CN103957041A

CN103957041A - 3D wave beam shaping method for large-scale MIMO TDD system

Info

Publication number: CN103957041A
Application number: CN201410101271.6A
Authority: CN
Inventors: 黄永明; 范立行; 杨绿溪; 姜蕾; 王刚
Original assignee: NEC China Co Ltd; Southeast University
Current assignee: NEC China Co Ltd; Southeast University
Priority date: 2014-03-18
Filing date: 2014-03-18
Publication date: 2014-07-30
Anticipated expiration: 2034-03-18
Also published as: CN103957041B

Abstract

The invention discloses a low-complexity 3D beamforming method for large-scale MIMO TDD systems. The 3D beamforming is the direct product of horizontal and vertical beamforming, and both the horizontal and vertical beamforming are double-layer precoding. DFT pre-beamforming and SLNR precoding; according to the channel sparsity, DFT pre-beamforming divides the cell into several circular areas in the vertical dimension and multiple sub-sectors in the horizontal dimension; according to the statistical CSIT, separate The user grouping algorithm assigns users into groups. Users in non-adjacent groups in the same sub-sector can multiplex pilots, and users in non-adjacent groups in the same ring area can also multiplex pilots. SLNR precoding is obtained based on low-order instantaneous equivalent CSIT . The present invention has lower computational complexity, and also reduces pilot overhead, and the reduced computational complexity and pilot overhead depend on the number of divided groups; when there are many base station antennas, the sum rate of this method is close to no grouping system.

Description

3D Beamforming Method for Massive MIMO TDD System

技术领域 technical field

本发明涉及一种低复杂度的面向大规模MIMO（多输入多输出）TDD（时分双工）系统的3D（三维）波束成形方法，属于无线通信技术。 The invention relates to a low-complexity 3D (three-dimensional) beamforming method for large-scale MIMO (multiple input multiple output) TDD (time division duplex) systems, which belongs to wireless communication technology. the

背景技术 Background technique

大规模MIMO是实现下一代移动通信的关键技术。但是在基站的水平维度配置过多天线不易实现，由此提出3D MIMO。3D MIMO克服了基站的这种限制，将天线放置在二维的平面上。垂直维度的基站天线形成垂直波束，可以提供更多的自由度。垂直波束和水平波束共同构成了3D波束。3D波束可以识别具有相同水平角的用户。目前相关文献中提出的基于Kronecker积的码书适用于3D MIMO FDD系统。然而，在该系统中，每次只能调用2个用户，没有充分发挥3D MIMO系统的长处。 Massive MIMO is the key technology to realize the next generation of mobile communication. However, it is not easy to configure too many antennas in the horizontal dimension of the base station, so 3D MIMO is proposed. 3D MIMO overcomes this limitation of the base station by placing the antenna on a two-dimensional plane. The base station antenna in the vertical dimension forms a vertical beam, which can provide more degrees of freedom. Together, the vertical beam and the horizontal beam form a 3D beam. 3D beams can identify users with the same horizontal angle. The codebook based on the Kronecker product proposed in the relevant literature is suitable for the 3D MIMO FDD system. However, in this system, only 2 users can be called at a time, which does not give full play to the strengths of the 3D MIMO system. the

为了有效实现3D波束成形，基站需要知道准确的CSIT。在FDD系统中，瞬时的CSIT是通过用户反馈至基站得到的。而且FDD系统所需要的下行训练量正比于基站天线数。当基站天线数很多时，利用FDD估计CSIT就会变得困难。因此，大规模MIMO更倾向于使用TDD模式来得到CSIT，TDD模式的导频开销正比于用户数，在大规模MIMO系统中，该导频开销远小于FDD模式。然而3D MIMO系统主要受益于多用户分集增益，因此降低3D大规模MIMO TDD系统的导频开销也十分必要。 To effectively implement 3D beamforming, the base station needs to know the accurate CSIT. In the FDD system, the instantaneous CSIT is obtained through user feedback to the base station. Moreover, the amount of downlink training required by the FDD system is proportional to the number of base station antennas. When there are many base station antennas, it becomes difficult to estimate CSIT using FDD. Therefore, massive MIMO prefers to use TDD mode to obtain CSIT. The pilot overhead of TDD mode is proportional to the number of users. In massive MIMO system, the pilot overhead is much smaller than that of FDD mode. However, the 3D MIMO system mainly benefits from the multi-user diversity gain, so it is necessary to reduce the pilot overhead of the 3D massive MIMO TDD system. the

发明内容 Contents of the invention

发明目的：为了克服现有技术中存在的不足，本发明提供一种低复杂度的面向大规模MIMO TDD系统的3D波束成形方法，是一种基于信道稀疏性设计的一种大规模MIMO下的3D波束形成方案，能够有效降低计算复杂度和导频开销。 Purpose of the invention: In order to overcome the deficiencies in the prior art, the present invention provides a low-complexity 3D beamforming method for large-scale MIMO TDD systems, which is a large-scale MIMO beamforming method based on channel sparsity design. The 3D beamforming scheme can effectively reduce computational complexity and pilot overhead. the

技术方案：为实现上述目的，本发明采用的技术方案为： Technical scheme: in order to achieve the above object, the technical scheme adopted in the present invention is:

面向大规模MIMO TDD系统的3D波束成形方法，基于信道稀疏性进行大规模MIMO下的3D波束形成，包括如下步骤： A 3D beamforming method for massive MIMO TDD systems, based on channel sparsity for 3D beamforming under massive MIMO, includes the following steps:

步骤一、利用信道稀疏性，DFT（离散傅里叶变换）预波束成形将小区在水平方向分割成若干子扇区、在垂直方向分割成若干环形区域； Step 1. Using channel sparsity, DFT (Discrete Fourier Transform) pre-beamforming divides the cell into several sub-sectors in the horizontal direction and several ring-shaped areas in the vertical direction;

步骤二、 Step two,

采用分离的用户分组算法对用户进行三维分组，先把用户分到垂直维度的环形区域内，再将垂直维度的各环形区域内的用户分到相应的水平维度的子扇区内，用户的所在小组的代号由环形区域的代号和子扇区的代号共同构成；处于同一子扇区的非相邻组内的用户可以复用导频，处于同一环形区域的非相邻组内的用户也可以复用导频，基站获取CSIT； Use the separate user grouping algorithm to carry out three-dimensional grouping of users. First, users are divided into circular areas in the vertical dimension, and then users in each circular area in the vertical dimension are divided into sub-sectors in the corresponding horizontal dimension. The code name of the group is composed of the code name of the ring area and the code name of the sub-sector; users in non-adjacent groups in the same sub-sector can reuse pilots, and users in non-adjacent groups in the same ring area can also reuse pilots. Using the pilot, the base station obtains CSIT;

步骤三、在每组用户内，采用SLNR（信漏噪比）最大策略调度用户； Step 3. Within each group of users, use the SLNR (Signal Leakage to Noise Ratio) maximum strategy to schedule users;

步骤四、结合各组的DFT预波束成形，首先得到低阶的瞬时等效CSIT，再计算SLNR预编码，在计算SLNR预编码时综合考虑组内干扰和相邻组间干扰； Step 4. Combining the DFT pre-beamforming of each group, first obtain the low-order instantaneous equivalent CSIT, and then calculate the SLNR precoding, and comprehensively consider the intra-group interference and adjacent inter-group interference when calculating the SLNR precoding;

步骤五、3D波束成形方案是水平和垂直波束成形的直积，水平和垂直维度的波束成形为双层预编码，由DFT预波束成形和SLNR预编码构成。 Step 5. The 3D beamforming scheme is the direct product of horizontal and vertical beamforming, and the horizontal and vertical beamforming is a double-layer precoding, consisting of DFT prebeamforming and SLNR precoding. the

进一步地，所述基站位于小区的中心，服务多个用户，所述用户为单天线用户；基站的位置高于周围的建筑物高度，用户位于街道平面上；基站天线是N×M维的矩形阵列天线，每行配置M根天线单元，共有N行； Further, the base station is located in the center of the cell and serves multiple users, and the users are single-antenna users; the position of the base station is higher than the height of the surrounding buildings, and the users are located on the street plane; the base station antenna is an N×M dimensional rectangle Array antenna, each row is equipped with M antenna units, a total of N rows;

所述步骤一中，将小区在水平方向分割成G个子扇区、在垂直方向分割成L个环形区域，则各组的水平维度的DFT预波束成形为B_g,H=F_H[:,(g-1)r_H+(1:r_H)],(g=1,…,G)，其中F_H∈为DFT矩阵，各组的垂直维度的DFT预波束成形为B_l,V=F_V[:,(l-1)r_V+(1:r_V)],(l=1,…,L)，其中F_V∈为DFT矩阵， In the first step, the cell is divided into G sub-sectors in the horizontal direction and L ring-shaped areas in the vertical direction, then the DFT pre-beamforming in the horizontal dimension of each group is B _g,H =F _H [:, (g-1)r _H +(1:r _H )],(g=1,…,G), where F _H ∈ is the DFT matrix, The DFT pre-beamforming of the vertical dimension of each group is B _l,V =F _V [:,(l-1)r _V +(1:r _V )],(l=1,…,L), where F _V ∈ is the DFT matrix,

进一步地，所述步骤二中，记小区内所有用户数为K，分离的用户分组算法对用户进行三维分组，具体描述为： Further, in the step 2, record the number of all users in the cell as K, and separate the user grouping algorithm to perform three-dimensional grouping of users, specifically described as:

Step2-1：对于l=1,…,L，位于环形区域l内的所有用户的集合对于l=1,…,L、g=1,…,G，位于环形区域l、子扇区g内的所有用户记为第(l,g)组用户，第(l,g)组用户的集合 Step2-1: For l=1,...,L, the set of all users located in the ring area l For l=1,...,L, g=1,...,G, all users located in the annular area l and sub-sector g are recorded as the (l,g)th group users, and the (l,g)th group users gather

Step2-2：对于用户k=1,…,K，对水平和垂直信道协方差阵R_k,H和R_k,V进行特征值分解，得到U_k,H和U_k,V； Step2-2: For users k=1,...,K, perform eigenvalue decomposition on the horizontal and vertical channel covariance matrix R _k,H and R _k,V to obtain U _k,H and U _k,V ;

Step2-3：对于用户k=1,…,K，找到环形区域l、 $l = \arg \min_{l^{'}} {| | U_{k, V} U_{k, V}^{H} - B_{l^{'}, V} B_{l^{'}, V}^{H} | |}_{F}^{2},$ 将用户k加入到集合中，即 Step2-3: For users k=1,...,K, find the circular area l, $l = \arg \min_{l^{'}} {| | u_{k, V} u_{k, V}^{h} - B_{l^{'}, V} B_{l^{'}, V}^{h} | |}_{f}^{2},$ Add user k to the set in, namely

Step2-4：l=1,…,L，找到子扇区g， $g = \arg \min_{g^{'}} {| | U_{t, H} U_{t, H}^{H} - B_{g^{'}, H} B_{g^{'}, H}^{H} | |}_{F}^{2},$ 将用户t加入到集合中，即 Step2-4: l=1,...,L, find the subsector g, $g = \arg \min_{g^{'}} {| | u_{t, h} u_{t, h}^{h} - B_{g^{'}, h} B_{g^{'}, h}^{h} | |}_{f}^{2},$ Add user t to the set in, namely

进一步地，所述步骤三中，采用SLNR最大策略调度用户时，令第(l,g)组的调度用户数为S_l,g=min(r_H,r_V)，若第(l,g)组的被调度用户数为K_l,g≤min(r_H,r_V)，则全部调度；该调度算法具体描述为： Further, in the step 3, when using the SLNR maximum policy to schedule users, let the number of scheduled users in the (l,g)th group be S _l,g =min(r _H ,r _V ), if the (l,g)th group ) group with the number of scheduled users K _l,g ≤min(r _H ,r _V ), all are scheduled; the scheduling algorithm is specifically described as:

Step3-1：初始化如果K_l,g≤min(r_H,r_V)，则同时退出调度算法，否则执行Step3-2； Step3-1: Initialization If K _l,g ≤min(r _H ,r _V ), then Exit the scheduling algorithm at the same time, otherwise execute Step3-2;

Step3-2：循环遍历第(l,g)组内的每一个用户，求解出每个用户对其他用户的SLNR，对于用户k=1,…,K_l,g， ${\tilde{h}}_{{(l, g)}_{k}} = [h_{{(l, g)}_{1}}, . . ., h_{{(l, g)}_{k - 1}}, h_{{(l, g)}_{k + 1}}, . . ., h_{{(l, g)}_{K_{l, g}}}]$ 表示除了的扩展信道矩阵，计算 ${SLNR}_{{(l, g)}_{k}} = λ_{\max} [{(\frac{1}{ρ} I + {\tilde{h}}_{{(l, g)}_{k}} {\tilde{h}}_{{(l, g)}_{k}}^{H})}^{- 1} h_{{(l, g)}_{k}} h_{{(l, g)}_{k}}^{H}],$ 其中ρ是信噪比SNR的度量，正比于发送功率除以噪声方差，λ_max(A)表示矩阵A的最大特征值；执行Step3-3； Step3-2: Cycle through each user in the (l,g)th group, and find out the SLNR of each user to other users, for users k=1,...,K _l,g , ${\tilde{h}}_{{(l, g)}_{k}} = [h_{{(l, g)}_{1}}, . . ., h_{{(l, g)}_{k - 1}}, h_{{(l, g)}_{k + 1}}, . . ., h_{{(l, g)}_{K_{l, g}}}]$ means except The extended channel matrix of ${SLNR}_{{(l, g)}_{k}} = λ_{\max} [{(\frac{1}{ρ} I + {\tilde{h}}_{{(l, g)}_{k}} {\tilde{h}}_{{(l, g)}_{k}}^{h})}^{- 1} h_{{(l, g)}_{k}} h_{{(l, g)}_{k}}^{h}],$ Among them, ρ is a measure of the signal-to-noise ratio SNR, which is proportional to the transmission power divided by the noise variance, and λ _max (A) represents the maximum eigenvalue of the matrix A; execute Step3-3;

Step3-3：对所有待选用户的SLNR进行降序排列，输出所有待选用户的降序排列情况： Step3-3: Arrange the SLNRs of all users to be selected in descending order, and output the descending order of all users to be selected:

[SLNR_value,Index]=sort(SLNR,'descend')，其中SLNR_value表示按降序排列的SLNR值组成的数组，Index表示相应的用户代号，sort是排序操作，'descend'表示降序；执行Step3-4； [SLNR_value,Index]=sort(SLNR,'descend'), where SLNR_value represents an array of SLNR values arranged in descending order, Index represents the corresponding user ID, sort is a sorting operation, and 'descend' represents descending order; execute Step3-4 ;

Step3-4：选择SLNR最大的前S_l,g个用户作为服务对象，得到最终选择的用户集 Step3-4: Select the top S _l,g users with the largest SLNR as service objects, and get the final selected user set

进一步地，所述步骤四中： Further, in the step four:

第(l,g)组的第s个用户的水平SLNR预编码为： The horizontal SLNR precoding of the sth user in the (l, g)th group is:

${p p}_{{((l l,, g g))}_{s the s},, H h} = = {δ δ}_{{((l l,, g g))}_{s the s},, H h} {((\frac{11}{ρ ρ} I I + + {B B}_{g g,, H h}^{H h} {\overset{~ ~}{h h}}_{{((l l,, g g))}_{s the s},, H h} {\overset{~ ~}{h h}}_{{((l l . . g g))}_{s the s},, H h}^{H h} {B B}_{g g,, H h} + + {B B}_{g g,, H h}^{H h} {\overset{~ ~}{H h}}_{((l l,, g g)) . . H h} {\overset{~ ~}{H h}}_{((l l,, g g)),, H h}^{H h} {B B}_{g g,, H h}))}^{- - 11} {B B}_{g g,, H h}^{H h} {h h}_{{((l l . . g g))}_{s the s}} H h$

其中： ${\tilde{h}}_{{(l, g)}_{s}, H} = [h_{{(l, g)}_{1}, H}, . . ., h_{{(l, g)}_{s - 1}, H}, h_{{(l, g)}_{s + 1}, H}, . . ., h_{{(l, g)}_{S_{l, g}}, H}]$ 是除了的扩展水平信道矩阵；是相邻子扇区的扩展水平信道矩阵；是功率归一化因子，满足 in: ${\tilde{h}}_{{(l, g)}_{the s}, h} = [h_{{(l, g)}_{1}, h}, . . ., h_{{(l, g)}_{the s - 1}, h}, h_{{(l, g)}_{the s + 1}, h}, . . ., h_{{(l, g)}_{S_{l, g}}, h}]$ except The extended horizontal channel matrix of ; is the extended horizontal channel matrix of adjacent sub-sectors; is the power normalization factor, satisfying

第(l,g)组的第s个用户的垂直SLNR预编码为： The vertical SLNR precoding of the sth user in the (l, g)th group is:

${p p}_{{((l l,, g g))}_{s the s},, V V} = = {δ δ}_{{((l l,, g g))}_{s the s},, V V} {((\frac{11}{ρ ρ} I I + + {B B}_{g g,, V V}^{H h} {\overset{~ ~}{h h}}_{{((l l,, g g))}_{s the s},, V V} {\overset{~ ~}{h h}}_{{((l l,, g g))}_{s the s},, V V}^{H h} + + {B B}_{g g,, V V} + + {B B}_{g g,, V V}^{H h} {\overset{~ ~}{H h}}_{((l l,, g g)),, V V} {\overset{~ ~}{H h}}_{((l l,, g g)),, V V}^{H h} {B B}_{g g,, V V}))}^{- - 11} {B B}_{g g,, V V}^{H h} {h h}_{{((l l,, g g))}_{s the s},, V V}$

其中： ${\tilde{h}}_{{(l, g)}_{s}, V} = [h_{{(l, g)}_{1}, V} {, . . ., h}_{{(l, g)}_{s - 1}, V}, h_{{(l, g)}_{s + 1}, V} {, . . ., h}_{{(l, g)}_{S_{l, g}}, V}]$ 是除了的扩展垂直信道矩阵，是相邻环形区域的扩展垂直信道矩阵；是功率归一化因子，满足 in: ${\tilde{h}}_{{(l, g)}_{the s}, V} = [h_{{(l, g)}_{1}, V} {, . . ., h}_{{(l, g)}_{the s - 1}, V}, h_{{(l, g)}_{the s + 1}, V} {, . . ., h}_{{(l, g)}_{S_{l, g}}, V}]$ except The extended vertical channel matrix of , is the extended vertical channel matrix of the adjacent ring area; is the power normalization factor, satisfying

进一步地，所述步骤五中：基站发射的3D波束成形为 $x = Σ_{l = 1}^{L} Σ_{g = 1}^{G} Σ_{s = 1}^{S_{l, g}} [(B_{g, H} p_{{(l . g)}_{s}, H}) &CircleTimes; (B_{l, V} p_{{(l, g)}_{s}, V})] d_{{(l, g)}_{s},}$ 其中为发送给第(l,g)组的第s个用户的数据。 Further, in step five: the 3D beamforming transmitted by the base station is $x = Σ_{l = 1}^{L} Σ_{g = 1}^{G} Σ_{the s = 1}^{S_{l, g}} [(B_{g, h} p_{{(l . g)}_{the s}, h}) &CircleTimes; (B_{l, V} p_{{(l, g)}_{the s}, V})] d_{{(l, g)}_{the s},}$ in is the data sent to the sth user of the (l,g)th group.

有益效果：本发明提供的面向大规模MIMO TDD系统的3D波束成形方法，能够降低大规模MIMO的计算预编码的复杂度，同时能够节省导频开销。 Beneficial effects: the 3D beamforming method for massive MIMO TDD systems provided by the present invention can reduce the complexity of computational precoding for massive MIMO and save pilot overhead at the same time. the

SLNR预编码是基于降低的瞬时等效信道得到，在计算水平SLNR预编码时，矩阵求逆的复杂度为在计算垂直SLNR预编码时矩阵求逆的复杂度为在这里SLNR预编码只考虑相邻小组的泄露，这是因为如果考虑所有组的泄露，会增加乘法复杂度；在不分组的系统中，水平SLNR预编码的矩阵求逆复杂度为O(M³)，垂直SLNR预编码的矩阵求逆复杂度为O(N³)；因此，本发明的SLNR预编码的矩阵求逆复杂度降低了，降低的复杂度取决于环形区域数和子扇区数。 SLNR precoding is obtained based on the reduced instantaneous equivalent channel, at the calculation level SLNR precoding When , the complexity of matrix inversion is When calculating the vertical SLNR precoding The complexity of matrix inversion is Here SLNR precoding only considers the leakage of adjacent groups, because if the leakage of all groups is considered, the multiplication complexity will increase; in the system without grouping, the matrix inversion complexity of horizontal SLNR precoding is O(M ³ ), the matrix inverse complexity of vertical SLNR precoding is O(N ³ ); therefore, the matrix inverse complexity of SLNR precoding of the present invention is reduced, and the reduced complexity depends on the number of ring regions and the number of sub-sectors .

第(l,g)组第s个用户的信道协方差阵为其中和R分别是用户的水平和垂直信道协方差阵，令（l'≠l或者g'≠g）表示第(l',g')组的第s个用户，有 $R_{{(l, g)}_{s}} R_{{(l^{'}, g^{'})}_{s}} = (R_{{(l, g)}_{s}} &CircleTimes; R_{{(l, g)}_{s}}) (R_{{(l^{'}, g^{'})}_{s}} &CircleTimes; R_{{(l^{'}, g^{'})}_{s}}) = (R_{{(l, g)}_{s}} R_{{(l^{'}, g^{'})}_{s}}) &CircleTimes; (R_{{(l, g)}_{s,} V} R_{{(l^{'}, g^{'})}_{s,} V})$ ；因此只要任一维度的信道协方差阵乘积为0，那么3D信道协方差阵就为0，根据相关文献可知这两个用户可以复用导频；只要用户任一维度的角度范围不重叠，那么他们的信道协方差阵的特征矩阵正交，3D信道协方差阵的乘积为0，意味着这两个用户可以复用导频；而处于同一子扇区的非相邻组内的用户，以及处于同一环形区域的非相邻组内的用户的角度范围不重叠，从而可以复用导频，降低导频开销。 The channel covariance matrix of the sth user in group (l, g) is in and R are the user’s horizontal and vertical channel covariance matrix respectively, so that (l'≠l or g'≠g) means the sth user of the (l',g')th group, have $R_{{(l, g)}_{the s}} R_{{(l^{'}, g^{'})}_{the s}} = (R_{{(l, g)}_{the s}} &CircleTimes; R_{{(l, g)}_{the s}}) (R_{{(l^{'}, g^{'})}_{the s}} &CircleTimes; R_{{(l^{'}, g^{'})}_{the s}}) = (R_{{(l, g)}_{the s}} R_{{(l^{'}, g^{'})}_{the s}}) &CircleTimes; (R_{{(l, g)}_{the s,} V} R_{{(l^{'}, g^{'})}_{the s,} V})$ ; Therefore, as long as the channel covariance matrix product of any dimension is 0, then the 3D channel covariance matrix is 0. According to relevant literature, the two users can reuse pilots; as long as the angle ranges of any dimension of the users do not overlap, Then the eigenmatrix of their channel covariance matrix is orthogonal, and the product of the 3D channel covariance matrix is 0, which means that the two users can reuse pilots; and users in non-adjacent groups in the same sub-sector, And the angular ranges of users in non-adjacent groups in the same ring area do not overlap, so pilots can be reused to reduce pilot overhead.

附图说明 Description of drawings

图1为本发明的模块图； Fig. 1 is a block diagram of the present invention;

图2为小区分割图； Fig. 2 is a cell division diagram;

图3为改变发送功率，三种系统的和速率比较图； Figure 3 is a comparison diagram of the sum rate of the three systems when the transmission power is changed;

图4为改变水平维度和垂直维度组数，三种系统的和速率比较图； Figure 4 is a comparison diagram of the sum rate of the three systems by changing the number of groups in the horizontal dimension and vertical dimension;

图5为改变基站天线数，三种系统的和速率比较图。 Figure 5 is a comparison diagram of the sum and rate of the three systems when the number of base station antennas is changed. the

具体实施方式 Detailed ways

下面结合附图对本发明作更进一步的说明。 The present invention will be further described below in conjunction with the accompanying drawings. the

一种面向大规模MIMO TDD系统的3D波束成形方法，在布置时将基站位于小区的中心，服务多个用户，所述用户为单天线用户；基站的位置高于周围的建筑物高度，用户位于街道平面上；基站天线是N×M维的矩形阵列天线，每行配置M根天线单元，共有N行。 A 3D beamforming method for large-scale MIMO TDD systems, the base station is located in the center of the cell when it is arranged, serving multiple users, and the users are single-antenna users; the position of the base station is higher than the surrounding building height, and the users are located On the street plane; the base station antenna is an N×M dimensional rectangular array antenna, with M antenna units configured in each row, and there are N rows in total. the

利用信道稀疏性，DFT预波束成形将小区在水平维度分为G个子扇区，在垂直维度分为L个环形区域，这样：各组的水平维度的DFT预波束成形为B_g,H=F_H[:,(g-1)r_H+(1:r_H)],(g=1,…,G)，其中F_H∈为DFT矩阵，为水平各组的虚拟天线数；各组的垂直维度的DFT预波束成形为 B_l,V=F_V[:,(l-1)r_V+(1:r_V)],(l=1,…,L)，其中F_V∈为DFT矩阵，为垂直各组的虚拟天线数。 Utilizing the channel sparsity, DFT pre-beamforming divides the cell into G sub-sectors in the horizontal dimension and L ring-shaped areas in the vertical dimension, so: the DFT pre-beamforming in the horizontal dimension of each group is B _g,H =F _H [:,(g-1)r _H +(1:r _H )],(g=1,…,G), where F _H ∈ is the DFT matrix, is the number of virtual antennas in each horizontal group; the DFT pre-beamforming in the vertical dimension of each group is B _l,V =F _V [:,(l-1)r _V +(1:r _V )],(l=1 ,…,L), where F _V ∈ is the DFT matrix, is the number of virtual antennas in each vertical group.

记小区内所有用户数为K，采用分离的用户分组算法对用户进行三维分组，先把用户分到垂直维度的环形区域内，再将垂直维度的各环形区域内的用户分到水平维度的子扇区内，用户的所在小组的代号由环形区域的代号和子扇区的代号共同构成；处于同一子扇区的非相邻组内的用户可以复用导频，处于同一环形区域的非相邻组内的用户也可以复用导频，基站获取CSIT；具体描述为： Note that the number of all users in the community is K, and use the separate user grouping algorithm to carry out three-dimensional grouping of users. First, users are divided into circular areas in the vertical dimension, and then users in each circular area in the vertical dimension are divided into sub-groups in the horizontal dimension. In a sector, the group code of the user is composed of the code name of the ring area and the code name of the sub-sector; users in non-adjacent groups in the same sub-sector can reuse pilots, and users in non-adjacent groups in the same ring area Users in the group can also multiplex pilots, and the base station obtains CSIT; the specific description is as follows:

Step2-1：对于l=1,…,L，位于环形区域l内的所有用户的集合；对于l=1,…,L、g=1,…,G，位于环形区域l、子扇区g内的所有用户记为第(l,g)组用户，第(l,g)组用户的集合 Step2-1: For l=1,...,L, the set of all users located in the ring area l ; For l=1,...,L, g=1,...,G, all users located in ring area l and sub-sector g are recorded as (l,g)th group users, (l,g)th group users collection of

在每组用户内，采用SLNR最大策略调度用户，令第(l,g)组的调度用户数为S_l,g=min(r_H,r_V)，若第(l,g)组的被调度用户数为K_l,g≤min(r_H,r_V)，则全部调度；该调度算法具体描述为： In each group of users, the SLNR maximum policy is used to schedule users, so that the number of scheduled users in the (l,g)th group is S _l,g =min(r _H ,r _V ), if the (l,g)th group is When the number of scheduled users is K _l,g ≤min(r _H ,r _V ), all are scheduled; the scheduling algorithm is specifically described as:

结合各组的DFT预波束成形，首先得到低阶的瞬时等效CSIT，再计算SLNR预编码，在计算SLNR预编码时综合考虑组内干扰和相邻组间干扰；其中第(l,g)组的第s个用户的水平SLNR预编码为： Combined with the DFT pre-beamforming of each group, the low-order instantaneous equivalent CSIT is firstly obtained, and then the SLNR precoding is calculated. When calculating the SLNR precoding, the interference within the group and the interference between adjacent groups are considered comprehensively; where (l, g) The horizontal SLNR precoding of the sth user of the group is:

其中： ${\tilde{h}}_{{(l, g)}_{k}} = [h_{{(l, g)}_{1}}, . . ., h_{{(l, g)}_{k - 1}}, h_{{(l, g)}_{k + 1}}, . . ., h_{{(l, g)}_{K_{l, g}}}]$ 是除了的扩展水平信道矩阵；是相邻子扇区的扩展水平信道矩阵；是功率归一化因子，满足 in: ${\tilde{h}}_{{(l, g)}_{k}} = [h_{{(l, g)}_{1}}, . . ., h_{{(l, g)}_{k - 1}}, h_{{(l, g)}_{k + 1}}, . . ., h_{{(l, g)}_{K_{l, g}}}]$ except The extended horizontal channel matrix of ; is the extended horizontal channel matrix of adjacent sub-sectors; is the power normalization factor, satisfying

3D波束成形方案是水平和垂直波束成形的直积，水平和垂直维度的波束成形为双层预编码，由DFT预波束成形和SLNR预编码构成；基站发射的3D波束成形为 $x = Σ_{l = 1}^{L} Σ_{g = 1}^{G} Σ_{s = 1}^{S_{l, g}} [(B_{g, H} p_{{(l . g)}_{s}, H}) &CircleTimes; (B_{l, V} p_{{(l, g)}_{s}, V})] d_{{(l, g)}_{s},}$ 其中为发送给第(l,g)组的第s个用户的数据。 The 3D beamforming scheme is the direct product of horizontal and vertical beamforming. The beamforming in the horizontal and vertical dimensions is a double-layer precoding, consisting of DFT pre-beamforming and SLNR precoding; the 3D beamforming transmitted by the base station is $x = Σ_{l = 1}^{L} Σ_{g = 1}^{G} Σ_{the s = 1}^{S_{l, g}} [(B_{g, h} p_{{(l . g)}_{the s}, h}) &CircleTimes; (B_{l, V} p_{{(l, g)}_{the s}, V})] d_{{(l, g)}_{the s},}$ in is the data sent to the sth user of the (l,g)th group.

下面对本发明方法与其他方法的性能对比做出说明： The performance comparison of the inventive method and other methods is described below:

图3到图5比较了三种3D大规模MIMO系统的和速率。仿真图中的标示“传统的ZF预编码”是指系统的多用户预编码使用的是只消除组内干扰的ZF预编码；标示“本发明的SLNR预编码”是指使用的多用户预编码是考虑组内和相邻组之间干扰的SLNR预编码。标示“不分组系统的SLNR预编码”是指系统不对小区进行分组，每个用户使用SLNR预编码。图3到图5的仿真场景设置如下：基站高度为50m。用户端天线为单天线接收。用户到基站的距离为30m到519.6m。用户均匀分布在120°的扇区内，共有10个用户等待调度。 Figures 3 to 5 compare the sum rates of the three 3D massive MIMO systems. The mark "traditional ZF precoding" in the simulation figure means that the multi-user precoding of the system uses ZF precoding that only eliminates intra-group interference; the mark "SLNR precoding of the present invention" means that the multi-user precoding used is the SLNR precoding that considers the interference within a group and between adjacent groups. The label "SLNR precoding in a non-grouping system" means that the system does not group cells, and each user uses SLNR precoding. The simulation scenarios in Figures 3 to 5 are set as follows: the height of the base station is 50m. The user end antenna is single-antenna reception. The distance from the user to the base station is 30m to 519.6m. Users are evenly distributed in a 120° sector, and a total of 10 users are waiting for scheduling. the

图3为改变发送功率，三种系统的和速率比较图。具体仿真场景设置如下：基站天线为64×64矩形天线，小区在水平维度分为8个子扇区，垂直维度分为8个环形区域，改变发送功率从0dB到20dB。由仿真结果可以看出，在整个发送功率范围内，本发明的性能比传统的ZF预编码的性能更接近不分组系统，这是因为本发明除了考虑组内干扰，也考虑了相邻组间干扰。同时本发明的水平和垂直SLNR预编码的矩阵求逆的计算复杂度均为O(8³)，不分组系统的计算复杂度为O(64³)，分组系统的计算复杂度远远低于不分组系统的计算复杂度。在分组系统中用户只需要使用4组导频，而不分组系统需使用全部64组导频。因此，导频开销也大大降低了。 Figure 3 is a comparison chart of the sum rate of the three systems when the transmission power is changed. The specific simulation scene settings are as follows: the base station antenna is a 64×64 rectangular antenna, the cell is divided into 8 sub-sectors in the horizontal dimension, and 8 ring-shaped areas in the vertical dimension, and the transmission power is changed from 0dB to 20dB. It can be seen from the simulation results that in the entire transmission power range, the performance of the present invention is closer to that of the non-grouping system than that of the traditional ZF precoding, because the present invention also considers the interference between adjacent groups in addition to the intra-group interference. interference. At the same time, the computational complexity of the matrix inversion of the horizontal and vertical SLNR precoding of the present invention is O(8 ³ ), the computational complexity of the non-grouping system is O(64 ³ ), and the computational complexity of the grouping system is far lower than Computational complexity of ungrouped systems. In the packet system, the user only needs to use 4 sets of pilots, while the non-packet system needs to use all 64 sets of pilots. Therefore, the pilot overhead is also greatly reduced.

图4为改变水平维度和垂直维度组数，三种系统的和速率比较图。具体仿真场景设置如下：基站天线为64×64矩形天线，固定基站的发送功率为10dB，同时改变环形区域数和子扇区数，水平和垂直组数取值为4组、8组、12组和16组。从仿真结果可以看出，组数越多，和速率越低，这是因为保持基站天线数不变，增加组数，每组的虚拟天线数降低，这样分集增益降低，且每组可服务的用户数也降低。但另一方面，增加组数，多用户预编码的计算复杂度大大降低。因此组数是权衡计算复杂度与和速率的重要因素。 Figure 4 is a comparison diagram of the sum rate of the three systems by changing the number of groups in the horizontal dimension and vertical dimension. The specific simulation scene settings are as follows: the base station antenna is a 64×64 rectangular antenna, the transmission power of the fixed base station is 10dB, and the number of circular areas and sub-sectors is changed at the same time, and the values of the horizontal and vertical groups are 4 groups, 8 groups, 12 groups and 16 groups. It can be seen from the simulation results that the more the number of groups, the lower the sum rate. This is because keeping the number of base station antennas constant and increasing the number of groups reduces the number of virtual antennas in each group, so that the diversity gain decreases, and each group can serve The number of users also decreased. But on the other hand, increasing the number of groups, the computational complexity of multi-user precoding is greatly reduced. Therefore, the number of groups is an important factor to balance the computational complexity and speed. the

图5为改变基站天线数，三种系统的和速率比较图。除了不分组系统的性能，绘制了两组曲线。仿真场景设置如下：固定基站的发送功率为10dB，改变基站的水平天线数和垂直天线数，取值为32×32、64×64、128×128和256×256。第一组仿真固定分割组数，小区在水平维度分为8个子扇区，垂直维度分为8个环形区域。从仿真结果可以看出，天线数越多，和速率越高，且随着天线数增多，本发明的和速率逐渐逼近不分组系统。第二组仿真保持各组的虚拟天线个数为8，由此改变环形区域数和子扇区数，水平和垂直组数取值为4组、8组、16组和32组。从仿真结果可以看出，ZF预编码的性能逐渐逼近SLNR预编码的性能。这是因为保持每组的虚拟天线数不变，ZF预编码和本发明的SLNR预编码的组内干扰相同，但改变基站的总天线数，DFT的波束变窄，组外干扰变小，因此ZF预编码的性能逐渐逼近SLNR预编码的性能。随着基站天线数目的不断增加，第二组仿真实验中每组的分集增益保持不变，第一组情况的每组虚拟天线数随着基站天线数的增加而增加。而且第二组实验中每组可选择的用户数小于第一组实验，所以第二组曲线的和速率提升不如第一组。 Figure 5 is a comparison diagram of the sum and rate of the three systems when the number of base station antennas is changed. In addition to the performance of the ungrouped systems, two sets of curves are plotted. The simulation scene is set as follows: the transmit power of the base station is fixed at 10dB, and the number of horizontal antennas and vertical antennas of the base station is changed, and the values are 32×32, 64×64, 128×128 and 256×256. The first group simulates a fixed number of division groups. The cell is divided into 8 sub-sectors in the horizontal dimension and 8 ring-shaped areas in the vertical dimension. It can be seen from the simulation results that the more the number of antennas, the higher the sum rate, and as the number of antennas increases, the sum rate of the present invention gradually approaches that of the non-grouping system. The second group of simulation keeps the number of virtual antennas in each group at 8, thereby changing the number of ring regions and sub-sectors, and the values of horizontal and vertical groups are 4 groups, 8 groups, 16 groups and 32 groups. It can be seen from the simulation results that the performance of ZF precoding is gradually approaching that of SLNR precoding. This is because the number of virtual antennas in each group is kept constant, ZF precoding and SLNR precoding of the present invention have the same intragroup interference, but changing the total number of antennas in the base station, the beam of DFT becomes narrower, and the interference outside the group becomes smaller, so The performance of ZF precoding gradually approaches that of SLNR precoding. As the number of base station antennas increases, the diversity gain of each group in the second group of simulation experiments remains unchanged, and the number of virtual antennas in each group increases with the increase of the number of base station antennas in the first group. Moreover, the number of selectable users in each group in the second group of experiments is smaller than that of the first group of experiments, so the sum rate improvement of the second group of curves is not as good as that of the first group. the

以上所述仅是本发明的优选实施方式，应当指出：对于本技术领域的普通技术人员来说，在不脱离本发明原理的前提下，还可以做出若干改进和润饰，这些改进和润饰也应视为本发明的保护范围。 The above is only a preferred embodiment of the present invention, it should be pointed out that for those of ordinary skill in the art, without departing from the principle of the present invention, some improvements and modifications can also be made, and these improvements and modifications are also possible. It should be regarded as the protection scope of the present invention. the

Claims

1. towards the 3D beam-forming method of extensive MIMO TDD system, it is characterized in that: the 3D wave beam carrying out under extensive MIMO based on the sparse property of channel forms, and comprises the steps:

Step 1, utilize the sparse property of channel, DFT switched-beam is divided into community in the horizontal direction some sub-sectors, is divided into some annular regions in the vertical direction;

Step 2, adopt separated user grouping algorithm to carry out three-dimensional grouping to user, first user is assigned in the annular region of vertical dimensions, the user in each annular region of vertical dimensions is assigned in the sub-sector of corresponding horizontal dimensions, the code name of user's place group consists of jointly the code name of annular region and the code name of sub-sector again; User in non-adjacent group in same sub-sector can multiplexed pilot, and the user in non-adjacent group in same annular region also can multiplexed pilot, and base station obtains CSIT;

Step 3, at every group with indoor, adopt the maximum tactful dispatched users of SLNR;

Step 4, in conjunction with the DFT switched-beam of each group, first obtain the instantaneous equivalent CSIT of low order, then calculate SLNR precoding, when calculating SLNR precoding, consider in group and disturb between interference and adjacent set;

Step 5,3D beam forming scheme are the direct products of horizontal and vertical beam forming, and the beam forming of horizontal and vertical dimension is double-deck precoding, DFT switched-beam and SLNR precoding, consist of.

2. the 3D beam-forming method towards extensive MIMO TDD system according to claim 1, is characterized in that: described base station is positioned at the center of community, serves a plurality of users, and described user is single antenna user; The projecting depth of building in position of base station, user is positioned in the plane of street; Antenna for base station is the Rectangular array antenna of N * M dimension, every row configuration M root antenna element, and total N is capable;

In described step 1, community is divided in the horizontal direction to G sub-sector, is divided into L annular region in the vertical direction, the DFT switched-beam of the horizontal dimensions of each group is B _g,H=F _h[:, (g-1) r _h+ (1:r _h)], (g=1 ..., G), F wherein _h∈ for DFT matrix, the DFT switched-beam of the vertical dimensions of each group is B _l,V=F _v[:, (l-1) r _v+ (1:r _v)], (l=1 ..., L), F wherein _v∈ for DFT matrix,

3. the 3D beam-forming method towards extensive MIMO TDD system according to claim 2, it is characterized in that: in described step 2, in note community, all numbers of users are K, use separated user grouping algorithm to carry out three-dimensional grouping to user, specifically describe to be:

Step2-1: for l=1 ..., L, is positioned at all users' of annular region l set for l=1 ..., L, g=1 ..., G, all users that are positioned at annular region l, sub-sector g are designated as (l, g) group user, (l, g) group user's set

Step2-2: for user k=1 ..., K, to horizontal and vertical channel covariance matrix R _k,Hand R _k,Vcarry out Eigenvalues Decomposition, obtain U _k,Hand U _k,V;

Step2-3: for user k=1 ..., K, find annular region l,

l = \arg \min_{l^{'}} {| | U_{k, V} U_{k, V}^{H} - B_{l^{'}, V} B_{l^{'}, V}^{H} | |}_{F}^{2},

User k is joined to set in,

Step2-4: l=1 ..., L, finds sub-sector g,

g = \arg \min_{g^{'}} {| | U_{t, H} U_{t, H}^{H} - B_{g^{'}, H} B_{g^{'}, H}^{H} | |}_{F}^{2},

User t is joined to set in,

4. the 3D beam-forming method towards extensive MIMO TDD system according to claim 3, is characterized in that: in described step 3, while adopting the maximum tactful dispatched users of SLNR, making the dispatched users number of (l, g) group is S _l,g=min (r _h, r _v), if scheduled user's number of (l, g) group is K _l,g≤ min (r _h, r _v), all scheduling; This dispatching algorithm specifically describes:

Step3-1: initialization if K _l,g≤ min (r _h, r _v), D _l,g=T _l,g, exit dispatching algorithm simultaneously, otherwise carry out Step3-2;

Step3-2: each user in searching loop (l, g) group, solve the SLNR of each user to other users, for user k=1 ...., K _{l, g}

{\tilde{h}}_{{(l, g)}_{k}} = [h_{{(l, g)}_{1}}, . . ., h_{{(l, g)}_{k - 1}}, h_{{(l, g)}_{k + 1}}, . . ., h_{{(l, g)}_{K_{l, g}}}]

Expression except extended channel matrices, calculate

{SLNR}_{{(l, g)}_{k}} = λ_{\max} [{(\frac{1}{ρ} I + {\tilde{h}}_{{(l, g)}_{k}} {\tilde{h}}_{{(l, g)}_{k}}^{H})}^{- 1} h_{{(l, g)}_{k}} h_{{(l, g)}_{k}}^{H}],

Wherein ρ is the tolerance of signal to noise ratio snr, is proportional to transmitted power divided by noise variance, λ _max(A) eigenvalue of maximum of representing matrix A; Carry out Step3-3;

Step3-3: all users' to be selected SLNR is carried out to descending, export all users' to be selected descending situation: [SLNR_value, Index]=sort (SLNR, ' descend'), wherein SLNR_value represents the array forming by the SLNR value of descending, Index represents corresponding User ID, and sort is sorting operation, ' descend' represents descending; Carry out Step3-4;

Step3-4: the front S that selects SLNR maximum _l,gindividual user, as service object, obtains final user's collection of selecting

5. the 3D beam-forming method towards extensive MIMO TDD system according to claim 4, is characterized in that: in described step 4:

S the user's of (l, g) group horizontal SLNR precoding is:

p_{{(l, g)}_{s}, H} = δ_{{(l, g)}_{s}, H} {(\frac{1}{ρ} I + B_{g, H}^{H} {\tilde{h}}_{{(l, g)}_{s}, H} {\tilde{h}}_{{(l . g)}_{s}, H}^{H} B_{g, H} + B_{g, H}^{H} {\tilde{H}}_{(l, g) . H} {\tilde{H}}_{(l, g), H}^{H} B_{g, H})}^{- 1} B_{g, H}^{H} h_{{(l . g)}_{s}} H

Wherein:

{\tilde{h}}_{{(l, g)}_{s}, H} = [h_{{(l, g)}_{1}, H}, . . ., h_{{(l, g)}_{s - 1}, H}, h_{{(l, g)}_{s + 1}, H}, . . ., h_{{(l, g)}_{S_{l, g}}, H}]

Be except expansion horizontal channel matrix; it is the expansion horizontal channel matrix of adjacent sub-sector; be the power normalization factor, meet

S the user's of (l, g) group vertical SLNR precoding is:

p_{{(l, g)}_{s}, V} = δ_{{(l, g)}_{s}, V} {(\frac{1}{ρ} I + B_{g, V}^{H} {\tilde{h}}_{{(l, g)}_{s}, V} {\tilde{h}}_{{(l, g)}_{s}, V}^{H} + B_{g, V} + B_{g, V}^{H} {\tilde{H}}_{(l, g), V} {\tilde{H}}_{(l, g), V}^{H} B_{g, V})}^{- 1} B_{g, V}^{H} h_{{(l, g)}_{s}, V}

Wherein:

{\tilde{h}}_{{(l, g)}_{s}, V} = [h_{{(l, g)}_{1}, V} {, . . ., h}_{{(l, g)}_{s - 1}, V}, h_{{(l, g)}_{s + 1}, V} {, . . ., h}_{{(l, g)}_{S_{l, g}}, V}]

Be except expansion vertical channel matrix, it is the expansion vertical channel matrix in adjacent annular region; be the power normalization factor, meet

6. the 3D beam-forming method towards extensive MIMO TDD system according to claim 5, is characterized in that: in described step 5: the 3D beam forming of base station transmitting is

x = Σ_{l = 1}^{L} Σ_{g = 1}^{G} Σ_{s = 1}^{S_{l, g}} [(B_{g, H} p_{{(l . g)}_{s}, H}) &CircleTimes; (B_{l, V} p_{{(l, g)}_{s}, V})] d_{{(l, g)}_{s},}

Wherein for sending to s the user's of (l, g) group data.