An RIS-assisted communication system channel estimation method based on MMSE optimization compressed sensing

Abstract

The application discloses an RIS-assisted communication system channel estimation method based on MMSE optimized compressed sensing, and relates to the technical field of wireless communication. The method comprises the following steps: acquiring an uplink signal received by a base station in an RIS-assisted communication system; transforming the received signal from a spatial domain to an angle domain to obtain a transformed received signal matrix; adopting a double-structure-orthogonal matching pursuit algorithm to roughly estimate an angle domain cascaded sparse channel matrix; adopting a minimum mean square error method to calculate signal average power; setting an adaptive regularization coefficient according to the signal average power, reconstructing and transforming the roughly estimated angle domain cascaded sparse channel matrix to obtain a channel estimation result. The application can adaptively adjust the regularization coefficient to replace fixed parameter setting, and realize higher-precision channel estimation.

Description

Technical Field

[0001] This invention relates to the field of wireless communication technology, and more specifically to a channel estimation method for a Reconfigurable Intelligent Surface (RIS)-assisted communication system. Background Technology

[0002] Traditional communication relies on the inherent characteristics of the environment for signal transmission. Faced with unfavorable channel conditions (such as deep fading and congestion), existing methods mainly compensate by increasing transmission power or complex receiving processing, but the effects are limited and energy consumption is high. The environment itself, as a "passive" scatterer, cannot be actively controlled to optimize signal propagation, while the introduction of smart metasurfaces provides a revolutionary approach to solving these problems.

[0003] By intelligently controlling the electromagnetic response of these units in real time, RIS can dynamically reflect incident signals in the desired direction. This is equivalent to embedding a programmable "router" and "signal amplifier" in the wireless environment, actively shaping and optimizing the wireless propagation environment itself, which can significantly improve the quality of received signals and expand coverage. Theoretically, by utilizing RIS's active channel control capabilities, it is possible to overcome deep fading, penetrate obstacles, and concentrate signal energy, thereby breaking through the performance bottlenecks of traditional communication in complex scenarios and providing a key enabling technology for 6G and future networks.

[0004] However, the immense potential of RIS (Radio-Reflection System) heavily relies on accurate estimation of the cascaded channels (transmitter-RIS-receiver) constructed with its assistance. Without accurate Channel State Information (CSI), the active control capabilities of RIS cannot be realized. Furthermore, due to its numerous passive reflective elements, RIS requires consideration of RIS unit responses and their links with the transmitter and receiver, significantly increasing the dimensionality of the channel model. This also leads to substantial pilot overhead resources for channel estimation in RIS-assisted communication systems, consuming valuable spectrum resources and transmission time / pilot overhead, thus reducing the system's effective throughput. Traditional channel estimation algorithms, in large-scale RIS scenarios, either lack sufficient accuracy or suffer from unacceptably high computational complexity and pilot overhead. Therefore, efficient, high-precision, and low-overhead channel estimation is the foundation and prerequisite for subsequent advanced signal processing such as RIS phase optimization, beamforming, and resource allocation.

[0005] In recent years, researchers have applied compressed sensing theory from the field of signal processing to channel estimation, utilizing the angular domain sparsity of wireless channels to complete channel estimation with fewer pilot resources. However, current compressed sensing channel estimation algorithms, such as Orthogonal Matching Pursuit (OMP), Row-Structured Orthogonal Matching Pursuit (RS-OMP), and Double-Structured Orthogonal Matching Pursuit (DS-OMP), do not fully consider the impact of noise when estimating the channel, resulting in less than ideal channel estimation results and insufficient estimation accuracy. Summary of the Invention

[0006] In view of this, the present invention provides a channel estimation method for RIS-assisted communication systems based on MMSE-optimized compressed sensing. The present invention utilizes compressed sensing and least squares (LS) methods to perform a coarse estimation of the channel matrix, and then reconstructs the channel matrix using the minimum mean square error (MMSE) method, thus solving the problem of insufficient estimation accuracy in existing algorithms.

[0007] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0008] A channel estimation method for a RIS-assisted communication system based on MMSE-optimized compressed sensing includes the following steps:

[0009] Step 1: Obtain the uplink signal received by the base station in the RIS-assisted communication system, i.e., the received signal;

[0010] Step 2: Transform the received signal from the spatial domain to the angular domain to obtain the transformed received signal matrix;

[0011] Step 3: Use the dual-structure-orthogonal matching pursuit algorithm to roughly estimate the angle domain cascaded sparse channel matrix;

[0012] Step 4: Using the minimum mean square error method, calculate the average signal power based on the non-zero elements in the roughly estimated angle-domain cascaded sparse channel matrix; set the adaptive regularization coefficient based on the average signal power, reconstruct and transform the roughly estimated angle-domain cascaded sparse channel matrix, and obtain the channel estimation result.

[0013] Furthermore, the RIS-assisted communication system includes a multi-antenna base station equipped with M antennas, a RIS smart metasurface with N reflector elements, and K single-antenna users, where M, N, and K are all greater than 1. The RIS-assisted communication system adopts an orthogonal pilot transmission strategy, where all users transmit known pilot symbols to the base station via the RIS in the uplink within Q time slots, and the base station collects and receives the signals. In the q-th time slot, the signal received by the base station is:

[0014]

[0015] in, This represents the reflection channel matrix from the RIS to the base station. Represents a complex field matrix of dimension M×N; This represents the channel vector from the k-th user to the RIS. Represents a complex field vector of dimension N×1; diag(•) represents a diagonal matrix; This represents the reflection coefficient vector of RIS, and the nth element of the reflection coefficient vector. , , These represent the amplitude and phase of the nth RIS reflector unit, respectively. It is the pilot signal sent by the kth user. It is the noise received at the base station;

[0016] Both the RIS-to-base station channel and the user-to-RIS channel adopt the SV channel model, which includes... and Multipath components, of which and These represent the number of paths from the RIS to the base station and from the k-th user to the RIS, respectively, and are called row sparsity and column sparsity.

[0017] Furthermore, the specific method for step 2 is as follows:

[0018] Step 201: Stack the signals received by the base station in each time slot to obtain the received signal matrix:

[0019]

[0020] in, Here is the reflection coefficient matrix of RIS, and Q is the number of time slots. This is a cascaded channel matrix, where the superscript H denotes the conjugate transpose. This is the noise matrix;

[0021] Step 202, using the dictionary unitary matrix on the base station side. The received signal matrix is ​​transformed from the spatial domain to the angular domain to obtain the transformed received signal matrix:

[0022] .

[0023] Furthermore, step 3 is performed as follows:

[0024] Step 301, Input the received signal matrix Sparsity ,conduct In each iteration of the loop, the column energy of the received signal is calculated to find... The column with the highest energy; after After the loop, we get The columns with the highest energy are used to form a matrix. In addition, records The one with the highest energy is listed in The position in the middle constitutes the row support set. ;

[0025] Step 302, settings Initialize column support set Initialize residuals Set the maximum number of iterations to ;in This represents the k-th user's angle-domain concatenated sparse channel matrix. A set of all non-zero columns in a non-zero row, the number of elements of which is equal to the column sparsity. ;

[0026] Step 303, Calculate the measurement matrix The inner product of each column with the current residual is used as the matching degree between each column and the current residual. The column with the highest matching degree is selected and added to the column support set. In; among them, This is the dictionary unitary matrix on the RIS side;

[0027] Step 304: Use the least squares method to find the th element in the angle-domain concatenated sparse channel matrix. A rough estimate of the non-zero column element with the highest current matching degree in the non-zero rows:

[0028] ,

[0029] in, The symbol represents the calculation of the pseudoinverse. Indicates a set supported by columns The position information of the non-zero columns in the measurement matrix is ​​used to extract the measurement matrix, and the pseudo-inverse of the extracted submatrix is ​​calculated. For storing the first A vector containing the values ​​of all non-zero columns in a row. express The estimate;

[0030] Step 305, Update Residuals: ;

[0031] Step 306: Repeat steps 303 to 305 until the number of iterations reaches [number missing]. ,Will Increment the value by 1;

[0032] Step 307: Repeat steps 303 to 306 until... The angle-domain concatenated sparse channel matrix is ​​roughly estimated. ,satisfy:

[0033] .

[0034] Furthermore, step 4 is specifically implemented as follows:

[0035] Step 401, set the maximum number of iterations to... and set Calculate the input noise variance:

[0036]

[0037] in, For linear values ​​of transmission power, and These represent the channel power attenuation from the base station to the RIS and the channel power attenuation from the RIS to the user, respectively. This is a linear value for the signal-to-noise ratio;

[0038] Step 402, calculate the average signal power:

[0039]

[0040] in, This indicates taking a complex amplitude. Re and Im represent the real and imaginary parts, respectively. Indicates the first A vector containing all non-zero column elements in each of the non-zero rows;

[0041] Step 403, set adaptive regularization parameters ;

[0042] Step 404: Reconstruct the angle-domain concatenated sparse channel matrix using the MMSE estimation method. The values ​​of all non-zero columns in a row of non-zero rows:

[0043]

[0044] in, Represents the identity matrix; The angular domain cascaded sparse channel matrix represents the first... The values ​​of all non-zero columns in a non-zero row. This represents a submatrix formed by extracting the relevant columns of the non-zero rows in the angle-domain concatenated sparse channel matrix of the received signal through row support sets;

[0045] Step 405, will Increment the value by 1;

[0046] Step 406: Repeat steps 402 to 405 until... The reconstructed angle-domain concatenated sparse channel matrix is ​​obtained; the reconstructed angle-domain concatenated sparse channel matrix is ​​then transformed to the spatial domain to obtain the channel estimation result optimized by the MMSE method. .

[0047] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0048] 1. This invention uses the DS-OMP algorithm to first roughly estimate the RIS-assisted angle-domain concatenated sparse channel matrix, and then uses the adaptive MMSE method based on signal average power to reconstruct the roughly estimated channel, thereby achieving higher-precision channel estimation.

[0049] 2. This invention can adaptively adjust the regularization coefficient to replace the fixed parameter setting, avoiding the error accumulation problem caused by the fixed regularization coefficient in the traditional MMSE method, and has stronger anti-noise performance.

[0050] 3. The MMSE channel matrix reconstruction step added in this invention is only performed on the row and column support set determined by the DS-OMP algorithm. The computational load is much smaller than that of the overall compressed sensing reconstruction process. The overall computational complexity of the algorithm proposed in this invention is on the same order of magnitude as the original algorithm, and does not bring significant computational burden. Attached Figure Description

[0051] Figure 1 This is a schematic diagram of a scenario according to an embodiment of the present invention.

[0052] Figure 2 This is a comparison chart showing the normalized mean square error of the channel estimation algorithm proposed in this invention with that of several reference algorithms under a signal-to-noise ratio (SNR) of -10dB to 10dB. Detailed Implementation

[0053] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the present invention will be described in detail below with reference to the accompanying drawings.

[0054] A channel estimation method for a RIS-assisted communication system based on MMSE-optimized compressed sensing includes the following steps:

[0055] Step 1: Obtain the uplink signal received by the base station in the RIS-assisted communication system, i.e., the received signal;

[0056] Step 2: Transform the received signal from the spatial domain to the angular domain to obtain the transformed received signal matrix;

[0057] Step 3: Use the dual-structure-orthogonal matching pursuit algorithm to roughly estimate the angle domain cascaded sparse channel matrix;

[0058] Step 4: Using the minimum mean square error method, calculate the average signal power based on the non-zero elements in the roughly estimated angle-domain cascaded sparse channel matrix; set the adaptive regularization coefficient based on the average signal power, reconstruct and transform the roughly estimated angle-domain cascaded sparse channel matrix, and obtain the channel estimation result.

[0059] In one specific embodiment, the RIS-assisted communication system includes a multi-antenna base station equipped with M=64 antennas, a RIS smart metasurface with N=256 reflective elements, and K=16 single-antenna users. The distance between the base station and the RIS is... The distance between RIS and the user A schematic diagram of a communication system scenario is shown below. Figure 1 As shown, the communication system operates in Time Division Duplex (TDD) mode and employs an orthogonal pilot transmission strategy. All users transmit known pilot symbols to the base station via RIS within Q=64 time slots in the uplink. The base station collects and receives signals; in the q-th time slot, the signal received by the base station is:

[0060]

[0061] in, This represents the reflection channel matrix from the RIS to the base station. Represents a complex field matrix of dimension M×N; This represents the channel vector from the k-th user to the RIS. Represents a complex field vector of dimension N×1; diag(•) represents a diagonal matrix; This represents the reflection coefficient vector of RIS, and the nth element of the reflection coefficient vector. , , These represent the amplitude and phase of the nth RIS reflector unit, respectively. It is the pilot signal sent by the kth user. It is the noise received at the base station;

[0062] Both the RIS-to-base station channel and the user-to-RIS channel adopt the SV channel model, which includes... and Multipath components, of which and These represent the number of paths from the RIS to the base station and from the k-th user to the RIS, respectively, and are called row sparsity and column sparsity.

[0063] The specific method for step 2 is as follows:

[0064] Step 201: Stack the signals received by the base station in each time slot to obtain the received signal matrix:

[0065]

[0066] in, Here is the reflection coefficient matrix of RIS, and Q is the number of time slots. This is a cascaded channel matrix, where the superscript H denotes the conjugate transpose. This is the noise matrix;

[0067] Step 202, using the dictionary unitary matrix on the base station side. The received signal matrix is ​​transformed from the spatial domain to the angular domain to obtain the transformed received signal matrix:

[0068] .

[0069] The specific method for step 3 is as follows:

[0070] Step 301, Input the received signal matrix Sparsity ,conduct In each iteration of the loop, the column energy of the received signal is calculated to find... The column with the highest energy; after After the loop, we get The columns with the highest energy are used to form a matrix. In addition, records The one with the highest energy is listed in The position in the middle constitutes the row support set. ;

[0071] Step 302, settings Initialize column support set Initialize residuals Set the maximum number of iterations to ;in This represents the k-th user's angle-domain concatenated sparse channel matrix. A set of all non-zero columns in a non-zero row, the number of elements of which is equal to the column sparsity. ;

[0072] Step 303, Calculate the measurement matrix The inner product of each column with the current residual is used as the matching degree between each column and the current residual. The column with the highest matching degree is selected and added to the column support set. In; among them, This is the dictionary unitary matrix on the RIS side;

[0073] Step 304: Use the least squares method to find the th element in the angle-domain concatenated sparse channel matrix. A rough estimate of the non-zero column element with the highest current matching degree in the non-zero rows:

[0074] ,

[0075] in, The symbol represents the calculation of the pseudoinverse. Indicates a set supported by columns The position information of the non-zero columns in the measurement matrix is ​​used to extract the measurement matrix, and the pseudo-inverse of the extracted submatrix is ​​calculated. For storing the first A vector containing the values ​​of all non-zero columns in a row. express The estimate;

[0076] Step 305, Update Residuals: Remove the contribution of the non-zero column with the highest current matching degree;

[0077] Step 306: Repeat steps 303 to 305 until the number of iterations reaches [number missing]. ,Will Increment the value by 1;

[0078] Step 307: Repeat steps 303 to 306 until... The angle-domain concatenated sparse channel matrix is ​​roughly estimated. ,satisfy:

[0079] .

[0080] The specific method for step 4 is as follows:

[0081] Step 401, set the maximum number of iterations to... and set Calculate the input noise variance:

[0082]

[0083] in, For linear values ​​of transmission power, and These represent the channel power attenuation from the base station to the RIS and the channel power attenuation from the RIS to the user, respectively. This is a linear value for the signal-to-noise ratio;

[0084] Step 402, calculate the average signal power:

[0085]

[0086] in, This indicates taking a complex amplitude. Re and Im represent the real and imaginary parts, respectively. Indicates the first A vector containing all non-zero column elements in each of the non-zero rows;

[0087] Step 403, set adaptive regularization parameters The regularization parameter can be automatically adjusted according to the channel power, especially when the signal-to-noise ratio is high or the power is high. When the signal approaches zero, the algorithm's performance is similar to that of the LS method, especially at low signal-to-noise ratios or low power. The value will increase, suppressing the influence of noise; if The value is less than Then let ;

[0088] Step 404: Reconstruct the angle-domain concatenated sparse channel matrix using the MMSE estimation method. The values ​​of all non-zero columns in a row of non-zero rows:

[0089]

[0090] in, Represents the identity matrix; The angular domain cascaded sparse channel matrix represents the first... The values ​​of all non-zero columns in a non-zero row. This represents a submatrix formed by extracting the relevant columns of the non-zero rows in the angle-domain concatenated sparse channel matrix of the received signal through row support sets;

[0091] Step 405, will Increment the value by 1;

[0092] Step 406: Repeat steps 402 to 405 until... The reconstructed angle-domain concatenated sparse channel matrix is ​​obtained; the reconstructed angle-domain concatenated sparse channel matrix is ​​then transformed to the spatial domain to obtain the channel estimation result optimized by the MMSE method. .

[0093] Figure 2 Two algorithms were selected for comparison: the traditional compressed sensing method and the dual-structure orthogonal matching pursuit algorithm. It can be seen that under signal-to-noise ratio (SNR) conditions of -10dB to 0dB, all three channel estimation algorithms are severely affected by noise, resulting in high normalized mean square errors (MSEs). However, the algorithm proposed in this invention, utilizing prior knowledge such as noise variance, has better noise resistance and exhibits excellent estimation performance at low SNRs, with an estimation accuracy approximately 5dB higher than the traditional compressed sensing algorithm and approximately 2dB higher than the dual-structure orthogonal matching pursuit algorithm. As the SNR increases, the performance of all estimation algorithms improves and tends to converge. Under SNR conditions of 0dB to 10dB, the proposed algorithm has the lowest MSE, with an estimation accuracy approximately 3dB higher than the traditional compressed sensing algorithm and 0.2dB to 0.5dB higher than the dual-structure orthogonal matching pursuit algorithm. The proposed algorithm has advantages over other reference algorithms.

[0094] For those skilled in the art, various modifications and variations can be made to this invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this invention should be included within the scope of the claims of this invention.

Claims

1. A channel estimation method for a RIS-assisted communication system based on MMSE-optimized compressed sensing, characterized in that, Includes the following steps: Step 1: Obtain the uplink signal received by the base station in the RIS-assisted communication system, i.e., the received signal; Step 2: Transform the received signal from the spatial domain to the angular domain to obtain the transformed received signal matrix; Step 3: Use the dual-structure-orthogonal matching pursuit algorithm to roughly estimate the angle domain cascaded sparse channel matrix; Step 4: Using the minimum mean square error method, calculate the average signal power based on the non-zero elements in the roughly estimated angle-domain cascaded sparse channel matrix; set the adaptive regularization coefficient based on the average signal power, reconstruct and transform the roughly estimated angle-domain cascaded sparse channel matrix, and obtain the channel estimation result.

2. The channel estimation method for a RIS-assisted communication system based on MMSE-optimized compressed sensing according to claim 1, characterized in that, The RIS-assisted communication system includes a multi-antenna base station equipped with M antennas, a RIS smart metasurface with N reflective elements, and K single-antenna users, where M, N, and K are all greater than 1. The RIS-assisted communication system adopts an orthogonal pilot transmission strategy. All users transmit known pilot symbols to the base station through the RIS in Q time slots in the uplink, and collect the received signals at the base station. In the q-th time slot, the signal received by the base station is: in, This represents the reflection channel matrix from the RIS to the base station. Represents a complex field matrix of dimension M×N; This represents the channel vector from the k-th user to the RIS. Represents a complex field vector of dimension N×1; diag(•) represents a diagonal matrix; This represents the reflection coefficient vector of RIS, and the nth element of the reflection coefficient vector. , , These represent the amplitude and phase of the nth RIS reflector unit, respectively. It is the pilot signal sent by the kth user. It is the noise received at the base station; Both the RIS-to-base station channel and the user-to-RIS channel adopt the SV channel model, which includes... and Multipath components, of which and These represent the number of paths from the RIS to the base station and from the k-th user to the RIS, respectively, and are called row sparsity and column sparsity.

3. The channel estimation method for a RIS-assisted communication system based on MMSE-optimized compressed sensing according to claim 2, characterized in that, The specific method for step 2 is as follows: Step 201: Stack the signals received by the base station in each time slot to obtain the received signal matrix: in, Here is the reflection coefficient matrix of RIS, and Q is the number of time slots. This is a cascaded channel matrix, where the superscript H denotes the conjugate transpose. This is the noise matrix; Step 202, using the dictionary unitary matrix on the base station side. The received signal matrix is ​​transformed from the spatial domain to the angular domain to obtain the transformed received signal matrix: 。 4. The channel estimation method for a RIS-assisted communication system based on MMSE-optimized compressed sensing according to claim 3, characterized in that, The specific method for step 3 is as follows: Step 301, Input the received signal matrix Sparsity ,conduct In each iteration of the loop, the column energy of the received signal is calculated to find... The column with the highest energy; after After the loop, we get The columns with the highest energy are used to form a matrix. In addition, records The one with the highest energy is listed in The position in the middle constitutes the row support set. ; Step 302, settings Initialize column support set Initialize residuals Set the maximum number of iterations to ;in This represents the k-th user's angle-domain concatenated sparse channel matrix. A set of all non-zero columns in a non-zero row, the number of elements of which is equal to the column sparsity. ; Step 303, Calculate the measurement matrix The inner product of each column with the current residual is used as the matching degree between each column and the current residual. The column with the highest matching degree is selected and added to the column support set. In; among them, This is the dictionary unitary matrix on the RIS side; Step 304: Use the least squares method to find the th element in the angle-domain concatenated sparse channel matrix. A rough estimate of the non-zero column element with the highest current matching degree in the non-zero rows: ， in, The symbol represents the calculation of the pseudoinverse. Indicates a set supported by columns The position information of the non-zero columns in the measurement matrix is ​​used to extract the measurement matrix, and the pseudo-inverse of the extracted submatrix is ​​calculated. For storing the first A vector containing the values ​​of all non-zero columns in a row. express The estimate; Step 305, Update Residuals: ; Step 306: Repeat steps 303 to 305 until the number of iterations reaches [number missing]. ,Will Increment the value by 1; Step 307: Repeat steps 303 to 306 until... The angle-domain concatenated sparse channel matrix is ​​roughly estimated. ,satisfy: 。 5. The channel estimation method for a RIS-assisted communication system based on MMSE-optimized compressed sensing according to claim 4, characterized in that, The specific method for step 4 is as follows: Step 401, set the maximum number of iterations to... and set Calculate the input noise variance: in, For linear values ​​of transmission power, and These represent the channel power attenuation from the base station to the RIS and the channel power attenuation from the RIS to the user, respectively. This is a linear value for the signal-to-noise ratio; Step 402, calculate the average signal power: in, This indicates taking a complex amplitude. Re and Im represent the real and imaginary parts, respectively. Indicates the first A vector containing all non-zero column elements in each of the non-zero rows; Step 403, set adaptive regularization parameters ; Step 404: Reconstruct the angle-domain concatenated sparse channel matrix using the MMSE estimation method. The values ​​of all non-zero columns in a row of non-zero rows: in, Represents the identity matrix; The angular domain cascaded sparse channel matrix represents the first... The values ​​of all non-zero columns in a non-zero row. This represents a submatrix formed by extracting the relevant columns of the non-zero rows in the angle-domain concatenated sparse channel matrix of the received signal through row support sets; Step 405, will Increment the value by 1; Step 406: Repeat steps 402 to 405 until... The reconstructed angle-domain concatenated sparse channel matrix is ​​obtained; the reconstructed angle-domain concatenated sparse channel matrix is ​​then transformed to the spatial domain to obtain the channel estimation result optimized by the MMSE method. .

Images