CN118054819A - Distributed large-scale MIMO low-complexity precoding method - Google Patents
Distributed large-scale MIMO low-complexity precoding method Download PDFInfo
- Publication number
- CN118054819A CN118054819A CN202410452583.5A CN202410452583A CN118054819A CN 118054819 A CN118054819 A CN 118054819A CN 202410452583 A CN202410452583 A CN 202410452583A CN 118054819 A CN118054819 A CN 118054819A
- Authority
- CN
- China
- Prior art keywords
- matrix
- precoding
- user
- users
- vector
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 76
- 239000011159 matrix material Substances 0.000 claims abstract description 98
- 239000013598 vector Substances 0.000 claims description 53
- 238000004364 calculation method Methods 0.000 claims description 10
- 238000004891 communication Methods 0.000 claims description 8
- 238000010187 selection method Methods 0.000 claims description 3
- 230000001133 acceleration Effects 0.000 abstract description 4
- 238000004422 calculation algorithm Methods 0.000 description 18
- 230000005540 biological transmission Effects 0.000 description 10
- 238000012545 processing Methods 0.000 description 10
- 238000004088 simulation Methods 0.000 description 9
- 238000001228 spectrum Methods 0.000 description 7
- 238000007792 addition Methods 0.000 description 4
- 238000010586 diagram Methods 0.000 description 4
- 230000001186 cumulative effect Effects 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 230000001427 coherent effect Effects 0.000 description 2
- 230000001010 compromised effect Effects 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 239000002699 waste material Substances 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000002939 conjugate gradient method Methods 0.000 description 1
- 238000005315 distribution function Methods 0.000 description 1
- 230000008030 elimination Effects 0.000 description 1
- 238000003379 elimination reaction Methods 0.000 description 1
- 238000005562 fading Methods 0.000 description 1
- 238000007667 floating Methods 0.000 description 1
- PCHJSUWPFVWCPO-UHFFFAOYSA-N gold Chemical compound [Au] PCHJSUWPFVWCPO-UHFFFAOYSA-N 0.000 description 1
- 239000010931 gold Substances 0.000 description 1
- 229910052737 gold Inorganic materials 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 230000008054 signal transmission Effects 0.000 description 1
- 230000011664 signaling Effects 0.000 description 1
- 230000003595 spectral effect Effects 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
- 238000012549 training Methods 0.000 description 1
Landscapes
- Radio Transmission System (AREA)
Abstract
Aiming at the problems of high computational complexity in the inverse process of a downlink precoding matrix and the characteristic of a user channel effectively matched with a distributed large-scale MIMO system, the invention provides a low-complexity precoding method based on the iterative Kaczmarz of parallel computation. The invention solves the problem of high expenditure of the centralized precoding backhaul link, and eliminates the mutual interference among partial RAUs by selecting partial RAUs and then using the centralized precoding, thereby improving the system performance. The invention does not go through direct matrix inversion solution, but adopts a Kaczmarz-based parallel iteration method to solveThe method realizes parallel acceleration in a multi-core environment, reduces the computational complexity and improves the performance.
Description
Technical Field
The invention relates to the technical field of wireless communication, in particular to a distributed large-scale MIMO low-complexity precoding method.
Background
Multiple input multiple output (Multiple Input Multiple Output, MIMO) is an antenna system that uses multiple antennas at both the transmitting and receiving ends to form multiple channels between transmissions and receptions in order to greatly increase channel capacity. The MIMO system has the obvious characteristics of extremely high spectrum utilization efficiency, and gains in reliability and effectiveness are obtained by utilizing space resources on the basis of fully utilizing the existing spectrum resources, and the cost is increased by the processing complexity of a transmitting end and a receiving end. The large-scale MIMO technology adopts a large number of antennas to serve users with relatively small quantity, and can effectively improve the frequency spectrum efficiency.
Compared with the traditional massive MIMO, the distributed massive MIMO has remarkable advantages in anti-interference performance and coverage range, and the system distributes and deploys a large number of RAUs at different geographic positions to perform signal processing and cooperation, so that stronger anti-interference capability and wider signal coverage are provided, and the reliability and performance of the communication system are improved. The continuously abundant service types also provide higher performance index requirements for the network, so that the transmission with high directivity and high gain is formed by utilizing the precoding technology under the restriction of factors such as complexity, cost, power consumption and the like, so that various non-ideal factors in signal transmission are overcome, and the coverage distance and transmission quality are ensured. Precoding can be classified into centralized precoding, distributed precoding according to whether signal processing is performed by the BBU based on channel state Information (CHANNEL STATE Information, CSI) of all RAUs or by the RAUs based on local CSI. The present invention contemplates centralized precoding, i.e., data coding and precoding is done over a centralized BBU, which benefits most from the interference cancellation capability of centralized signal processing.
Precoding techniques can generally be divided into two categories, non-linear precoding techniques and linear precoding techniques. When the CSI information is collected, some nonlinear algorithms may be used for precoding, such as DIRTY PAPER Coding (DPC), tomlinson-HARASHIMA PRECODING (THP), vector perturbation (Vector Perturbation, VP), etc. These nonlinear algorithms, while higher performing, are computationally too complex to be employed by practical systems. Practical systems typically employ linear precoding algorithms such as Zero Forcing (ZF), minimum mean square error (Minimum mean square error, MMSE), regularized Zero Forcing precoding (Regularized Zero Frocing, RZF), and the like.
However, the weight vectors of the pre-coding method cannot be effectively matched with the characteristics of the user channels, and the RAU is difficult to provide good service for all users due to factors such as signal fading, so that the RAU distribution in the system is scattered, and higher requirements are also provided for backhaul network resources; secondly, the RAU close to the user in the distributed massive MIMO system provides most of the gain, and whether excessive overhead is paid for ideal elimination of interference between all RAUs is also a considerable problem; in addition, the above precoding methods all involve a large amount of matrix operations, the matrix inversion process has too high computation complexity, and the search for low-complexity signal processing technology is one of the important research problems at present.
Disclosure of Invention
The invention aims to provide a distributed large-scale MIMO low-complexity precoding method, which solves the problems existing in the prior art and obtains better performance with lower system overhead and lower calculation complexity.
In order to achieve the above object, the solution of the present invention is:
A distributed massive MIMO low-complexity precoding method comprises the following steps:
Step 1, acquiring an input channel estimation matrix, and determining the number of users after communication scheduling Selecting the number/>, of RAUs serving the user；
Step 2, executing a maximum norm selection method based on a channel estimation matrix, and selectingThe RAU provides service and performs centralized precoding;
step 3, The individual users execute the iterative precoding method based on Kaczmarz in parallel according to the channel estimation matrix;
Step 4, merging The individual users solve the obtained results in parallel;
and step 5, outputting a final precoding matrix result.
Preferably, in the step 1, the actual system parameters are acquired, including a downlink channel estimation matrixNumber of users/>Iteration number/>Initializing precoding matrix/>Given RAU and user distribution, number of RAUs serving the user/>。
Preferably, the specific steps of the step 2 are as follows:
2-1 pair of Individual user,/>Represents the/>Individual user and/>The matrix dimension between RAUs is/>From the (th) >, downlink channel estimation matrixThe matrix dimension of individual users is/>Channel estimation matrix of (c)Find column vector L2 norm/>Maximum/>A column;
2-2 before being selected Individual RAU join set/>For matrix dimension/>Angular matrix of (a)To determine the/>Whether or not the individual RAUs serve the/>And the individual users.
Preferably, the specific steps of the step 3 are as follows:
3-1 first, initializing a state vector 、/>；
Next, a canonical basis vector is calculated as the original signalIts vector dimension is/>; Definition of basis vector per user/>And for other users/>Then/>；
3-2 According to the set iteration timesFor user/>Iterating until the iteration reaches the/>Next time, the randomly selected row is noted as/>：
From the effective channel estimation matrixRandomly choose/>The line is denoted as/>，/>The probability of being selected is:
wherein, Frobenius norms,/>, representing matricesIs a regularization factor;
3-3 residual calculation is performed on all row vectors:
Wherein from the first Basis vector selection/of individual usersThe line is denoted as/>Vector/>Vector/>Is denoted as/>；
3-4 Based on residual errorUpdating a state vector for a randomly selected row, wherein the state vector is/>、/>：
And repeat。
Preferably, in the step 4, solutions obtained by each user iteration are summarized, i.e. updated。
Preferably, in the step 5, the final precoding matrix is output as。
After the technical scheme is adopted, the invention has the following technical effects:
① The invention proposes that each user is not served by all RAUs but is served by selected partial RAUs, the downlink communication overhead is less, the system can more flexibly allocate resources, and the resource waste is avoided; although the number of RAUs served for a given user is reduced, the target signal of each user is significantly improved, thus reducing unnecessary interference and improving the overall performance and user satisfaction of the system; in addition, only relevant signals of partial users need to be processed at each RAU, so that the signal processing complexity is reduced;
② The low-complexity precoding method based on the parallel computing Kaczmarz iteration is different from other low-complexity precoding methods, does not need to assume covariance matrixes of known channel vectors in advance, allows data streams of a plurality of users to be processed in parallel at the same time, and realizes parallel acceleration in a multi-core environment, so that better performance is obtained with lower computing complexity; in addition, the convergence rate of the solving process of the invention only depends on the condition number of the coefficient matrix, but does not depend on the number of equations in a linear system (namely the dimension of the matrix), so that the method is more flexible, and the performance and the complexity can be compromised by adjusting the iteration times.
Drawings
FIG. 1 is a schematic flow chart of an embodiment of the present invention;
FIG. 2 is a system diagram of an embodiment of the present invention;
FIG. 3 is a detailed flow chart of an embodiment of the present invention;
Fig. 4 is a cumulative distribution diagram of the spectrum efficiency of the user in the scenario of providing different RAUs with different service numbers according to the present invention;
FIG. 5 is a graph showing the performance of the embodiment of the present invention compared with the prior art precoding method;
FIG. 6 is a second comparison chart of the performance of the present invention and the prior art precoding method;
fig. 7 is a comparison diagram of the computational complexity of the pre-coding method according to the embodiment of the present invention.
Detailed Description
In order to further explain the technical scheme of the invention, the invention is explained in detail by specific examples.
Based on channel matrixThe conventional RZF precoding method can be expressed as formula (1):
wherein, Representing the power control factor,/>Is an identity matrix/>Is a normalization factor, typically equal to the number of users/signal-to-noise ratio of the sender channel/>Data stream/sent to userAfter RZF precoding processing, a transmission signal vector/>Is formula (2):
the matrix portion that is difficult to find is defined as The way in which the inversion is performed becomes critical in reducing the computational complexity.
The traditional large-scale MIMO low-complexity precoding research works are quite numerous, and the main idea is to indirectly obtain the approximate solution of the inverse matrix through some algorithms, and the approximate solution can be divided into two types: firstly, a precoding algorithm based on matrix inversion approximation, such as a truncated polynomial expansion method, a Neumann Series Approximation (NSA), a Newton iteration method, a Chebyshev iteration method and a conjugate gradient method; secondly, based on fixed point iteration solutionLinear precoding equation/>Continuously modifying the components of the approximation solution in fixed steps starting from a given approximation solution, one or more at a time and in a certain order, to construct a sequence of approximation solutions that continuously approximates the true solution, rather than directly solving/>. The method based on fixed-point iterative solution can be divided into a matrix splitting method and a projection method: matrix splitting method i.e. matrix/>A split successive approximation true value such as Jacobi method, gaussSeidel method, successive super-relaxation SOR method, and symmetric successive super-relaxation SSOR method; whereas in projection, kaczmarz is one of the most classical and popular line projection methods.
The Kaczmarz method does not require the advance assumption of covariance matrices for known channel vectors, differing from the other methods described above. In a distributed massive MIMO system, the channel vector covariance matrix dimension of each user is large and the estimates need to be updated over time from channel measurements, which is not practical for practical systems; in addition, the low-complexity precoding method based on Kaczmarz has certain advantages in terms of computational complexity, robustness and adaptability, and the performance and complexity can be easily controlled by adjusting the iteration times.
Aiming at the problems of high computational complexity in the inverse process of a downlink precoding matrix and the characteristic of a user channel effectively matched with a distributed large-scale MIMO system, the invention provides a low-complexity precoding method based on the iterative Kaczmarz of parallel computation.
The RAU serves all users in a distributed massive MIMO system, which clearly puts higher demands on the coverage capability of the RAU, at the same time this means that the BBU needs to have more powerful computing power. The RAU serves remote users, which can result in enhanced interference of users nearby the RAU, and the signaling overhead and interference caused by the RAU are also larger, which is not perfect in practical systems. The invention solves the problem of high expenditure of the centralized precoding backhaul link, and eliminates the mutual interference among partial RAUs by selecting partial RAUs and then using the centralized precoding, thereby improving the system performance. The invention relates to a high-dimension and high-complexity custom hard matrix in a formula (2)Solving the problem, solving/> by adopting a Kaczmarz-based parallel iteration method instead of directly solving matrix inversionThe method realizes parallel acceleration in a multi-core environment, reduces the computational complexity and improves the performance.
Referring to fig. 1, the invention discloses a distributed large-scale MIMO low-complexity precoding method, comprising the following steps:
Step 1, acquiring an input channel estimation matrix, and determining the number of users after communication scheduling Selecting the number/>, of RAUs serving the user；
Step 2, executing a maximum norm selection method based on a channel estimation matrix, and selectingThe RAU provides service and performs centralized precoding;
step 3, The individual users execute the iterative precoding method based on Kaczmarz in parallel according to the channel estimation matrix;
Step 4, merging The individual users solve the obtained results in parallel;
and step 5, outputting a final precoding matrix result.
In step 1, the actual system parameters are obtained, including the downlink channel estimation matrixNumber of users/>Iteration number/>Initializing precoding matrix/>Given RAU and user distribution, number of RAUs serving the user/>。
Further, in step 2, for a given RAU and user profile and parametersCase selection/>The scheme for providing services by the RAUs is as follows:
2-1 pair of Individual user,/>Represents the/>Individual user and/>The matrix dimension between RAUs is/>From the (th) >, downlink channel estimation matrixThe matrix dimension of individual users is/>Channel estimation matrix of (c)Find column vector L2 norm/>Maximum/>A column;
2-2 front of the selection Individual RAU join set/>For matrix dimension/>Angular matrix of (a)To determine the/>Whether or not the individual RAUs serve the/>And the individual users.
Meanwhile, the specific steps of the step 3 are as follows:
3-1 first, initializing a state vector 、/>；
Next, a canonical basis vector is calculated as the original signalDefining a basis vector for each userAnd for other users/>Then/>；
The use of canonical basis vectors as the original signal here is to simplify the computation and effectively reduce the interference, and no additional scalar coefficients are needed when the precoding update is performed, nor is the update performed in each iteration;
3-2 according to the set iteration times For user/>Iterating until the iteration reaches the/>Next time, the randomly selected row is noted as/>：
From the effective channel estimation matrixRandomly choose/>The line is denoted as/>The probability chosen depends on the gain of the channel; for users, the APs are distributed in different geographic positions, and the AP which is close to the user has strong channel gain and the AP which is far away from the user has weak channel disputes due to the relation between the path loss and shadow and the channel estimation power; thus/>The probability of being selected is:
wherein, Frobenius norms,/>, representing matricesIs a regularization factor;
3-3 residual calculation is performed on all row vectors:
Wherein from the first Basis vector selection/of individual usersThe line is denoted as/>Vector/>Vector/>Is denoted as/>；
3-4 Based on residual errorUpdating a state vector for a randomly selected row, wherein the state vector is/>、/>：
And repeat。
In the step 4 of the process, the process is carried out,As a part requiring matrix inversion in the pre-coding process, to avoid the problem of excessive computational complexity caused by direct inversion, K users execute Kaczmarz iteration solution in parallel, and the solution obtained by each user iteration is summarized, namely, update/>; For each user/>Solving for the resulting vector/>, at step 3Is the inversion of the matrix sought/>Is defined in the specification.
In step 5, the final precoding matrix output is:
Through the scheme, the invention provides the partial RAU service that each user is not served by all RAUs but selected, the downlink communication overhead is less, the system can more flexibly allocate resources, and the resource waste is avoided; although the number of RAUs served for a given user is reduced, the target signal of each user is significantly improved, thus reducing unnecessary interference and improving the overall performance and user satisfaction of the system; in addition, only relevant signals of partial users need to be processed at each RAU, so that the complexity of signal processing is reduced.
The low-complexity precoding method based on the parallel computing Kaczmarz iteration is different from other low-complexity precoding methods, does not need to assume covariance matrixes of known channel vectors in advance, allows data streams of a plurality of users to be processed in parallel at the same time, and realizes parallel acceleration in a multi-core environment, so that better performance is obtained with lower computing complexity; in addition, the convergence rate of the solving process of the invention only depends on the condition number of the coefficient matrix, but does not depend on the number of equations in a linear system (namely the dimension of the matrix), so that the method is more flexible, and the performance and the complexity can be compromised by adjusting the iteration times.
Referring to fig. 2-7, specific embodiments of the present invention are shown.
Referring to fig. 2, taking downlink precoding of a distributed massive MIMO system as an example, it is assumed that a square area of 1km×1km is surrounded by eight identical areas, and a large network deployment without boundaries is simulated by a surround technique. The system hasRau, configuration/>A root antenna; /(I)Individual users, each user configuring/>The root antenna, the RAU and the user all obey random distribution, and all the RAU and the user receive interference from all directions. Information transmission between RAUs and users in the system operates by Time Division Duplexing (TDD) while assuming reciprocity of uplink and downlink channels. In each coherent time block, the communication system goes through uplink pilot training to obtain an uplink channel estimation matrix, an uplink data transmission stage and a downlink data transmission stage.Representing user/>And acquiring the downlink channel matrix using channel reciprocity. /(I)Representing user/>Is used for the downlink channel matrix of the (c),Representing the user/>And/>The matrix dimension between RAUs is/>The joint channel matrix between all users of the system and all RAUs can be expressed as/>. System transmission to the/>The signal of the individual user is denoted/>The centralized BBU designs a precoding matrix to be adopted by utilizing the channel state information, namely/>Precoding matrix of individual users is/>. After precoding, the user signal is distributed to each RAU to be sent, and then reaches the receiving end through a wireless transmission channel.
Referring to fig. 3, the specific flow of the present embodiment includes the following steps:
step 1, acquiring system parameters: downlink channel estimation matrix Number of users/>Iteration number/>Initializing precoding matrix/>Given RAU and user distribution, number of RAUs serving the user/>。
Preferably, in steps 2 and 3 described belowAnd the users execute the iterative precoding method in parallel, firstly, partial RAU is selected based on the channel estimation matrix to perform centralized precoding, and then the iterative precoding method based on Kaczmarz is executed in parallel.
Step 2, for a given RAU and user distribution and parametersIn the case that a part of RAU is selected to provide service for the user, the RAU selection scheme is as follows:
2-1 pair of Individual user,/>Represents the/>Individual user and/>The matrix dimension between RAUs is/>From the (th) >, downlink channel estimation matrixChannel estimation matrix/>, for individual usersIts matrix dimension is/>I.e.Find column vector L2 norm/>Maximum/>A column;
2-2 front of the selection Set/>, of corresponding RAU additionsFor matrix dimension/>Angular matrix of (a)To determine the/>Whether or not the individual RAUs serve the/>Individual users, namely:
Downlink No The received signals of the individual users can be expressed as:
wherein, And/>All represent target signals,/>AndAll represent inter-user interference;
At this time, part of RAUs provide services for users, and an effective precoding matrix is defined as:
the BBU uses the channel estimation matrix to calculate a downlink centralized precoding matrix:
For the first The precoding matrix form of each user is as follows:
the dimension of the centralized precoding inversion matrix is preliminarily reduced by selecting part of RAU; in the above By adding a matrix/>, comprising noise variance and transmit powerRegularized inversion process.
Step 3, the matrix inversion problem is also converted into an indirect solution matrix inversion part, which adoptsThe iterative precoding method based on Kaczmarz is executed by each user in parallel:
3-1 first, initializing a state vector 、/>；
Next, a canonical basis vector is calculated as the original signalDefine the basis vector/>, per userAnd for other users/>Then/>；
3-2 According to the set iteration timesFor user/>Iterating until the iteration reaches the/>Next time, the randomly selected row is noted as/>：
From the effective channel estimation matrixRandomly choose/>The line is denoted as/>，/>The probability of being selected is:
wherein, Frobenius norms,/>, representing matricesIs a regularization factor;
3-3 residual calculation is performed on all row vectors:
Wherein from the first Basis vector selection/of individual usersThe line is denoted as/>Vector/>Vector/>Is denoted as/>。
3-4 Based on residual errorUpdating a state vector for a randomly selected row, wherein the state vector is/>、/>：
Repeating:。
Step 4, summarizing solutions obtained by each user iteration, namely updating 。
And 5, outputting a final precoding matrix as follows:
Based on the downlink data transmission model, the use-discard (UatF) technique is utilized to deduce the frequency spectrum efficiency of the downlink as a performance index for comparing different precoding schemes.
The beneficial effects achieved by the invention are explained through simulation experimental data.
Simulation content 1:
Referring to fig. 4, a cumulative distribution diagram of the spectrum efficiency of the user in a scenario of providing different numbers of services by using RAUs according to the present invention is compared, where the abscissa is the downlink spectrum efficiency and the ordinate is the cumulative distribution function CDF. Simulation parameter setting: total received noise power of RAU and user, performed at bandwidth b=20 MHz For-94 dBm, the maximum uplink transmitting power of each user is 100MW, the maximum downlink transmitting power of each AP is 200MW, and each coherent block interval/>200 Pilot length/>Equal to 10, the pilots are randomly allocated to the users. Number of system set RAUs/>For 64, each RAU is configured with the number of antennas/>4; Number of users/>For 16, each user is configured with a single antenna/>=1, The iteration number/>, of the present inventionSet to 400, compare the number of services offered to the user at RAU/>=20、/>=40、/>Performance of the RZF precoding method in the 3 scenarios of =64.
As can be seen from fig. 4, the curve effect graph of the present invention and RZF in the same scene is almost fitted; in addition, it can be observed that=20 Compared to/>=40、/>The performance of the =64 scenario is worse, limited by inter-user interference, and/>Compared with the scene of 40 =40The performance of the =64 scene is slightly better due to/>The number of services provided by 40 is moderate, which can provide enough precoding flexibility while avoiding some interference and complexity from a larger number of RAUs, and this balance helps to improve spectral efficiency.
Simulation content 2:
Referring to fig. 5 and 6, the performance of the low complexity precoding algorithm of the 3 approximate matrix inversions of the present invention is compared with MR, RZF and Gauss-Seidel (GS), jacobi Over-Relaxation (JOR). The simulation parameter settings are the same as the simulation content 1. For comparison, the iteration number of the three approximate matrix inversion precoding algorithms of GS, JOR and the invention is set to 10. Number of RAUs in fig. 5 For 100, each RAU is configured with the number of antennas/>4, Number of users/>For 40, each user is configured with a single antenna/>=1, And setting the number of RAUs serving the user/>=40; Number of RAUs/>, fig. 6For 64, each RAU is configured with the number of antennas/>4, Number of users/>For 16, each user is configured with a single antenna/>=1, And setting the number of RAUs serving the user/>=40。
It can be seen from fig. 5 and 6 that the MR performance is worst JOR times, whereas the GS and the approximate matrix inversion algorithm of the present invention are close to the CDF curve of the RZF direct inversion algorithm.
Simulation content 3:
referring to fig. 7, the computational complexity of the low complexity precoding algorithm of the invention, which is the inverse of the 3 approximate matrices, is compared to RZF with GS, JOR. Considering the mixed operation on real and complex numbers, the computational complexity of matrix inversion needs to be calculated according to the number of real-valued floating point operations (flops).
RZF direct matrix inversion requirementFlops, where/>, respectivelySum/>, of real multiplication operationsAnd performing real summation operation.
The GS iterative algorithm is first required in the initialization processFlops, wherein experiences/>, respectively, are requiredReal multiplication operation,/>Performing real number summation operation; furthermore, GS requires/>, per iteration processFlops, where/>, respectivelySum/>, of real multiplication operationsThe calculation complexity of the GS iterative algorithm can be observed to be higher by the addition operation of the real numbers.
The JOR iterative algorithm is first required during the initialization processFlops, in which the experiences are respectively requiredReal multiplication operation,/>Performing real number summation operation; furthermore, JOR requires/>, per iteration processFlops, where/>, respectivelySum/>, of real multiplication operationsA real addition operation.
The pre-coding algorithm provided by the invention needs to be initializedFlops, wherein experiences/>, respectively, are requiredReal multiplication operation,/>Performing real number summation operation; furthermore, each iteration process of the invention requires/>Flops, where/>, respectivelySum/>, of real multiplication operationsA real addition operation. As shown in fig. 7, the low-complexity precoding algorithm provided by the invention has lower computational complexity than GS, and the invention can well fit the approximate RZF precoding performance curve by combining the results of fig. 5 and fig. 6, which makes practical application possible.
Simulation content 4:
The GS iterative precoding algorithm is a serial algorithm, the result of the last iteration is needed in each iteration, the process is sequentially carried out, and a plurality of iteration steps cannot be carried out at the same time, so that parallelization processing cannot be carried out. In contrast, the method is an algorithm capable of solving in parallel, and the calculation efficiency can be remarkably improved and the convergence speed can be accelerated through parallelization processing. The method of the invention can better utilize the multi-core processor and the distributed computing environment of the modern computer, thereby bringing higher parallel gain. In order to illustrate the parallel gain brought by the invention, compared with the configuration of different distributed large-scale MIMO systems, the invention indirectly solves the calculation overhead of matrix inversion in a CPU multi-core environment, and the calculation overhead is shown in the following table. The simulation content 4 is realized on a second-generation intelligent Intel-to-strong extensible processor Intel (R) Xeon (R) Gold 6230R [email protected] GHz 26 core 52 thread x 2.
The calculation overhead test result table of the invention under the multi-core CPU
The above examples and drawings are not intended to limit the form or form of the present invention, and any suitable variations or modifications thereof by those skilled in the art should be construed as not departing from the scope of the present invention.
Claims (6)
1. The distributed large-scale MIMO low-complexity precoding method is characterized by comprising the following steps of:
Step 1, acquiring an input channel estimation matrix, and determining the number of users after communication scheduling Selecting the number/>, of RAUs serving the user；
Step 2, executing a maximum norm selection method based on a channel estimation matrix, and selectingThe RAU provides service and performs centralized precoding;
step 3, The individual users execute the iterative precoding method based on Kaczmarz in parallel according to the channel estimation matrix;
Step 4, merging The individual users solve the obtained results in parallel;
and step 5, outputting a final precoding matrix result.
2. The distributed massive MIMO low-complexity precoding method of claim 1, wherein:
In the step 1, the actual system parameters including the downlink channel estimation matrix are obtained Number of users/>Iteration number/>Initializing precoding matrix/>Given RAU and user distribution, number of RAUs serving the user/>。
3. The distributed massive MIMO low-complexity precoding method of claim 2, wherein the specific steps of step 2 are:
2-1 pair of Individual user,/>Represents the/>Individual user and/>The matrix dimension between RAUs is/>From the (th) >, downlink channel estimation matrixThe matrix dimension of individual users is/>Channel estimation matrix of (c)Find column vector L2 norm/>Maximum/>A column;
2-2 front of the selection Individual RAU join set/>For matrix dimension/>Angular matrix of (a)To determine the/>Whether or not the individual RAUs serve the/>And the individual users.
4. The distributed massive MIMO low-complexity precoding method of claim 3, wherein the specific steps of step 3 are:
3-1 first, initializing a state vector 、/>；
Next, a canonical basis vector is calculated as the original signalIts vector dimension is/>; Definition of basis vector per user/>And for other users/>Then/>；
3-2 According to the set iteration timesFor user/>Iterating until the iteration reaches the/>Next time, the randomly selected row is recorded as：
From the effective channel estimation matrixRandomly choose/>The line is denoted as/>，/>The probability of being selected is:
wherein, Frobenius norms,/>, representing matricesIs a regularization factor;
3-3 residual calculation is performed on all row vectors:
Wherein from the first Basis vector selection/of individual usersThe line is denoted as/>Vector/>Vector/>Is denoted as/>；
3-4 Based on residual errorUpdating a state vector for a randomly selected row, wherein the state vector is/>、/>：
And repeat。
5. The distributed massive MIMO low-complexity precoding method of claim 4, wherein:
In the step 4 of the above-mentioned process, As a part requiring matrix inversion in the precoding process, performing Kaczmarz iterative solution in parallel through K users, namely merging and summarizing/>Updating/>, by solving the obtained results in parallel by the individual users; For each user/>Solving for the resulting vector/>, at step 3Is the inversion of the matrix sought/>Is defined in the specification.
6. The distributed massive MIMO low-complexity precoding method of claim 5, wherein:
In the step 5, the final precoding matrix is output as 。
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410452583.5A CN118054819A (en) | 2024-04-16 | 2024-04-16 | Distributed large-scale MIMO low-complexity precoding method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410452583.5A CN118054819A (en) | 2024-04-16 | 2024-04-16 | Distributed large-scale MIMO low-complexity precoding method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN118054819A true CN118054819A (en) | 2024-05-17 |
Family
ID=91053871
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410452583.5A Pending CN118054819A (en) | 2024-04-16 | 2024-04-16 | Distributed large-scale MIMO low-complexity precoding method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN118054819A (en) |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2021029727A1 (en) * | 2019-08-14 | 2021-02-18 | 엘지전자 주식회사 | Method for transmitting/receiving signal in wireless communication system and device supporting same |
-
2024
- 2024-04-16 CN CN202410452583.5A patent/CN118054819A/en active Pending
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2021029727A1 (en) * | 2019-08-14 | 2021-02-18 | 엘지전자 주식회사 | Method for transmitting/receiving signal in wireless communication system and device supporting same |
Non-Patent Citations (7)
| Title |
|---|
| BOKAI XU 等: "Low-Complexity Precoding for Extremely Large-Scale MIMO Over Non-stationary Channels", ICC 2023 - IEEE INTERNATIONAL CONFERENCE ON COMMUNICATIONS, 23 October 2023 (2023-10-23), pages 2 - 3 * |
| MAHDI NOURI BOROUJERDI 等: "Low-Complexity Statistically Robust Precoder/Detector Computation for Massive MIMO Systems", IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, vol. 17, no. 10, 6 August 2018 (2018-08-06) * |
| VICTOR CROISFELT RODRIGUES 等: "Kaczmarz Precoding and Detection for Massive MIMO Systems", 2019 IEEE WIRELESS COMMUNICATIONS AND NETWORKING CONFERENCE (WCNC), 31 October 2019 (2019-10-31) * |
| 杨军;张正孝;蒋占军;: "分布式CoMP***中联合预处理方法研究", 计算机工程, no. 01, 15 January 2013 (2013-01-15), pages 2 - 3 * |
| 秦浩飞: "大规模MIMO***中低复杂度信号检测算法研究", 中国优秀硕士学位论文全文数据库, 15 March 2022 (2022-03-15) * |
| 蒋晔;孙文胜;: "一种基于Kaczmarz算法的大规模MIMO信号检测方法", 杭州电子科技大学学报(自然科学版), no. 05, 15 September 2020 (2020-09-15) * |
| 黄欢 等: "XL-MIMO***中随机Kaczmarz算法的仿真与实现", 山西大同大学学报, vol. 40, no. 2, 11 April 2024 (2024-04-11) * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109005551B (en) | Multi-user NOMA downlink power distribution method of non-ideal channel state information | |
| CN111953391A (en) | Intelligent reflector assisted multi-user MIMO uplink energy efficiency and spectrum efficiency combined optimization method | |
| CN110212959B (en) | Hybrid precoding energy efficiency optimization method of millimeter wave MIMO-OFDM communication system | |
| CN107086886B (en) | Double-layer precoding design for large-scale MIMO system fusion zero forcing and Taylor series expansion | |
| CN113162665B (en) | Pre-coding method based on deep learning channel prediction | |
| CN105450274B (en) | Based on the extensive multiple antennas relay system number of users optimization method that efficiency is optimal | |
| CN114285444B (en) | Power optimization method for large-scale de-cellular MIMO system | |
| CN109347529B (en) | Channel estimation and hybrid beam forming method for resisting non-ideality of phase shifter | |
| CN102025462B (en) | Block diagonalization precoding method used in MU-MIMO (Multiuser-Multiple Input Multiple Output) system down link | |
| CN107994933B (en) | Method for optimizing system capacity of secondary user in cognitive MIMO network | |
| CN101765123A (en) | Method for distributing uplink resource of multi-user MIMO-OFDM | |
| CN107707284B (en) | Mixed precoding method based on channel statistic codebook quantization feedback | |
| CN105812032B (en) | Channel estimation methods based on wave beam block structure compressed sensing | |
| CN104253638B (en) | MIMO interference alignment algorithm based on Stiefel manifold conjugate gradient method | |
| CN112600593A (en) | NOMA-based beam selection method | |
| CN105429687B (en) | A kind of interference alignment schemes minimizing jamming power and dimension | |
| CN112770398A (en) | Far-end radio frequency end power control method based on convolutional neural network | |
| CN107733487B (en) | Signal detection method and device for large-scale multi-input multi-output system | |
| CN113258985B (en) | Energy efficiency optimization method for single-station multi-satellite MIMO (multiple input multiple output) upper injection system | |
| CN118054819A (en) | Distributed large-scale MIMO low-complexity precoding method | |
| CN112929062B (en) | MIMO interference channel network interference alignment method based on group sparsity | |
| CN114760647A (en) | First-order optimization method for large-scale cellular MIMO uplink total rate | |
| CN113691287B (en) | Large-scale MIMO system adaptive modulation method based on average bit error rate constraint | |
| CN115428345A (en) | Multi-user precoding | |
| Liu et al. | A distributed detection algorithm for uplink massive MIMO systems |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |