7, characterized in that: The implementation method of step S6 is: For the factor matrix Each column is represented by the array response and blockage corresponding to the angle at RIS. The lth column of the factor matrix R can be expressed as: Based on formula (30), since the residual channel parameters are coupled with the blocking vector, a joint estimation framework is required to alternately optimize the angular domain and the blocking domain; First, the RIS spatial angle is estimated by a normalized correlation maximization method: in, After obtaining the cascade angle at the RIS, reconstruct a RIS array response matrix: definition is represented as a sparse blocking vector, and can be expressed as Without noise influence So the path gain can be calculated as: use can be re-expressed as: Among them, N R Represents the estimation error, and the column transformation on both sides of Equation (33) is: Among them, n r =vec(N R ); in addition, define: Therefore, formula (34) can be transformed into: r=Φk+n r . (36); Therefore, the problem of solving the blocking vector is expressed as a sparse signal recovery problem; the sparse Bayesian algorithm is selected to solve it. In this case, assuming that the noise n r Obey complex Gaussian distribution β follows the gamma distribution Γ(β; c, d); the sparse vector k follows the complex Gaussian distribution Where Λ = diag(α), α follows the gamma distribution Γ(α; 1, ρ); the posterior probability distribution of the sparse vector k is: The mean and variance of the posterior probability distribution of the sparse vector k are: The update formula for hyperparameters α and β is:

9. The RIS-assisted channel estimation and blocking diagnosis method for MIMO-OFDM system according to claim 8, characterized in that: Design a threshold-based detection rule, which is expressed as follows: The predefined threshold δ determines the sparsity level.

Images