Very large-scale intelligent reflecting surface assisted channel estimation method based on sparse representation

Abstract

The present invention discloses a super-large-scale intelligent reflection surface assisted channel estimation method based on sparse representation. First, a lightweight dual-domain filter is implemented on the super-large-scale intelligent reflection surface, and the time domain signal is transformed into a sparse representation in the angle domain and the polarization domain, and then the redundant part close to zero is forcibly set to zero to achieve noise filtering in the non-transmitting area. Then, according to the received filtered pilot signal, a failure-aware orthogonal matching pursuit algorithm is used for fault perception, sparse channel reconstruction and intelligent reflection surface component failure fault diagnosis are achieved, thereby improving the accuracy of channel estimation.

Description

Technical Field

[0001] The invention relates to the field of communication technology, and in particular to a super-large-scale intelligent reflection surface auxiliary channel estimation method based on sparse representation. Background Art

[0002] With the increasing demand for communication, building a wider and more efficient wireless communication network has become an inevitable trend in communication research. In the future prospects of 6G communication systems, scholars have widely discussed the possibility of deploying extremely large-scale intelligent reflective surfaces (XL-RIS) to enhance communication capacity and service area. Compared with traditional intelligent reflective surfaces (RIS), XL-RIS consists of more and more widely arranged sub-elements, which can significantly enhance communication capabilities, but also faces more complex signal processing and system design challenges.

[0003] Compared with traditional communication scenarios, the electromagnetic characteristics between the signal transmitter and the receiver change in the ultra-large-scale transmission / reflection element deployment scenario. When the deployment scale of communication elements increases or the distance between the user and the RIS decreases, the Rayleigh distance of the communication environment will decrease, and the corresponding far-field communication process will be converted to a near-field communication process. Unlike the assumption of plane wave transmission signals in traditional far-field communications, near-field communications assume that signals propagate in the form of spherical waves. This results in the angular domain sparseness commonly used for far-field channel estimation in near-field communications no longer existing. When the signal is transmitted in the form of a spherical wave, the near-field channel information contains not only angle information, but also the distance information between the signal source and the receiver.

[0004] At present, most of the research on channel estimation for RIS-assisted systems focuses on far-field scenarios, while research on channel estimation for near-field communication scenarios between users (UEs) and XL-RIS is still in its infancy. Channel estimation methods in far-field scenarios cannot be directly applied to near-field communications for two main reasons: on the one hand, the sparse angle domain characteristics usually used in far-field channels are weakened in near-field channel scenarios, resulting in a decrease in the performance of estimation methods based on sparse reconstruction. On the other hand, the channel parameters of XL-RIS-assisted systems are large in scale, making channel estimation based on limited pilot resources more challenging. In order to solve this dual complexity, the existing technology proposes a low-pilot-overhead near-field beam training scheme for XL-RIS-assisted communication, or a sparse representation scheme based on the polar coordinate domain and a second-order Fresnel approximation method to achieve near-field channel estimation and UE positioning. However, existing research on near-field communications assumes that all RIS components can work perfectly, which is not realistic. Since RIS components are densely connected and deployed unprotected on walls or ceilings, they are prone to damage or occlusion. The present invention collectively refers to these phenomena as XL-RIS component failures, and aims to further ensure the accuracy of channel estimation by estimating failed components. Summary of the invention

[0005] In order to solve the limitations and defects of the prior art, the present invention provides a super-large-scale intelligent reflection surface assisted channel estimation method based on sparse representation, comprising:

[0006] Mathematical modeling is performed to obtain the pilot signal received by the BS end. The expression is as follows:

[0007] ,

[0008] in, , is the sparse representation of the near-field channel, is the sparse transformation matrix, is the failure indication matrix, is the control vector formed by connecting the reflection parameters of each element in series, , is the far-field channel between the BS and the RIS, , for The original signal sent from the UE to the RIS in a time slot, , It is the signal processing operation of RIS;

[0009] The pilot signal received by the BS is simplified as follows:

[0010] ,

[0011] For receiving signal , a dual-domain filter is used to transform the time domain signal into a sparse representation of the near-field channel in the angle domain and the polarization domain, and the redundant part is forced to be set to zero according to a preset threshold to achieve noise filtering in the non-transmitting area;

[0012] For the filtered signal, the failure-aware orthogonal matching pursuit algorithm is used to estimate the sparse channel at the BS end, and the failure fault diagnosis of the smart reflective surface element is performed based on the estimated residual.

[0013] Optionally, the step of transforming the time domain signal into a sparse representation of the near-field channel in the angle domain and the polarization domain using a dual-domain filter comprises:

[0014] Multiply the signal received by the VLSS by the inverse transformation matrix in the angular domain Convert to sparse representation in angle domain , multiply the signal received by the ultra-large-scale smart reflective surface by the inverse transformation matrix in the polarization domain Convert to sparse representation in polarization domain .

[0015] Optionally, the step of forcibly setting the redundant part to zero according to a preset threshold value to achieve noise filtering in the non-emitting area includes:

[0016] Set the filter window size in the angle domain to , set the filter window size in the polarization domain to ;

[0017] The sparse expression of the angle domain Split into length The sub-blocks of the polarization domain are sparsely expressed Split into length , calculate the norm of each sub-block;

[0018] Determine the sparse representation of the polarization domain based on the norm value of the sub-block The size of the element value in the middle is selected, and the amplitude is less than the preset threshold The corresponding angle index is summarized into the set In the example, the sparse expression of the angle domain is determined according to the norm value of the sub-block. The size of the element value in the middle is selected, and the amplitude is less than the preset threshold The corresponding distance-angle grids are summarized into the set middle;

[0019] Will gather The angle index in is mapped to the polarization domain grid to obtain the index set ,pass ∪ Get the inactive grid set ;

[0020] The sparse expression of the angle domain Set of inactive and inactive grids All corresponding element values ​​are set to zero to filter out noise.

[0021] Optionally, the step of estimating the sparse channel at the BS end using the failure-aware orthogonal matching pursuit algorithm and performing intelligent reflective surface element failure fault diagnosis according to the estimated residual includes:

[0022] The joint optimization expression for obtaining the pilot signal is as follows:

[0023] ,

[0024] Failure Indication Matrix Set all element values ​​in to zero and solve the channel sparse representation The reconstruction problem is expressed as follows:

[0025] ,

[0026] For the BS Antenna, when receiving When the pilot signal of time slots is 1, the channel estimation problem is modeled as a sparse reconstruction problem, and the expression is as follows:

[0027] ,

[0028] in, express No. List, For the The channel between the transmitting antenna and the smart reflecting surface;

[0029] set up , while using Sparse representation of the near-field channel by the pilot signal received by each BS antenna To estimate, the expression is as follows:

[0030] ,

[0031] in, , ;

[0032] According to the sparse reconstruction results, we obtain The residual error of the channel estimation of the received signal of each antenna is expressed as follows:

[0033] ,

[0034] in, is the channel sparse representation, for noise;

[0035] The improved residual matrix is ​​obtained by matrix transformation, and the expression is as follows:

[0036] ,

[0037] Use BS at the same time The received signal of the antenna is used for fault diagnosis. Improved residual matrix of receiving antennas Arrange as ;

[0038] The matrix Add columns to get the residual vector , and obtain the residual vector The amplitude is greater than the preset threshold The peak positions are sorted according to the absolute values ​​of the amplitudes to confirm the The position corresponding to the peak is the candidate position of the smart reflective surface element failure.

[0039] Optionally, also include:

[0040] Replace candidate peak positions one by one for ;

[0041] According to the support set Re-estimate using the least squares algorithm ;

[0042] Compare the residual of the reconstructed value before replacement with the residual of the reconstructed value after replacement;

[0043] If the residual of the reconstructed value after replacement is less than the residual of the reconstructed value before replacement, the candidate peak is confirmed to be the failure point of the intelligent reflective surface. If the residual of the reconstructed value after replacement is greater than or equal to the residual of the reconstructed value before replacement, the replacement operation is canceled. Corresponding row The elements of are reset to zero.

[0044] Optionally, also include:

[0045] Orthogonal matching pursuit reconstruction and intelligent reflective surface element failure fault detection are iteratively performed according to a failure-aware orthogonal matching pursuit algorithm until a stop condition is met, wherein the stop condition includes:

[0046] The residual is less than the preset value; or

[0047] No candidate peak can be selected, and the counter ;or

[0048] The maximum number of iterations has been reached.

[0049] The present invention has the following beneficial effects:

[0050] The present invention provides a super-large-scale intelligent reflection surface assisted channel estimation method based on sparse representation, realizes a lightweight dual-domain filter on a super-large-scale intelligent reflection surface, transforms the time domain signal into a sparse representation in the angle domain and the polarization domain, and then forcibly sets the redundant part close to zero to zero, thereby realizing noise filtering in the non-transmitting area. According to the received filtered pilot signal, a failure-aware orthogonal matching pursuit algorithm is used for fault perception, realizing sparse channel reconstruction and intelligent reflection surface component failure fault diagnosis, thereby improving the accuracy of channel estimation. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 This is a schematic diagram of the structure of a near-field communication system assisted by an ultra-large-scale intelligent reflective surface provided in Embodiment 1 of the present invention.

[0052] Figure 2 The standard mean square error between the original signal sent by the UE and the signal received by the ultra-large-scale intelligent reflecting surface when a polarization domain filter, a dual-domain filter, and no filter are used respectively provided in the first embodiment of the present invention.

[0053] Figure 3a This is a schematic diagram of a first performance comparison between the FA-OMP algorithm and the OMP_K algorithm provided in the first embodiment of the present invention.

[0054] Figure 3b This is a second performance comparison diagram of the FA-OMP algorithm and the OMP_K algorithm provided in the first embodiment of the present invention.

[0055] Figure 3c A third performance comparison diagram of the FA-OMP algorithm and the OMP_K algorithm provided in the first embodiment of the present invention.

[0056] Figure 3d A fourth performance comparison diagram of the FA-OMP algorithm and the OMP_K algorithm provided in the first embodiment of the present invention.

[0057] Figure 3e This is a fifth performance comparison diagram of the FA-OMP algorithm and the OMP_K algorithm provided in the first embodiment of the present invention.

[0058] Figure 3f This is a sixth performance comparison diagram of the FA-OMP algorithm and the OMP_K algorithm provided in the first embodiment of the present invention. DETAILED DESCRIPTION

[0059] In order to enable those skilled in the art to better understand the technical solution of the present invention, the ultra-large-scale intelligent reflecting surface-assisted channel estimation method based on sparse representation provided by the present invention is described in detail below with reference to the accompanying drawings. Embodiment 1

[0060] 1. Overview

[0061] In order to solve the complex problems of channel estimation and failure diagnosis of intelligent reflective surface (Reconfigurable Intelligent Surface, RIS) components, this embodiment proposes a signal-assisted channel estimation (Signal-Assisted Sparse Representation-based Channel Estimation, SA-SRCE) based on sparse representation to achieve joint near-field channel state reconstruction and RIS failure component diagnosis. The SA-SRCE method is divided into two stages. In the first stage, a lightweight double domain filter (DDF) is implemented on the extremely large-scale intelligent reflective surface (Extremely Large-Reconfigurable Intelligent Surface, XL-RIS). By transforming the time domain signal into a sparse representation in the angle domain or polarization domain, and then forcibly setting the redundant part close to zero to zero, noise filtering in the non-transmitting area is achieved. In the second stage, the BS end proposes a fault-aware orthogonal matching pursuit (FA-OMP) algorithm based on the received filtered pilot signal, which realizes sparse channel reconstruction based on polar coordinate transformation and RIS component failure fault diagnosis.

[0062] 2. SA-SRCE Method

[0063] 1. System model

[0064] This embodiment studies the uplink time division duplex XL-RIS assisted communication system, which is equipped with a base station (BS) The distance between two adjacent antennas at the BS is set to ,in is the wavelength of the transmission carrier. This embodiment uses the XL-RIS technology to establish an additional communication link between the user (UE) and the BS. Assume that the RIS is set up in an area close to the user. The distance between two adjacent RIS elements is set to In this model, the RIS channel observed by the BS is a hybrid channel consisting of a cascade of a multipath near-field channel from the UE to the RIS and a multipath far-field channel from the RIS to the BS.

[0065] like Figure 1 As shown in Figure 1, the near-field channel is not only related to the angle of the signal transmitted from the UE to the XL-RIS, but also to the distance between the UE and the XL-RIS. Assume that the UE is located at the coordinate point ,in Indicates the angle, is the distance between the UE and the center point of XL-RIS. The vector of the XL-RIS uniform linear array beamforming is at an angle ,distance It can be expressed as , specifically:

[0066] ,

[0067] In this equation, Indicates the center carrier The wave number at . It is The distance between a RIS element and the UE is

[0068]

[0069] in, , By summing all possible paths, we can get the spherical wave channel model in the near field region.

[0070]

[0071] in, Indicates The communication space where the RIS element is located is divided into Different angle distance grids, the sparse transformation matrix is ​​recorded as , the sparse representation of the near field communication channel is denoted as , and define The sparsity is , Channel It can be sparsely represented as

[0072]

[0073] In actual communication scenarios, RIS components do not always reflect signals perfectly as expected. In the XL-RIS scenario, due to the complexity and dynamics of the RIS deployment environment, the UE sometimes cannot receive the reflected signals of certain RIS components, that is, the RIS components fail. In order to model this failure phenomenon, this embodiment establishes a failure fault indication matrix , where the elements correspond to the grid transformed in the polar coordinate domain. The locations where non-ideal reception occurs due to RIS component failure are considered as matrix perturbations. It is assumed that when a failure occurs, the corresponding components will lose all reflection capabilities, that is, they are The corresponding element value on is recorded as ,but No. A row can be modeled as:

[0074]

[0075] Among them, the collection Include sub-items, each of which represents the index of a failed component. When the values ​​of a row of elements are all 0, it means that the RIS components corresponding to this row are working normally, while other non-ideal rows The element value of Considering the failure of RIS components, the near-field channel between UE and RIS is obtained

[0076] ,

[0077] In addition, the amplitude and phase of the BS received signal can be adjusted by the controller on the RIS. In order to describe the control status of all components in the RIS, this section concatenates the reflection parameters of each component into a control vector, denoted as For the elements, Indicates the amplitude adjustment made by RIS. represents the corresponding phase adjustment. Assuming that In the time slot, the RIS auxiliary channel remains unchanged. By multiplying each channel information with the corresponding RIS control vector, the near-field channel state of the tth time slot is obtained as

[0078] ,

[0079] in, It is a time slot The control vector of XL-RIS is Represents the Hadamard product.

[0080] Assume that there is The far-field channel between RIS and BS can be modeled as

[0081] ,

[0082] in, , It is The complex gain of the paths, denote the arrival angle and departure angle respectively, Refers to the pointing vector for far-field communication with plane wave assumption

[0083] ,

[0084] No. OK( ) indicates the The channel between the transmitting antenna and the RIS is denoted as .

[0085] Combining the above near-field channel and far-field channel models, The original signal sent from UE to RIS in a time slot is expressed as . The signal processing operation is defined as ,get Time BS end The signal received by the antenna is

[0086] ,

[0087] in, Represents the noise between RIS and BS.

[0088] definition and , then the overall receiving matrix at the BS is

[0089] ,

[0090] in, , In order to simplify the problem, this embodiment assumes that all element values ​​of the transmitted pilot signal are 1. When channel estimation is performed based on the pilot signal, the pilot signal received by the BS can be simplified as

[0091] ;

[0092] Optimization issues

[0093] This embodiment aims to jointly realize channel estimation and RIS component failure diagnosis. Assume that the channel between the base station and the RIS is known and is a slowly varying channel. In the scenario based on pilot estimation, this embodiment uses the received Reconstructing sparse channel representation and the perturbation matrix Considering The inherent sparsity and The row sparsity of this embodiment establishes the following joint optimization problem:

[0094] ;

[0095] (III) Detailed explanation of the optimization plan

[0096] To solve the target problem, this embodiment is carried out in two stages.

[0097] 1. Phase 1

[0098] In the first stage, this embodiment proposes a DDF method, which filters out the noise between the UE and the RIS by designing a filter at the RIS end to improve the system robustness. The DDF method is based on the sparse representation of the near-field channel in the angle domain and the polar coordinate domain, and uses the angle domain transformation matrix and the polar domain transformation matrix , the RIS communication area is divided into angle grids and distance-angle grids, and targeted filtering is performed in the sparse domain to achieve signal noise reduction. The DDF method consists of three steps:

[0099] In the first step, the signal received by XL-RIS is multiplied by the inverse transformation matrices in the angle domain and polarization domain (inverse matrix or pseudo-inverse matrix in the case of non-square matrix, respectively). and ) is converted into a sparse representation in the angle domain and polarization domain. That is, the first step is to convert the time slot The received signal is converted into polarization domain representation

[0100] ,

[0101] Similarly, for time slot The received signal is transformed into the angle domain to obtain .

[0102] In the second step, a dual-domain window filter was designed on RIS to and Some elements in are forced to be zero. The specific process includes the following:

[0103] Set the filter window sizes in the angle domain and polarization domain to Then, and Split into length and calculate the 1-norm of each sub-block.

[0104] Set two thresholds As a threshold for determining the sub-block amplitude, it is used to determine whether the window block should be set to zero.

[0105] Based on the block 1 norm value, judge The size of the element value in the The sub-blocks are selected and their corresponding angle indexes are summarized into the "inactive grid" set middle.

[0106] Similarly, select The amplitude is less than and summarizes its corresponding inactive distance-angle grids into the set middle.

[0107] Will gather The index in the polarization domain is mapped to the polarization domain grid (select the polarization domain with All grid indices corresponding to the angles in the graph), get a new index set Then, by calculating ∪ Get the collection of "inactive grids" .

[0108] The set of “inactive grids” selected according to the above strategy ,Will In and Collection All element values ​​corresponding to the grid are set to zero to filter out noise.

[0109] The operation of the second step is expressed as a function , then the filtered signal is expressed as

[0110] ,

[0111] The third step is to multiply the filtered signal by the polarization domain sparse basis , re-transformed into time domain signal

[0112] Then transmit it to the BS.

[0113] Ideally, the entire filtering process of RIS, i.e. operation, should only act on noise Therefore, the joint optimization problem based on the pilot signal can be simplified as

[0114] ;

[0115] 2. Second stage

[0116] The second stage is channel reconstruction and component failure diagnosis based on pilot signals. This embodiment proposes an iterative solution method for RIS component failure fault perception based on orthogonal matching pursuit (Orthogonal Matching Pursuit, OMP), called the Failure-Aware Orthogonal Matching Pursuit (FA-OMP) algorithm. Since RIS component failure is relatively rare, FA-OMP first ignores the impact of RIS component failure and roughly estimates the state of the channel. That is, assuming All element values ​​in are 0, and the channel sparse representation is roughly solved directly. Reconstruction problem:

[0117] ,

[0118] In order to simplify the algorithm, the above problem is split into a channel estimation problem based on the signal received by each receiving antenna. Antenna, when it receives When the pilot signal of time slots is

[0119] ,

[0120] in, express No. Column. Definition , in order to simultaneously utilize The pilot signal received by each BS antenna is used to estimate , the above problem is optimized and expanded into

[0121] ,

[0122] in, , In this regard, the original optimization problem is transformed into a traditional CS reconstruction problem, which can be solved using the classic OMP algorithm.

[0123] Next, based on the roughly reconstructed channel sparse representation and support set , the failure state of RIS components is detected through the following three steps. Based on the reconstruction results, the first The residual error of the channel estimation of the received signal of each antenna is:

[0124] ,

[0125] The first step is to exclude and The impact of residuals on detection judgment is concentrated on The improved residual matrix is ​​obtained by matrix transformation

[0126] ,

[0127] Improved residual matrix Can better characterize failure. Reconstruction error and noise When both are small, parts (a) and (c) in the above formula will not have obvious high-amplitude peaks, while part (b) corresponds to The non-zero row element positions will have peaks with larger magnitudes.

[0128] The second step is to identify The significant peak points with larger amplitudes are candidate locations where RIS component failure may occur. The received signal of the antenna is used for fault diagnosis. Improved residual matrix of receiving antennas Arrange as When selecting candidate positions, firstly transform the matrix Add columns together to get a residual vector Then, based on the preset threshold , find The amplitude is greater than Then, these peaks are sorted by their absolute amplitudes. The corresponding positions of the peaks are regarded as candidate positions of RIS failure.

[0129] In the third step, in order to avoid the iterative algorithm from not converging due to incorrect RIS failure diagnosis, a backtest operation is set to determine whether the above candidate positions can truly represent the phenomenon of RIS failure at the corresponding point. Specifically, replace the candidate peak positions one by one for , then based on the support set Re-estimate using the least squares algorithm If the residual of the reconstructed value before and after the replacement is reduced, the candidate peak is considered to be the RIS failure point; otherwise, the replacement operation will be canceled. Corresponding row The elements of are reset to all zeros.

[0130] This embodiment provides a super-large-scale intelligent reflection surface assisted channel estimation method based on sparse representation, implements a lightweight dual-domain filter on the super-large-scale intelligent reflection surface, and realizes noise filtering in the non-transmitting area by transforming the time domain signal into a sparse representation in the angle domain and polarization domain, and then forcibly setting the redundant part close to zero to zero. According to the received filtered pilot signal, the failure-aware orthogonal matching pursuit algorithm is used to perform fault perception, realize sparse channel reconstruction and intelligent reflection surface component failure fault diagnosis, thereby improving the accuracy of channel estimation. This embodiment first performs a lightweight dual-domain filter operation on the super-large-scale intelligent reflection surface according to the received pilot signal, and realizes dual-domain filtering in the polarization domain and the angle domain through the operation of the transformation layer and the filter layer. Then, the failure-aware orthogonal matching pursuit algorithm is used to estimate the sparse channel, and the intelligent reflection surface component failure fault diagnosis is performed based on the estimated residual.

[0131] Table 1 FA-OMP algorithm

[0132]

[0133] FA-OMP iterates to perform OMP reconstruction and RIS component failure detection until the stop condition is met. This condition consists of three sub-conditions: a) the residual is small enough; b) no candidate peak can be selected through the third step, that is, the counter ; c) The maximum number of iterations is reached. If any of the above conditions is met, the iteration is stopped.

[0134] In summary, the SA-SRCE method first performs the first-stage lightweight DDF operation on XL-RIS based on the received pilot signal, and realizes dual-domain filtering in polarization domain and angle domain through transformation layer and filter layer operations. Then, in the second stage, the sparse channel is estimated at the BS end using the FA-OMP algorithm, and RIS component failure fault diagnosis is performed based on the estimated residual.

[0135] 3. Simulation Results

[0136] The default simulation parameters of this embodiment are shown in Table 2, where is uniformly distributed. During the simulation process, some simulation parameters may be changed to verify the impact of the parameters on the algorithm performance. Unless otherwise specified, the default settings of these parameters are as follows: Number of RIS failure elements , the noise signal-to-noise ratio SNR between UE and RIS = 10dB, the noise signal-to-noise ratio SNR between RIS and BS = 20dB, the pilot signal length is equal to 512, the sparsity of the sparse channel vector K = 6, the original signal length is equal to 1024, and the original signal compression ratio is equal to 0.5.

[0137] Table 2 Parameter settings in simulation

[0138]

[0139] In order to verify the reliability and effectiveness of the solution proposed in this embodiment, two benchmark algorithms are set as comparison algorithms. The first is the OMP_K algorithm, which assumes that the location of the RIS failure is known and reconstructs the channel sparse representation based on the OMP algorithm. The results of the OMP_K algorithm are used as the true value of the proposed algorithm's target approximation to verify the performance of sparse vector reconstruction. In addition, the OMP_I algorithm ignores RIS damage, that is, All element values ​​in are set to 0, and the signal is reconstructed using the OMP algorithm to verify the necessity of RIS failure fault diagnosis for channel estimation. Each experimental result in this section is the average result of 1000 Monte Carlo experiments.

[0140] First, this section evaluates the feasibility of the DDF method. The simulation results are shown in Figure 2 As shown. This figure compares the Normalized Mean Square Error (NMSE) between the original signal transmitted by the UE, the filtered and unfiltered signals received at the RIS. In addition, this figure also compares and analyzes the experimental results of filtering only in the polarization domain instead of dual-domain DDF filtering to verify the importance of dual-domain design in the DDF method. The research results show that the sparse domain filtering method (i.e., polarization domain filtering and DDF method) can effectively eliminate part of the noise between the UE and the RIS, and the DDF filter has a more significant performance improvement than the method of filtering only in the polarization domain. In addition, the improvement effect of DDF is particularly obvious under low signal-to-noise ratio conditions, further confirming its strong noise resistance.

[0141] Figure 3a-3f The simulation results of the FA-OMP algorithm are shown. It can be observed that the SA-SRCE scheme performs best in most scenarios. It is worth noting that when the pilot length is equal to 128 or the signal-to-noise ratio is lower than 10dB, the performance of the FA-OMP algorithm based on the filtered pilot is slightly worse than that of the OMP algorithm. This phenomenon is due to the poor reconstruction performance of the sparse channel, which has a negative impact on the RIS failure fault detection, thereby reducing the overall performance of the system. Figure 3b and Figure 3c The channel reconstruction performance is demonstrated, where the pilot signal transmitted by the UE is affected by noise with different signal-to-noise ratios between the UE and the RIS and between the RIS and the BS. Figure 2 The results are consistent with those of , when the noise between UE and RIS increases, the performance advantage of FA-OMP is more significant. Similar conclusions can be drawn when there are different power noises between RIS and BS. Figure 3d and Figure 3e Given in different and The comparative experimental results under the setting prove the universality and robustness of the FA-OMP algorithm. Figure 3f It shows that when the pilot transmission length is moderate, the running time of the FA-OMP algorithm does not increase significantly compared with the OMP algorithm, which verifies the feasibility of FA-OMP.

[0142] This embodiment provides a super-large-scale intelligent reflection surface assisted channel estimation method based on sparse representation, implements a lightweight dual-domain filter on the super-large-scale intelligent reflection surface, and realizes noise filtering in the non-transmitting area by transforming the time domain signal into a sparse representation in the angle domain and polarization domain, and then forcibly setting the redundant part close to zero to zero. According to the received filtered pilot signal, the failure-aware orthogonal matching pursuit algorithm is used to perform fault perception, realize sparse channel reconstruction and intelligent reflection surface component failure fault diagnosis, thereby improving the accuracy of channel estimation. This embodiment first performs a lightweight dual-domain filter operation on the super-large-scale intelligent reflection surface according to the received pilot signal, and realizes dual-domain filtering in the polarization domain and the angle domain through the operation of the transformation layer and the filter layer. Then, the failure-aware orthogonal matching pursuit algorithm is used to estimate the sparse channel, and the intelligent reflection surface component failure fault diagnosis is performed based on the estimated residual.

[0143] It is to be understood that the above embodiments are merely exemplary embodiments used to illustrate the principles of the present invention, but the present invention is not limited thereto. For those of ordinary skill in the art, various modifications and improvements can be made without departing from the spirit and essence of the present invention, and these modifications and improvements are also considered to be within the scope of protection of the present invention.

Claims

1. A super-large-scale intelligent reflecting surface assisted channel estimation method based on sparse representation, characterized in that: include: Mathematical modeling is performed to obtain the pilot signal received by the BS end. The expression is as follows: ， in, , is the sparse representation of the near-field channel, is the sparse transformation matrix, is the failure indication matrix, is the control vector formed by connecting the reflection parameters of each element in series, , is the far-field channel between the BS and the RIS, , for The original signal sent from the UE to the RIS in a time slot, , It is the signal processing operation of RIS; The pilot signal received by the BS is simplified as follows: ， For receiving signals , a dual-domain filter is used to transform the time domain signal into a sparse representation of the near-field channel in the angle domain and the polarization domain, and the redundant part is forced to be set to zero according to a preset threshold to achieve noise filtering in the non-transmitting area; For the filtered signal, the failure-aware orthogonal matching pursuit algorithm is used to estimate the sparse channel at the BS end, and the failure diagnosis of the intelligent reflective surface element is performed based on the estimated residual; The step of using a dual-domain filter to transform the time domain signal into a sparse representation of the near-field channel in the angle domain and the polarization domain comprises: Multiply the signal received by the VLSS by the inverse transformation matrix in the angular domain Convert to sparse representation in angle domain , multiply the signal received by the ultra-large-scale smart reflective surface by the inverse transformation matrix in the polarization domain Convert to sparse representation in polarization domain ; The step of forcibly setting the redundant part to zero according to a preset threshold value to filter out noise in the non-emitting area includes: Set the filter window size in the angle domain to , set the filter window size in the polarization domain to ; The sparse expression of the angle domain Split into length The sub-blocks of the polarization domain are sparsely expressed Split into length , calculate the norm of each sub-block; Determine the sparse representation of the polarization domain based on the norm value of the sub-block The size of the element value in the middle is selected, and the amplitude is less than the preset threshold The corresponding angle index is summarized into the set In the example, the sparse expression of the angle domain is determined according to the norm value of the sub-block. The size of the element value in the middle is selected, and the amplitude is less than the preset threshold The corresponding distance-angle grids are summarized into the set middle; Will gather The angle index in is mapped to the polarization domain grid to obtain the index set ,pass ∪ Get the inactive grid set ; The sparse expression of the angle domain Set of inactive and inactive grids All corresponding element values ​​are set to zero to filter out noise; The steps of estimating the sparse channel at the BS end using the failure-aware orthogonal matching pursuit algorithm and performing intelligent reflective surface element failure fault diagnosis according to the estimated residual include: The joint optimization expression for obtaining the pilot signal is as follows: ， Failure Indication Matrix Set all element values ​​in to zero and solve the channel sparse representation The reconstruction problem is expressed as follows: ， For the BS Antenna, when receiving When the pilot signal of time slots is 1, the channel estimation problem is modeled as a sparse reconstruction problem, and the expression is as follows: ， in, express No. List, For the The channel between the transmitting antenna and the smart reflecting surface; set up , while using Sparse representation of the near-field channel by the pilot signal received by each BS antenna To estimate, the expression is as follows: ‘ in, , ; According to the sparse reconstruction results, we obtain The residual error of the channel estimation of the received signal of each antenna is expressed as follows: ， in, is the channel sparse representation, for noise; The improved residual matrix is ​​obtained by matrix transformation, and the expression is as follows: ， Use BS at the same time The received signal of the antenna is used for fault diagnosis. Improved residual matrix of receiving antennas Arrange as ; The matrix Add columns to get the residual vector , and obtain the residual vector The amplitude is greater than the preset threshold The peak positions are sorted according to the absolute values ​​of the amplitudes to confirm the The position corresponding to the peak is the candidate position of the smart reflective surface element failure.

2. The method for super-large-scale intelligent reflecting surface-assisted channel estimation based on sparse representation according to claim 1, characterized in that: Also includes: Replace candidate peak positions one by one for ; According to the support set Re-estimate using the least squares algorithm ; Compare the residual of the reconstructed value before replacement with the residual of the reconstructed value after replacement; If the residual of the reconstructed value after replacement is less than the residual of the reconstructed value before replacement, the candidate peak is confirmed to be the failure point of the intelligent reflective surface. If the residual of the reconstructed value after replacement is greater than or equal to the residual of the reconstructed value before replacement, the replacement operation is canceled. Corresponding row The elements of are reset to zero.

3. The method for super-large-scale intelligent reflecting surface-assisted channel estimation based on sparse representation according to claim 2, characterized in that: Also includes: Orthogonal matching pursuit reconstruction and intelligent reflective surface element failure fault detection are iteratively performed according to a failure-aware orthogonal matching pursuit algorithm until a stop condition is met, wherein the stop condition includes: The residual is less than the preset value; or No candidate peak can be selected, and the counter ;or The maximum number of iterations has been reached.

Images