CN100550686C

CN100550686C - The beam forming method that a kind of receiver is auxiliary

Info

Publication number: CN100550686C
Application number: CNB2005100385861A
Authority: CN
Inventors: 黄森华; 邱玲
Original assignee: University of Science and Technology of China USTC
Current assignee: University of Science and Technology of China USTC
Priority date: 2005-03-24
Filing date: 2005-03-24
Publication date: 2009-10-14
Anticipated expiration: 2025-03-24
Also published as: CN1838559A

Abstract

The receiver-assisted beamforming method of the present invention is characterized in that a beam codebook that is generated by computer simulation or given by theoretical analysis and includes multiple beam matrices is stored at the base station and the user end; the channel state information of the user receiver In addition to being used for data demodulation, it is also used to select the beam matrix and beam vector that enable the user to obtain the maximum rate in the beam codebook; the user end transmits the beam matrix sequence number that obtains the maximum rate in the beam codebook to the base station through the feedback link ; The base station first calculates the throughput rate that each beam matrix can achieve according to the feedback information of the user, and then selects the beam matrix with the largest throughput rate as the beam matrix multiplied with the transmitted signal vector; the base station passes the data of the users in the effective user set The method of amplitude expansion distributes the total power evenly to the used beams. Compared with the existing algorithm, the present invention obtains higher throughput rate with smaller feedback amount and scheduling algorithm complexity.

Description

A receiver-assisted beamforming method

技术领域： Technical field:

本发明属于移动通信单小区多天线广播信道容量技术领域，特别是涉及多天线广播信道中基站根据用户的反馈信息进行调度和预处理的波束成型方法。The invention belongs to the technical field of mobile communication single-cell multi-antenna broadcast channel capacity, and in particular relates to a beamforming method in which a base station performs scheduling and preprocessing according to user feedback information in a multi-antenna broadcast channel.

背景技术： Background technique:

多用户分集是目前移动通信单小区多天线广播信道容量技术领域广泛研究的、提高无线系统频谱效率的有效方法。Multi-user diversity is an effective method to improve the spectral efficiency of wireless systems, which has been extensively studied in the field of mobile communication single-cell multi-antenna broadcast channel capacity technology.

《国际电子与电气工程师协会信息学报》(“On the achievable throughput of amultiantenna Gaussian broadcast channel”，IEEE Transactions on Information Theory，Vol.49No.7，July 2003)介绍了一种在基站端有多天线、用户端有单天线的广播信道中利用脏纸编码技术达到下行广播信道和容量的方法，但由于脏纸编码是非线性的，而且需要知道别的用户的干扰信息，使该方法复杂度太高，且基站端需要完全的信道状态信息，这在实际中是难以实现的。此文还介绍了迫零波束成型方法，由于其须在基站端对用户信道进行反转，使得该方法需要基站端具有完全的信道状态信息，且其最优的用户调度由于需要在所有可能的用户组合中搜索最优的用户组合，计算复杂度会随着用户数的增加而呈指数增长。"On the achievable throughput of amultiantenna Gaussian broadcast channel", IEEE Transactions on Information Theory, Vol.49No.7, July 2003, introduced a multi-antenna, user In the broadcast channel with a single antenna at the end, the method of using dirty paper coding technology to achieve the downlink broadcast channel and capacity, but because the dirty paper coding is nonlinear and needs to know the interference information of other users, the complexity of this method is too high, and The base station needs complete channel state information, which is difficult to realize in practice. This paper also introduces the zero-forcing beamforming method. Because it needs to invert the user channel at the base station, this method requires the base station to have complete channel state information, and its optimal user scheduling needs to be in all possible Searching for the optimal user combination in the user combination, the computational complexity will increase exponentially with the increase of the number of users.

《国际电子与电气工程师协会信息学报》(“On the capacity of MIMO broadcast channelwith partial side information”，IEEE Transactions on Information Theory，)介绍了一种实现多用户分集的只需要部分信道信息的随机波束成型方法，但由于用户反馈的信息太少，接收机的信道状态信息没有得到充分的利用，该方法的吞吐率损失太大。"International Institute of Electronics and Electrical Engineers Information Journal" ("On the capacity of MIMO broadcast channel with partial side information", IEEE Transactions on Information Theory,) introduces a random beamforming method that only requires partial channel information to achieve multi-user diversity , but because the information fed back by the user is too little, the channel state information of the receiver is not fully utilized, and the throughput loss of this method is too large.

发明内容： Invention content:

本发明提出一种接收机辅助的波束成型方法，可以充分利用用户端接收机的信道状态信息，减少吞吐率的损失，而且在基站端只需要部分的信道状态信息，其最优的用户调度计算复杂度随用户数的增加为线性增长。The present invention proposes a receiver-assisted beamforming method, which can make full use of the channel state information of the receiver at the user end to reduce the loss of throughput, and only part of the channel state information is needed at the base station end, and its optimal user scheduling calculation The complexity grows linearly with the number of users.

本发明接收机辅助的波束成型方法，在基站将发送信号矢量与波束矩阵相乘，所述波束矩阵为酉阵；每个用户通过反馈链路将其得到的所有可能的信号干扰噪声比中最大的信号干扰噪声比和对应的波束矢量的序号反馈给基站；基站根据反馈信息将每个波束矢量分配给在那个波束矢量上取得最大信号干扰噪声比的用户；用户接收机根据基站在训练阶段发射的导频信号和本接收机在训练阶段的接收信号进行反向求解获得信道状态信息；In the receiver-assisted beamforming method of the present invention, the base station multiplies the transmitted signal vector with the beam matrix, and the beam matrix is a unitary matrix; each user obtains the largest signal-to-interference-to-noise ratio through a feedback link The signal-to-interference-noise ratio and the serial number of the corresponding beam vector are fed back to the base station; the base station assigns each beam vector to the user who obtains the maximum signal-to-interference-noise ratio on that beam vector according to the feedback information; The pilot signal and the received signal of the receiver in the training phase are reversely solved to obtain the channel state information;

其特征在于：It is characterized by:

在基站和用户端分别存储一个由计算机仿真生成或由理论分析给出的、包含多个波束矩阵的波束码本；用户接收机的信道状态信息除了用于数据解调，还用于在波束码本中选择使本用户获得最大信号干扰噪声比的波束矩阵和波束矢量；用户端将波束码本中获得最大信号干扰噪声比的波束矩阵序号通过反馈链路传送给基站；基站根据反馈信息将每个波束矢量分配给在那个波束矢量上取得最大信号干扰噪声比的用户；基站根据用户的反馈信息先计算每个波束矩阵可以达到的吞吐率，然后从中选择吞吐率最大的波束矩阵作为与发送信号矢量相乘的波束矩阵；选取此波束矩阵中的波束向量对应的用户作为选中的用户；基站通过对有效用户集合中的用户的数据进行幅度扩大的方式将总功率平均分配到用到的波束上。A beam codebook containing multiple beam matrices, which is generated by computer simulation or given by theoretical analysis, is stored in the base station and the user end respectively; the channel state information of the user receiver is not only used for data demodulation, but also used in the beam code book In this book, the beam matrix and beam vector that enable the user to obtain the maximum SINR are selected; the user end transmits the sequence number of the beam matrix that obtains the maximum SINR in the beam codebook to the base station through the feedback link; the base station sends each A beam vector is assigned to the user with the largest SINR on that beam vector; the base station first calculates the throughput rate that each beam matrix can achieve according to the feedback information of the user, and then selects the beam matrix with the highest throughput rate as the transmission signal The beam matrix of vector multiplication; select the user corresponding to the beam vector in this beam matrix as the selected user; the base station distributes the total power evenly to the used beams by expanding the amplitude of the user data in the effective user set .

以下通过分析本发明方法所依据的原理，并与现有技术相比较来说明本发明的优点。The advantages of the present invention will be described below by analyzing the principle on which the method of the present invention is based and comparing it with the prior art.

本发明适用的单小区系统中包括一个基站和K个用户(编号为U_1，...，U_K)，基站装配有M个发射天线，每个用户装配有一个接收天线，且K＞M，此系统的输入输出关系可以描述为：The single-cell system applicable to the present invention includes a base station and K users (numbered U_1, ..., U_K), the base station is equipped with M transmitting antennas, and each user is equipped with a receiving antenna, and K>M, where The input-output relationship of the system can be described as:

Y＝HSX+N (F1)Y＝HSX+N (F1)

其中，Y为K×1的用户接收信号矢量，H为K×M的信道矩阵，S为M×M的预处理矩阵，X为M×1的用户数据，N为K×1的用户接收的加性高斯白噪声矢量。Among them, Y is the K×1 user received signal vector, H is the K×M channel matrix, S is the M×M preprocessing matrix, X is the M×1 user data, and N is the K×1 user received signal vector. Additive white Gaussian noise vector.

为了以较小的反馈量获得较高的吞吐率，与现有的随机波束成型技术每次随机地在基站端生成预处理矩阵S或迫零波束成型在基站端对用户信道进行反转的预处理矩阵S不同，本发明中的预处理矩阵S是从存储在基站和用户端的含有L个独立的M×M波束矩阵的波束码本中以最大化吞吐率为目标选择出来的，其中每个波束矩阵S_l，l＝1，…，L都是酉阵，并且满足In order to obtain a higher throughput rate with a smaller amount of feedback, the existing random beamforming technology randomly generates a preprocessing matrix S at the base station each time or zero-forcing beamforming performs preprocessing on the user channel inversion at the base station. The processing matrix S is different. The preprocessing matrix S in the present invention is selected from the beam codebook containing L independent M×M beam matrices stored in the base station and the user end with the goal of maximizing the throughput rate, wherein each The beam matrix S _l , l=1,..., L are all unitary arrays, and satisfy

E{|s_lm|²}＝1，l＝1，…，L；m＝1，…，M，E{|s _lm | ² }=1, l=1,...,L; m=1,...,M,

其中s_l，m是波束矩阵S_l的第m列。这样在基站端不进行信道反转操作，避免了由于信道反转操作带来的在基站端需要完全信道知识反馈的缺点；同时通过用户在波束码本中选择波束矩阵并将选择的结果反馈回基站的方式，基站比随机波束成型方法获得更多的信道知识，从而提高了吞吐率。where sl _,m is the mth column of the beam matrix _Sl . In this way, the channel inversion operation is not performed at the base station, which avoids the disadvantage of requiring complete channel knowledge feedback at the base station due to the channel inversion operation; at the same time, the user selects the beam matrix in the beam codebook and feeds back the selected result to the By way of the base station, the base station obtains more channel knowledge than the random beamforming method, thus improving the throughput.

用户为了在波束码本中选择出最优的波束矩阵，需要获得信道状态信息，这是通过在用户端利用基站端发送的导频信号和本接收机在训练阶段的接收信号进行反向求解得到的。基站的操作分为两个阶段：训练阶段和传送阶段。在训练阶段，基站在每个天线上只发送导频信号，不发送数据；在传送阶段，基站在每个天线上只发送预处理模块的输出数据，不发送导频。导频和预处理模块的输出数据通过复用器进行时分复用。In order to select the optimal beam matrix in the beam codebook, the user needs to obtain channel state information, which is obtained by reversely solving the pilot signal sent by the base station at the user end and the received signal of the receiver during the training phase of. The operation of the base station is divided into two phases: the training phase and the transmission phase. In the training phase, the base station only sends the pilot signal on each antenna, and does not send data; in the transmission phase, the base station only sends the output data of the preprocessing module on each antenna, and does not send the pilot. The output data of the pilot and preprocessing modules are time-division multiplexed through the multiplexer.

在基站的训练阶段，基站在每个天线上发送相互正交的导频，每个用户通过信道估计模块估计出从基站的M个发射天线到本用户的接收天线的信道增益矢量h_i＝[h_i1…h_iM]^T，其中i是用户的序号。然后用户将估计出来的信道增益矢量h_i＝[h_i1…h_iM]^T输入到信号噪声干扰比SINR计算模块。在SINR计算模块用户根据h′_i＝[h_i1…h_iM]^T结合波束码本中的波束矩阵用下式计算其所有可能达到的L×M个SINR：In the training phase of the base station, the base station sends mutually orthogonal pilots on each antenna, and each user estimates the channel gain vector h _i =[ h _i1 ... h _iM ] ^T , where i is the serial number of the user. Then the user inputs the estimated channel gain vector h _i =[h _i1 ... h _iM ] ^T to the SINR calculation module. In the SINR calculation module, the user calculates all possible L×M SINRs according to h′ _i =[h _i1 ...h _iM ] ^T combined with the beam matrix in the beam codebook as follows:

${SINR SINR}_{i i,, l l,, m m} = = \frac{{| | {h h}_{i i} {s the s}_{l l,, m m} | |}^{22}}{11 / / SNR SNR + + \underset{j j &NotEqual; &NotEqual; i i}{Σ Σ} {| | {h h}_{i i} {s the s}_{l l,, j j} | |}^{22}},, m m = = 11,, . . . . . .,, M m;; l l = = 11,, . . . . . .,, L L - - - - - - ((F f 22))$

其中SNR是每个用户的接收信号噪声比；假设基站的总的发送功率为ρ，而且每个波束上的功率一样，那么这时候总的信号噪声比TSNR＝ρ/σ²而每个用户的SNR＝ρ/(Mσ²)，其中σ²为用户的接收信号中加性高斯白噪声的方差。Among them, SNR is the received signal-to-noise ratio of each user; assuming that the total transmission power of the base station is ρ, and the power on each beam is the same, then the total signal-to-noise ratio TSNR=ρ/σ ² and each user’s SNR=ρ/(Mσ ² ), where σ ² is the variance of additive white Gaussian noise in the user's received signal.

在计算出所有可能的SINR之后，用户i可以求出一个波束矩阵序号l_i ^*以及波束矩阵中的某个波束矢量的序号m_i ^*使得它们对应的

是最大的，即：After calculating all possible SINRs, user i can find a beam matrix sequence number l _i ^* and a certain beam vector sequence number m _i ^* in the beam matrix so that their corresponding

is the largest, that is:

${l l}_{i i}^{* *} = = \underset{11 \leq \leq l l \leq \leq L L}{arg arg} {{\underset{11 \leq \leq l l \leq \leq L L}{max max} {{max max {{\underset{11 \leq \leq m m \leq \leq M m}{{SINR SINR}_{i i,, l l,, m m}}}}}}}}$

(F3)(F3)

${m m}_{i i}^{* *} = = \underset{11 \leq \leq m m \leq \leq M m}{arg arg} {{\underset{11 \leq \leq l l \leq \leq L L}{max max} {{max max {{\underset{11 \leq \leq m m \leq \leq M m}{{SINR SINR}_{i i,, l l,, m m}}}}}}}}$

然后每个用户将

m_i ^*和本发明特有的波束矩阵序号l_i ^*组成反馈信息，通过反馈链路传送给基站，这样每个用户的反馈量为一个实数和两个整数，从而避免了现有的脏纸编码和迫零波束成型方法中需要在基站端具有全部信道状态信息的方法所带来的巨大的反馈开销。为了表示的方便，定义：Then each user will

m _i ^* and the unique beam matrix number l _i ^* of the present invention form the feedback information, which is transmitted to the base station through the feedback link, so that the feedback amount of each user is a real number and two integers, thus avoiding the existing dirty paper coding And the huge feedback overhead brought by the method that needs to have all the channel state information at the base station in the zero-forcing beamforming method. For the convenience of expression, define:

${γ γ}_{i i,, l l,, m m} = = \{\begin{matrix} {SINR SINR}_{i i,, {l l}^{* *},, {m m}^{* *}}^{* *},, & l l = = {l l}^{* *},, m m = = {m m}^{* *} \\ 00,, & else else \end{matrix},, i i = = 11,, . . . . . .,, K K - - - - - - ((F f 44))$

在本发明中，由于引入了波束码本，从而使得基站通过反馈链路得到的反馈信息是一个三维的矩阵 ${&aleph;}_{K \times L \times M} = {γ_{i, l, m}} .$ 本发明中的基站端在得到反馈信息 ${&aleph;}_{K \times L \times M} = {γ_{i, l, m}}$ 之后的操作遵循以下的步骤：In the present invention, since the beam codebook is introduced, the feedback information obtained by the base station through the feedback link is a three-dimensional matrix ${&aleph;}_{K \times L \times m} = {γ_{i, l, m}} .$ The base station in the present invention obtains the feedback information ${&aleph;}_{K \times L \times m} = {γ_{i, l, m}}$ Subsequent operations follow the steps below:

第一步，对波束码本中的每个波束矩阵和每个波束矢量，求出具有最大γ_i,l，m的那个用户，即求出 $i_{l, m}^{*} = \underset{1 \leq i \leq K}{\arg} {\max_{1 \leq i \leq K} {γ_{i, l, m}}}, l = 1, . . ., L; m = 1, . . ., M;$ In the first step, for each beam matrix and each beam vector in the beam codebook, find the user with the largest γ _i,l,m , that is, find $i_{l, m}^{*} = \underset{1 \leq i \leq K}{\arg} {\max_{1 \leq i \leq K} {γ_{i, l, m}}}, l = 1, . . ., L; m = 1, . . ., m;$

第二步，对波束码本中的每个波束矩阵，用下列公式计算其可达的吞吐率：In the second step, for each beam matrix in the beam codebook, use the following formula to calculate its achievable throughput rate:

${R R}_{l l} = = {Σ Σ}_{m m = = 11}^{M m} ((log log ((11 + + \underset{11 \leq \leq i i \leq \leq K K}{max max} {{{γ γ}_{i i,, l l,, m m}}})),, l l = = 11,, . . . . . .,, L L - - - - - - ((F f 55))$

第三步，求出具有最大可达吞吐率的波束矩阵

也就是说求出The third step is to find the beam matrix with the maximum achievable throughput

That is to say find out

${l l}^{* *} = = \underset{11 \leq \leq l l \leq \leq L L}{arg arg} \underset{11 \leq \leq l l \leq \leq L L}{max max} {{{R R}_{l l}}} - - - - - - ((F f 66))$

第四步，最终选择出来的用户集合可以表示为 $A = {k (m) | k (m) = i_{l^{*}, m}^{*}, m = 1, . . ., M},$ 选择出来的用户的M×1数据矢量X＝[x_k(1)，…，x_k(M)]^T。假设A中包含的M个用户中，满足条件 $γ_{k (m), l^{*}, m} &NotEqual; 0$ 的用户数为M′，显然M′≤M，那么对每一个m＝1，…，M，进行如下操作：如果 $γ_{k (m), l^{*}, m} = 0,$ 则数据矢量X的第m路数据乘以0，否则乘以

操作后得到调度器的数据输出M×1数据矢量X′。In the fourth step, the finally selected user set can be expressed as

A = {k (m) | k (m) = i_{l^{*}, m}^{*}, m = 1, . . ., m},

The selected user's M×1 data vector X=[x _k(1) , . . . , x _k(M) ] ^T . Assume that among the M users contained in A, the condition is satisfied

γ_{k (m), l^{*}, m} &NotEqual; 0

The number of users is M', obviously M'≤M, then for each m=1,...,M, the following operations are performed: if

γ_{k (m), l^{*}, m} = 0,

Then the m-th data of the data vector X is multiplied by 0, otherwise multiplied by

After the operation, the data output M×1 data vector X' of the scheduler is obtained.

第五步，将调度器的输出数据X′输入预处理模块；将l^*输入到波束码本模块，通过索引波束码本中存储的波束矩阵得到所需的波束矩阵

在预处理模块中将X′左乘波束矩阵得到预处理器模块的输出数据。然后通过发射天线发射出去。The fifth step is to input the output data X′ of the scheduler into the preprocessing module; input l ^* to the beam codebook module, and obtain the required beam matrix by indexing the beam matrix stored in the beam codebook

Left-multiply X′ by the beam matrix in the preprocessing module Get the output data of the preprocessor module. Then transmit it through the transmitting antenna.

定义有效用户集合

在本发明中，基站通过在调度器中对有效用户集合中的用户数据进行幅度扩大，对每个m′∈B波束矢量上分配功率ρ/M′。这样，对每个有效用户k∈A′，其实际的SINR可以表示为：Define the set of valid users

In the present invention, the base station allocates power ρ/M' to each m'∈B beam vector by expanding the range of user data in the effective user set in the scheduler. In this way, for each effective user k∈A′, its actual SINR can be expressed as:

${SINR SINR}_{k k,, {l l}^{* *},, {m m}^{' '}}^{' '} = = \frac{{| | {h h}_{k k} {s the s}_{{m m}^{' '}} | |}^{22}}{11 / / {SNR SNR}^{' '} + + \underset{j j &NotEqual; &NotEqual; m m,, j j &Element; &Element; B B}{Σ Σ} {| | {h h}_{k k} {s the s}_{j j} | |}^{22}},, k k &Element; &Element; {A A}^{' '} - - - - - - ((F f 77))$

用户k∈A′的接收信号(9)为：The received signal (9) of user k∈A′ is:

${y the y}_{k k} = = {h h}_{k k} {s the s}_{{l l}^{* *},, {m m}^{' '}} {x x}_{k k} + + \underset{j j &NotEqual; &NotEqual; {m m}^{' '},, j j &Element; &Element; B B}{Σ Σ} {h h}_{i i} {s the s}_{{l l}^{* *},, j j} {x x}_{j j} + + {n no}_{k k},, k k &Element; &Element; {A A}^{' '} - - - - - - ((F f 88))$

用户将信道增益矢量h_k＝[h_k1…h_kM]^T和

输入到数据解调模块对接收信号进行相干解调，即可以获得发送的数据信息。The user sets the channel gain vector h _k =[h _k1 …h _kM ] ^T and

Input to the data demodulation module to coherently demodulate the received signal to obtain the transmitted data information.

利用本发明可以达到的吞吐率为：Utilize the throughput that the present invention can achieve:

${R R}_{sum sum} = = \underset{k k &Element; &Element; A A,, {m m}^{' '} &Element; &Element; B B}{Σ Σ} log log ((11 + + {SINR SINR}_{k k,, {l l}^{* *},, {m m}^{' '}}^{' '})) - - - - - - ((F f 99))$

与现有的随机波束成型技术每次都随机的生成一个随机波束矩阵，然后每个用户反馈最大的SINR和对应的波束矢量的序号相比，由于本发明在基站端、用户端引入波束码本，在用户接收机处通过在波束码本中选择跟本用户信道最匹配的波束矩阵，也就是可以获得最大的SINR的波束矩阵，然后将此SINR、对应的波束矩阵的序号和波束矢量的序号反馈给基站。Compared with the existing random beamforming technology that randomly generates a random beam matrix each time, and then each user feeds back the maximum SINR and the corresponding beam vector sequence number, since the present invention introduces a beam codebook at the base station and user end , at the user receiver, select the beam matrix that best matches the user channel in the beam codebook, that is, the beam matrix that can obtain the largest SINR, and then use this SINR, the serial number of the corresponding beam matrix and the serial number of the beam vector feedback to the base station.

与现有的脏纸编码和迫零波束成型技术需要反馈完全的信道状态信息即M个复数不同，由于不需要在基站端进行信道反转，本发明只需反馈一个实数和两个整数，有效地节省了反馈量。与现有的随机波束成型相比，本发明由于在用户和基站进行了波束矩阵选择的操作，基站在接收机提供的波束矩阵反馈信息的辅助下，可以获得更多的用户信道状态信息，从而更有效地利用接收机具有的信道状态信息，在只增加一个整数的反馈量的代价下显著地提高了吞吐率。Unlike the existing dirty paper coding and zero-forcing beamforming technologies that need to feed back complete channel state information, that is, M complex numbers, since there is no need to perform channel inversion at the base station, the present invention only needs to feed back one real number and two integers, effectively greatly saves the amount of feedback. Compared with the existing random beamforming, since the beam matrix selection operation is performed between the user and the base station, the base station can obtain more user channel state information with the assistance of the beam matrix feedback information provided by the receiver, thereby The more effective use of the channel state information that the receiver has can significantly improve the throughput at the cost of only increasing the amount of feedback by an integer.

与脏纸编码需要在所有的用户排列中搜索最优的用户排列不同以及迫零波束成型需要在所有的用户组合中搜索最优的用户组合不同，由于本发明中的调度算法是在每个波束矩阵中对每个波束矢量在所有用户中进行最大值的搜索，不涉及排列和组合运算，而且波束码本中波束矩阵的个数和波束矢量的个数是固定的，所以其计算复杂度是随着用户数K的增长而线性增长的。Unlike dirty paper coding, which needs to search for the optimal user arrangement in all user arrangements, and zero-forcing beamforming, which needs to search for the optimal user combination in all user combinations, since the scheduling algorithm in the present invention is in each beam In the matrix, the search for the maximum value of each beam vector among all users does not involve permutation and combination operations, and the number of beam matrices and beam vectors in the beam codebook is fixed, so its computational complexity is It grows linearly with the increase of the number of users K.

本发明提出的接收机辅助的波束成型方法通过引入波束码本对每次信道实现所用的预处理模块的波束矩阵进行选择而获得了增益，更有效地利用了接收机的信道状态信息，并利用这种状态信息进行更合理的调度从而以增加很小反馈开销的代价提高了吞吐率。The receiver-assisted beamforming method proposed by the present invention obtains gain by introducing a beam codebook to select the beam matrix of the preprocessing module used for each channel realization, more effectively utilizes the channel state information of the receiver, and utilizes This state information enables more reasonable scheduling to improve throughput at the cost of little feedback overhead.

附图说明： Description of drawings:

图1是本发明接收机辅助的波束成型方法系统示意图。FIG. 1 is a schematic diagram of the system of the receiver-assisted beamforming method of the present invention.

图2是接收机辅助的波束成型方法与随机波束成型方法的吞吐率比较示意图。Fig. 2 is a schematic diagram illustrating the comparison of the throughput between the receiver-assisted beamforming method and the random beamforming method.

具体实施方式： Detailed ways:

以下结合附图说明本方法的实施例。Embodiments of the method are described below in conjunction with the accompanying drawings.

实施例1：Example 1:

本实施例以基站端具有M＝3个发射天线、波束码本内有L＝8个波束矩阵的多天线广播信道为例，说明接收机辅助的波束成型方法的实施。In this embodiment, a multi-antenna broadcast channel with M=3 transmit antennas at the base station and L=8 beam matrices in the beam codebook is taken as an example to illustrate the implementation of the receiver-assisted beamforming method.

本发明的实现框图如图1所示：Realization block diagram of the present invention as shown in Figure 1:

在基站端，K个独立的用户数据1输入到调度模块2，调度模块根据用户通过反馈链路15反馈回基站端的反馈信息14进行调度(其中反馈信息14包括每个用户的最大SINR、对应的波束码本的序号和波束矢量的序号)，求出最大化吞吐率的波束矩阵的序号送到存储波束码本3的存储器索引出需要的波束矩阵，同时从K路用户数据中选择有效的用户数据并进行幅度扩大得到预处理模块4的输入数据。从波束码本3输出的波束矩阵也输入导预处理模块4，在预处理模块4将该模块的输入数据左乘从波束码本3得到的波束矩阵，获得最终要发送的数据然后通过发送天线8发射出去。At the base station, K independent user data 1 are input to the scheduling module 2, and the scheduling module performs scheduling according to the feedback information 14 that the user feeds back to the base station through the feedback link 15 (wherein the feedback information 14 includes the maximum SINR of each user, the corresponding Beam codebook serial number and beam vector serial number), obtain the serial number of the beam matrix that maximizes the throughput rate and send it to the memory that stores the beam codebook 3 to index the required beam matrix, and select effective users from the K user data at the same time The data is expanded to obtain the input data of the preprocessing module 4. The beam matrix output from the beam codebook 3 is also input to the preprocessing module 4. In the preprocessing module 4, the input data of the module is multiplied by the beam matrix obtained from the beam codebook 3 to obtain the final data to be sent and then passed through the transmitting antenna. 8 launch out.

在用户接收端，用户从接收天线9上获得接收数据，接收数据输入到信道估计模块10和数据解调模块11分别进行信道估计和数据解调。在信道估计模块，利用基站端发送的导频信号和本接收机在训练阶段的接收信号进行反向求解可以得到信道状态信息，并输出到SINR计算模块12。SINR计算模块12的另外一个输入为波束码本3的所有波束矩阵，用户端的波束码本跟基站端的是相同的。在SINR计算模块12，用户根据从信道估计模块10输入的信道状态信息和波束码本3输入的波束矩阵求出反馈信息14。数据解调模块11根据接收信号，利用波束码本3中的波束矢量和信道估计模块10的信道状态信息进行相干解调最终恢复基站端的用户数据13。波束码本3中预先存储着8个由计算机仿真产生的波束矩阵，如下所示：At the user receiving end, the user obtains received data from the receiving antenna 9, and the received data is input to the channel estimation module 10 and the data demodulation module 11 for channel estimation and data demodulation respectively. In the channel estimation module, the channel state information can be obtained by using the pilot signal sent by the base station and the received signal of the receiver in the training phase to perform reverse calculation, and output to the SINR calculation module 12 . Another input of the SINR calculation module 12 is all beam matrices of the beam codebook 3, and the beam codebook of the user end is the same as that of the base station end. In the SINR calculation module 12 , the user calculates the feedback information 14 according to the channel state information input from the channel estimation module 10 and the beam matrix input from the beam codebook 3 . The data demodulation module 11 uses the beam vector in the beam codebook 3 and the channel state information of the channel estimation module 10 to perform coherent demodulation according to the received signal, and finally restores the user data 13 at the base station. Eight beam matrices generated by computer simulation are pre-stored in beam codebook 3, as follows:

${S S}_{11} = = [\begin{matrix} - - 0.6354 0.6354 + + 0.1489 0.1489 i i & 0.6109 0.6109 - - 0.3518 0.3518 i i & 0.1551 0.1551 - - 0.2304 0.2304 i i \\ - - 0.5178 0.5178 + + 0.0370 0.0370 i i & - - 0.3681 0.3681 + + 0.4141 0.4141 i i & - - 0.4566 0.4566 - - 0.4637 0.4637 i i \\ - - 0.5217 0.5217 - - 0.1 0.1 801801 i i & - - 0.3615 0.3615 - - 0.2556 0.2556 i i & - - 0.0768 0.0768 + + 0.7025 0.7025 i i \end{matrix}];;$

${S S}_{22} = = [\begin{matrix} - - 0.4581 0.4581 + + 0.7896 0.7896 i i & 0.0871 0.0871 + + 0.0675 0.0675 i i & - - 0.1352 0.1352 - - 0.3690 0.3690 i i \\ - - 0.3689 0.3689 + + 0.1672 0.1672 i i & 0.0442 0.0442 - - 0.2056 0.2056 i i & - - 0.2082 0.2082 + + 0.8651 0.8651 i i \\ 0.0060 0.0060 + + 0.0497 0.0497 i i & - - 0.5302 0.5302 + + 0.8139 0.8139 i i & - - 0.1495 0.1495 + + 0.1775 0.1775 i i \end{matrix}];;$

${S S}_{33} = = [\begin{matrix} - - 0.1386 0.1386 - - 0.3426 0.3426 i i & 0.3754 0.3754 + + 0.1864 0.1864 i i & - - 0.8135 0.8135 - - 0.1612 0.1612 i i \\ - - 0.0215 0.0215 + + 0.4983 0.4983 i i & 0.7844 0.7844 - - 0.1230 0.1230 i i & 0.1859 0.1859 - - 0.2937 0.2937 i i \\ - - 0.4606 0.4606 + + 0.6345 0.6345 i i & - - 0.4400 0.4400 - - 0.0136 0.0136 i i & - - 0.3404 0.3404 - - 0.2749 0.2749 i i \end{matrix}];;$

${S S}_{44} = = [\begin{matrix} - - 0.3996 0.3996 + + 0.6219 0.6219 i i & 0.5637 0.5637 - - 0.0228 0.0228 i i & 0.2849 0.2849 - - 0.2326 0.2326 i i \\ - - 0.1784 0.1784 + + 0.4119 0.4119 i i & - - 0.4282 0.4282 - - 0.3232 0.3232 i i & 0.2255 0.2255 + + 0.6781 0.6781 i i \\ 0.2698 0.2698 + + 0.4234 0.4234 i i & 0.1367 0.1367 + + 0.6126 0.6126 i i & - - 0.4800 0.4800 + + 0.3517 0.3517 i i \end{matrix}];;$

${S S}_{55} = = [\begin{matrix} - - 0.0081 0.0081 - - 0.1267 0.1267 i i & 0.2317 0.2317 + + 0.9412 0.9412 i i & - - 0.1810 0.1810 - - 0.1080 0.1080 i i \\ 0.0524 0.0524 + + 0.5883 0.5883 i i & 0.1007 0.1007 - - 0.0789 0.0789 i i & - - 0.2427 0.2427 - - 0.7589 0.7589 i i \\ - - 0.7952 0.7952 - - 0.0523 0.0523 i i & - - 0.1997 0.1997 - - 0.0655 0.0655 i i & - - 0.5626 0.5626 + + 0.0648 0.0648 i i \end{matrix}];;$

${S S}_{66} = = [\begin{matrix} - - 0.4201 0.4201 - - 0.2545 0.2545 i i & 0.2526 0.2526 + + 0.5874 0.5874 i i & - - 0.3823 0.3823 - - 0.4513 0.4513 i i \\ 0.1120 0.1120 - - 0.6523 0.6523 i i & - - 0.2445 0.2445 - - 0.1475 0.1475 i i & - - 0.5596 0.5596 + + 0.4089 0.4089 i i \\ - - 0.4024 0.4024 - - 0.3984 0.3984 i i & 0.2110 0.2110 - - 0.6820 0.6820 i i & 0.2143 0.2143 - - 0.3519 0.3519 i i \end{matrix}];;$

${S S}_{77} = = [\begin{matrix} - - 0.1363 0.1363 + + 0.2811 0.2811 i i & - - 0.3004 0.3004 + + 0.2544 0.2544 i i & 0.6667 0.6667 + + 0.5505 0.5505 i i \\ 0.2678 0.2678 + + 0.0007 0.0007 i i & - - 0.4843 0.4843 + + 0.7329 0.7329 i i & - - 0.3905 0.3905 - - 0.0639 0.0639 i i \\ - - 0.3126 0.3126 + + 00 . . 85618561 i i & 0.2476 0.2476 + + 0.1100 0.1100 i i & - - 0.3032 0.3032 - - 0.0631 0.0631 i i \end{matrix}];;$

${S S}_{88} = = [\begin{matrix} - - 0.0001 0.0001 + + 0.3122 0.3122 i i & 0.6704 0.6704 - - 0.5923 0.5923 i i & 0.2478 0.2478 + + 0.2021 0.2021 i i \\ - - 0.8298 0.8298 + + 0.0444 0.0444 i i & - - 0.1951 0.1951 - - 0.3457 0.3457 i i & - - 0.3877 0.3877 + + 0.0400 0.0400 i i \\ 0.4436 0.4436 - - 0.1234 0.1234 i i & 0.0303 0.0303 - - 0.2030 0.2030 i i & - - 0.7544 0.7544 + + 0.4205 0.4205 i i \end{matrix}]$

S₁，…，S₈都是酉阵，并且满足E{|s_l，m|²}＝1，l＝1，…，8；m＝1，2，3，其中s_l，m是波束矩阵S_l的第m列。S ₁ ,..., S ₈ are all unitary arrays, and satisfy E{|s _{l, m} | ² }=1, l=1,...,8; m=1, 2, 3, where s _{l, m} are beams The mth column of matrix S _l .

基站的操作分为两个阶段：训练阶段和传送阶段。在训练阶段，基站在每个天线上只发送导频信号，不发送数据；在传送阶段，基站在每个天线上只发送预处理模块的输出数据，不发送导频。导频和预处理模块的输出数据通过复用器进行时分复用。The operation of the base station is divided into two phases: the training phase and the transmission phase. In the training phase, the base station only sends the pilot signal on each antenna, and does not send data; in the transmission phase, the base station only sends the output data of the preprocessing module on each antenna, and does not send the pilot. The output data of the pilot and preprocessing modules are time-division multiplexed through the multiplexer.

基站的操作分为两个阶段：训练阶段和传送阶段。在训练阶段，基站在每个天线上只发送导频5，不发送预处理模块3的输出数据6，其中导频是用来在用户的接收机估计信道状态信息的；在传送阶段，基站在每个天线上只发送预处理模块的输出数据，不发送导频。导频和预处理模块的输出数据通过复用器进行时分复用。The operation of the base station is divided into two phases: the training phase and the transmission phase. In the training phase, the base station only sends the pilot 5 on each antenna, and does not send the output data 6 of the preprocessing module 3, where the pilot is used to estimate the channel state information in the user's receiver; in the transmission phase, the base station Only the output data of the preprocessing module is sent on each antenna, and the pilot frequency is not sent. The output data of the pilot and preprocessing modules are time-division multiplexed through the multiplexer.

在基站的训练阶段，基站在每个天线上发送相互正交的导频，每个用户通过信道估计模块估计出从基站的M个发射天线到本用户的接收天线的信道增益矢量h_i＝[h_i1 h_i2 h_i3]^T，其中i是用户的序号。然后用户将估计出来的信道增益矢量h_i＝[h_i1 h_i2 h_i3]^T输入到信号噪声干扰比SINR计算模块13。在SINR计算模块13，用户根据h_i＝[h_i1 h_i2 h_i3]^T结合波束码本3中的波束矩阵S₁，…，S₈用公式F2计算所有可能达到的SINR，其中用户端的波束码本3跟基站端的波束码本是相同的。In the training phase of the base station, the base station sends mutually orthogonal pilots on each antenna, and each user estimates the channel gain vector h _i =[ h _i1 h _i2 h _i3 ] ^T , where i is the serial number of the user. Then the user inputs the estimated channel gain vector h _i =[h _i1 h _i2 h _i3 ] ^T to the SINR calculation module 13 . In the SINR calculation module 13, the user calculates all possible SINRs with the formula F2 according to h _i =[h _i1 h _i2 h _i3 ] ^T combined with the beam matrix S ₁ ,..., S ₈ in the beam codebook 3, where the beam of the user end The codebook 3 is the same as the beam codebook at the base station.

在计算出所有可能的SINR之后，用户i根据公式F3求出一个波束矩阵序号l_i ^*以及波束矩阵中的某个波束矢量的序号m_i ^*使得它们对应的是最大的，然后每个用户将

m_i ^*和本发明特有的波束矩阵序号l_i ^*组成反馈信息14，最后将14通过反馈链路15传送给基站。基站通过反馈链路15得到每个用户的反馈信息，组成一个三维的矩阵

{&aleph;}_{K \times 8 \times 3} = {γ_{i, l, m}},

其中γ_i，l，m如公式F4定义。然后基站进行如下的操作：After calculating all possible SINRs, user i calculates a beam matrix sequence number l _i ^* and a certain beam vector sequence number m _i ^* in the beam matrix according to formula F3 so that their corresponding is the largest, then each user will

m _i ^* and the unique beam matrix number l _i ^* of the present invention form the feedback information 14 , and finally transmit 14 to the base station through the feedback link 15 . The base station obtains the feedback information of each user through the feedback link 15 to form a three-dimensional matrix

{&aleph;}_{K \times 8 \times 3} = {γ_{i, l, m}},

Wherein γ _{i, l, m} are defined as formula F4. The base station then performs the following operations:

第一步：对波束码本中的每个波束矩阵和每个波束矢量，求出具有最大γ_i，l，m的那个用户，也就是说求出 $i_{l, m}^{*} = \underset{1 \leq i \leq K}{\arg} {\max_{1 \leq i \leq K} {γ_{i, l, m}}}, l = 1, . . ., 8; m = 1,2,3;$ The first step: For each beam matrix and each beam vector in the beam codebook, find the user with the largest γ _{i, l, m} , that is to say find $i_{l, m}^{*} = \underset{1 \leq i \leq K}{\arg} {\max_{1 \leq i \leq K} {γ_{i, l, m}}}, l = 1, . . ., 8; m = 1,2,3;$

第二步：对波束码本中的每个波束矩阵，用公式F5计算其可达的吞吐率R_l，l＝1，…，8；Step 2: For each beam matrix in the beam codebook, use formula F5 to calculate its achievable throughput rate R _l , l=1,...,8;

第三步：根据公式F6求出具有最大可达吞吐率的波束矩阵

Step 3: Calculate the beam matrix with the maximum achievable throughput according to formula F6

第四步，最终选择出来的用户集合可以表示为 $A = {k (m) | k (m) = i_{l^{*}, m}^{*}, m = 1, . . ., 3},$ 选择出来的用户的3×1数据矢量X＝[x_k(1)，x_k(2)，x_k(3)]^T。满足条件 $γ_{k (m), l^{*}, m} &NotEqual; 0$ 的用户数为M′，显然M′≤3，那么对每一个m＝1，…，3，进行如下操作：如果 $γ_{k (m), l^{*}, m} = 0,$ 则数据矢量X的第m路数据乘以0，否则乘以

操作后得到调度器的数据输出3×1数据矢量X′。In the fourth step, the finally selected user set can be expressed as

A = {k (m) | k (m) = i_{l^{*}, m}^{*}, m = 1, . . ., 3},

The selected user's 3×1 data vector X=[x _k(1) , x _k(2) , x _k(3) ] ^T . To meet the conditions

γ_{k (m), l^{*}, m} &NotEqual; 0

The number of users is M', obviously M'≤3, then for each m=1,...,3, the following operations are performed: if

γ_{k (m), l^{*}, m} = 0,

After the operation, the data output 3×1 data vector X′ of the scheduler is obtained.

在预处理模块中将X′左乘波束矩阵

得到预处理器模块的输出数据。然后通过发射天线发射出去。The fifth step is to input the output data X′ of the scheduler into the preprocessing module; input l ^* to the beam codebook module, and obtain the required beam matrix by indexing the beam matrix stored in the beam codebook

Left-multiply X′ by the beam matrix in the preprocessing module

Get the output data of the preprocessor module. Then transmit it through the transmitting antenna.

有效用户集合为

用户k∈A′的接收信号9可以用公式F8表示。用户将信道增益矢量h_i＝[h_i1 h_i2 h_i3]^T和

输入到数据解调模块11对接收信号进行相干解调，即可以获得发送的数据信息13。The valid user set is

The received signal 9 of user k∈A' can be expressed by formula F8. The user sets the channel gain vector h _i =[h _i1 h _i2 h _i3 ] ^T and

Input to the data demodulation module 11 to coherently demodulate the received signal to obtain the transmitted data information 13 .

图2给出了本实施例在信道H为瑞丽不相关信道下，假设用户为独立同分布的系统中，接收机辅助的波束成型和随机波束成型的吞吐率比较图。图中横坐标为用户数，用户数取从4到128以4为步长递增，纵坐标为用奈特每秒每赫兹表示的系统吞吐率。曲线A给出了在总的信噪比TSNR为0dB时随机波束成型的吞吐率；曲线B给出了在总的信噪比TSNR为0dB时接收机辅助的波束成型的吞吐率；曲线C给出了在总的信噪比TSNR为5dB时随机波束成型的吞吐率；曲线D给出了在总的信噪比TSNR为5dB时接收机辅助的波束成型的吞吐率；曲线E给出了在总的信噪比TSNR为10dB时随机波束成型的吞吐率；曲线F给出了在总的信噪比TSNR为10dB时接收机辅助的波束成型的吞吐率。由图2可见，在用户数从4到128的情况下，在总的信噪比TSNR为0，5，10dB的情况下，接收机辅助的波束成型方法所达到的吞吐率比随机波束成型所达到的吞吐率要高。而且信噪比越高的时候，这种提高越大。在用户数为60，总的信噪比为10dB的时候，接收机辅助的波束成型方法所达到的吞吐率比随机波束成型所达到的吞吐率要高将近20％。FIG. 2 shows a comparison diagram of throughput between receiver-assisted beamforming and random beamforming in a system in which channel H is a Rayleigh uncorrelated channel and users are assumed to be independent and identically distributed in this embodiment. The abscissa in the figure is the number of users, and the number of users is from 4 to 128 in increments of 4, and the ordinate is the system throughput expressed in Knights per second per Hertz. Curve A shows the throughput of random beamforming when the total signal-to-noise ratio TSNR is 0dB; curve B shows the throughput rate of receiver-assisted beamforming when the total signal-to-noise ratio TSNR is 0dB; curve C gives The throughput rate of random beamforming is shown when the total signal-to-noise ratio TSNR is 5dB; curve D shows the throughput rate of receiver-assisted beamforming when the total signal-to-noise ratio TSNR is 5dB; The throughput rate of random beamforming when the total signal-to-noise ratio TSNR is 10dB; Curve F shows the throughput rate of receiver-assisted beamforming when the total signal-to-noise ratio TSNR is 10dB. It can be seen from Fig. 2 that when the number of users ranges from 4 to 128 and the total SNR TSNR is 0, 5, and 10 dB, the throughput achieved by the receiver-assisted beamforming method is higher than that achieved by random beamforming. The throughput achieved is high. And the higher the signal-to-noise ratio, the greater the improvement. When the number of users is 60 and the total SNR is 10dB, the throughput achieved by the receiver-assisted beamforming method is nearly 20% higher than that achieved by random beamforming.

本发明提出的接收机辅助波束成型方法在基站端不需要完全的信道状态信息，每个用户只需要反馈一个实数和两个整数，较随机波束成型，只增加了3比特的反馈信息(用来指示波束码本中所选波束矩阵的序号)有效地改善系统吞吐率；调度算法的计算复杂度随用户的增长线性增长；在比随机波束成型方法只增加了3比特反馈量的代价下，获得了更高的系统吞吐率。本发明提出的接收机辅助的波束成型方法通过引入波束码本对每次信道实现所用的预处理模块的波束矩阵进行选择而获得了增益，更有效地利用了接收机的信道状态信息，并利用这种状态信息进行更合理的调度从而提高了吞吐率。The receiver-assisted beamforming method proposed by the present invention does not require complete channel state information at the base station, and each user only needs to feed back a real number and two integers. Compared with random beamforming, only 3 bits of feedback information (used for Indicates the serial number of the selected beam matrix in the beam codebook) effectively improves the system throughput; the computational complexity of the scheduling algorithm increases linearly with the growth of users; at the cost of only increasing the amount of 3-bit feedback compared with the random beamforming method, the obtained higher system throughput. The receiver-assisted beamforming method proposed by the present invention obtains gain by introducing a beam codebook to select the beam matrix of the preprocessing module used for each channel realization, more effectively utilizes the channel state information of the receiver, and utilizes This state information enables more reasonable scheduling to improve throughput.

Claims

1. A receiver-assisted beamforming method, in which the transmitted signal vector is multiplied by the beam matrix at the base station, and the beam matrix is a unitary array; each user uses a feedback link to maximize the signal-to-interference-noise ratio and the corresponding beam The serial number of the matrix and the serial number of the beam vector are fed back to the base station; the base station assigns each beam vector to the user who obtains the maximum signal-to-interference-noise ratio on that beam vector according to the feedback information; Perform reverse solution with the received signal of the receiver in the training phase to obtain channel state information;

It is characterized by:

A beam codebook containing multiple beam matrices, which is generated by computer simulation or given by theoretical analysis, is stored in the base station and the user end respectively; the channel state information of the user receiver is not only used for data demodulation, but also used in the beam code book In this book, the beam matrix and beam vector that enable the user to obtain the maximum SINR are selected; the user end obtains the maximum SINR obtained in the beam codebook, the corresponding sequence number of the beam matrix and the sequence number of the beam vector through the feedback link Send to the base station; the base station first calculates the throughput rate that each beam matrix can achieve according to the feedback information of the user, and then selects the beam matrix with the highest throughput rate as the beam matrix multiplied by the transmitted signal vector, and selects the beam vector in this beam matrix The corresponding user is selected as the user; the base station evenly distributes the total power to the used beams by expanding the data of the users in the effective user set.