A cluster-based uplink multi-user MIMO precoding transmission method

Abstract

The application designs an uplink multi-user MIMO precoding transmission method based on clustering, which can adaptively adjust the codebook structure according to the transmission environment by using the statistical characteristics of the channel. The method first needs to sample the channel state information in a period of time, calculates the optimal precoding matrix using the iterative precoding algorithm, and then uses the clustering method to train the precoding codebook suitable for the uplink multi-user MIMO system on the data set, and synchronously deploys the codebook at the base station and the user end. In the transmission, the method is based on the distance between the code word and the optimal precoding matrix to greedily search the codebook. The precoding codebook designed by the method can adapt to the change of channel characteristics, and has better performance than the classic codebook. The greedy strategy for codebook search reduces the complexity of online calculation, and is more suitable for use in actual systems.

Description

TECHNICAL FIELD

[0001] The application relates to a clustering-based uplink multi-user MIMO precoding transmission method and belongs to the technical field of wireless mobile communication. BACKGROUND

[0002] MIMO is a kind of communication system in which multiple antennas are equipped at both the receiving and transmitting ends to form a multipath channel between the receiving and transmitting ends, and can greatly improve the channel capacity of the system. In order to meet the requirements of continuously improving peak rate and spectrum utilization, MIMO technology is valued as a key technology of 4G LTE and 5G. The latest evolution of MIMO technology, large-scale MIMO, is suitable for various application scenarios such as eMBB, mMTC and uRLLC of 5G, is defined as a key technology of the physical layer in the R16 standard of 5G, and is also the focus of the research on the next generation of wireless communication technology.

[0003] Because the receiving and transmitting ends of the MIMO system are both equipped with multiple antennas, detection technology is needed at the receiving side to take the expected information stream from the target transmitting antenna as useful information while minimizing or eliminating the interference from other antennas. At the transmitting end, a corresponding precoding scheme can be designed according to a certain criterion to encode the original signal stream into antenna transmitting signals that can improve the transmission reliability of the communication system.

[0004] With the evolution of 5G technology standards, multi-user uplink MIMO scenarios gradually gain attention. The terminals in this scenario are isolated from each other and thus face many challenges different from the downlink scenario. The existing uplink MIMO uplink system precoding is mainly divided into two categories: centralized calculation scheme and codebook transmission scheme. The centralized calculation scheme completes the calculation of the uplink precoding matrix at the base station end, and then sends the complete information of the matrix to each terminal through the downlink data link or control link. This strategy will bring huge feedback overhead, and it is often impossible to implement in the actual system. Therefore, 3GPP 5G protocol recommends using a codebook scheme for uplink transmission, that is, the base station selects a precoding matrix suitable for the current communication state from a specified precoding matrix set, because the complete matrix set is pre-stored at the terminal side and the base station side, so the base station only needs to feedback the serial number of the selected matrix in the entire set, which greatly reduces the feedback overhead. The core problem in the codebook-based uplink precoding scheme is the codebook construction problem, because the number of code words is directly related to the feedback amount and the calculation complexity of selection. If a codebook design is efficient enough, it can achieve better performance with a limited number of code words. At present, researchers have proposed many classic codebook construction schemes, such as Grassmannian codebook construction algorithm, Kerdock codebook construction algorithm, DFT codebook construction algorithm, etc. However, these classic codebook construction algorithms are almost all proposed for standard Rayleigh fading channels, and cannot adapt to the complex and variable transmission environment in the actual communication system.

[0005] In summary, in the uplink multi-user MIMO system, the existing codebook design still has room for performance improvement. In this scheme, a codebook design method based on k-means clustering algorithm is proposed, which has better performance by using channel statistical characteristics. In addition, a distance-based greedy codebook search method suitable for clustering codebooks is also proposed, which has a lower calculation complexity. SUMMARY

[0006] The technical problem to be solved by the present application is that in the uplink multi-user MIMO system, the traditional codebook design scheme has weak adaptability to different channel environments and poor performance. A codebook design scheme with better performance is obtained by using a clustering algorithm, and a low-complexity codebook search method suitable for the codebook is given. The following will introduce the scheme in detail:

[0007] First, briefly introduce the channel model used in the simulation process of the present patent, and investigate the scenario of a single base station serving multiple users in the uplink multi-user MIMO system as shown in Figure 1 In the system, there is a base station, which serves a total of K users, and we denote the number of transmit antennas equipped by the kth user as N k,tIn the uplink transmission process, the user uses the transmit antenna to transmit the modulated and precoded signal stream. Before precoding, we call the modulated signal stream as modulated signal stream, and the number of modulated signal streams of user k is N k,s The transmitter needs to use the precoding matrix to convert the modulated signal stream into the transmit signal stream before transmission. Assuming that the base station configures N r antennas for receiving the data of all K users. In order to distinguish the signals between different users and different antennas, a detection method needs to be used, and using a linear detector is equivalent to that the base station prepares a detection matrix for each user. After the received signal is processed by the detector corresponding to the user, the modulated signal stream of the user can be restored. Finally, the detected signal stream of user k can be expressed as:

[0008]

[0009] In the formula, represents the modulated signal stream vector, represents the noise on the receiving antenna, which is modeled as a Gaussian white noise, and the power is denoted as σ 2 . After the modulated signal stream is precoded by the precoding matrix F k , it is transmitted through the transmit antenna. Because each user and the base station are configured with multiple antennas, the channel formed between user k and the base station is in the form of a two-dimensional matrix, and the transmission of signals of other users is the same. At the receiving end, the received signal of user k H k F k s k will be superimposed with the signals of other users and the thermal noise n at the receiving end, wherein the former is the target signal and the latter two are interference signals. Using the detection matrix G k , the estimation of the modulated signal stream vector of user k For the above model, first, the steps of using the clustering method for adaptive codebook design are as follows:

[0010] Step A. Constructing the optimal precoding matrix data set First, sample the channel state information in a period of time in the space, and denote the set of channel state information data as Then, we divide the channel state data in into groups every K, and denote the i-th group of channel state data as Using the following iterative algorithm, the optimal precoding matrix data set corresponding to this group of channel state data can be calculated

[0011] (1) Initialize the channel state data corresponding to the precoding matrix Under the user's transmit power constraint the following equation must be satisfied:

[0012]

[0013] where is the power of the uncoded modulated data stream of user k, and Under QAM modulation, it can be considered that is a unit matrix. In order to make the initial precoding matrix satisfy the above equation, the method constructs a diagonal matrix satisfying where Further Cholesky decomposition on can obtain the initial precoding matrix satisfying the single antenna power constraint where

[0014] (2) Iterative calculation of optimal precoding matrix data set In one iteration, first update the detection matrix using the following formula

[0015]

[0016] where:

[0017]

[0018] Then, update the Lagrange multiplier matrix has:

[0019]

[0020] Finally, update the precoding matrix using the following formula

[0021]

[0022] Repeat the above iteration steps until the maximum number of iterations ε is reached, and the precoding matrix obtained in the last iteration is recorded as , that is The optimal precoding matrix data set corresponding to the i-th group of channel state data can be constructed. For each group , perform the above steps, and the complete optimal precoding matrix data set

[0023] Step B. Construct the precoding codebook using the k-means clustering method. First, determine the number of codewords *m* in the codebook based on the actual environment; generally, *m* is chosen as a power of 2. Then, use the optimal precoding matrix dataset obtained in Step A... Perform clustering operations:

[0024] (1) Initialize the centroid. Randomly select m sample points as the initial codebook. Let the iteration number be l = 0.

[0025] (2) Begin algorithm iteration. For the dataset Each optimal precoding matrix sample point x j Calculate its sum codebook The distance x is calculated by finding the distance between each codeword. j The latest writing sequence number x j Classified by code word Determined clusters middle.

[0026] (3) Codebook Determined clusters Recalculate the codewords using the following formula.

[0027]

[0028] (4) Update the codebook. If all codewords obtained in the (l+1)th iteration are updated... The codewords corresponding to 1≤j≤m and the l-th iteration result If all are the same, stop the iteration and record. For clustering codebook Otherwise, return to step (2) and continue iterating.

[0029] The above are the steps of the clustering-based adaptive codebook design method, which occurs before transmission begins. During transmission, the base station needs to search for suitable codewords for each user. The steps of the distance-based greedy codebook search method in this scheme are as follows:

[0030] Step C. Treat the user's uplink channel data during transmission as a set of channel state data from Step A. The corresponding optimal precoding matrix set is obtained by using the same steps as in step A.

[0031] Step D. Search for precoded codewords based on matrix distance. Let the optimal precoded matrix for user k be denoted as . Traversing the codebook The optimal code for it is: This operation is performed for each user k, and the corresponding best codeword is searched for each user. These codewords are the precoding matrices used in actual transmission for the corresponding user. BRIEF DESCRIPTION OF DRAWINGS

[0032] Figure 1 The structure diagram of the uplink multi-user MIMO system of the present application is shown.

[0033] Figure 2 The spatial distribution diagram of a codeword and an optimal precoding matrix is shown.

[0034] Figure 3 The flowchart of the codebook design method based on clustering is shown.

[0035] Figure 4 The flowchart of the distance-based greedy codebook search method is shown.

[0036] Figure 5 The codebook transmission performance diagram under full-flow transmission of simulation example 1 is shown.

[0037] Figure 6 The codebook transmission performance diagram under non-full-flow transmission of simulation example 2 is shown. DETAILED DESCRIPTION

[0038] The specific embodiments of the present application are further described below in conjunction with the accompanying drawings of the specification.

[0039] The present application provides an uplink multi-user MIMO precoding transmission method based on clustering, which can solve the technical problem of poor performance of the conventional codebook design scheme in different channel environments. A codebook design scheme with better performance is obtained by using a clustering algorithm, and a low-complexity codebook search method suitable for the codebook is given. The algorithm derivation of each part is described as follows:

[0040] First, the algorithm derivation of constructing the optimal precoding matrix data set is described. In the present application, in order to measure the correctness of the modulation signal stream estimation, the minimum mean square error evaluation criterion is introduced in the present application, and the following formula is used to evaluate the correctness of the final system estimation:

[0041]

[0042] In the above formula, η is called the mean square error between the actual modulation signal stream vector and the estimation. It considers the statistical expectation value of the mean square error under the condition that the channel matrix H is random. The smaller η is, the smaller the error is, and the better the performance of the system is.

[0043] In actual communication system, the transmit power of user equipment or the transmit power of each antenna on the equipment needs to be limited due to the energy consumption limit of the user terminal. The transmit power limit of each user equipment is referred to as user power limit, which can be shown by the following expression:

[0044]

[0045] Φk sk represents the power of the uncoded modulation data stream of user k, which can be considered as a unit matrix under QAM modulation. k represents the transmit power of user k after coding. Based on the minimum mean square error criterion, the optimization expression and constraint condition of the optimal uplink multi-user MIMO system precoding problem with user power constraint can be given as follows:

[0046]

[0047]

[0048] The Lagrange multiplier method is used to solve this problem. First, the following Lagrange function is constructed:

[0049]

[0050] P k is the power constraint matrix, and P k = (P k / N k,t )I. Then the optimal precoding matrix expression is solved by using the cross-optimization method. First, set the partial derivative of ξ k to the detection matrix G k to zero, and consider that G k is irrelevant to F k and Λ k when solving the partial derivative, to obtain the expression:

[0051]

[0052] Wherein:

[0053]

[0054] Then set the partial derivative of ξ k to the precoding matrix F k to zero, and use the power constraint condition to obtain the expression of the Lagrange multiplier matrix , which is a diagonal matrix, and each diagonal element λ k,j is defined by the following formula:

[0055]

[0056] Finally, the detection matrix G k and the multiplier matrix Λ k are put into the expression of the partial derivative zero of F k , and the following is obtained:

[0057]

[0058] After the above steps, one iteration of solving is completed, and after multiple iterations, F k tends to be stable, that is, the optimal solution is reached. For each user, the above solving process can be completed independently. Through this step, the statistical characteristics of the channel are converted into the distribution characteristics of the optimal precoding matrix thereon, and then a clustering algorithm can be used to construct a more efficient codebook structure according to the characteristics.

[0059] Next, the reason why a clustering algorithm can be used to construct an uplink precoding codebook with stronger adaptability and better performance is explained. As shown in Figure 2 , a spatial distribution diagram of a uniform quantization codebook is shown. The hollow circles in the figure represent the positions of the code words in the matrix space, and the solid circles represent the positions of the optimal precoding matrices under a specific channel. As can be seen from the figure, the optimal precoding matrices in the matrix space are not necessarily uniformly distributed, which leads to the fact that some code words in the lower right corner of the matrix space where the optimal precoding matrices are sparsely distributed are "wasted". If these code words are moved to the upper left corner where the optimal precoding matrices are densely distributed, the quantization accuracy of the region can be improved, and better performance can be achieved.

[0060] Therefore, we hope to design a codebook construction algorithm that can adapt to changes in channel characteristics. The codebook generated by such an algorithm should provide higher quantization accuracy in regions where the optimal precoding matrices are densely distributed. The k-means clustering algorithm in the field of machine learning is a better choice, which is an unsupervised clustering algorithm and can be used to solve the following problem: for a given input sample set , the elements are divided into a total of m clusters, denoted as (C1, C2,..., C m ), so that the average distance from the sample points x (i) belonging to the i-th cluster to the centroid μ i of the cluster is minimized. That is:

[0061]

[0062] where μ i is defined as the centroid of the cluster C i , and is shown by the following formula:

[0063]

[0064] The codebook construction problem and the clustering problem have some commonality in conversion: if the codebook can adapt to the spatial distribution of the optimal precoding matrix, the quantization precision in the dense distribution area is improved, which can improve the system performance under the limitation of limited quantization bits. The optimization problem solved by k-means has a convergent structure: under the condition of a specified quantization precision of m centroids, for a discrete sample set with uneven distribution, the centroid distribution obtained by k-means clustering is also irregular: it can make the distance between the code word and the optimal precoding matrix as close as possible, thereby reducing the quantization error.

[0065] Finally, the principle of the distance-based greedy codebook search method is explained. It is noted that the clustering input sample set of the clustering codebook is constructed by the global precoding algorithm, so selecting the precoding matrix combination that minimizes the system mean square error in the clustering codebook is equivalent to finding a group of code words with the minimum distance from the result of the global precoding algorithm. Since the distances between the algorithm outputs of different users and the code words are not affected by each other, for user k, after using the global precoding algorithm to obtain its optimal precoding matrix S k , only the code word F with the minimum distance ||F k -S k ||2is selected from the codebook k .

[0066] The present scheme is applicable to a cell model in which a single base station serves multiple users. Taking a typical single-cell model as an example, the cell is configured with 1 base station serving 4 users, the base station has 32 antennas, each user has 4 antennas, and the number of user modulation signal streams is 2. Before transmission, the system constructs a clustering codebook according to the channel state information collected in a period of time, and synchronizes it to the base station and the user. During transmission, the base station estimates the uplink channel state information of each user according to the pilot signal sent by the user, and then uses the distance-based greedy codebook search algorithm to solve the code word that should be used by each user, and then feeds back the code word corresponding to the number to the user. After receiving, the user uses the code word to map the modulation signal stream layer to the sending antenna and sends it out, and the receiving end can use the minimum mean square error detector to detect and restore the modulation signal stream.

[0067] In summary, the present scheme designs a clustering-based uplink multi-user MIMO precoding transmission method, and gives a clustering-based codebook design method and a distance-based codebook search method adapted thereto. The specific implementation steps of each link are shown in Figure 3 and Figure 4 , and the textual description is as follows:

[0068] First, the steps of using the clustering method for adaptive codebook design are as follows:

[0069] Step A. Construct the optimal precoding matrix dataset First, sample the channel state information over a period of time in the sampling space, and denote this set of channel state information dataset as... After that, we will The channel state data is divided into groups of K data points each, and the i-th group of channel state data is denoted as . The optimal precoding matrix dataset corresponding to this set of channel state data can be calculated using the following iterative algorithm.

[0070] (1) Initialize channel state data corresponding precoding matrix Under the user's transmit power constraint In this case, the following equation must be satisfied:

[0071]

[0072] In the formula The power of the uncoded modulated data stream of user k is given by: Under QAM modulation, it can be considered that It is an identity matrix. To ensure that the initial precoding matrix satisfies the above equation, this method constructs a matrix that satisfies the following equation for any antenna label j. diagonal matrix In the formula Again Applying Cholesky decomposition yields the initial precoding matrix that satisfies the single-antenna power constraint. in

[0073] (2) Iteratively calculate the optimal precoding matrix dataset In one iteration, the detection matrix is ​​first updated using the following formula.

[0074]

[0075] in:

[0076]

[0077] Then, update the Lagrange multiplier matrix. have:

[0078]

[0079] Finally, update the precoding matrix using the following formula.

[0080]

[0081] Repeat the above iterative steps until the set maximum number of iterations ε is reached, and then use the result from the last iteration. Recorded as This allows us to construct the optimal precoding matrix dataset corresponding to the i-th group of channel state data.

[0082] For each group

[0083] By performing the above steps, you can obtain the complete optimal precoding matrix dataset.

[0084] Step B. Construct the precoding codebook using the k-means clustering method. First, determine the number of codewords *m* in the codebook based on the actual environment; generally, *m* is chosen as a power of 2. Then, use the optimal precoding matrix dataset obtained in Step A... Perform clustering operations:

[0085] (1) Initialize the centroid. Randomly select m sample points as the initial codebook. Let the iteration number be l = 0.

[0086] (2) Begin algorithm iteration. For the dataset Each optimal precoding matrix sample point x j Calculate its sum codebook The distance x is calculated by finding the distance between each codeword. j The latest writing sequence number x j Classified by code word Determined clusters middle.

[0087] (3) Codebook Determined clusters Recalculate the codewords using the following formula.

[0088]

[0089] (4) Update the codebook. If all codewords obtained in the (l+1)th iteration are updated... The codewords corresponding to 1≤j≤m and the l-th iteration result If all are the same, stop the iteration and record. For clustering codebook Otherwise, return to step (2) and continue iterating.

[0090] The above are the steps of the clustering-based adaptive codebook design method, which occurs before transmission begins. During transmission, the base station needs to search for suitable codewords for each user. The steps of the distance-based greedy codebook search method in this scheme are as follows:

[0091] Step C. Take the uplink channel data of the user at the transmission time as a set of channel state data in Step A Get the corresponding optimal precoding matrix set and using the same steps as Step A

[0092] Step D. Search for the precoding code word according to the matrix distance. Let the optimal precoding matrix of the user k be Traverse the codebook Search for the best code word corresponding to it: Perform this operation for each user k, and the best code word corresponding to each user can be searched. These code words are the precoding matrices used by the corresponding user in the actual transmission.

[0093] Next, the clustering-based uplink multi-user MIMO precoding transmission method designed according to the present application is described in combination with simulation as follows:

[0094] Simulation Example 1: The simulation conditions are shown in Table 1 below:

[0095] Table 1 Simulation conditions

[0096] Number of users K 3 Number of code words m 8 Number of user antennas N k,t ]]> 4 Receiver MMSE receiver Base station antenna number N r ]] 32 User power constraint 4W Modulation stream number N k,s ]]> 4 Modulation scheme 16QAM

[0097] Figure 3 The performance simulation comparison of the clustering codebook transmission scheme proposed in the present application with the DFT codebook transmission scheme, the Grassmannian codebook transmission scheme, and the Kerdock codebook transmission scheme under the above conditions and full-flow transmission is shown. The clustering codebook transmission scheme proposed in the present application is improved by about 1.7 dB compared with the Kerdock codebook transmission scheme with the best performance, and the reliability of the uplink MIMO communication system is enhanced.

[0098] Simulation Example 2: The simulation conditions are shown in Table 2 below:

[0099] Table 2 Simulation conditions

[0100] Number of users K 4 Number of code words m 16 Number of user antennas N k,t ]]> 4 Receiver MMSE receiver Base station antenna number N r ]] 32 User power constraint 4W Modulation stream number N k,s ]] 2 Modulation scheme 16QAM

[0101] Figure 4 The performance simulation comparison of the clustering codebook transmission scheme proposed in the present application with the DFT codebook transmission scheme, the Grassmannian codebook transmission scheme, and the Kerdock codebook transmission scheme under the above conditions and non-full-flow transmission is shown. The clustering codebook transmission scheme proposed in the present application is improved by about 2.2 dB compared with the Kerdock codebook transmission scheme with the best performance, and the reliability of the uplink MIMO communication system is enhanced.

[0102] Embodiments of the present application will be described in detail with reference to the drawings, but the present application is not limited to the embodiments described below, and various changes can be made within the knowledge of those skilled in the art without departing from the spirit of the present application.

Claims

1.A cluster-based uplink multi-user MIMO precoding transmission method, characterized in that, It comprises the following steps: constructing an optimal precoding codebook dataset, training a precoding codebook suitable for an uplink multi-user MIMO system on the optimal precoding matrix dataset using a clustering method, and synchronously deploying the codebook at a base station and a user end; in transmission, a codebook is searched based on a distance between a code word and an optimal precoding matrix; Constructing the optimal precoding codebook dataset The specific steps include: first, sampling channel state information over a period of time in the space, and then recording this set of channel state information data as... ,Will The channel state data is divided into groups of K, denoted as the first group. Group channel status data is The optimal precoding matrix dataset corresponding to this set of channel state data is calculated using the following iterative algorithm. : (1) initialization of channel state data corresponding precoding matrix under the user's transmit power constraint the following equation needs to be satisfied: ； In the formula Indicates user The power of an uncoded modulated data stream has Under QAM modulation, it is believed that It is an identity matrix; for any antenna label Construct a satisfying diagonal matrix In the formula ; then Applying Cholesky decomposition yields the initial precoding matrix that satisfies the single-antenna power constraint. ,in ; For the first The number of transmitting antennas equipped per user; Represents the modulated signal stream vector; (2) iteratively computing an optimal precoding matrix data set The detection matrix is first updated in one iteration using the following equation : ； Wherein: ； denotes the noise power; Then, the Lagrange multiplier matrix is updated has: ； Finally, the precoding matrix is updated using the following equation : ； Repeat the above iterative steps until the set maximum number of iterations is reached. The result of the last iteration Recorded as This allows us to construct the first... The optimal precoding matrix dataset corresponding to each group of channel state data; for each group By performing the above steps, you can obtain the complete optimal precoding matrix dataset. ; The specific steps of constructing the precoding codebook using the clustering method include: firstly determining the number of code words contained in the codebook according to the actual environment , and then performing clustering operation on the optimal precoding matrix data set : (1) initialize the centroid; randomly select a sample point as the initial codebook and set the iteration round to ; (2) Start algorithm iteration; for each optimal precoding matrix sample point in the data set calculate the distance of each codeword in the codebook , obtain the distance of the nearest codeword , divide into the cluster decided by the codeword ; (3) on the codebook decided cluster , the code word is recalculated using the following formula : ； (4) update the codebook if all the code words of the last iteration are the same as the code words of the last iteration, then stop the iteration and record the codebook as the clustering codebook, otherwise go back to step (2) and continue the iteration; ​​​ The steps of the distance-based greedy codebook searching method are as follows: 1) considering the uplink channel data of the user at the transmission time as channel state data , obtaining the corresponding optimal precoding matrix set and ; 2) Search the precoding code word according to the matrix distance; record the user The optimal precoding matrix of the user is Search the best code word corresponding to it by traversing the code book ; perform this operation for each user , and search the best code word corresponding to each user.

Images