IRS-assisted cloud access network downlink beamforming method with non-ideal CSI

Through the joint optimization of the transmit beamforming matrix of the BBU pool and RRH, the phase shift matrix of the reflective surface and the covariance matrix of the preamble quantized noise, the communication quality reduction caused by inaccurate channel state information is solved, and the intelligent reflection surface assistance is significantly improved The downlink transmission rate of the time division duplex cloud access network.

CN114900398BActive Publication Date: 2024-11-26ZHEJIANG UNIV OF TECH
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202210460209.0
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-04-28
Publication Date
2024-11-26
Estimated Expiration
2042-04-28

AI Technical Summary

Technical Problem

When the existing wireless communication system faces a large number of device access and high efficiency needs, inaccurate channel status information leads to a decline in communication quality, especially when the intelligent reflection surface assists the wireless downlink preamble link and connection of the intelligent reflection surface. It is difficult to effectively increase the rate when entering the link.

Method used

By jointly optimizing the transmit beamforming matrix of the BBU pool and RRH, the phase shift matrix of the reflective surface, and the preamble quantized noise covariance matrix, the downlink transmission rate of the entire system is maximized. The specific steps include defining and optimizing the multivariate compression quantized noise covariance matrix, transmit beamforming matrix and phase shift matrix to meet various constraints, and solving them using alternating optimization methods and semi-definite relaxation methods.

Benefits of technology

The downlink transmission rate of the intelligent reflective surface assisted time division duplex cloud access network is significantly improved, surpassing the performance of traditional cloud access networks without intelligent reflective surface assist.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN114900398B_ABST
    Figure CN114900398B_ABST
Patent Text Reader

Abstract

The present invention discloses a beamforming method for IRS-assisted cloud access network downlink with non-ideal CSI, wherein a multi-antenna baseband processing unit BBU pool communicates with multiple single-antenna users through a multi-antenna remote radio head RRH, wherein the BBU pool processes baseband signals through point-to-point compression or multi-element compression, and sends quantized bits to the RRH through a fronthaul link, and multiple intelligent reflective surface IRSs are deployed near the RRH to assist the wireless fronthaul link and access link in different time slots. The present invention aims to improve the system and rate, and jointly optimizes the transmit beamforming matrix of the BBU pool and RRH, the phase shift matrix of the reflective surface, and the fronthaul quantization noise covariance matrix, so as to effectively improve the downlink transmission rate of the cloud access network communication system when the channel state information CSI is inaccurate.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of wireless communications, and in particular to a communication system for a wireless downlink fronthaul link and an access link of a time-division duplex cloud access network assisted by an intelligent reflector with inaccurate channel state information (CSI), and a method for enhancing communication quality by improving downlink sum rate. Background Art

[0002] With the rapid development of wireless communication technology, the number of devices connected to the communication network is also growing rapidly, and the requirements for communication efficiency are also higher. Traditional wireless communication systems can no longer meet the needs.

[0003] Smart reflective surface is an emerging technology that can effectively enhance wireless communication systems. Smart reflective surface is a passive array with a large number of reprogrammable elements. The reflection amplitude and phase of each reflective element can be adjusted by a controller. It does not transmit the RF chain, so it does not introduce additional noise on the reflected signal. Smart reflective surface can more effectively improve the performance of wireless networks than traditional relays. Cloud access network (C-RAN) is a promising mobile network architecture that can improve the spectrum efficiency and energy efficiency of communication networks. C-RAN moves the baseband processing function of traditional base stations to the BBU pool. The RRH is deployed close to the mobile user. The multi-antenna BBU pool communicates to multiple single-antenna users through multiple multi-antenna RRHs. The cloud access network wireless downlink fronthaul link and access link are assisted by smart reflective surfaces. The RRH not only receives the signal of the direct link sent by the BBU pool, but also receives the signal of the IRS reflection path. The user also receives the signal from the RRH direct link and the signal reflected by the IRS, which is more advantageous than the cloud access network communication without the assistance of smart reflective surfaces. The cloud wireless access network communication system assisted by the intelligent reflective surface can improve the efficiency of the communication system and realize the intelligence of the communication network.

[0004] The performance of the cloud access network communication system assisted by the intelligent reflector can be further improved by jointly optimizing the transmit beamforming matrix of the BBU pool and RRH, the phase shift matrix of the reflector, and the fronthaul quantization noise covariance matrix. Summary of the invention

[0005] The purpose of the present invention is to provide a communication system for a wireless downlink fronthaul link and access link of a time-division duplex cloud access network assisted by an intelligent reflector with inaccurate channel state information (CSI), and to optimize the precoding matrix, the phase shift matrix of the reflector, and the fronthaul quantization noise covariance matrix to improve the system downlink and rate method. That is, by jointly optimizing the transmit beamforming matrix of the BBU pool and RRH, the phase shift matrix of the reflector, and the fronthaul quantization noise covariance matrix, the purpose of maximizing the downlink and rate of the entire system is achieved.

[0006] The technical solution of the present invention is:

[0007] The IRS-assisted cloud access network downlink beamforming method with non-ideal CSI adopts an intelligent reflector to assist the communication system of the wireless downlink fronthaul link and access link of the cloud access network. The system jointly optimizes the fronthaul link quantization noise covariance matrix, the transmit beamforming matrix of the BBU pool and RRH, and the reflector phase shift matrix in order to maximize the system downlink and rate. The specific steps include the following:

[0008] 1.1) In the downlink cloud access network communication system with intelligent reflective surface assisting the fronthaul link and access link, the multi-antenna BBU pool communicates with K single-antenna users through L multi-antenna RRHs, where the BBU pool processes the baseband signal through point-to-point compression or multi-element compression, and sends the quantized bits to the RRHs through the fronthaul link. I intelligent reflective surfaces are deployed near the RRHs, and the number of IRSs assisting the wireless fronthaul link and access link in different time slots is I respectively. F and I A The above three groups of IRS are expressed as and The number of antennas for each BBU and RRH is N B and N R , the number of reflection units of IRS is M. RRH is a half-duplex node, working in time division duplex (TDD) mode, each time slot is divided into two parts (1-α0) and α0, which are used for the transmission of BBU-RRH fronthaul link and RRH-user access link respectively.

[0009] 1.2) In the fronthaul link, the BBU pool first encodes the downlink message of user k into a baseband signal s k , and then linearly precode the signals of all users into where v k is the transmit beamforming matrix of user k on all RRHs, is the signal transmitted by RRH 1. Before being transmitted to the RRH Perform quantization compression: E l It is LN R ×N R The matrix of (l-1)N R +1 line to lN R The row is of size N R Except for the identity matrix, the remaining elements are 0; Indicates independence from The quantization noise of all RRHs is defined as q = [q1; ...; q L ], Ω is the complex Gaussian distribution covariance matrix of q, Ωl,l It is the lth dimension N on the diagonal of Ω R The identity matrix of x l Transmitted to RRH1, the BBU pool encodes its corresponding compression index to generate a baseband signal t l , the signal transmitted by the BBU pool is where F l is the transmit beamforming matrix, which is affected by the BBU transmit power P B constraint:

[0010] 1.3) Define H l,B , G i,B and G l,i They are the channel matrices from BBU pool to RRH 1, BBU pool to IRS i, and IRS i to RRH 1, respectively. The received signal of RRH 1 is:

[0011]

[0012] in From BBU pool to IRS set and from the IRS The channel matrix to RRH1. express The passive beamforming matrix of assuming that the IRS only adjusts the phase shift, The phase shift of the mth element in is the additive white Gaussian noise of all channels from BBU to RRH l, Yes F,l Complex Gaussian distribution covariance matrix. The received signal of the rth antenna of RRH l is:

[0013]

[0014] where h l,B,r and represents the rth antenna from the BBU pool to RRH l and from the IRS set The channel vectors to the rth antenna of RRH l are H l,B and The rth row vector of is collected from the BBU pool through the IRS The cascaded channel reflected to the rth antenna of RRH l, is included Vector of diagonal elements, n F,l,r Indicates n F,l The rth element of .

[0015] 1.4) In the access link, RRH l converts the compressed signal x l Forwarded to all users. The maximum transmission power of RRH is P R , according to 1.2), the transmit power constraint of RRH 1 is: h k,l and g k,i denote the distance from RRH l to user k and from The channel vector to user k, G i,l is the channel matrix from RRH l to IRS i, is the phase shift matrix of IRS i, where express The phase shift of the mth element, the signal received by user k is:

[0016]

[0017] in is included A vector of diagonal elements, RRH set The channel matrix to user k, is the concatenated channel from the RRH set to user k, From RRH l through IRS The cascaded channel to user k is and Respectively represent the IRS set to user k and from RRH l to the IRS set The channel matrix, q is the quantization noise vector of all RRHs, is the additive white Gaussian noise at user k.

[0018] 1.5) The direct channels from the BBU pool to RRH l and from RRH l to user k are expressed as: in and is the estimated CSI, and is the corresponding channel estimation error, For the cascaded channels of BBU-IRS-RRH and RRH-IRS-user, the channel matrix is: and is the estimated cascade is the corresponding channel estimation error, and Respectively represent and The covariance matrix of the complex Gaussian distribution, the CSI error of each channel is independent of each other.

[0019] Furthermore, the transmit beamforming matrix F of the BBU pool and RRH l and v k , the IRS phase shift matrix Θ of the auxiliary access link and the fronthaul link A and θ F And the forward quantization noise covariance matrix Ω is optimized to maximize the system and rate. The specific steps are as follows:

[0020] 2.1) The communication system for the intelligent reflector-assisted downlink time-division duplex cloud access network wireless fronthaul link and access link described in steps 1.1)-1.5) is characterized by:

[0021] 2.1.1) The lower limit of the user's achievable rate is:

[0022]

[0023] in,

[0024]

[0025]

[0026] IRS set The total number of reflective elements.

[0027] 2.1.2) The achievable rate of the wireless fronthaul link should meet the following requirements:

[0028]

[0029] in,

[0030]

[0031]

[0032]

[0033]

[0034] 2.1.3) BBU pool precoding signal Compression is performed, and the output rate of the compressor cannot exceed the achievable rate of the fronthaul link. Two compression strategies are considered: point-to-point compression and multi-dimensional compression.

[0035] The forward constraints for point-to-point compression are:

[0036]

[0037] Point-to-point compression generates independent quantization noise on the RRHs, so the quantization noise covariance matrix of all RRHs is a block diagonal matrix, that is,

[0038] The forward constraints for multivariate compression are:

[0039]

[0040] Multivariate compression makes the quantization noise of each RRH correlated, so the overall quantization noise covariance matrix Ω is a complete matrix.

[0041] 2.2) With the goal of maximizing the downlink and rate, the optimization problem of optimizing the above system parameters when the BBU pool adopts multivariate compression can be expressed as:

[0042]

[0043]

[0044]

[0045]

[0046]

[0047]

[0048]

[0049] Ω≥0 (1h)

[0050] ω k represents the weight of each user, (1b) represents the achievable rate constraint of the fronthaul link, (1c) is the fronthaul compression constraint, (1d) and (1e) represent the transmit power constraints of the BBU pool and each RRH, respectively. (1f) and (1g) are the unit modulo 1 constraints of the passive beamforming matrices of the IRS-assisted fronthaul link and access link, respectively, and (1h) indicates that the fronthaul quantization noise covariance matrix is ​​a semi-positive definite matrix.

[0051] 2.3) Transform problem (P1) and then solve it.

[0052] 2.3.1) Using the mean square error (MSE) method, the objective function of (P1) is transformed into:

[0053]

[0054] w A,k is the auxiliary variable introduced, u A,k is from y at user k A,k Medium estimatek The linear estimator of where y A,k is the signal received by user k, s k is the baseband signal obtained after the downlink message of user k is encoded by BBU, 1 is s k The dimension of is the mean square error:

[0055]

[0056] When A,k and u A,k When the following values ​​are obtained, R sum Get the optimal value:

[0057]

[0058]

[0059] 2.3.2) Similar to 2.3.1), constraint (1b) can be approximated as follows using the MSE method:

[0060]

[0061] W F,l It is required to be semi-positive definite, which is an auxiliary variable introduced. is RRH l from y F,l The estimated signal t l The linear receiver, that is, the predicted signal is where y F,l is the signal received by RRH l, t l Is BBU to x l The baseband signal obtained by encoding the compressed index is d R Yes l Dimension. is the mean square error matrix:

[0062]

[0063] When W F,l and U F,l When the following values ​​are taken, the right side of formula (6) reaches the maximum value:

[0064]

[0065]

[0066] 2.3.3) For constraint (1c), let (1c) can be rewritten as:

[0067]

[0068] Constraint (1c) can be further transformed into the following approximate constraints:

[0069]

[0070] ∑ l It is required to be semi-positive definite, which is an auxiliary variable introduced. Is a collection The number of RRHs in the l When taking the following values, (10) is equivalent to (1c):

[0071]

[0072] 2.3.4) Transform the optimization problem (P1) into:

[0073]

[0074]

[0075]

[0076]

[0077]

[0078]

[0079]

[0080] Ω≥0 (12h)

[0081] Update the auxiliary variable w through (4), (5), (8), (9) and (11) A,k 、u A,k , W F,l , U F,l and∑ l For a fixed w A,k 、u A,k , W F,l , U F,l and∑ l , we optimize F by solving the following problem l 、v k ,Ω,Θ A and θ F , where ω k Represents the weight of each user:

[0082]

[0083] st(12b)~(12h) (13b)

[0084] 2.4) Next, problem (P3) is decomposed into three sub-problems and solved alternately.

[0085] 2.4.1) First, fix Ω and Θ in problem (P3) A and θ F To optimize the transmit beamforming matrix F of the BBU pool and RRH l and v k The first sub-problem is given by:

[0086]

[0087]

[0088]

[0089]

[0090]

[0091] Problem (P3.1) is convex and can be solved using some standard optimization tools (e.g., CVX).

[0092] 2.4.2) The second sub-problem is to fix F l 、v k and θ A , optimize the fronthaul quantization noise covariance matrix Ω and the IRS phase shift matrix Θ for the auxiliary fronthaul link F The second sub-question is as follows:

[0093]

[0094]

[0095]

[0096]

[0097]

[0098] Ω≥0 (15f)

[0099] It can be seen that the objective (15a), constraints (15c) and (15d) are convex with respect to Ω. Next, constraint (15b) is transformed according to And formula (6) is The definition of , (15b) can be rewritten as follows:

[0100]

[0101] in,

[0102]

[0103]

[0104] d R is the data dimension of each RRH, in the above formulas:

[0105]

[0106]

[0107] Problem (P3.2) is transformed into the following question:

[0108]

[0109]

[0110]

[0111]

[0112]

[0113]

[0114]

[0115] Constraint (17f) is still non-convex. Apply the semidefinite relaxation method (SDR) to remove constraint (17f) and use the standard optimization tool CVX to obtain the optimal solution. If The rank of the optimal solution is not 1, so Perform randomization to produce feasible suboptimal solutions, and update Ω and Θ only when the objective function value of (P3.2.1) increases. F .

[0116] 2.4.3) Fixed F l 、v k , Ω and Θ F , for Θ A After optimization, we get the third sub-problem:

[0117]

[0118]

[0119] The objective function is rewritten as follows:

[0120]

[0121] in,

[0122]

[0123] In the above formula,

[0124]

[0125]

[0126] Subproblem (P3.3) can be rewritten as:

[0127]

[0128]

[0129]

[0130] Problem (P3.3.1) is similar to (P3.2.1). We can also use the semidefinite relaxation method (SDR) to obtain a feasible suboptimal solution using a randomized method, and update Θ only when the objective function value of (P3.3.1) increases. A .

[0131] 2.5) Repeat steps 2.4.1)-2.4.3) until convergence.

[0132] 2.6) When the BBU pool adopts the point-to-point compression method, the transmit beamforming matrix of the BBU pool and RRH, the phase shift matrix of the IRS, and the forward quantization noise covariance matrix are jointly optimized. Considering the forward compression constraint of the point-to-point compression, the problem can be expressed as:

[0133]

[0134] st(1b),(1d)~(1g), (19b)

[0135]

[0136]

[0137] The difference between problems (P4) and (P1) lies in constraints (19c) and (19d), where (19c) is relative to Ω l,l and v k is non-convex. Let is rewritten as:

[0138]

[0139] log|Ol The upper limit of | satisfies: S l ≥0 is an auxiliary variable, so (19c) can be replaced by the following convex constraint:

[0140]

[0141] The objective function and other constraints in problem (P4) are treated in the same way as problem (P1), and then similar methods can be applied to solve the problem using continuous convex approximation methods and alternating optimization methods.

[0142] 2.7) For the discrete phase of the IRS reflector, first follow the steps in 2.1)-2.3) until convergence, and obtain Θ A and θ F , its diagonal elements θ A,i,m and θ F,i,m Mapped to discrete phase points, determine Θ F Then we scale Ω to get the solution that satisfies the constraints of problem (P3.2.1).

[0143] The beneficial effect of the present invention is that for the intelligent reflector-assisted time-division duplex cloud access network communication system, by jointly optimizing the transmit beamforming matrix of the BBU pool and RRH, the phase shift matrix of the reflector, and the forward quantization noise covariance matrix, the total downlink transmission rate is significantly higher than that of the traditional cloud access network without the assistance of the intelligent reflector. BRIEF DESCRIPTION OF THE DRAWINGS

[0144] Figure 1 It is a model diagram of the intelligent reflector-assisted time-division duplex cloud access network wireless downlink communication system in the present invention;

[0145] Figure 2 The diagram is a relationship between the downlink sum rate and the number of reflection units of the intelligent reflection surface of the intelligent reflection surface-assisted time-division duplex cloud access network wireless downlink fronthaul link and access link communication system after adopting the joint optimization method of the present invention. DETAILED DESCRIPTION

[0146] The present invention will be further described below in conjunction with the accompanying drawings.

[0147] like Figure 1-2 As shown, Figure 1 The multi-antenna BBU pool communicates with K single-antenna users through L multi-antenna RRHs. The BBU pool processes the baseband signal through point-to-point compression or multi-element compression, and sends the quantized bits to the RRHs through the fronthaul link. I smart reflectors are deployed near the RRHs. The number of IRSs assisting the wireless fronthaul link and access link in different time slots is I respectively. F and IA The RRH works in time division duplex mode, and each time slot is divided into two parts: (1-α0) and α0, which are used for the transmission of the BBU-RRH fronthaul link and the RRH-user access link respectively; Figure 2 Comparison Scheme 1 indicates that when the BBU pool adopts multivariate compression, the transmit beamforming matrix of the BBU pool and RRH and the forward quantization noise covariance matrix are jointly optimized, and the phase of the reflection surface is randomly selected; "Comparison Scheme 2" indicates that there is no reflection surface assistance; "This patented method, multivariate, continuous" indicates that the BBU pool adopts multivariate compression, and the phase of the reflection surface is between 0-2π; "This patented method, multivariate, 2bit" indicates that the BBU pool adopts multivariate compression, and the phase level of the reflection surface is 4; "This patented method, point-to-point, continuous" and "This patented method, point-to-point, 2bit" are respectively the cases where the BBU pool adopts point-to-point compression, the phase of the reflection surface is between 0-2π and the phase level of the reflection surface is 4.

[0148] For the communication system transmission process of the wireless downlink fronthaul link and access link of the intelligent reflector-assisted time-division duplex cloud access network with inaccurate channel state information, before the transmission starts, the BBU pool first obtains the inaccurate system channel information, and then jointly optimizes the fronthaul link quantization noise covariance matrix, the BBU pool and RRH precoding matrix, and the reflector phase shift matrix.

[0149] The transmission process is as follows: the multi-antenna BBU pool communicates with K single-antenna users through L multi-antenna RRHs. I smart reflectors are deployed near the RRHs. The number of IRSs assisting the wireless fronthaul link and access link in different time slots is I and F and I A , so that the RRH not only receives the signal of the direct link sent by the BBU pool, but also receives the signal of the IRS reflection path. Similarly, the user also receives the sum of the signal from the RRH direct link and the IRS reflection. The specific steps are as follows:

[0150] 1.1) In the downlink cloud access network communication system with intelligent reflective surface assisting the fronthaul link and access link, the multi-antenna BBU pool communicates with K single-antenna users through L multi-antenna RRHs, where the BBU pool processes the baseband signal through point-to-point compression or multi-element compression, and sends the quantized bits to the RRHs through the fronthaul link. I intelligent reflective surfaces are deployed near the RRHs, and the number of IRSs assisting the wireless fronthaul link and access link in different time slots is I respectively. F and I A The above three groups of IRS are expressed as and The number of antennas for each BBU and RRH is N B and N R, the number of reflection units of IRS is M. RRH is a half-duplex node, working in time division duplex (TDD) mode, each time slot is divided into two parts (1-α0) and α0, which are used for the transmission of BBU-RRH fronthaul link and RRH-user access link respectively.

[0151] 1.2) In the fronthaul link, the BBU pool first encodes the downlink message of user k into a baseband signal s k , and then linearly precode the signals of all users into where v k is the transmit beamforming matrix of user k on all RRHs, is the signal transmitted by RRH 1. Before being transmitted to the RRH Perform quantization compression: E l It is LN R ×N R The matrix of (l-1)N R +1 line to lN R The row is of size N R Except for the identity matrix, the remaining elements are 0; Indicates independence from The quantization noise of all RRHs is defined as q = [q1; ...; q L ], Ω is the complex Gaussian distribution covariance matrix of q, Ω l,l It is the lth dimension N on the diagonal of Ω R The identity matrix of x l Transmitted to RRH1, the BBU pool encodes its corresponding compression index to generate a baseband signal t l , the signal transmitted by the BBU pool is where F l is the transmit beamforming matrix, which is affected by the BBU transmit power P B constraint:

[0152] 1.3) Define H l,B , G i,B and G l,i They are the channel matrices from BBU pool to RRH1, BBU pool to IRS i, and IRS i to RRH1, and the received signal of RRH1 is:

[0153]

[0154] in From BBU pool to IRS set and from the IRS The channel matrix to RRH1. express The passive beamforming matrix of assuming that the IRS only adjusts the phase shift, The phase shift of the mth element in is the additive white Gaussian noise of all channels from BBU to RRH l, Yes F,l Complex Gaussian distribution covariance matrix. The received signal of the rth antenna of RRH l is:

[0155]

[0156] where h l,B,r and represents the rth antenna from the BBU pool to RRH1 and from the IRS set The channel vectors to the rth antenna of RRHl are H l,B and The rth row vector of is collected from the BBU pool through the IRS The cascaded channel reflected to the rth antenna of RRH l, is included Vector of diagonal elements, n F,l,r Indicates n F,l The rth element of .

[0157] 1.4) In the access link, RRH1 converts the compressed signal x l Forwarded to all users. The maximum transmission power of RRH is P R , from 1.2), the transmit power constraint of RRH1 is: h k,l and g k,i denote the distance from RRH l to user k and from The channel vector to user k, G i,l is the channel matrix from RRH l to IRS i, is the phase shift matrix of IRS i, where express The phase shift of the mth element, the signal received by user k is:

[0158]

[0159] in is included A vector of diagonal elements, RRH set The channel matrix to user k, is the concatenated channel from the RRH set to user k, It is collected from RRH1 through IRS The cascaded channel to user k is and Respectively represent the IRS set to user k and from RRH l to the IRS set The channel matrix, q is the quantization noise vector of all RRHs, is the additive white Gaussian noise at user k.

[0160] 1.5) The direct channels from the BBU pool to RRH l and from RRH l to user k are expressed as: in and is the estimated CSI, and is the corresponding channel estimation error, For the cascaded channels of BBU-IRS-RRH and RRH-IRS-user, the channel matrix is: and is the estimated cascade is the corresponding channel estimation error, and Respectively represent and The covariance matrix of the complex Gaussian distribution, the CSI error of each channel is independent of each other.

[0161] Transmit beamforming matrix F of BBU pool and RRH l and v k , the IRS phase shift matrix Θ of the auxiliary access link and the fronthaul link A and θ F And the forward quantization noise covariance matrix Ω is optimized to maximize the system and rate. The specific steps are as follows:

[0162] 2.1) The communication system for the intelligent reflector-assisted downlink time-division duplex cloud access network wireless fronthaul link and access link described in steps 1.1)-1.5) is characterized by:

[0163] 2.1.1) The lower limit of the user's achievable rate is:

[0164]

[0165] in,

[0166]

[0167]

[0168] IRS set The total number of reflective elements.

[0169] 2.1.2) The achievable rate of the wireless fronthaul link should meet the following requirements:

[0170]

[0171] in,

[0172]

[0173]

[0174]

[0175]

[0176] 2.1.3) BBU pool precoding signal Compression is performed, and the output rate of the compressor cannot exceed the achievable rate of the fronthaul link. Two compression strategies are considered: point-to-point compression and multi-dimensional compression.

[0177] The forward constraints for point-to-point compression are:

[0178]

[0179] Point-to-point compression generates independent quantization noise on the RRHs, so the quantization noise covariance matrix of all RRHs is a block diagonal matrix, that is,

[0180] The forward constraints for multivariate compression are:

[0181]

[0182] Multivariate compression makes the quantization noise of each RRH correlated, so the overall quantization noise covariance matrix Ω is a complete matrix.

[0183] 2.2) With the goal of maximizing the downlink and rate, the optimization problem of optimizing the above system parameters when the BBU pool adopts multivariate compression can be expressed as:

[0184]

[0185]

[0186]

[0187]

[0188]

[0189]

[0190]

[0191] Ω≥0 (1h)

[0192] ω k represents the weight of each user, (1b) represents the achievable rate constraint of the fronthaul link, (1c) is the fronthaul compression constraint, (1d) and (1e) represent the transmit power constraints of the BBU pool and each RRH, respectively. (1f) and (1g) are the unit modulo 1 constraints of the passive beamforming matrices of the IRS-assisted fronthaul link and access link, respectively, and (1h) indicates that the fronthaul quantization noise covariance matrix is ​​a semi-positive definite matrix.

[0193] 2.3) Transform problem (P1) and then solve it.

[0194] 2.3.1) Using the mean square error (MSE) method, the objective function of (P1) is transformed into:

[0195]

[0196] w A,k is the auxiliary variable introduced, u A,k is from y at user k A,k Medium estimate k The linear estimator of where y A,k is the signal received by user k, s k is the baseband signal obtained after the downlink message of user k is encoded by BBU, 1 is s k The dimension of is the mean square error:

[0197]

[0198] When A,k and u A,k When the following values ​​are obtained, R sum Get the optimal value:

[0199]

[0200]

[0201] 2.3.2) Similar to 2.3.1), constraint (1b) can be approximated as follows using the MSE method:

[0202]

[0203] W F,l It is required to be semi-positive definite, which is an auxiliary variable introduced. is RRH l from y F,l The estimated signal t l The linear receiver, that is, the predicted signal is where y F,l is the signal received by RRH l, t l Is BBU to x l The baseband signal obtained by encoding the compressed index is d R Yes l Dimension. is the mean square error matrix:

[0204]

[0205] When W F,l and U F,l When the following values ​​are taken, the right side of formula (6) reaches the maximum value:

[0206]

[0207]

[0208] 2.3.3) For constraint (1c), let (1c) can be rewritten as:

[0209]

[0210] Constraint (1c) can be further transformed into the following approximate constraints:

[0211]

[0212] ∑ l It is required to be semi-positive definite, which is an auxiliary variable introduced. Is a collection The number of RRHs in the l When taking the following values, (10) is equivalent to (1c):

[0213]

[0214] 2.3.4) Transform the optimization problem (P1) into:

[0215]

[0216]

[0217]

[0218]

[0219]

[0220]

[0221]

[0222] Ω≥0 (12h)

[0223] Update the auxiliary variable w through (4), (5), (8), (9) and (11) A,k 、u A,k , W F,l , U F,l and∑ l For a fixed w A,k 、u A,k , W F,l , U F,l and∑ l , we optimize F by solving the following problem l 、v k ,Ω,Θ A and θ F , where ω k Represents the weight of each user:

[0224]

[0225] st(12b)~(12h) (13b)

[0226] 2.4) Next, problem (P3) is decomposed into three sub-problems and solved alternately.

[0227] 2.4.1) First, fix Ω and Θ in problem (P3) A and θ F To optimize the transmit beamforming matrix F of the BBU pool and RRH l and v k The first sub-problem is given by:

[0228]

[0229]

[0230]

[0231]

[0232]

[0233] Problem (P3.1) is convex and can be solved using some standard optimization tools (e.g., CVX).

[0234] 2.4.2) The second sub-problem is to fix F l 、v k and θ A , optimize the fronthaul quantization noise covariance matrix Ω and the IRS phase shift matrix Θ for the auxiliary fronthaul link F The second sub-question is as follows:

[0235]

[0236]

[0237]

[0238]

[0239]

[0240] Ω≥0 (15f)

[0241] It can be seen that the objective (15a), constraints (15c) and (15d) are convex with respect to Ω. Next, constraint (15b) is transformed according to And formula (6) is The definition of , (15b) can be rewritten as follows:

[0242]

[0243] in,

[0244]

[0245]

[0246] d R is the data dimension of each RRH, in the above formulas:

[0247]

[0248]

[0249] Problem (P3.2) is transformed into the following question:

[0250]

[0251]

[0252]

[0253]

[0254]

[0255]

[0256]

[0257] Constraint (17f) is still non-convex. Apply the semidefinite relaxation method (SDR) to remove constraint (17f) and use the standard optimization tool CVX to obtain the optimal solution. If The rank of the optimal solution is not 1, so Perform randomization to produce feasible suboptimal solutions, and update Ω and Θ only when the objective function value of (P3.2.1) increases. F .

[0258] 2.4.3) Fixed F l 、v k , Ω and Θ F , for Θ A After optimization, we get the third sub-problem:

[0259]

[0260]

[0261] The objective function is rewritten as follows:

[0262]

[0263] in,

[0264]

[0265] In the above formula,

[0266]

[0267]

[0268] Subproblem (P3.3) can be rewritten as:

[0269]

[0270]

[0271]

[0272] Problem (P3.3.1) is similar to (P3.2.1). We can also use the semidefinite relaxation method (SDR) to obtain a feasible suboptimal solution using a randomized method, and update Θ only when the objective function value of (P3.3.1) increases. A .

[0273] 2.5) Repeat steps 2.4.1)-2.4.3) until convergence.

[0274] 2.6) When the BBU pool adopts the point-to-point compression method, the transmit beamforming matrix of the BBU pool and RRH, the phase shift matrix of the IRS, and the forward quantization noise covariance matrix are jointly optimized. Considering the forward compression constraint of the point-to-point compression, the problem can be expressed as:

[0275]

[0276] st(1b),(1d)~(1g), (19b)

[0277]

[0278]

[0279] The difference between problems (P4) and (P1) lies in constraints (19c) and (19d), where (19c) is relative to Ω l,l and v k is non-convex. Let (19c) is rewritten as:

[0280]

[0281] log|O l The upper limit of | satisfies: S l ≥0 is an auxiliary variable, so (19c) can be replaced by the following convex constraint:

[0282]

[0283] The objective function and other constraints in problem (P4) are treated in the same way as problem (P1), and then similar methods can be applied to solve the problem using continuous convex approximation methods and alternating optimization methods.

[0284] 2.7) For the discrete phase of the IRS reflector, first follow the steps in 2.1)-2.3) until convergence, and obtain Θ A and θ F , its diagonal elements θ A,i,m and θ F,i,m Mapped to discrete phase points, determine Θ F Then we scale Ω to get the solution that satisfies the constraints of problem (P3.2.1).

[0285] Computer simulation shows that for the communication system of the wireless downlink fronthaul link and access link of the intelligent reflector-assisted time-division duplex cloud access network with inaccurate channel state information, after adopting the joint optimization method of this patent, its downlink and rate have achieved effective gain, which is significantly higher than the traditional cloud access network without intelligent reflector assistance.

Claims

1. An IRS-assisted cloud access network downlink beamforming method for non-ideal CSI, characterized by: When the channel state information CSI is inaccurate, for the communication system of the downlink fronthaul link and access link of the intelligent reflector IRS-assisted cloud access network, the transmit beamforming matrix of the BBU pool and RRH, the phase shift matrix of the reflector, and the fronthaul quantization noise covariance matrix are jointly optimized for the purpose of improving the system rate, including the following steps: 1.1) In the communication system of the downlink fronthaul link and access link of the cloud access network assisted by the intelligent reflector, the multi-antenna BBU pool communicates with K single-antenna users through L multi-antenna RRHs, where the BBU pool processes the baseband signal through point-to-point compression or multi-element compression, and sends the quantized bits to the RRHs through the fronthaul link. I intelligent reflectors are deployed near the RRHs, and the number of IRSs assisting the wireless fronthaul link and access link in different time slots is I respectively. F and I A The above three groups of IRS are expressed as and The number of antennas for each BBU and RRH is N B and N R , the number of reflection units of IRS is M; RRH is a half-duplex node, working in time division duplex TDD mode. Each time slot is divided into two parts: (1-α0) and α0, which are used for the transmission of BBU-RRH fronthaul link and RRH-user access link respectively; 1.2) In the fronthaul link, the BBU pool first encodes the downlink message of user k into a baseband signal s k , and then linearly precode the signals of all users into where v k is the transmit beamforming matrix of user k on all RRHs, is the signal transmitted by RRH1; Before being transmitted to the RRH Perform quantization compression: E l It is LN R ×N R The matrix, except (l-1)N R +1 line to lN R The row is of size N R Except for the identity matrix, the remaining elements are 0; Indicates independence from The quantization noise of all RRHs is defined as q = [q1; ...; q L ], q~CN(0,Ω), Ω is the complex Gaussian distribution covariance matrix of q, Ω l,l It is the lth dimension N on the diagonal of Ω R The identity matrix of x l Transmitted to RRH1, the BBU pool encodes its corresponding compression index to generate a baseband signal t l , the signal transmitted by the BBU pool is where F l is the transmit beamforming matrix, which is affected by the BBU transmit power P B constraint: 1.3) Define H l,B , G i,B and G l,i They are the channel matrices from BBU pool to RRH1, BBU pool to IRSi, and IRSi to RRH1. The received signal of RRH1 is: in From BBU pool to IRS set and from the IRS The channel matrix to RRH1; represents the passive beamforming matrix of IRSi, Assuming that IRS only adjusts the phase shift, the phase shift of the mth element of IRSi is in is the additive white Gaussian noise of all channels from BBU to RRH1, Yes F,l Complex Gaussian distribution covariance matrix; The received signal of the rth antenna of RRH1 is: where h l,B,r and represents the rth antenna from the BBU pool to RRH1 and from the IRS set The channel vectors to the rth antenna of RRHl are H l,B and The rth row vector of is collected from the BBU pool through the IRS The cascaded channel reflected to the rth antenna of RRH l, is included vector of diagonal elements, n F,l,r Indicates n F,l The rth element of ; 1.4) In the access link, RRH l converts the compressed signal x l Forwarded to all users; the maximum transmit power of RRH is P R , from step 1.2), the transmit power constraint of RRH1 is: h k,l and g k,i They represent from RRH1 to user k and from IRSi, The channel vector to user k, G i,l is the channel matrix from RRH 1 to IRSi, is the phase shift matrix of IRSi, where represents IRS i, The phase shift of the mth element of , the signal received by user k is: in is included A vector of diagonal elements, RRH set The channel matrix to user k, is the concatenated channel from the RRH set to user k, From RRH l through IRS The cascaded channel to user k is and Respectively represent the IRS set to user k and from RRH l to the IRS set The channel matrix, q is the quantization noise vector of all RRHs, is the additive white Gaussian noise at user k; 1.5) The direct channels from the BBU pool to RRH1 and from RRH1 to user k are expressed as: in and is the estimated CSI, and is the corresponding channel estimation error, For the cascaded channels of BBU-IRS-RRH and RRH-IRS-user, the channel matrix is: and is the estimated cascade is the corresponding channel estimation error, and Respectively represent and The covariance matrix of the complex Gaussian distribution, the CSI error of each channel is independent of each other; The transmit beamforming matrix F of the BBU pool and RRH is l and v k , the IRS phase shift matrix Θ of the auxiliary access link and the fronthaul link A and θ F And the forward quantization noise covariance matrix Ω is optimized to maximize the system rate. The specific steps are as follows: 2.1) For the communication system of the intelligent reflector-assisted downlink time-division duplex cloud access network wireless fronthaul link and access link described in steps 1.1)-1.5): 2.1.1) The lower limit of the user's achievable rate is: in, IRS set The total number of reflective elements; 2.1.2) The achievable rate of the wireless fronthaul link should meet the following requirements: in, 2.1.3) BBU pool precoding signal Compression is performed, and the output rate of the compressor cannot exceed the achievable rate of the fronthaul link; two compression strategies, point-to-point compression and multi-element compression, are considered; The forward constraints for point-to-point compression are: Point-to-point compression generates independent quantization noise on the RRHs, so the quantization noise covariance matrix of all RRHs is a block diagonal matrix, that is, The forward constraints for multivariate compression are: Multivariate compression makes the quantization noise of each RRH correlated, so the overall quantization noise covariance matrix Ω is a complete matrix; 2.2) With the goal of maximizing the downlink and rate, the optimization problem of optimizing the above system parameters when the BBU pool adopts multivariate compression is expressed as: Ω≥0(1h) ω k represents the weight of each user, (1b) represents the achievable rate constraint of the fronthaul link, (1c) is the fronthaul compression constraint, (1d) and (1e) represent the transmit power constraints of the BBU pool and each RRH respectively; (1f) and (1g) are the unit modulo 1 constraints of the passive beamforming matrices of the IRS-assisted fronthaul link and access link respectively, (1h) indicates that the fronthaul quantization noise covariance matrix is ​​a semi-positive definite matrix; 2.3) Transform the problem (P1) and then solve it; 2.3.1) Using the mean square error (MSE) method, the objective function of (P1) is transformed into: w A,k is the auxiliary variable introduced, u A,k is from y at user k A,k Medium estimate k The linear estimator of where y A,k is the signal received by user k, s k is the baseband signal obtained after the downlink message of user k is encoded by BBU, 1 is s k The dimension of is the mean square error: When A,k and u A,k When the following values ​​are obtained, R sum Get the optimal value: 2.3.2) is similar to 2.3.1). By using the MSE method, constraint (1b) is approximated as: W F,l It is required to be semi-positive definite, which is an auxiliary variable introduced. It is RRH1 from y F,l The estimated signal t l The linear receiver, that is, the predicted signal is where y F,l is the signal received by RRH l, t l Is BBU to x l The baseband signal obtained by encoding the compressed index is d R Yesl Dimensions; is the mean square error matrix: When W F,l and U F,l When the following values ​​are taken, the right side of formula (6) reaches the maximum value: 2.3.3) For constraint (1c), let (1c) can be rewritten as: Constraint (1c) is further transformed into the following approximate constraint: Σ l It is required to be semi-positive definite, which is an auxiliary variable introduced. Is a collection The number of RRHs in l When taking the following values, (10) is equivalent to (1c): 2.3.4) Transform the optimization problem (P1) into: Ω≥0(12h) Update the auxiliary variable w through (4), (5), (8), (9) and (11) A,k 、u A,k , W F,l , U F,l and Σ l ; For a fixed w A,k 、u A,k , W F,l , U F,l and Σ l , we optimize F by solving the following problem l 、v k ,Ω,Θ A and θ F , where ω k Represents the weight of each user: st(12b)~(12h)(13b) 2.4) Next, problem (P3) is decomposed into three sub-problems to be solved alternately; 2.4.1) First, fix Ω and Θ in problem (P3) A and θ F To optimize the transmit beamforming matrix F of the BBU pool and RRHl and v k ; The first sub-problem is given by: Problem (P3.1) is convex and is solved using the standard optimization tool CVX; 2.4.2) The second sub-problem is to fix F l 、v k and θ A , optimize the fronthaul quantization noise covariance matrix Ω and the IRS phase shift matrix Θ for the auxiliary fronthaul link F ; The second sub-question is as follows: Ω≥0(15f) It can be seen that the objective (15a), constraints (15c) and (15d) are convex with respect to Ω; the constraint (15b) is transformed according to And formula (6) is The definition of , (15b) can be rewritten as follows: in, d R is the data dimension of each RRH, in the above formulas: Problem (P3.2) is transformed into the following question: Constraint (17f) is still non-convex. Apply the semidefinite relaxation method SDR. After removing constraint (17f), use the standard optimization tool CVX to obtain the optimal solution. If The rank of the optimal solution is not 1, so Perform randomization to produce feasible suboptimal solutions, and update Ω and Θ only when the objective function value of (P3.2.1) increases. F ; 2.4.3) Fixed F l 、v k , Ω and Θ F , for Θ A After optimization, we get the third sub-problem: The objective function is rewritten as follows: in, In the above formula, Subproblem (P3.3) can be rewritten as: Problem (P3.3.1) is similar to (P3.2.1). It also uses the semidefinite relaxation method SDR to obtain a feasible suboptimal solution using a randomized method, and updates Θ only when the objective function value of (P3.3.1) increases. A ; 2.5) Repeat steps 2.4.1)-2.4.3) until convergence; 2.6) When the BBU pool adopts the point-to-point compression method, the transmit beamforming matrix of the BBU pool and RRH, the phase shift matrix of the IRS, and the forward quantization noise covariance matrix are jointly optimized, considering the forward compression constraint of the point-to-point compression. The problem is expressed as: st(1b),(1d)~(1g),(19b) The difference between problems (P4) and (P1) lies in constraints (19c) and (19d), where (19c) is relative to Ω l,l and v k is non-convex; let (19c) is rewritten as: log|O l The upper limit of | satisfies: S l ±0 is an auxiliary variable, so the following convex constraint is used to replace (19c): The objective function and other constraints in problem (P4) are treated in the same way as problem (P1), and then similar methods are applied to solve the problem using continuous convex approximation method and alternating optimization method; 2.7) For the discrete phase of the IRS reflector, first follow the steps in 2.1)-2.3) until convergence, and obtain Θ A and θ F , its diagonal elements θ A,i,m and θ F,i,m Mapped to discrete phase points, determine Θ F Then we scale Ω to obtain the solution that satisfies the constraints of problem (P3.2.1).