System complexity optimization method for low-altitude large-scale MIMO fusion network

Abstract

The invention discloses a system complexity optimization method for a low-altitude large-scale MIMO (Multiple Input Multiple Output) fusion network, which is characterized in that a base station with M antennas and randomly distributed single antenna users are arranged in a large-scale MIMO system based on a downlink angular domain. All users are divided into K groups after user scheduling, and each group has # imgabs0 # users. It is assumed that a base station knows a channel set and corresponding channel response information in advance. An approximate information transfer algorithm is adopted to obtain a precoding design for reducing high PAPR (Peak to Average Power Ratio). In addition, the invention further provides a novel optimization iteration scheme based on the optimal mean square error minimization principle, and the scheme is used for reducing the PAPR in the angular domain under the condition that the finite set is considered. The method has an important application value for reducing the peak-to-average power ratio of a large-scale multiple-input-multiple-output communication system.

Description

Technical Field

[0001] The present invention belongs to the technical field of communication networks, and in particular relates to a system complexity optimization method for a low-altitude large-scale MIMO fusion network. Background Art

[0002] In massive MIMO communication systems, the transmitted signals of orthogonal frequency division multiplexing (OFDM) systems often face the problem of excessively high peak-to-average power ratio (PAPR). High PAPR poses great challenges to digital-to-analog converters (DACs) and power amplifiers, reducing system power efficiency. Although various methods have been tried to solve the PAPR problem in OFDM systems, the PAPR problem has not been effectively solved in complex scenarios such as massive MIMO angle division multiple access (ADMA) systems. At the same time, the system must also take into account multiple performance requirements such as low bit error rate (SER), low inter-user interference (IUI), and computational complexity. Therefore, the development of efficient signal processing technology is crucial to improving the performance of massive MIMO communication systems. Summary of the invention

[0003] Purpose of the invention: To solve the problem of reducing the peak-to-average power ratio (PAPR) of large-scale multiple-input multiple-output (MIMO) communication systems, an approximate information transfer (AMP) algorithm based on Bayesian theory was developed. In addition, based on the characteristics of user sparsity, a precoding-based signal recovery technology in an angular division multiple access (ADMA) large-scale multiple-input multiple-output system was studied.

[0004] Technical solution: A system complexity optimization method for low-altitude massive MIMO fusion network, including the following steps: Step S1: constructing a system model, setting a system architecture and a signal transmission model under the assumption that the base station knows the channel set and the corresponding channel response information in advance in a downlink angle-domain-based massive multiple-input multiple-output system; Step S2: Construct a precoding scheme based on generalized approximate message passing and use the generalized approximate message passing precoding scheme to estimate the posterior distribution parameters and , based on the posterior likelihood probability , estimate the posterior message to obtain the posterior estimator , is an intermediate variable related to signal processing; the noise variance estimator is calculated using the expectation maximization process and updated to minimize the Euclidean distance ; Repeat the above iterative process until the iteration termination condition is met; Step S3: construct a precoding technology solution based on optimizing the minimum mean square error to obtain a linear minimum mean square error estimation result; introduce a penalty factor to calculate the relevant quantity, including calculating the iterative update signal estimation value; calculate the precoding factor to simplify the optimal MMSE.

[0005] Preferably, the specific steps of step S1 are: Step 101: In a large-scale MIMO system based on downlink angular domain, a base station with M antennas and randomly distributed single-antenna users, all users are divided into K groups after user scheduling, each group has User, assuming that the base station knows the channel set and the corresponding channel response information in advance, defines the angular domain channel set as , No. The channel set of the group is In addition, represents the channel state information in the downlink, and the corresponding angular domain CSI is expressed as , and finally, the angle rotation factor is , Then the received signal of group k is

[0006] in is the downlink channel coefficient of each group, is the downlink channel coefficient of the kth group. It is assumed that the channel responses in the downlink follow independent and identical distributions with zero expected value and unit variance. In addition, the received signal is , the transmission signal at the base station is The noise vector is represented as For the sake of brevity; Step 102: Since multiple access is considered in angle division, and The relationship between

[0007] in, and is the rotation matrix with the rotation factors , is the discrete Fourier transform matrix; Then, send it to The symbol vector of group users is represented as in , For the kth group of transmitted symbols, a PAPR reduction mapping function based on precoding is given

[0008] is the downlink CSI in the angular domain, Can be written as

[0009] in, and is the unit DFT matrix Column, rotation factor is the inverse DFT matrix Column, which can be expressed as

[0010] Given the precoding constraints to eliminate multi-user interference, the The group receive signal can be rewritten as

[0011] in, .

[0012] Preferably, the specific steps according to step S2 are: inputting a received signal y, a downlink channel coefficient H, a transmission symbol vector r, and a number of iterations T; initializing, setting the value of the number of iterations t to 0, and Initialize, the initialization method is to Set to 0 to meet the subsequent calculation requirements; Step 201: To reduce PAPR, use the following formula:

[0013] In the formula for The Frobenius norm of ; Reducing high PAPR in angular domain based massive MIMO systems, a posteriori likelihood probability The posterior estimator is obtained by using the GAMP framework, namely , using the embedded EM algorithm Programs and The precoding scheme based on generalized approximate message passing AMP is constructed, and the iterative steps of iterative reduction of PAPR in the angular domain based on precoding are demonstrated. Initialize to ,satisfy ,therefore can be minimized, where is the signal estimation value, according to the derivation of the precoding scheme based on generalized approximate message passing AMP, Updated to

[0014] in is the number of iterations, the value of the step size It can be calculated as

[0015] in .

[0016] Preferably, the iteration termination condition in step S2 is that a preset maximum number of iterations is reached or the difference between two adjacent iteration results is less than a set threshold. After the termination condition is met, the final result is output to complete the calculation process of reducing PAPR based on the GAMP precoding scheme.

[0017] Preferably, the specific steps according to step S3 are: defining relevant parameters including precoding parameters and noise variance, and constructing a channel model, wherein n is a noise vector, and the precoding parameters are limited to a finite set; calculating a correlation matrix and a vector; Step 301: Due to , according to the following optimization scheme, Another approximate estimate can be obtained;

[0018] in , is the precoding parameter, is the noise variance, restricted to a finite set Defined as ; First, define the channel symbol vector model as

[0019] in ; Step 302: The linear MMSE estimate is

[0020] in,

[0021] and

[0022] Then there is

[0023] and

[0024] Where E is the expected value in probability calculation; Step 303: Set is the penalty factor, so we can get in is the iteration index; Therefore, the optimal linear MMSE iterative estimate can be written as

[0025] Precoding Factor for

[0026] And the optimal estimate can be simplified to

[0027] In the precoding technology scheme based on optimizing the minimum mean square error (MMSE), the mapping function is used The damping coefficient is expressed as , it can be seen from the precoding technology scheme based on optimizing the minimum mean square error that the computational complexity of the calculation is The overall computational complexity is

[0028] Wherein, M is the total number of base station antennas; M1 is the number of user groups, i.e., the antennas are divided into M1 groups; T1 is the number of iterations, i.e., the total number of iterations performed during the optimization process; T2 is the number of sub-iterations, i.e., the number of sub-iterations performed in each iteration.

[0029] Compared with the prior art, the RIS-assisted channel transmission characteristic analysis method for macro cell wireless communication of the Internet of Things provided by the present invention has the following advantages: (1) PAPR reduction The AMP algorithm based on Bayesian theory and the signal recovery technology based on precoding can effectively reduce the PAPR in large-scale MIMO systems. The AMP algorithm optimizes the precoding scheme through iteration, while the precoding technology reduces the pressure on DACs and power amplifiers and improves the system power efficiency by reasonably designing precoding factors and other operations.

[0030] (2) Performance improvement While reducing PAPR, the two technologies help achieve low SER and improve signal transmission accuracy; reduce interference between users, improve overall system performance, and enable multiple users to communicate efficiently in the same system.

[0031] (3) Reduced computational complexity Compared with some traditional methods, the two technologies of the present invention have lower computational complexity while ensuring system performance. The AMP algorithm uses Bayesian theory and approximate message passing mechanism, and reduces the system computational burden and resource consumption by optimizing the iterative scheme based on precoding technology, which is more suitable for the complex computational requirements of large-scale MIMO systems. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 : SER and different signal-to-noise ratios of different PAPR reduction schemes; Figure 2 : PAPR performance of different precoding-based PAPR reduction schemes; Figure 3 : IUI performance of Algorithm 1 and Algorithm 2, where (a) Algorithm 1, (b) Algorithm 2; Figure 4 : SER performance of different signal-to-noise ratios under different conditions; Figure 5 :When the modulation mode is 16QAM and M=128, different The peak-to-average power ratio (PAPR) performance under , (b) is ; Figure 6 :When the modulation mode is 16QAM and The inter-user interference (IUI) performance under , (b) is . DETAILED DESCRIPTION

[0033] The present invention is further explained below in conjunction with specific embodiments. It should be understood that these embodiments are only used to illustrate the present invention and are not used to limit the scope of the present invention. After reading the present invention, various equivalent forms of modifications to the present invention by those skilled in the art all fall within the scope defined by the claims attached to this application.

[0034] In the solution of the present invention, a system complexity optimization method for low-altitude large-scale MIMO fusion network is provided.

[0035] It includes the following three steps: The first step is to build a system model. Under the two assumptions that the downlink is a large-scale multi-input multi-output system based on the angle domain and the base station knows the channel set and the corresponding channel response information in advance, the system architecture is set and the signal transmission model is established.

[0036] The channel coefficients are obtained based on the previously known channel information and always follow certain statistical properties (such as independent and identical distribution with zero expectation and unit variance) for subsequent calculations.

[0037] Step 101: Assume that in a downlink angular domain based massive multiple-input multiple-output (MIMO) system, there is a base station with M antennas and randomly distributed single-antenna users. All users are divided into K groups after user scheduling, and each group has User. Assume that the base station knows the channel set and the corresponding channel response information in advance. Define the angular domain channel set as , No. The channel set of the group is In addition, represents the channel state information (CSI) in the downlink, and the corresponding angular domain CSI is expressed as Finally, the angle rotation factor is .

[0038] The received signal of group k is

[0039] in is the downlink channel coefficient for each group. It is assumed that the channel responses in the downlink follow independent and identical distributions with zero expected value and unit variance. In addition, the received signal is , the transmission signal at the base station is The noise vector is represented as For brevity, Subscript omitted.

[0040] Step 102: Since multiple access is considered in angle division, and The relationship between

[0041] in, and is the rotation matrix with the rotation factors , is the discrete Fourier transform (DFT) matrix.

[0042] Then, the symbol vector sent to the kth group of users is expressed as

[0043] in , given a precoding-based PAPR reduction mapping function

[0044] is the downlink CSI in the angular domain, Can be written as

[0045] in, and is the mth column of the identity DFT matrix. The rotation factor is the nth column of the inverse DFT (IDFT) matrix and can be expressed as

[0046] Next, the precoding constraints to eliminate multi-user interference are given. Therefore, the k-th group of received signals can be rewritten as

[0047] in .

[0048] The second step is to build a precoding scheme based on generalized AMP (GAMP), input the received signal y, the downlink channel coefficient H, the transmit symbol vector r, and the number of iterations T; initialize, set the value of the number of iterations t to 0, and change the variable Initialize, the initialization method is to Set to 0 to meet the requirements of subsequent calculations; use the generalized approximate message passing (GAMP) scheme to estimate the posterior distribution and , based on the likelihood approximation , the estimated a posteriori message is obtained, where is the signal estimator, and is the intermediate variable related to signal processing; the noise variance estimator is calculated using the expectation maximization (EM) process and updated to minimize ; Repeat the above iterative process until the iteration termination condition is met; the iteration termination condition is that the preset maximum number of iterations is reached or the difference between the results of two adjacent iterations is less than the set threshold; after the termination condition is met, the final result is output to complete the calculation process of the GAMP-based precoding scheme to reduce PAPR.

[0049] Specifically, the application of the approximate message passing (AMP) algorithm based on Bayesian theory in large-scale MIMO communication systems can effectively reduce the peak-to-average power ratio (PAPR) while achieving low bit error rate (SER) and low inter-user interference (IUI). Compared with traditional methods, it has lower computational complexity and is suitable for signal processing requirements in large-scale MIMO systems.

[0050] Step 201: To reduce PAPR, we have:

[0051] In the formula for The convex optimization problem in the above formula can be solved by designing .

[0052] The generalized AMP (GAMP) algorithm is a Bayesian iterative algorithm based on the factor graph principle to obtain likelihood estimates, while the expectation maximization (EM) scheme can be used to estimate certain variables and parameters.

[0053] Here, the EM-TGM-GAMP technique is extended to reduce the high PAPR in angular domain based massive MIMO systems. The posterior estimator is obtained by using the GAMP framework, namely In addition, the embedded EM algorithm is also used Programs and Steps are used to update the parameters. Algorithm 1 demonstrates the iterative steps of iteratively reducing the PAPR in the angular domain based on precoding. Algorithm 1 constructs a precoding scheme based on generalized approximate message passing AMP (GAMP). It should be noted that the variable Initialize to To satisfy ,therefore can be minimized, where is the signal estimation value. According to the derivation of Algorithm 1, Updated to

[0054] in is the number of iterations, the value of the step size It can be calculated as

[0055] in .

[0056] The third step is to build a precoding technology solution based on optimizing the minimum mean square error (MMSE). Define relevant parameters, including precoding parameters and noise variance, and build a channel model, where n is the noise vector, and the precoding parameters are limited to a finite set; calculate the correlation matrix and vector to obtain the linear minimum mean square error (MMSE) estimation result; introduce a penalty factor to calculate the correlation quantity, including calculating the iterative update signal estimation value; calculate the precoding factor, and simplify the optimal MMSE under specific conditions (such as when the value is large, it is a parameter related to the system).

[0057] Specifically, the optimization iterative scheme based on the optimal MMSE principle further optimizes the signal recovery effect and improves the overall system performance by reasonably selecting and updating relevant parameters (such as precoding parameters, penalty factors, etc.) during the calculation process. Then, the correlation matrix and vector are calculated to obtain the linear minimum mean square error (MMSE) estimation result; then, the penalty factor is introduced to calculate the related quantity, including calculating the iterative update signal estimation value.

[0058] Step 301: Due to , according to the following optimization scheme, Another approximate estimate can be obtained.

[0059]

[0060] in , is the precoding parameter, is the noise variance, restricted to a finite set Defined as .

[0061] First, we define the channel model as

[0062] in ; Step 302: The linear MMSE estimate is ; in

[0063] and ; Then there is

[0064] and ; Step 303: Set is the penalty factor, so we can get

[0065] is the iteration index.

[0066] Therefore, the optimal linear MMSE iterative estimate can be written as

[0067] Precoding Factor for

[0068] Furthermore, the optimal estimate of the optimal linear MMSE iteration can be simplified to ; exist When larger: The simplified MMSE-based optimal precoding PAPR reduction algorithm is shown in Algorithm 2. Algorithm 2 represents the construction of a precoding technology solution based on the optimization of the minimum mean square error (MMSE). In Algorithm 2, the mapping function is used The damping coefficient is expressed as ; From Algorithm 2, we can see that the computational complexity of the calculation is The overall computational complexity is .

[0069] Based on the above derivation, Figure 1 The results show that all SER curves decrease with increasing SNR. When QPSK or 16-QAM is adopted and antennas are equipped at the BS, the SER of Algorithm 1 changes slowly when the SNR changes from -10 dB to 30 dB. Specifically, when Algorithm 1 uses 16-QAM at high SNR values, the SER value decreases from about 0.1 dB to about dB. Figure 1 It can also be seen that the SER value of Algorithm 2 is relatively small at low SNR. For example, when SNR = 10 dB and QPSK, the SER value using Algorithm 2 is less than dB; when the SNR value is greater than 20 dB, when the BS has or antennas, the SER curve of 16-QAM using Algorithm 2 remains almost unchanged; in addition, the clipping ratio of the MF and ZF methods is equal to 1.5. It is observed that the SER performance of MF and ZF is slightly better than that of Algorithm 1 when the SNR is less than 30 dB. In addition, the SER performance of the traditional MMSE scheme is close to that of Algorithm 2. However, it is obtained at the expense of higher PAPR suppression performance.

[0070] Figure 2 The ADMA scheme without PAPR suppression is shown in the ADMA legend; it can be seen that Algorithms 1 and 2 achieve satisfactory PAPR reduction performance; Figure 3 The IUI performance of Algorithm 1 and Algorithm 2 is shown when the modulation mode is 16-QAM and equal to 256; it can be seen that the IUI value of Algorithm 1 is approximately between -55 dB and -25 dB. On the contrary, in the same simulation environment, the IUI value of Algorithm 2 is between -16 dB and -7 dB. It is worth noting that Figure 1-3 It is achieved based on constant envelope precoding of finite symbols.

[0071] Figure 4 It can be seen that as the radius difference increases, the performance of SER is improved.

[0072] Figure 5 and Figure 6 Shown with Figure 4 PAPR performance and IUI performance under the same simulation environment. Figure 5 It can be seen from (a) and (b) that when the radius difference increases from 0.05 to 0.1, the PAPR value expands from (0 dB, 0.045 dB) to (0 dB, 0.06 dB). Similarly, Figure 6 The IUI results in (a) are about -22 dB to -9 dB, while Figure 6 In (b), it is approximately -24 dB to -9 dB.

[0073] Although the embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are exemplary and are not to be construed as limitations of the present invention. A person skilled in the art may change, modify, replace and vary the above embodiments within the scope of the present invention.

Claims

1. A system complexity optimization method for low-altitude large-scale MIMO fusion network, characterized in that: The following steps are involved: Step S1: constructing a system model, setting a system architecture and a signal transmission model under the assumption that the base station knows the channel set and the corresponding channel response information in advance in a downlink angle-domain-based massive multiple-input multiple-output system; Step S2: Construct a precoding scheme based on generalized approximate message passing and use the generalized approximate message passing precoding scheme to estimate the posterior distribution parameters and , based on the posterior likelihood probability , estimate the posterior message to obtain the posterior estimator , is an intermediate variable related to signal processing; the noise variance estimator is calculated using the expectation maximization process and updated to minimize the Euclidean distance ; Repeat the above iterative process until the iteration termination condition is met; Step S3: construct a precoding technology solution based on optimizing the minimum mean square error to obtain a linear minimum mean square error estimation result; introduce a penalty factor to calculate the correlation quantity, including calculating an iterative update signal estimation value; Calculate the precoding factor and simplify the optimal MMSE.

2. The system complexity optimization method for low-altitude large-scale MIMO fusion network according to claim 1, characterized in that: The specific steps of step S1 are: Step 101: In a large-scale MIMO system based on downlink angular domain, a base station with M antennas and randomly distributed single-antenna users, all users are divided into K groups after user scheduling, each group has User, assuming that the base station knows the channel set and the corresponding channel response information in advance, defines the angular domain channel set as , No. The channel set of the group is In addition, represents the channel state information in the downlink, and the corresponding angular domain CSI is expressed as , and finally, the angle rotation factor is , Then the received signal of group k is ； in is the downlink channel coefficient of each group, is the downlink channel coefficient of the kth group. It is assumed that the channel responses in the downlink follow independent and identical distributions with zero expected value and unit variance. In addition, the received signal is , the transmission signal at the base station is The noise vector is represented as For the sake of brevity; Step 102: Since multiple access is considered in angle division, and The relationship between ； in, and is the rotation matrix with the rotation factors , is the discrete Fourier transform matrix; Then, send it to The symbol vector of group users is represented as ； in , For the kth group of transmitted symbols, a PAPR reduction mapping function based on precoding is given ； is the downlink CSI in the angular domain, Can be written as ； in, and is the unit DFT matrix Column, rotation factor is the inverse DFT matrix Column, which can be expressed as ； Given the precoding constraints to eliminate multi-user interference, the The group receive signal can be rewritten as ； in, .

3. The system complexity optimization method for low-altitude large-scale MIMO fusion network according to claim 1, characterized in that: The specific steps according to step S2 are: inputting the received signal y, the downlink channel coefficient H, the transmitted symbol vector r, and the number of iterations T; initializing, setting the value of the number of iterations t to 0, and Initialize, the initialization method is to Set to 0 to meet the subsequent calculation requirements; Step 201: To reduce PAPR, use the following formula: ； In the formula for The Frobenius norm of ; Reducing high PAPR in angular domain based massive MIMO systems, a posteriori likelihood probability The posterior estimator is obtained by using the GAMP framework, namely , using the embedded EM algorithm Programs and The precoding scheme based on generalized approximate message passing AMP is constructed, and the iterative steps of iterative reduction of PAPR in the angular domain based on precoding are demonstrated. Initialize to ,satisfy ,therefore Minimize, where is the signal estimation value, according to the derivation of the precoding scheme based on generalized approximate message passing AMP, Updated to ； in is the number of iterations, the value of the step size It can be calculated as ； in .

4. The system complexity optimization method for low-altitude large-scale MIMO fusion network according to claim 1, characterized in that: The iteration termination condition in step S2 is that the preset maximum number of iterations is reached or the difference between two adjacent iteration results is less than a set threshold. After the termination condition is met, the final result is output to complete the calculation process of reducing PAPR based on the GAMP precoding scheme.

5. The system complexity optimization method for low-altitude large-scale MIMO fusion network according to claim 1, characterized in that: The specific steps according to step S3 are: defining relevant parameters including precoding parameters and noise variance, and constructing a channel model, wherein n is a noise vector, and the precoding parameters are limited to a finite set; calculating a correlation matrix and a vector; Step 301: Due to , according to the following optimization scheme, Get another approximate estimate; ； in , is the precoding parameter, is the noise variance, restricted to a finite set Defined as ; First, define the channel symbol vector model as ； in ; Step 302: The linear MMSE estimate is ； in, ； and ； Then there is ； and ； Where E is the expected value in probability calculation; Step 303: Set is the penalty factor, so we can get ； in is the iteration index; Therefore, the optimal linear MMSE iterative estimate can be written as ； Precoding Factor for ； And the optimal estimate can be simplified to ； In the precoding technology scheme based on optimizing the minimum mean square error (MMSE), the mapping function is used The damping coefficient is expressed as , from the precoding technology scheme based on optimizing the minimum mean square error, the computational complexity of the calculation is The overall computational complexity is ； Wherein, M is the total number of base station antennas; M1 is the number of user groups, i.e., the antennas are divided into M1 groups; T1 is the number of iterations, i.e., the total number of iterations performed during the optimization process; T2 is the number of sub-iterations, i.e., the number of sub-iterations performed in each iteration.

Images