IRS-assisted cellular-removal large-scale MIMO resource allocation method based on RSMA

Abstract

The invention discloses an RSMA-based IRS-assisted cellular removal large-scale MIMO resource allocation method. The method comprises the following steps: constructing an RSMA-based IRS-assisted cellular removal large-scale MIMO system model and an RSMA framework; constructing an SR maximization problem of the IRS-assisted cellular removal large-scale MIMO system based on the RSMA; the fraction is converted into a convex function through Lagrange dual transformation and secondary transformation, and the SR maximization problem is converted into an active beam forming and rate allocation problem and a passive beam forming problem through block coordinate descent (BCD); solving an active beamforming and rate allocation problem and a passive beamforming problem; and deducing an optimal user and rate based on a joint beam forming and rate allocation algorithm. According to the method, the problem of SR maximization in an IRS-assisted CF-mMIMO system based on RSMA is solved by designing a joint beam forming and rate allocation algorithm.

Description

Technical Field

[0001] The present invention relates to the field of communication technology, and in particular to an IRS-assisted de-cellularized large-scale MIMO resource allocation method based on RSMA. Background Art

[0002] 1. Current IRS-assisted decellularized massive MIMO system architecture based on rate division multiple access: Cell-free massive multiple input multiple output (CF-mMIMO) deploys a large number of access points (APs) in an area, and serves users through all access points in a coordinated manner, solving the problem of serious inter-cell interference in traditional cellular architecture and becoming one of the key technologies in the sixth generation of mobile communications. However, with the large number of users accessing, in order to ensure the quality of communication, it is inevitable to deploy more APs, which will bring huge energy overhead and complex hardware implementation, which is not conducive to the implementation of CF-mMIMO. With the increase in the number of APs, the wireless channel environment is filled with a large number of interference signals, which reduces the communication rate of the system. Intelligent Reflecting Surface (IRS) is a low-cost, low-power, and easy-to-deploy technology that can significantly increase data rates and enhance signal coverage while reducing energy consumption. IRS is a planar array composed of many configurable passive devices. Unlike traditional amplification and forwarding relays, IRS does not have the ability to receive and transmit signals, but can only reflect signals, so it consumes less energy. By adjusting the IRS controller, the IRS can change the phase or amplitude of the reflected signal at low power consumption, thereby effectively controlling the propagation path of the reflected signal and achieving the enhancement of the desired signal and the suppression of the interference signal at the receiving end. The introduction of IRS in CF-mMIMO can not only reduce the system energy consumption and expand the system coverage, but also reduce the impact of shadow fading. However, due to the existence of multiple APs and IRSs, the wireless channel environment becomes complex and there are a large number of interference signals. Traditional Space Division Multiple Access (SDMA) needs to rely on perfect channel state information at the transmitter and needs to be in an underloaded network to effectively suppress or eliminate multi-user interference. Since Rate-Splitting Multiple Access (RSMA) can improve system performance under various network loads, user deployments and CSI quality conditions, it has become a promising non-orthogonal transmission multiple access scheme in multi-antenna networks. The IRS-assisted CF-mMIMO system based on RSMA can effectively reduce inter-user interference and improve system robustness. This patent is aimed at the IRS-assisted CF-mMIMO system based on RSMA, with the optimization goal of maximizing the sum rate (SR) of all users, and proposes a low-complexity joint beamforming and rate allocation algorithm to solve the proposed optimization problem.

[0003] The existing RSMA-based IRS-assisted decellularized massive MIMO system architecture has the following shortcomings: In the traditional RSMA-based IRS-assisted decellularized massive MIMO system, the IRS phase model is considered as a continuous form for easy optimization. However, in actual hardware implementation, continuous phase setting cannot be achieved. The IRS phase can only be a discrete multi-bit phase. Considering the RSMA-based IRS-assisted decellularized massive MIMO system under discrete phase is of great significance.

[0004] 2. Joint beamforming and rate allocation algorithm for IRS-assisted decellularized massive MIMO based on RSMA: In the IRS-assisted CF-mMIMO system based on RSMA, how to design the precoding matrix, rate allocation factor and phase shift matrix of the AP side and maximize the objective function, i.e., SR, is a key issue. Since the optimization problem constraints include the non-convex fractional signal-to-interference-plus-noise ratio (SINR) expression, IRS phase constraints and common stream decodability constraints, the optimization problem is difficult to solve and has high complexity.

[0005] The main joint beamforming and rate allocation algorithms currently used are block coordinate descent (BCD), which solves the original problem by decomposing it into optimization problems for each block. The joint beamforming and rate allocation problem can be decomposed into active beamforming and rate allocation problems at the AP and passive beamforming problems at the IRS. The active beamforming and rate allocation problem and passive beamforming are often solved using Successive Convex Approximation (SCA) and Weighted Minimum Mean Square Error (WMMSE). SCA approximates non-convex functions to convex functions, which can then be solved using traditional convex optimization solvers. WMMSE transforms non-convex optimization problems into weighted minimum mean square error minimization problems by introducing auxiliary variables, and solves the original problem by iteratively optimizing the convex optimization problem of each variable. Passive beamforming problems often use SCA or penalty methods. The penalty method eliminates the non-convex constraint by adding the IRS phase constraint to the objective function in the form of a penalty term.

[0006] The shortcomings of the existing joint beamforming and rate allocation algorithm for RSMA-based IRS-assisted decellularized massive MIMO: In order to solve the non-convex SR maximization problem, the current mainstream method decomposes the original problem into active beamforming and rate allocation problems at the AP and passive beamforming at the IRS through the BCD algorithm. For the active beamforming and rate allocation problem at the AP, there are currently two mainstream methods: the first is to relax the original problem into a convex problem and then use a convex optimization solver to solve it; the second is to convert the original problem into a weighted minimum mean square error minimization problem and solve it by iteratively optimizing the convex optimization problem of each variable. However, all of the above methods need to be solved by a convex optimization solver, which has high complexity and long numerical calculation time. In addition, only the numerical solution of the original problem can be obtained, but not the closed form solution. The existing active beamforming and rate allocation algorithms are highly complex and cannot obtain closed solutions. It is necessary to design a low-complexity active beamforming and rate allocation closed solution algorithm. As for the passive beamforming problem at the IRS, the current mainstream methods consider the passive beamforming problem of the continuous IRS more, but rarely consider the passive beamforming problem of the discrete IRS. The rare discrete IRS passive beamforming problem is only obtained by quantizing the continuous IRS phase solution. It is necessary to design a discrete IRS passive beamforming algorithm. Summary of the invention

[0007] The main purpose of the present invention is to propose an IRS-assisted de-cellularized massive MIMO resource allocation method based on RSMA, aiming to solve the SR maximization problem in the IRS-assisted CF-mMIMO system based on RSMA by designing a joint beamforming and rate allocation algorithm.

[0008] To achieve the above object, the present invention provides an IRS-assisted de-cellularized massive MIMO resource allocation method based on RSMA, the method comprising the following steps: Step S10, constructing an RSMA-based IRS-assisted de-cellularized massive MIMO system model and an RSMA framework, wherein the RSMA-based IRS-assisted de-cellularized massive MIMO system model includes M APs and K single-antenna users, wherein each AP has N antennas, and in order to reduce energy consumption, an IRS is introduced, wherein the number of units of the IRS is U, and the IRS can control the phase of each unit through a controller; Step S20, constructing the SR maximization problem of the IRS-assisted decellularized massive MIMO system based on RSMA with the maximum transmission power of each AP, the public rate decodable constraint and the IRS discrete phase constraint as constraints; Step S30, converting the fraction into a convex function through Lagrange dual transformation and quadratic transformation, and converting the SR maximization problem into an active beamforming and rate allocation problem and a passive beamforming problem by using block coordinate descent; Step S40, using a low-complexity active beamforming and rate allocation algorithm based on fixed-point iteration to solve the active beamforming and rate allocation problem, and using a discrete passive beamforming algorithm based on successive refinement to solve the passive beamforming problem; Step S50: derive the optimal user and rate R based on the joint beamforming and rate allocation algorithm.

[0009] A further technical solution of the present invention is that the step S10 of constructing an RSMA-based IRS-assisted de-cellularized massive MIMO system model comprises: make , , Represents the index set of AP, user, and IRS unit respectively; Considering a more realistic discrete IRS phase, assuming Represents the number of quantization bits for each IRS unit, that is, Indicates the number of configurable phase shift quantization levels for each IRS unit; in order to simplify the model, consider Uniform quantization on , the set of possible discrete phase shifts for each IRS unit can be expressed as: (1); in Represents the quantization interval, the phase shift matrix of IRS It can be expressed as: (2); in It is expressed as the reflection coefficient of the u-th IRS unit, where is a natural constant, is an imaginary unit, is the discrete phase shift of the u-th IRS unit, and diag is the diagonalization operation; make It is represented as the channel matrix of the direct link between the mth AP and the kth user, Expressed as the channel matrix of the mth AP and IRS, It is represented as the channel matrix of IRS and k-th user; therefore, the cascade channel matrix of m-th AP and k-th user can be expressed as (3).

[0010] A further technical solution of the present invention is that in step S10, the multiple access transmission mode adopts the RSMA framework, and at the transmitter, each user message Split into two parts: Public sub-message and private sub-messages ; All users’ public sub-messages are merged into one public message through the message merger ; Public Message Encoded into a public stream through a codebook shared by all users ; Private sub-messages for all users Encoded separately as a private stream ; Public streams can be decoded by all users while private streams can only be decoded by the corresponding users; In the mth AP, all streams pass through the linear precoding matrix of the mth AP Therefore, the transmission signal of the mth AP is: (4); The signal received by the kth user can be expressed as: (5); in , Represent the equivalent channels from all APs to the kth user and the precoding vectors of the private streams, represents the precoding vector of the common stream from all APs to the kth user, represents the noise energy received by the kth user, is the noise variance, and the precoding vector of each AP needs to meet the maximum transmission power limit of each AP, let represents the maximum downlink transmission power of each AP, then: (6); Each user first receives the signal by treating the private stream as noise The public stream s0 is decoded in . Therefore, the signal-to-interference-noise ratio of the public stream of the kth user is: (7); After removing the public stream from the received signal through continuous interference cancellation, the signal-to-interference-noise ratio of the private stream of the kth user is: (8); Each user decodes and splits its own public sub-message from the public stream , decode the private sub-message from the private stream , public sub-messages and private sub-messages are merged into a complete user message .

[0011] A further technical solution of the present invention is that, in step S20, the steps of constructing the SR maximization problem of the IRS-assisted decellularized massive MIMO system based on RSMA with the maximum transmission power of each AP end, the common rate decodable constraint and the IRS discrete phase constraint as constraints include: The instantaneous rate at which the kth user decodes the public stream is: (9); In order to ensure that the public stream can be decoded by all users, the rate of the public stream cannot exceed the minimum instantaneous rate of all users decoding the public stream, that is: (10); Since the public flow rate is derived from the contribution of each user's public sub-message, the following formula is available: (11); in represents the public rate allocated to the kth user by the public flow rate; The instantaneous rate at which the kth user decodes the private stream is: (12); The users and rates are: (13); The final SR maximum problem for all users can be expressed as: (14); The decision variables , , They represent the precoding matrix of all APs, the common rate of all users, and the phase matrix of IRS respectively. Both the objective function and the constraint conditions are non-convex fractions and difficult to solve. Since the objective function is the user and rate, there is no need to specifically calculate the common rate of each user, only the common rate sum of all users is required. Therefore, P1 can be expressed as: (15); in ; Introduce auxiliary variables , for the kth user public flow rate Using the Lagrange dual transformation method, we can get: (16); in is the public flow rate of the kth user The lower bound of season hour, Get the kth user's public flow rate The lower bound is: (17); symbol represents the optimal variable, The third term is still a fraction, so we introduce an auxiliary variable , using the quadratic transformation we get: (18); in For expression The lower bound of ,pass hour, Get The lower bound is: (19); Similarly, introducing and A similar transformation can be made to the private flow rate, and the following formula can be obtained: (20); (twenty one); (twenty two); By fixing , P2 can be expressed as: (twenty three); P3 is still a non-convex problem. The BCD method is used to decompose P3 into active beamforming and rate allocation problem and passive beamforming problem to solve the original problem.

[0012] A further technical solution of the present invention is that in step S40, the step of using a low-complexity active beamforming and rate allocation algorithm based on fixed-point iteration to solve the active beamforming and rate allocation problem includes: By fixing , P3 can be expressed as: (twenty four); P4 is a convex optimization problem and can be solved by a convex optimization solver. However, a closed solution cannot be obtained using a convex optimization solver and the complexity is high. Therefore, a low-complexity active beamforming and rate allocation algorithm is used to solve this problem. In order to solve P4, a low-complexity active beamforming and rate allocation algorithm based on fixed-point iteration is proposed. First, P4 is analyzed by the Carlo-Kuhn-Tucker condition. The Lagrangian function of this problem is: (25); in and is the Lagrangian dual variable of the corresponding constraint; By introducing: (26); (27); (28); (29); Then the KKT condition is: (30); (31); (32); (33); (34); in , is a block diagonal matrix operation, ; By transforming KKT1 and KKT2 equations, the optimal solution of the precoding vector of the common stream is: (35); The optimal solution for the precoding vector of the private stream is: (36); The optimal solution for the sum of public rates for , assuming that in each iteration the user index n is such that minimum; next we need to find the Lagrange dual variable The optimal solution of ; find the Lagrange dual variable by fixed-point iteration method The optimal solution; according to KKT4, there is the following formula: (37); in is the step length; By transforming the equation, we can get the expression of fixed-point iteration: (38); That is, The iterations have the following iteration expressions: (39); Note that the updated It does not necessarily meet KKT3. The desired Projection onto a plane Right now: (40); Similarly, KKT5 has: (41); In the In iteration There are the following iteration expressions: (42); Therefore, in The iteration order of updating the Lagrangian dual variables is: (43); (44); (45).

[0013] A further technical solution of the present invention is that in step S40, the step of using a discrete passive beamforming algorithm based on successive refinement to solve the passive beamforming problem includes: By fixing , P3 can be expressed as: (46); The passive beamforming algorithm based on successive refinement finds the IRS phase matrix that maximizes the sum rate by changing the phase of only one IRS unit in each iteration while keeping the phases of other IRS units unchanged. After changing the phase, it is necessary to determine the decodability conditions of the common stream. Is it satisfied? Only when this condition is met, the IRS phase matrix is ​​desirable; when the public stream decodable condition is met, the following formula is available: (47).

[0014] A further technical solution of the present invention is that, in the step S50, in the step of deriving the optimal user and rate R based on the joint beamforming and rate allocation algorithm: At each iteration time t, the BCD algorithm follows the following optimization order: (47).

[0015] The present invention studies the SR maximization problem of the IRS-assisted CF-mMIMO system based on RSMA, with the constraints of the maximum transmission power of each AP, the public rate decodable constraint and the IRS discrete phase constraint. This patent first transforms the original optimization problem through Lagrangian dual transformation and quadratic transformation. The original optimization problem is decomposed into: active beamforming and rate allocation problem, passive beamforming problem. In the active beamforming and rate allocation problem, considering the high computational complexity and high computational time of solving the optimization problem using a convex optimization solver, a low-complexity active beamforming and rate allocation algorithm based on fixed-point iteration is proposed. The proposed active beamforming and rate allocation algorithm based on fixed-point iteration can achieve almost the same performance as the traditional algorithm, but with lower computational complexity and numerical calculation time. In passive beamforming, a discrete passive beamforming algorithm based on successive refinement is proposed for the IRS discrete phase. Finally, a joint beamforming and rate allocation algorithm is proposed based on the above algorithm. Simulation verifies the effectiveness of the proposed algorithm compared with the benchmark algorithm. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the structures shown in these drawings without paying creative work.

[0017] Figure 1 It is a flow chart of the RSMA-based IRS-assisted de-cellularized large-scale MIMO resource allocation method of the present invention; Figure 2 This is a schematic diagram of the CF-mMIMO system model based on RSMA; Figure 3 This is a schematic diagram of the RSMA framework; Figure 4 This is a comparison diagram of the sum rate and CPU calculation time under different algorithms; Figure 5 It is a schematic diagram of the relationship between the sum rate and the number of IRS units under different algorithms; Figure 6 This is a schematic diagram of the relationship between the sum rate and the maximum transmission power of each AP under different algorithms.

[0018] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0019] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0020] The present invention proposes an IRS-assisted de-cellularized massive MIMO resource allocation method based on RSMA. In an IRS-assisted CF-mMIMO system based on RSMA, the SR maximization problem is solved by designing a joint beamforming and rate allocation algorithm.

[0021] The present invention first proposes the SR maximization optimization problem, where the objective function is SR, and the constraints are the maximum transmit energy of each AP, the common stream decodability constraint, and the IRS discrete phase constraint. Then, a low-complexity joint beamforming and rate allocation algorithm is proposed to solve the above non-convex SR maximization optimization problem. Compared with the traditional joint beamforming and rate allocation algorithm, the joint beamforming and rate allocation algorithm proposed in the present invention can not only obtain the closed form solution of the AP precoding matrix, but also obtain the discrete IRS phase solution in a non-quantized form, while requiring only lower complexity and numerical calculation time.

[0022] Please refer to Figure 1 The IRS-assisted decellularized large-scale MIMO resource allocation method based on RSMA of the present invention comprises the following steps: Step S10, construct an RSMA-based IRS-assisted de-cellularized large-scale MIMO system model and an RSMA framework, wherein the RSMA-based IRS-assisted de-cellularized large-scale MIMO system model includes M APs and K single-antenna users, wherein each AP has N antennas. In order to reduce energy consumption, an IRS is introduced, wherein the number of units of the IRS is U, and the IRS can control the phase of each unit through a controller.

[0023] The RSMA-based IRS-assisted decellularized massive MIMO system model in the present invention is as follows: The system model of the present invention considers M APs and K single-antenna users, where each AP has N antennas. In order to reduce energy consumption, an IRS is introduced, where the number of units of the IRS is U, and the IRS can control the phase of each unit through a controller. The schematic diagram of the IRS-assisted decellularized massive MIMO system model based on RSMA is shown in the figure. Figure 2 As shown. , , Represents the index set of AP, user, and IRS unit respectively. The system works in time division duplex (TDD) mode, and each coherent interval includes three stages: uplink training, downlink data transmission, and uplink data transmission. The present invention focuses on the downlink data transmission stage, assuming that the channel state information at the transmitting end is known.

[0024] In the present invention, a more practical discrete IRS phase is considered. Assume that l represents the number of quantization bits of each IRS unit, that is, Indicates the number of configurable phase shift quantization levels for each IRS unit. To simplify the model, we consider Uniform quantization on . The set of possible discrete phase shifts for each IRS unit can be expressed as: (1); in Represents the quantization interval. The phase shift matrix of IRS It can be expressed as: (2); in It is expressed as the reflection coefficient of the u-th IRS unit, where is a natural constant, is an imaginary unit, is the discrete phase shift of the u-th IRS unit, and diag is the diagonalization operation.

[0025] make It is represented as the channel matrix of the direct link between the mth AP and the kth user, It is represented as the channel matrix of the mth AP and IRS, It is represented as the channel matrix of IRS and k-th user. Therefore, the cascade channel matrix of m-th AP and k-th user can be expressed as (3).

[0026] The RSMA framework involved in the present invention is as follows: The multiple access transmission method of the present invention adopts the RSMA framework, as shown in the schematic diagram Figure 3 At the transmitter, each user message Split into two parts: Public sub-message and private sub-messages All users’ public sub-messages are merged into one public message through the message merger. . Public Message Encoded into a public stream through a codebook shared by all users . Private sub-messages for all users Encoded separately as a private stream The public stream can be decoded by all users while the private stream can only be decoded by the corresponding user.

[0027] In the mth AP, all streams pass through the linear precoding matrix of the mth AP Therefore, the transmission signal of the mth AP is: (4); The signal received by the kth user can be expressed as: (5); in , Represent the equivalent channels from all APs to the kth user and the precoding vectors of the private streams, represents the precoding vector of the common stream from all APs to the kth user, represents the noise energy received by the kth user, is the noise variance. And the precoding vector of each AP needs to meet the maximum transmission power limit of each AP, let represents the maximum downlink transmission power of each AP, then: (6); Each user first receives the signal by treating other private flows as noise The public stream s0 is decoded in . Therefore, the signal-to-interference-plus-noise ratio (SINR) of the public stream of the kth user is: (7); After removing the public stream from the received signal through Successive Interference Cancellation (SIC), the SINR of the private stream of the kth user is: (8); Each user decodes and splits its own public sub-message from the public stream , decode the private sub-message from the private stream , public sub-messages and private sub-messages are merged into a complete user message .

[0028] Step S20, constructing the SR maximization problem of the IRS-assisted decellularized massive MIMO system based on RSMA with the maximum transmission power of each AP, the public rate decodable constraint and the IRS discrete phase constraint as constraints.

[0029] The step S20 specifically includes: The instantaneous rate at which the kth user decodes the public stream is: (9); In order to ensure that the public stream can be decoded by all users, the rate of the public stream cannot exceed the minimum instantaneous rate of all users decoding the public stream, that is: (10); Since the public flow rate is derived from the contribution of each user's public sub-message, the following formula is available: (11); in Represents the public rate allocated to the k-th user.

[0030] The instantaneous rate at which the kth user decodes the private stream is: (12); The users and rates are: (13); The final SR maximum problem for all users can be expressed as: (14); The decision variables , , Represent the precoding matrix of all APs, the common rate of all users, and the phase matrix of IRS respectively. Both the objective function and the constraint conditions are non-convex fractions, which are difficult to solve. Since the objective function is the user and rate, there is no need to specifically calculate the common rate of each user, only the common rate sum of all users is required. Therefore, P1 can be expressed as: (15); in . Introducing auxiliary variables , for the kth user public flow rate Using the Lagrange dual transformation method, we can get: (16); in is the public flow rate of the kth user When fixed season hour, Get the kth user's public flow rate The lower bound is: (17); symbol represents the optimal variable, The third term is still a fraction, so we introduce an auxiliary variable , using the quadratic transformation we get: (18); in For expression The lower bound of ,pass hour, Get The lower bound is: (19); Similarly, introducing and A similar transformation can be made to the private flow rate, and the following formula can be obtained: (20); (twenty one); (twenty two); By fixing , P2 can be expressed as: (twenty three); P3 is still a non-convex problem. The BCD method is used to decompose P3 into active beamforming and rate allocation problem and passive beamforming problem to solve the original problem.

[0031] Step S30, converting the fraction into a convex function through Lagrange dual transformation and quadratic transformation, and converting the SR maximization problem into an active beamforming and rate allocation problem and a passive beamforming problem by using block coordinate descent.

[0032] Step S40, using a low-complexity active beamforming and rate allocation algorithm based on fixed-point iteration to solve the active beamforming and rate allocation problem, and using a discrete passive beamforming algorithm based on successive refinement to solve the passive beamforming problem.

[0033] Specifically, in step S40, the step of using a low-complexity active beamforming and rate allocation algorithm based on fixed-point iteration to solve the active beamforming and rate allocation problem includes: By fixing , P3 can be expressed as: (twenty four); P4 is a convex optimization problem, which can be solved by a convex optimization solver; however, a closed solution cannot be obtained using a convex optimization solver and the complexity is high, so the present invention adopts a low-complexity active beamforming and rate allocation algorithm to solve the problem; In order to solve P4, a low-complexity active beamforming and rate allocation algorithm based on fixed-point iteration is proposed. First, P4 is analyzed by the Karush-Kuhn-Tucker condition (KKT). The Lagrangian function of this problem is: (25); in and is the Lagrangian dual variable of the corresponding constraint; By introducing: (26); (27); (28); (29); Then the KKT condition is: (30); (31); (32); (33); (34); in , is a block diagonal matrix operation, ; By transforming KKT1 and KKT2 equations, the optimal solution of the precoding vector of the common stream is: (35); The optimal solution for the precoding vector of the private stream is: (36); The optimal solution for the sum of public rates for , assuming that in each iteration the user index n is such that minimum; next we need to find the Lagrange dual variable Due to the existence of common rate constraints, the traditional binary search-based method is not applicable to the RSMA-based decellularized massive MIMO system. The present invention uses a fixed-point iteration method to find the Lagrangian dual variable The optimal solution; according to KKT4, there is the following formula: (37); in is the step length; By transforming the equation, we can get the expression of fixed-point iteration: (38); That is, The iterations have the following iteration expressions: (39); Note that the updated KKT3 is not necessarily satisfied. The present invention changes The desired Projection onto a plane Right now: (40); Similarly, KKT5 has: (41); In the In iteration There are the following iteration expressions: (42); Therefore, in The iteration order of updating the Lagrangian dual variables is: (43); (44); (45).

[0034] The low-complexity active beamforming and rate allocation algorithm based on fixed-point iteration is summarized in Table 1 below.

[0035] Table 1

[0036] In step S40, the step of solving the passive beamforming problem by using a discrete passive beamforming algorithm based on successive refinement includes: By fixing , P3 can be expressed as: (46); The discrete passive beamforming algorithm based on successive refinement finds the IRS phase matrix that maximizes the sum rate by changing the phase of only one IRS unit in each iteration while keeping the phases of other IRS units unchanged. It is worth noting that after changing the phase, it is necessary to determine the decodability condition of the common stream. Whether it is satisfied, this IRS phase matrix is ​​desirable only if it meets this condition; when the public stream decodable condition is established, there is the following formula: (47).

[0037] The discrete passive beamforming algorithm proposed in the present invention is summarized in Table 2.

[0038] Table 2

[0039] Step S50: derive the optimal user and rate R based on the joint beamforming and rate allocation algorithm.

[0040] In the step S50, in which the optimal user and rate R are derived based on the joint beamforming and rate allocation algorithm: At each iteration time t, the optimization sequence is as follows: (47).

[0041] The proposed joint beamforming and rate allocation algorithm is summarized in Table 3.

[0042] Table 3

[0043] The present invention verifies the proposed joint beamforming and rate allocation algorithm to solve the SR maximization problem through simulation. We assume that there are 6 APs in this scenario, each AP has 4 antennas and 4 users. Assume that the maximum transmission power of each AP is 10dbm. In the simulation, we simulate the comparison between the proposed algorithm and some baseline algorithms.

[0044] Figure 4 The relationship between the sum rate and CPU time of the binary search-based beamforming algorithm under SDMA without IRS, the SCA algorithm under RSMA, and the fixed-point iteration-based beamforming and rate allocation algorithm under RSMA is shown. It can be observed that compared with space division multiple access, rate division multiple access can overcome the interference between users and thus improve the sum rate of users. In addition, the fixed-point iteration-based beamforming and rate allocation algorithm we proposed can achieve almost the same performance as the SCA-based beamforming and rate allocation algorithm, but significantly reduces the CPU calculation time. From a theoretical analysis, the computational complexity of the SCA-based beamforming and rate allocation algorithm is , and the computational complexity of the beamforming and rate allocation algorithm based on fixed-point iteration is ,When the number of APs and AP antennas is large, the proposed beamforming and rate allocation algorithm based on ,fixed-point iteration can effectively reduce the computational complexity.

[0045] Figure 5The relationship between the sum rate of the continuous refinement discrete passive beamforming under RSMA and the optimization-minimization (MM) passive beamforming under SDMA and the number of IRS units in the case of IRS is shown. It can be observed that when the number of IRS units increases, the sum rate of both algorithms increases. This is because as the number of IRS units increases, IRS can achieve more precise phase control. It can also be observed that as the number of IRS units increases, the performance gap between RSMA and SDMA algorithms is gradually narrowing. This is because more IRS units can provide better equivalent cascade channels, and the role of RSMA common flows is reduced.

[0046] Figure 6 The relationship between the sum rate and the maximum transmission energy of each AP under different algorithms is shown. First, as the maximum transmit power of each AP increases, the SR performance of all algorithms will improve. Second, compared with the algorithm without IRS, the algorithm containing IRS shows significantly better SR performance. This enhancement is attributed to the fact that IRS creates additional line-of-sight links and improves the user's anti-interference ability by optimizing the phase shift matrix. In addition, the algorithm under RSMA outperforms the algorithm under SDMA. It is worth noting that as the maximum transmit power of each AP increases, the performance advantage of RSMA over SDMA becomes more obvious. This improvement is due to the public message of RSMA, which is similar to the concept of multicast and can be decoded by all users, resulting in additional SR gain.

[0047] In summary, the present invention studies the SR maximization problem of the IRS-assisted CF-mMIMO system based on RSMA, with the constraints of the maximum transmission power of each AP, the public rate decodable constraint and the IRS discrete phase constraint. This patent first transforms the original optimization problem through Lagrangian dual transformation and quadratic transformation. The original optimization problem is decomposed into: active beamforming and rate allocation problem, passive beamforming problem. In the active beamforming and rate allocation problem, considering the high computational complexity and high computational time of using convex optimization solver to solve the optimization problem, a low-complexity active beamforming and rate allocation algorithm based on fixed-point iteration is proposed. The proposed active beamforming and rate allocation algorithm based on fixed-point iteration can achieve almost the same performance as the traditional algorithm, but with lower computational complexity and numerical calculation time. In passive beamforming, a discrete passive beamforming algorithm based on successive refinement is proposed for the IRS discrete phase. Finally, a joint beamforming and rate allocation algorithm is proposed based on the above algorithm. Simulations verify the effectiveness of the proposed algorithm compared with the benchmark algorithm.

[0048] The above description is only a preferred embodiment of the present invention, and does not limit the patent scope of the present invention. All equivalent structural changes made by using the contents of the present invention specification and drawings under the concept of the present invention, or directly / indirectly applied in other related technical fields are included in the patent protection scope of the present invention.

Claims

1. An IRS-assisted decellularized massive MIMO resource allocation method based on RSMA, characterized in that: The method comprises the following steps: Step S10, constructing an RSMA-based IRS-assisted de-cellularized massive MIMO system model and an RSMA framework, wherein the RSMA-based IRS-assisted de-cellularized massive MIMO system model includes M APs and K single-antenna users, wherein each AP has N antennas, and in order to reduce energy consumption, an IRS is introduced, wherein the number of units of the IRS is U, and the IRS can control the phase of each unit through a controller; Step S20, constructing the SR maximization problem of the IRS-assisted decellularized massive MIMO system based on RSMA with the maximum transmission power of each AP, the public rate decodable constraint and the IRS discrete phase constraint as constraints; Step S30, converting the fraction into a convex function through Lagrange dual transformation and quadratic transformation, and converting the SR maximization problem into an active beamforming and rate allocation problem and a passive beamforming problem by using block coordinate descent; Step S40, using a low-complexity active beamforming and rate allocation algorithm based on fixed-point iteration to solve the active beamforming and rate allocation problem, and using a discrete passive beamforming algorithm based on successive refinement to solve the passive beamforming problem; Step S50: derive the optimal user and rate R based on the joint beamforming and rate allocation algorithm.

2. The RSMA-based IRS-assisted de-cellularized massive MIMO resource allocation method according to claim 1, characterized in that: The step S10 of constructing an RSMA-based IRS-assisted de-cellularized massive MIMO system model includes: make , , Represents the index set of AP, user, and IRS unit respectively; Considering a more realistic discrete IRS phase, assuming Represents the number of quantization bits for each IRS unit, that is, Indicates the number of configurable phase shift quantization levels for each IRS unit; in order to simplify the model, consider Uniform quantization on , the set of possible discrete phase shifts for each IRS unit can be expressed as: (1)； in Represents the quantization interval, the phase shift matrix of IRS It can be expressed as: (2)； in It is expressed as the reflection coefficient of the u-th IRS unit, where is a natural constant, is an imaginary unit, is the discrete phase shift of the u-th IRS unit, and diag is the diagonalization operation; make It is represented as the channel matrix of the direct link between the mth AP and the kth user, It is represented as the channel matrix of the mth AP and IRS, It is represented as the channel matrix of IRS and k-th user; therefore, the cascade channel matrix of m-th AP and k-th user can be expressed as (3)。 3. The RSMA-based IRS-assisted de-cellularized massive MIMO resource allocation method according to claim 2, characterized in that: In step S10, the multiple access transmission mode adopts the RSMA framework. At the transmitter, each user message Split into two parts: Public sub-message and private sub-messages ; All users’ public sub-messages are merged into one public message through the message merger ; Public Message Encoded into a public stream through a codebook shared by all users ; Private sub-messages for all users Encoded separately as a private stream ; Public streams can be decoded by all users while private streams can only be decoded by the corresponding users; In the mth AP, all streams pass through the linear precoding matrix of the mth AP Therefore, the transmission signal of the mth AP is: (4)； The signal received by the kth user can be expressed as: (5)； in , Represent the equivalent channels from all APs to the kth user and the precoding vectors of the private streams, represents the precoding vector of the common stream from all APs to the kth user, represents the noise energy received by the kth user, is the noise variance, and the precoding vector of each AP needs to meet the maximum transmission power limit of each AP, let represents the maximum downlink transmission power of each AP, then: (6)； Each user first receives the signal by treating the private stream as noise The public stream s0 is decoded in ; therefore, the signal to interference noise ratio of the public stream of the kth user is: (7)； After removing the public stream from the received signal through continuous interference cancellation, the signal-to-interference-noise ratio of the private stream of the kth user is: (8)； Each user decodes and splits its own public sub-message from the public stream , decode the private sub-message from the private stream , public sub-messages and private sub-messages are merged into a complete user message .

4. The RSMA-based IRS-assisted de-cellularized massive MIMO resource allocation method according to claim 3, characterized in that: The step S20, taking the maximum transmission power of each AP end, the public rate decodable constraint and the IRS discrete phase constraint as constraint conditions, constructing the SR maximization problem of the IRS-assisted decellularized massive MIMO system based on RSMA includes: The instantaneous rate at which the kth user decodes the public stream is: (9)； In order to ensure that the public stream can be decoded by all users, the rate of the public stream cannot exceed the minimum instantaneous rate of all users decoding the public stream, that is: (10)； Since the public flow rate is derived from the contribution of each user's public sub-message, the following formula is available: (11)； in represents the public rate allocated to the kth user by the public flow rate; The instantaneous rate at which the kth user decodes the private stream is: (12)； The users and rates are: (13)； The final SR maximum problem for all users can be expressed as: (14)； The decision variables , , They represent the precoding matrix of all APs, the common rate of all users, and the phase matrix of IRS respectively. Both the objective function and the constraint conditions are non-convex fractions and difficult to solve. Since the objective function is the user and rate, there is no need to specifically calculate the common rate of each user, only the common rate sum of all users is required. Therefore, P1 can be expressed as: (15)； in ; Introduce auxiliary variables , for the kth user public flow rate Using the Lagrange dual transformation method, we can get: (16)； in is the public flow rate of the kth user The lower bound of season hour, Get the kth user's public flow rate The lower bound is: (17)； symbol represents the optimal variable, The third term is still a fraction, so we introduce an auxiliary variable , using the quadratic transformation we get: (18)； in For expression The lower bound of ,pass hour, Get The lower bound is: (19)； Similarly, introducing and A similar transformation can be made to the private flow rate, and the following formula can be obtained: (20)； (21)； (22)； By fixing , P2 can be expressed as: (23)； P3 is still a non-convex problem. The BCD method is used to decompose P3 into active beamforming and rate allocation problem and passive beamforming problem to solve the original problem.

5. The RSMA-based IRS-assisted de-cellularized massive MIMO resource allocation method according to claim 4, characterized in that: In step S40, the steps of solving the active beamforming and rate allocation problem by using a low-complexity active beamforming and rate allocation algorithm based on fixed-point iteration include: By fixing , P3 can be expressed as: (24)； P4 is a convex optimization problem and can be solved by a convex optimization solver. However, a closed solution cannot be obtained using a convex optimization solver and the complexity is high. Therefore, a low-complexity active beamforming and rate allocation algorithm is used to solve this problem. In order to solve P4, a low-complexity active beamforming and rate allocation algorithm based on fixed-point iteration is proposed. First, P4 is analyzed by the Carlo-Kuhn-Tucker condition. The Lagrangian function of this problem is: (25)； in and is the Lagrangian dual variable of the corresponding constraint; By introducing: (26)； (27)； (28)； (29)； Then the KKT condition is: (30)； (31)； (32)； (33)； (34)； in , is a block diagonal matrix operation, ; By transforming KKT1 and KKT2 equations, the optimal solution of the precoding vector of the common stream is: (35)； The optimal solution for the precoding vector of the private stream is: (36)； The optimal solution for the sum of public rates for , assuming that in each iteration the user index n is such that minimum; next we need to find the Lagrange dual variable The optimal solution of ; find the Lagrange dual variable by fixed-point iteration method The optimal solution; according to KKT4, there is the following formula: (37)； in is the step length; By transforming the equation, we can get the expression of fixed-point iteration: (38)； That is, The iterations have the following iteration expressions: (39)； Note that the updated It does not necessarily meet KKT3. The desired Projection onto a plane Right now: (40)； Similarly, KKT5 has: (41)； In the In iteration There are the following iteration expressions: (42)； Therefore, in The iteration order of updating the Lagrangian dual variables is: (43)； (44)； (45)。 6. The RSMA-based IRS-assisted de-cellularized massive MIMO resource allocation method according to claim 5, characterized in that: In step S40, the step of solving the passive beamforming problem by using a discrete passive beamforming algorithm based on successive refinement includes: By fixing , P3 can be expressed as: (46)； The passive beamforming algorithm based on successive refinement finds the IRS phase matrix that maximizes the sum rate by changing the phase of only one IRS unit in each iteration while keeping the phases of other IRS units unchanged. After changing the phase, it is necessary to determine the decodability conditions of the common stream. Whether it is satisfied, this IRS phase matrix is ​​desirable only if it meets this condition; when the public stream decodable condition is established, there is the following formula: (47)。 7. The RSMA-based IRS-assisted de-cellularized massive MIMO resource allocation method according to claim 6, characterized in that: In the step S50, in which the optimal user and rate R are derived based on the joint beamforming and rate allocation algorithm: At each iteration time t, the BCD algorithm follows the following optimization order: (47)。

Images