[ background of the invention ]
The Massive MIMO technology can obviously improve the performance and the capacity of the 5G wireless communication system. The number of the receiving and transmitting antennas of the Massive MIMO system is large, and the digital beam forming technology is used in the system, so that the system implementation cost and the energy consumption are high, and the obstruction is brought to the application of the Massive MIMO beam forming technology.
In order to overcome the problem, a Massive MIMO hybrid beamforming scheme (i.e., a partial connection architecture) is presented, and compared with a digital beamforming technology, using the hybrid beamforming technology can reduce system implementation cost and energy consumption, that is, in the partial connection architecture, each rf chain is only connected to a partial antenna, although the system sacrifices a gain of partial beamforming, the implementation complexity of hardware is greatly reduced, but in the partial connection architecture, performance loss is relatively serious based on some algorithms designed by an encoder.
As shown in fig. 1, it is a downlink communication transmission model of Massive MIMO system under a partial connection structure. In this system, the transmitting end is equipped with N
tA receiving end equipped with a single antenna, and a transmitting end for transmitting N
sThe stripe data flows to the receiving end. For realizing multi-data stream transmission, the transmitting end is equipped
A radio frequency chain, each radio frequency chain is connected with L
tThe receiving end of the root antenna is provided with a single radio frequency chain.
Assuming the array antenna model is ULA, the spacing of the antennas is half the wavelength. Defining a dimension vector f
iAnd a dimension vector f
iEach element in (1) satisfies
Then the whole analog precoding matrix at the transmitting end is
To simplify the following description, an index set Ω is defined
1From F
RFThe position composition of zero element is (i, j) ∈ Ω
1When the temperature of the water is higher than the set temperature,
where i and j represent the number of rows and columns, respectively, of the analog precoding matrix,
elements representing ith row and jth column in analog precoding matrix, digital precoding matrix F in transmitting end structure
BBHas the dimension of
The total power constraint of the transmitting end is normalized by F
BBRealization, satisfies | | F
RFF
BB||=N
s。
Considering a narrowband block flat fading channel, its received signal may be denoted as y ═ HF
RFF
BBs + N, y is dimension N
r× 1, H is the dimension N
r×N
tChannel matrix of, N
rThe number of antennas at the receiving end is shown, s is a transmitted symbol vector, and the mathematical period is satisfied
With a representation dimension of N
s×N
sN is a noise vector, follows an independent identically distributed gaussian distribution with a mean of 0 and a variance of σ
2。
In the above partial connection architecture, since the analog domain precoding matrix is constant modulus, the optimization problem is non-convex, so that solving the optimization problem at the transmitting end is very complicated, and a large amount of time is consumed to solve the globally optimal precoding matrix of the optimization problem, which results in increasing the energy loss of the existing partial connection architecture system and greatly improving the network delay.
[ summary of the invention ]
The invention aims to provide a large-scale multi-input multi-output hybrid beam forming algorithm based on partial connection, which improves the performance of the algorithm on the premise of controlling the cost and the energy consumption, so that the performance of the algorithm is close to that of a full digital beam forming system, and the energy consumption of a partial connection architecture system is reduced.
The invention adopts the following technical scheme: a large-scale multiple-input multiple-output hybrid beam forming algorithm based on partial connection comprises the following steps:
giving an optimal unconstrained precoder, an initial analog precoding matrix and a convergence tolerance of a part of connection architecture system, and calculating an initial digital precoding matrix according to the unconstrained precoder and the initial analog precoding matrix so as to calculate an initial error;
when the initial error is smaller than or equal to the convergence tolerance, taking the initial analog precoding matrix and the initial digital precoding matrix as the optimal analog precoding matrix and the optimal digital precoding matrix of the partial connection architecture system;
when the initial error is larger than the convergence tolerance, carrying out iterative calculation by taking the minimized error as a target until the error is smaller than or equal to the convergence tolerance or the maximum iteration times is finished, and taking the corresponding analog precoding matrix and the corresponding digital precoding matrix as an optimal analog precoding matrix and an optimal digital precoding matrix;
and generating and sending out mixed beams by the part of the connection architecture system according to the optimal analog pre-coding matrix and the optimal digital pre-coding matrix.
Further, the optimal unconstrained precoder is obtained by the following steps:
obtaining a channel matrix according to the channel state information of the partial connection architecture system
Wherein N is
cIndicates the number of clusters, N
pIndicating the number of paths in each cluster, α
ilThe gain factor of the ith transmission path in the ith reflector cluster is expressed by theta for the (i, l) th sub-path
r,ilAnd
respectively representing azimuth and elevation angles, theta, of the departure angle
t,ilAnd
respectively representing the azimuth and elevation angles of the angle of arrival,
and
respectively representing the azimuth angle theta
r,ilAnd
pitch angle theta
t,ilAnd
corresponding receive array responses and transmit array responses;
by H ═ U ∑ V
HPerforming singular value decomposition on the channel matrix H to obtain an identity matrix V, and extracting the first N from the identity matrix V
sColumn derived matrix V
1(ii) a By passing
Calculate the diagonal matrix and pass F
*=V
1Obtaining the optimal unconstrained precoder F
*(ii) a Where U is an identity matrix and Σ is a diagonal matrix.
Further, by
Deriving an initial digital precoding matrix
And pass through
Deriving an initial error
0。
Further, the specific process of iterative computation is as follows:
setting an objective function
Transformation of | | | F
*-F
RFF
BB||
FCan obtain the product
Wherein, T
nIs F
*N-th matrix block of (1), f
nDenotes the composition F
RFThe nth matrix block of (a) is,
is represented by F
BBThe nth row vector of (1);
the objective function is optimized to
Wherein, t
n,mTo represent
The (c) th column (c) of (c),
to represent
The m-th element of (a) is,
f for the k-th iteration
RFThe nth matrix block of (a) is,
is shown as
The increment of the phase of the mth element of (c),
after the k iteration F
BBThe nth row vector of (1);
to represent
An average value of the phase increment;
obtaining a simulation pre-coding matrix after optimizing the objective function
And further deriving a digital precoding matrix
And error of
k;
Will be wrong
kCompared with the convergence tolerance when
Then, the corresponding analog precoding matrix is used
And a digital precoding matrix
As an optimal analog pre-coding matrix and an optimal digital pre-coding matrix; when in use
Then, the steps are repeated until the error is less than or equal to the convergence tolerance or the maximum iteration times are finished, and the corresponding simulation precoding matrix is used
And a digital precoding matrix
As an optimal analog precoding matrix and an optimal digital precoding matrix.
The invention has the beneficial effects that: the invention is based on a Massive MIMO mixed beam forming system, aims at improving the frequency spectrum efficiency, adopts an alternate optimization algorithm based on matrix decomposition to design a mixed pre-coding matrix, carries out singular value decomposition on a channel matrix, designs an optimal unconstrained digital pre-coder and a synthesizer, designs a final pre-coder through alternate optimization, improves the performance of the algorithm on the premise of controlling the cost and energy consumption as much as possible, enables the performance of the algorithm to be close to that of a full-digital beam forming system, solves the problem of performance loss, and effectively verifies the performance of the algorithm.
[ detailed description ] embodiments
The present invention will be described in detail below with reference to the accompanying drawings and specific embodiments.
The invention discloses a large-scale multi-input multi-output hybrid beam forming algorithm based on partial connection, which is used for acquiring channel state information from a base station to all user terminals
Wherein H
kAnd the information fading from the base station to a user K is shown, wherein the K is the total number of users in the cell. An optimal digital precoding matrix and a synthesizer matrix of a base station end are designed based on singular value decomposition, and a digital precoding matrix and an analog precoding matrix are designed through alternate optimization. Information data transmission begins by first passing the transmitted signal through a digital precoder F
BBThen pass through
A radio frequency link, and then
An analog precoder, then L
tThe root antenna simultaneously feeds the signal to the radio channel. At the receiving end, an analog precoder receives the signal in the channel, via a radio frequency chain, and via a digital precoder W
BBA signal is received.
As shown in fig. 2, the method of the present invention specifically comprises:
and giving an optimal unconstrained precoder, an initial analog precoding matrix and a convergence tolerance of the partial connection architecture system, and calculating an initial digital precoding matrix according to the unconstrained precoder and the initial analog precoding matrix so as to calculate an initial error.
The optimal unconstrained precoder is obtained by the following steps:
obtaining a channel matrix according to the channel state information of the partial connection architecture system
Wherein N is
cIndicates the number of clusters, N
pIndicating the number of paths in each cluster, α
ilThe gain factor of the ith transmission path in the ith reflector cluster is expressed by theta for the (i, l) th sub-path
r,ilAnd
respectively representing azimuth and elevation angles, theta, of the departure angle
t,ilAnd
respectively representing the azimuth and elevation angles of the angle of arrival,
and
respectively representing the azimuth angle theta
r,ilAnd
pitch angle theta
t,ilAnd
corresponding receive array responses and transmit array responses;
by H ═ U ∑ V
HSingular value decomposition of a channel matrix HObtaining an identity matrix V, and extracting the first N from the identity matrix V
sColumn derived matrix V
1(ii) a By passing
Calculate the diagonal matrix and pass F
*=V
1Obtaining the optimal unconstrained precoder F
*(ii) a Where U is an identity matrix and Σ is a diagonal matrix.
By passing
Deriving an initial digital precoding matrix
And pass through
Deriving an initial error
0。
And when the initial error is smaller than or equal to the convergence tolerance, taking the initial analog precoding matrix and the initial digital precoding matrix as the optimal analog precoding matrix and the optimal digital precoding matrix of the partial connection architecture system.
And when the initial error is larger than the convergence tolerance, carrying out iterative calculation by taking the minimized error as a target until the error is smaller than or equal to the convergence tolerance or the maximum iteration times is finished, and taking the corresponding analog precoding matrix and the corresponding digital precoding matrix as the optimal analog precoding matrix and the optimal digital precoding matrix.
The specific process of iterative computation is as follows:
setting an objective function
And iteratively optimizing two parameters, namely an analog precoding matrix and a digital precoding matrix.
Transformation of | | | F
*-F
RFF
BB||
FCan obtain the product
Wherein,T
nIs F
*N-th matrix block of (1), f
nDenotes the composition F
RFThe nth matrix block of (a) is,
is represented by F
BBThe nth row vector of (1), the maximum number of iterations in the present invention can be set to K
u100, convergence tolerance is set to
Then the objective function is optimized since it should be satisfied in the system
So during the optimization process, the normalization constraints are temporarily removed
The simplified optimization problem is as follows:
defining the kth iteration of hybrid precoding as
Assuming that the initial is known
Then
Closed form solution of
On the contrary, when it is known
Can update
In the process of updating
When F is greater than F, the following derivation is performed
BBAt the time of giving, define
Is v
iThen can obtain
Considering F
RFThe derivation process is as follows:
wherein, TnIs F*Of dimension Lt×Ns。
The optimization problem described above can be translated into
Sub-questions, wherein the nth sub-question is
To solve the optimization problem, assume that
Search pair in a small range
Update and define
The phase of the m-th element is
Can be expressed as
Then it is determined that,
can be expressed as
Wherein the content of the first and second substances,
to represent
Increment of phase of m-th element when
Making Taylor expansion approximation very small
Wherein the content of the first and second substances,
is a vector, a symbol
Defined as the Hadamard product, the optimization problem can be reconstructed as
The above-mentioned optimization problem is a convex quadratic objective function, and has considered the over-constant modulus constraintThe above formula is based on approximation
Is reconstructed only when
Is only satisfied when it is very small, therefore
Must be added, the optimization problem then translates into
This is now a convex optimization problem, once obtained
Can obtain
Further define the
The m-th element of (a) is
Is the m-th column of (1) as t
n,mThe above equation can be derived as:
the optimization problem can be translated into a solution LtA sub-question, wherein the mth sub-question is
Obtaining a simulation pre-coding matrix after optimizing the objective function
And further deriving a digital precoding matrix
And error of
k;
Will be wrong
kCompared with the convergence tolerance when
Then, the corresponding analog precoding matrix is used
And a digital precoding matrix
As an optimal analog pre-coding matrix and an optimal digital pre-coding matrix; when in use
Then, the steps are repeated until the error is less than or equal to the convergence tolerance or the maximum iteration times are finished, and the corresponding simulation precoding matrix is used
And a digital precoding matrix
As an optimal analog precoding matrix and an optimal digital precoding matrix.
And generating and sending out mixed beams by the part of the connection architecture system according to the optimal analog pre-coding matrix and the optimal digital pre-coding matrix.
The first verification embodiment:
for the partial connection architecture system of the present invention, the corresponding spectral efficiency is
Wherein the content of the first and second substances,
is the signal-to-noise ratio, P represents the average transmission power between the base station and the mobile station, sigma
1The first dimension representing the diagonal matrix Σ is N
s×N
sIs defined as
When Gaussian symbols are transmitted in the channel, then the spectral efficiency is
Wherein the content of the first and second substances,
is N
s×N
sP is the average transmission power, N
sIs the number of data streams, R
nIs a noise covariance matrix, R
n=σ
2F
RFF
BBσ denotes the variance, H
kIs a channel matrix;
hybrid precoding design can be achieved by solving an optimal solution to the optimization problem of the transmitting end, which aims at maximizing spectral efficiency, i.e.
Wherein the content of the first and second substances,
is the element of the ith row and jth column in the analog precoding matrix,
is the element in the ith row and jth column of the digital precoding matrix.
The system adopts a geometric Saleh-Vallenzuela narrowband cluster channel model. The antenna arrays of the base station and the user terminal both adopt uniform linear arrays, and the array response vector of the array is
If a uniform linear array is used, the receive array response and the transmit array response in the channel matrix of the above formula are replaced with αULA(θ)Where N denotes the number of array elements in the linear array, λ denotes the carrier wavelength, d denotes the spacing between antennas, and θ denotes the off-angle from/arrival angle.
In the simulation part, not only a millimeter wave cluster channel model but also a Rayleigh channel model are considered. In the rayleigh fading channel model, each element in the normalized channel matrix H follows an independent equal distribution with a mean of 0 and a variance of σ2Complex gaussian distribution.
The numerical result shows that the performance of the hybrid precoding method can be close to that of a high-dimensional full-digital beamforming system and is far higher than that of an analog beamforming system. The analog precoding design of the invention is based on equal gain transmission, and the digital precoding matrix design is based on singular value decomposition. Finally, the design of the hybrid precoder is based on alternate optimization, so that the hybrid precoder is close to an optimal unconstrained matrix, and the system performance is close to the performance of a full digital beamforming algorithm.
Verification example two:
in the embodiment, the advantages of the spectrum efficiency and the performance of other algorithms of the matrix decomposition-based alternating optimization algorithm are verified through Matlab simulation.
In this embodiment, the number of antennas respectively equipped at the transmitting and receiving ends is 128 and 1, the number of radio frequency chains is 4, the number of data streams is 2, and a millimeter wave channel model of the ULA array model is used, and specific values of parameters in the model are as described in table 1.
Parameter assignment in the example of Table 1
As shown in fig. 3, which shows the results of a comparison of spectral efficiencies for various algorithms. The performance of the analog beamforming system is the worst, and the performance of the digital beamforming system is the best. The performance of the algorithm provided by the invention is obviously superior to the performance of analog beam forming, and is superior to the performance of the algorithm in the prior art, and is closest to the performance of a full digital beam forming system.
As shown in fig. 4, the effect of the number of data streams on the spectral efficiency of the algorithm is shown. Compared with the spectral efficiency of the proposed algorithm with the number of data streams of 1, 2, 4 and 8, as the number of data streams increases, the spectral efficiency of the system increases and the system performance becomes better.