WO2017126541A1

WO2017126541A1 - Calculation method, wireless station, and storage medium

Info

Publication number: WO2017126541A1
Application number: PCT/JP2017/001521
Authority: WO
Inventors: 伸一田島; ブンサーンピタックダンロンキジャー
Original assignee: 日本電気株式会社
Priority date: 2016-01-20
Filing date: 2017-01-18
Publication date: 2017-07-27
Also published as: JPWO2017126541A1

Abstract

The present invention provides a wireless station for reducing the number of iterations of an iteration method in eigenvalue decomposition and reducing the amount of computation. The wireless station includes a determination unit for determining an initial vector on the basis of channel information in a propagation path from a wireless station having a plurality of antennas to other wireless stations, and a calculation unit for performing the eigenvalue decomposition of a transmission covariance matrix using an iteration method on the basis of the initial vector.

Description

Calculation method, radio station, storage medium

This disclosure relates to at least one of a calculation method, a radio station, and a storage medium.

In MIMO (Multiple Input Multiple Output), multiple antennas are arranged on both the transmitter and the receiver, and spatial multiplexing transmission is performed, thereby improving the frequency utilization efficiency.

Furthermore, in recent years, with the progress of miniaturization technology for wireless circuits, the practical use of Massive MIMO in which a base station includes several tens to several hundreds of antenna elements is being studied. In Massive MIMO, the received power of the terminal is greatly improved due to the high array gain by a large number of antenna elements. In addition, since the degree of freedom of the antenna is high, a large number of data can be multiplexed and transmitted. For this reason, the channel capacity can be significantly improved as compared with general MIMO.

First, general background technology related to MIMO is shown. The issues in realizing this with Massive MIMO are described.

In MIMO, when a base station multiplies transmission data by a precoder, the receiving side can spatially separate data transmitted at the same frequency and the same time, and spatial multiplexing is realized.

In LTE (Long Term Evolution), a subband (FIG. 11) is defined as a radio resource area to which the same precoder is applied. A subband is composed of a plurality of RBs (Resource Blocks) on the same time and frequency axis. The RB is the minimum unit of radio resources on the time and frequency axes in LTE, and is composed of 7 OFDM (Orthogonal Frequency Division Multiplexing) symbols on the time axis and 12 subcarriers on the frequency axis.

As a precoder generation method for spatial multiplexing, there is a method using singular value decomposition (including eigenvalue decomposition) of channel information. In the following, an SVD (Single Value Decomposition, Singular Value Decomposition) precoder generation method in downlink MIMO communication when the number of base stations is one and the number of terminals is one is shown as an example.

As a premise, a TDD (Time Division Division Duplex) system using the same carrier frequency in the uplink and downlink is assumed. The base station estimates the channel matrix of the uplink channel based on the known signal transmitted by the terminal for uplink channel estimation. However, in TDD, channel reciprocity is established, and the channel matrix of the channel is used as downlink channel estimation information. Can be used as

The SVD precoder described in Non-Patent Document 1 is a channel matrix representing a channel response of downlink communication between a base station and a terminal.

Is calculated from the following equation (1).

Here, √λ _i is the i-th singular value of the channel matrix H. Σ is a diagonal matrix with singular values as diagonal components. diag {} is a diagonal matrix with elements in {} as diagonal elements. min () means the minimum value among the elements in parentheses.

Is the left singular vector.

Is the right singular vector. The shoulder H is a complex conjugate transpose.

The right singular vector V is used as a precoder that multiplies transmission data on the base station side, and the left singular vector U is used as a postcoder that multiplies reception data on the receiving terminal side. As a result, the terminal can diagonalize the channel matrix H as shown in the following equation (2) and separate a plurality of data.

Here, x is a transmission signal vector. y is a received signal vector. n is a noise vector.

On the other hand, there is a channel estimation error in the real environment, but the influence of the channel estimation error can be reduced by averaging the information about the channel matrix over a plurality of RBs on the same time and frequency axis called a subband. . Furthermore, by calculating the SVD precoder from the information averaged within the subband, it is possible to calculate a precoder having channel estimation error tolerance.

As channel information used for averaging, a transmission covariance matrix R _Tx that is power information of the channel matrix is used. The transmission covariance matrix R _Tx is calculated by the following equation (3).

When performing averaging on the frequency axis using a channel matrix that is amplitude and phase information, cancellation may occur due to phase changes between average samples, and averaging information may not be calculated correctly.

The SVD precoder can also be obtained from the eigenvalue decomposition of the transmission covariance matrix. By substituting equation (1) into the defining equation of the transmission covariance matrix, the following equation (4) is obtained. The last equation of the following equation (4) is in the form of eigenvalue decomposition of the transmission covariance matrix. It can also be seen that the eigenvector and eigenvalue of the transmission covariance matrix are the square of the right singular vector and singular value of the channel matrix, respectively.

Here, D is a diagonal matrix having the eigenvalues of the transmission covariance matrix as diagonal elements. Details of the relationship between eigenvalues and eigenvectors between the transmission covariance matrix and the channel matrix will be described later.

Actually, when reducing the effect of channel estimation error, the transmission covariance matrix of each RB in the subband is averaged over multiple RBs to obtain the average transmission covariance matrix and eigenvalue decomposition of the average transmission covariance matrix To obtain the SVD precoder. Average transmit covariance matrix

Is calculated by the following equation (5).

here

Is the transmission covariance matrix of the i _{RBth RB} in the subband. N _RB is the number of RBs in the subband.

In a Massive MIMO system in which a base station has several tens to several hundreds of antennas, the transmission covariance matrix to be subjected to eigenvalue decomposition becomes large, and the amount of computation of eigenvalue decomposition of the transmission covariance matrix increases.

Generally, there are a direct method and an iterative method for implementing eigenvalue decomposition in hardware.

The direct method is a method of obtaining eigenvalues and eigenvectors by a finite number of operations, and includes a method of applying a Gaussian elimination method to eigen equations and relational expressions between eigenvalues and eigenvectors.

On the other hand, the iterative method is a method of obtaining an approximate solution by converging initial values by iterative calculation. An example of the iterative method is the power method described in Non-Patent Document 2.

Generally, the direct method has higher calculation accuracy than the iterative method, but has a problem in that it requires a large amount of calculation.

The inventor considering this point has found that in a Massive MIMO system that requires eigenvalue decomposition of a large-scale transmission covariance matrix, an iterative method is used to reduce the amount of calculation of eigenvalue decomposition.

When performing eigenvalue decomposition of a transmission covariance matrix by an iterative method, a randomly generated vector is usually used as an initial vector.

However, if the difference between this random vector and the true eigenvector is large, the number of iterations until convergence increases, and the amount of computation of the precoder increases.

Therefore, one of the objects of the exemplary embodiment is to provide a new mechanism for reducing the number of iterations of the iterative method in eigenvalue decomposition and reducing the amount of calculation. It should be noted that this object is only one of a plurality of objects that the embodiments disclosed herein intend to achieve. Other objects or problems and novel features will become apparent from the description of the present specification or the accompanying drawings.

The calculation method of the exemplary embodiment includes determining an initial vector based on channel information in a propagation path from a radio station having a plurality of antennas to another radio station, and performing an iterative method based on the initial vector. And performing eigenvalue decomposition of the transmission covariance matrix.

The wireless station of the exemplary embodiment includes a determination unit that determines an initial vector based on channel information in a propagation path from a wireless station having a plurality of antennas to another wireless station, and an iterative method based on the initial vector. And a calculation unit that performs eigenvalue decomposition of the transmission covariance matrix.

The program of the exemplary embodiment determines an initial vector based on channel information in a propagation path from a radio station having a plurality of antennas to another radio station, and uses an iterative method based on the initial vector. And performing eigenvalue decomposition of the transmission covariance matrix.

According to the exemplary embodiment, the number of iterations of the iterative method in eigenvalue decomposition can be reduced, and the amount of computation can be reduced.

It is a figure which illustrates the structure of the eigenvalue calculating apparatus of 1st exemplary embodiment. It is a figure which illustrates operation | movement of the eigenvalue calculating apparatus of 1st exemplary embodiment. 3 is a diagram for explaining a configuration of an initial vector calculation unit 103. FIG. It is a figure for demonstrating the detail of operation | movement S102. FIG. 5 is a diagram for explaining a configuration of an eigenvalue decomposition execution unit 104. It is a figure for demonstrating the structure of operation | movement S103. It is a figure for demonstrating the structure of operation | movement S103. It is a figure which illustrates the structure of the eigenvalue calculating apparatus of 2nd exemplary embodiment. It is a figure which illustrates operation | movement of the eigenvalue calculating apparatus of 2nd exemplary embodiment. It is a figure which illustrates the structure of the eigenvalue calculating apparatus of 3rd exemplary embodiment. It is a figure which illustrates operation | movement of the eigenvalue calculating apparatus of 3rd exemplary embodiment. It is the figure which showed the concept of the subband. It is the figure which showed the concept of the radio | wireless resource utilization in 1st exemplary embodiment. It is the figure which showed the concept of the indirect calculation method of the eigenvector of the transmission covariance matrix in 1st RB in 1st illustrative embodiment. It is the figure which showed the concept of the radio | wireless resource utilization in 2nd exemplary embodiment. FIG. 10 illustrates a wireless station in a fourth exemplary embodiment.

An exemplary embodiment is shown below.

According to an exemplary embodiment, an average transmission covariance matrix (hereinafter, subband average transmission covariance) averaged over a plurality of RBs in a subband by a base station (also referred to as a radio station) having multiple antennas. (Referred to as a matrix). When performing this eigenvalue decomposition, an iterative method is performed with the first vector calculated based on the channel information as an initial vector. As an iterative method, for example, a power method is used. Details of the power method will be described later.

In order to simplify the explanation, it is assumed that the number of receiving antennas is two, but the present invention is not limited to this. For example, when the exemplary embodiment is applied to a base station in a Massive MIMO system, the number of transmission antennas can be several tens to several hundreds.

<First Exemplary Embodiment>
FIG. 12 shows an example of a method for using radio resources in the first embodiment. In FIG. 12, one RB in one subband is selected as the first RB. Based on the selected channel matrix of the first RB, the first vector and the second vector are calculated and set as initial vectors used for eigenvalue decomposition of the average transmission covariance matrix of each subband.

FIG. 1 shows a system configuration example of the first exemplary embodiment. The system shown in FIG. 1 includes a channel matrix storage unit 101, an average transmission covariance matrix calculation unit 102, an initial vector calculation unit 103, an eigenvalue decomposition execution unit 104, and an eigenvalue decomposition result storage unit 105.

The channel matrix storage unit 101 is connected to the average transmission covariance matrix calculation unit 102. Average transmission covariance matrix calculation section 102 is connected to eigenvalue decomposition execution section 104. The initial vector calculation unit 103 is connected to the eigenvalue decomposition execution unit 104. The eigenvalue decomposition result storage unit 105 is connected to the eigenvalue decomposition execution unit 104.

The channel matrix storage unit 101 is configured to store the channel matrix of each RB obtained by channel estimation.

The average transmission covariance matrix calculation unit 102 calculates the average transmission covariance matrix of each subband based on the channel matrix of each RB read from the channel matrix storage unit 101 and outputs the average transmission covariance matrix to the eigenvalue decomposition execution unit 104 It is configured.

The initial vector calculation unit 103 uses the first RB channel matrix read from the channel matrix storage unit 101 to calculate a first vector, which is an initial vector for use in the power method, and performs an eigenvalue decomposition execution unit 104. (S102).

The eigenvalue decomposition execution unit 104 is configured to input the average transmission covariance matrix in each subband from the average transmission covariance matrix calculation unit 102 and input the first vector from the initial vector calculation unit 103. The eigenvalue decomposition execution unit 104 is configured to perform a power method based on the average transmission covariance matrix and the first vector in each subband, and to perform eigenvalue decomposition on the average transmission covariance matrix in each subband. The The eigenvalue decomposition execution unit 104 is configured to output the calculation result of the eigenvalue decomposition to the eigenvalue decomposition result storage unit 105 (S103).

FIG. 2 shows an example of the operation of the first exemplary embodiment.

In S101, the average transmission covariance matrix of each subband is calculated based on the channel matrix of each RB obtained by channel estimation.

In S102, using the channel matrix of each RB obtained by channel estimation, a first vector that is an initial vector for use in the power method is calculated.

In S103, power multiplication is performed based on the average transmission covariance matrix in each subband and the first vector, and eigenvalue decomposition of the average transmission covariance matrix in each subband is performed.

Hereinafter, details of S101, S102, and S103 will be described.

<Calculation of average transmission covariance matrix>
The average transmission covariance matrix calculation unit 102 calculates an average transmission covariance matrix based on the channel matrix of each RB read from the channel matrix storage unit 101 (S101). average transmit covariance matrix of i _SB-th subband

Is calculated using the following equation (6)

Where N _RB is the number of RBs in the subband,

Is the channel matrix of the i _RB th RB in the i _SB th subband,

Is the transmission covariance matrix of the i _RB th RB in the i _SB th subband.

<Calculation of the first vector>
The initial vector calculation unit 103 uses the channel matrix for each RB read from the channel matrix storage unit 101 to calculate a first vector that is an initial vector for use in the power method (S102).

Here, an example of the initial vector calculation unit 103 is shown in FIG. FIG. 4 shows a detailed flow of the operation S102 of the initial vector calculation unit 103.

In FIG. 3, an initial vector calculation unit 103 includes an RB selection unit 103-1, a reception covariance matrix calculation unit 103-2, a reception covariance matrix eigenvalue calculation unit 103-3, and a reception covariance matrix eigenvector calculation unit 103. -4 and a transmission covariance matrix eigenvector calculation unit 103-5.

In FIG. 2, S102 includes operations S102-1 to S102-5.

The RB selection unit 103-1 selects a single RB as the first RB for calculating the first vector based on the channel matrix of each RB in the subband read from the channel matrix storage unit 101. The RB selection unit 103-1 outputs the channel matrix of the selected first RB to the reception covariance matrix calculation unit 103-2 (S102-1).

The reception covariance matrix calculation unit 103-2 calculates the reception covariance matrix of the first RB based on the channel matrix of the first RB input from the RB selection unit 103-1. Reception covariance matrix calculation section 103-2 outputs the calculated reception covariance matrix of the first RB to reception covariance matrix eigenvalue calculation section 103-3 (S102-2).

The reception covariance matrix eigenvalue calculation unit 103-3 calculates eigenvalues for the reception covariance matrix of the first RB based on the reception covariance matrix of the first RB input from the reception covariance matrix calculation unit 103-2. calculate. Receive covariance matrix eigenvalue calculation section 103-3 outputs the received covariance matrix of the eigenvalues of the calculation result and the first RB to receive covariance matrix eigenvector calculation section 103-4 (S102-3).

The reception covariance matrix eigenvector calculation unit 103-4 is based on the eigenvalues of the first RB reception covariance matrix and the first RB reception covariance matrix input from the reception covariance matrix eigenvalue calculation unit 103-3. The eigenvector of the reception covariance matrix of the first RB is calculated. Reception covariance matrix eigenvector calculation section 103-4 outputs the eigenvalue and eigenvector of the first RB reception covariance matrix to transmission covariance matrix eigenvector calculation section 103-5 (S102-4).

The transmission covariance matrix eigenvector calculation section 103-5 is configured to transmit the first RB transmission co-ordinate based on the eigenvalue and eigenvector of the first RB reception covariance matrix input from the reception covariance matrix eigenvector calculation section 103-4. Compute the eigenvectors of the variance matrix. Transmission covariance matrix eigenvector calculation section 103-5 outputs the calculated eigenvector of the first RB transmission covariance matrix to eigenvalue decomposition execution section 104 (S102-5).

Details of operations S102-1 to S102-5 are shown below.

<Selection of RB for initial vector calculation>
The RB selection unit 103-1 selects a single RB for calculating the first vector based on the channel matrix of each RB in the subband read from the channel matrix storage unit 101, and selects the selected RB. The channel matrix is output to reception covariance matrix calculation section 103-2 (S102-1).
The index of the first RB used for the initial vector calculation is calculated from the following equation (7).

Here, tr () is a matrix trace and represents the sum of the diagonal elements of the matrix in (). In equation (7), the RB with the highest channel power is selected from within the subband. The reason why the channel power can be calculated from the trace of the transmission covariance matrix is as follows.

The power of the i _RBth RB channel can be calculated from the sum of the squares of the singular values of the channel matrix. Further, from the relationship between the singular value decomposition of the channel matrix and the eigenvalue decomposition of the transmission covariance matrix, the sum of the squares of the singular values of the channel matrix is equal to the sum of the eigenvalues of the transmission covariance matrix. Furthermore, since the transmission covariance matrix is a Hermitian matrix, the sum of the eigenvalues of the transmission covariance matrix matches the trace of the transmission covariance matrix. Considering the above together, it can be seen that the power of each RB channel can be calculated from the sum of the traces of the transmission covariance matrix.

The eigenvector close to the average transmission covariance matrix in the subband can be obtained by selecting the RB whose power influence is dominant in the subband and obtaining the eigenvector of the transmission covariance matrix.

<Calculation of eigenvectors of reception covariance matrix>
Reception covariance matrix calculation section 103-2 calculates a reception covariance matrix based on the channel matrix input from RB selection section 103-1, and outputs it to reception covariance matrix eigenvalue calculation section 103-3 (S102-). 2).

The reception covariance matrix is calculated by the following equation (8).

here,

Is the channel matrix of the RB selected by the RB selection unit 103-1.

<Calculation of eigenvalues of reception covariance matrix>
The reception covariance matrix eigenvalue calculation unit 103-3 calculates the eigenvalue of the reception covariance matrix based on the reception covariance matrix input from the reception covariance matrix calculation unit 103-2, and calculates the eigenvalue calculation result and the reception covariance. The matrix is output to reception covariance matrix eigenvector calculation section 103-4. (S102-3)
The eigenvalue (9) of the reception covariance matrix is solved to obtain the eigenvalue.

However, det () is a determinant of the matrix in ().
Since the equation (8) is a quadratic equation related to the eigenvalue λ, it can be solved analytically.

<Calculation of eigenvectors of received covariance row matrix>
Reception covariance matrix eigenvector calculation section 103-4 calculates the eigenvector of the reception covariance matrix based on the reception covariance matrix and the eigenvalue of the reception covariance matrix input from reception covariance matrix eigenvalue calculation section 103-3. The eigenvalues and eigenvectors of the reception covariance matrix are output to the eigenvector calculation unit 103-5 of the transmission covariance matrix. (S102-4)
The eigenvector corresponding to the eigenvalue obtained by solving equation (9) can be calculated by solving the eigenvalue defining equation of the following equation (10).

Equation (10) are the N _R source simultaneous linear equations for each element of the eigenvector as a variable, analytically solution of the equation is obtained.

<Calculation of eigenvectors of transmission covariance matrix>
The transmission covariance matrix eigenvector calculation section 103-5 is configured to transmit the first RB transmission co-ordinate based on the eigenvalue and eigenvector of the first RB reception covariance matrix input from the reception covariance matrix eigenvector calculation section 103-4. The eigenvector of the variance matrix is calculated and output to the eigenvalue decomposition execution unit 104 (S102-5).

In the following, the principle of determining the eigenvector of the transmission covariance matrix indirectly from the eigenvalues and eigenvectors of the reception covariance matrix will be described.

<Relationship between singular values of channel matrix, singular vectors, eigenvalues of reception covariance matrix, and eigenvectors>
The relationship between the eigenvalues and eigenvectors of the reception covariance matrix and the singular values and singular vectors of the channel matrix will be described.
The receive covariance matrix is the singular value decomposition of the channel matrix

Can be expressed in the form of eigenvalue decomposition of the reception covariance matrix as in the following equation (11).

Since the last equation of Equation (11) is in the form of eigenvalue decomposition of the reception covariance matrix, the left singular vector U of the channel matrix matches the eigenvector of the reception covariance matrix, and the singular value of the channel matrix It can be seen that the square matches the eigenvalue of the reception covariance matrix.

As for the eigenvalue decomposition of the transmission covariance matrix and the singular value decomposition of the channel matrix, the right singular vector V of the channel matrix and the eigenvector of the transmission covariance matrix coincide with each other as shown in Equation (4). The square of the value matches the eigenvalue of the transmission covariance matrix.

<Method for indirectly obtaining eigenvector of transmission covariance matrix>
Transform the singular value decomposition equation of the channel matrix without directly performing the eigenvalue decomposition of the transmit covariance matrix, and indirectly calculate the right singular vector of the channel matrix from the singular value, left singular vector, and channel matrix of the channel matrix Indicates that
The definition formula of the singular value decomposition of the channel matrix H is modified as the following formula (12) for the right singular vector.

Note that the left singular vector U is a unitary matrix, and utilizes the fact that the product of its own complex conjugate transpose becomes a unit matrix. Note that Σ ⁻¹ is a diagonal matrix whose diagonal element is the reciprocal of the singular value of the channel matrix.

FIG. 13 shows the concept of a method for indirectly calculating the eigenvector of the transmission covariance matrix. As described above, the left singular vector and singular value of the channel matrix can be calculated from the eigenvector and eigenvalue of the reception covariance matrix, respectively. The right singular vector of the channel matrix matches the eigenvector of the transmission covariance matrix. Therefore, the eigenvector of the transmission covariance matrix can be calculated from the eigenvector and eigenvalue of the reception covariance matrix and the channel matrix from Equation (12).

The above is the principle by which the eigenvector of the transmission covariance matrix is obtained from the eigenvalue and eigenvector of the reception covariance matrix.

Using the above principle, the eigenvector of the transmission covariance matrix of the first RB is calculated from the following equation (13).

The calculated v ₁ and v ₂ are output to the eigenvalue decomposition execution unit 104 as the first vector x ₁ and the second vector x ₂ , respectively.

When the number of transmission antennas is larger than the number of reception antennas, the eigenvalue decomposition of the transmission covariance matrix is not performed directly, but indirectly through the reception covariance matrix as in S102-1 to S102-5. In particular, the amount of calculation can be reduced by calculating the eigenvector of the transmission covariance matrix.

Specifically, when eigenvalue decomposition of the transmission covariance matrix is performed directly, for example, the calculation amount of one iteration of the power method requires a calculation amount of the square size of the matrix size for both the multiplier and the adder. On the other hand, as described above, when the eigenvalue decomposition of the transmission covariance matrix is indirectly performed, the eigenvalues and eigenvectors of the reception covariance matrix may be algebraic calculations, and the equation for obtaining the eigenvalues of the transmission covariance matrix (13) But

The product of the matrix between each other is once,

Since the matrix product is one time, the amount of computation is the first order of the size of the transmission covariance matrix.
Therefore, when the number of receiving antennas is significantly smaller than that of transmitting antennas, eigenvalue decomposition is performed on the transmission covariance matrix having a large size through eigenvalue decomposition of the reception covariance matrix having a small size as described above. The amount of calculation for decomposition can be reduced.

<Eigenvalue decomposition of average transmission covariance matrix>
The eigenvalue decomposition execution unit 104 applies the average transmission covariance matrix in each subband input from the average transmission covariance matrix calculation unit 102, and the first vector and the second vector input from the initial vector calculation unit 103. Based on this, a power method is performed, and eigenvalue decomposition of the average transmission covariance matrix in each subband is performed (S103).

FIG. 5 shows a detailed configuration of the eigenvalue decomposition execution unit 104. 6A and 6B show a detailed flow of the operation S103 of the eigenvalue decomposition execution unit 104. FIG. The eigenvalue decomposition execution unit 104 includes a first eigenvalue calculation unit 104-1 and a second eigenvalue calculation unit 104-2. The operation S103 includes operations S103-1 to S103-13.

The first eigenvalue calculation unit 104-1 is a power method based on the average transmission covariance matrix in each subband input from the average transmission covariance matrix calculation unit and the first vector input from the initial vector calculation unit 103. To calculate the first eigenvalue and first eigenvector of the average transmission covariance matrix of each RB, and output them to the eigenvalue decomposition result storage unit 105 and the second eigenvalue calculation unit 104-2 (S103-1 to S103-7) .

The second eigenvalue calculator 104-2 receives the average transmission covariance matrix of each subband input from the average transmission covariance matrix calculator 102, the first eigenvalue input from the first eigenvalue calculator 104-1 and the first eigenvalue Based on one eigenvector, a second eigenvalue and a second eigenvector are calculated and output to the eigenvalue decomposition result storage unit 105 (S103-8 to S103-15).

First, the principle of the power method will be described. Here, it is assumed that eigenvalue decomposition of the symmetric matrix A is performed. In the power method, the initial vector is converged to the eigenvector by repeatedly multiplying the initial vector by the eigenvalue decomposition target vector.
An arbitrary vector can be expressed by a linear combination of all eigenvectors. Therefore, the first vector x ⁽⁰⁾ is expressed by the following equation (i) using the i-th eigenvector v _i of the average transmission covariance matrix of each subband and an arbitrary constant c _i (i = 1 to N _EV ). 14).

Here, N _EV is the number of eigenvalues.
The initial vector, that is, the estimated eigenvector v ⁽⁰⁾ before the iteration is set as the first vector x ⁽⁰⁾ . A vector obtained by multiplying the initial vector v ⁽⁰⁾ by the target matrix to be subjected to eigenvalue decomposition i times is expressed by the following equation (15) using an eigenvector definition equation Av _i = λv _i .

here,
Since λ ₁ > λ _n (n = 2 to N _EV ), λ _{n /} λ ₁ in the second term in {} of Equation (15) is smaller than 1. For this reason, if i is sufficiently large, v ⁽ⁱ⁾ is expressed by the following equation (16).

In equation (16)

This causes an overflow, but overflow can be prevented by normalizing v ⁽ⁱ⁾ at each step in the power method. In this way, since a vector that is a scalar multiple of the eigenvector is obtained, the final eigenvector can be obtained by normalizing the estimated eigenvector after sufficiently repeating the iteration as shown in the following equation (17).

As described above, the power method is a method for obtaining a maximum eigenvalue of a target matrix and an eigenvector corresponding to the maximum eigenvalue. However, the second and subsequent eigenvalues can be obtained by performing the following processing.

In order to obtain the second eigenvalue of matrix A, a matrix A ′ having the second eigenvalue of matrix A as the first eigenvalue is calculated using spectral decomposition of Hermitian matrix, and a power method is applied to matrix A ′ To do. The matrix A is represented by the following equation (18) for spectral decomposition for each eigenvector.

Therefore, the matrix A ′ can be obtained by subtracting the component of the eigenvector corresponding to the first eigenvalue from the matrix A, and the matrix A ′ is calculated by the following equation (19).

Hereinafter, details of the operation S103 of the eigenvalue decomposition execution unit 104 shown in FIGS. 6A and 6B will be described.

<Calculation of first eigenvalue and first eigenvector of average transmission covariance matrix of each subband>
The first eigenvalue calculation unit 104-1 is a power based on the average transmission covariance matrix in each subband input from the average transmission covariance matrix calculation unit 102 and the first vector input from the initial vector calculation unit 103. The first eigenvalue and the first eigenvector of the average transmission covariance matrix of each subband are calculated and output to the eigenvalue decomposition result storage unit 105 and the second eigenvalue calculation unit 104-2 (S103-1 to S103-). 7).

Before entering the iterative process, initial vector setting (S103-1) and index initialization are performed (S103-2).
The initial vector, the first vector x ₁ is used.

In the iterative process, as shown in the following equation (20),
Multiplying the _{R Tx} is calculated subject to the eigenvalue decomposition on ^v calculated in the previous iteration ^{(k-1) (S103-3)} .

Vector

In order to prevent the overflow described above, normalization is performed by the following equation (21) (S103-4).

Finally, an estimated value λ ^(k) of the eigenvalue in the k-th iterative process is calculated by the following equation (22) (S103-5).

Here, (,) is the inner product of the vectors in ().

S103-3 to S103-5 are one iteration process in the power method.
If the convergence condition of the following equation (23) is not satisfied after the iterative processing, the processing of S103-3 to S103-5 is executed again (S103-7).

Here, ε is an eigenvalue tolerance.
In the power method, the maximum error for the true eigenvalue of λ ^(k) for any k is λ ^(k) −λ ^(k−1) . Therefore, ε may be determined according to the accuracy of the eigenvalue required by the system.

Further, in order to perform the convergence determination described above, it is necessary to calculate the difference between the estimated eigenvalues between the two iterations. Therefore, in the first iteration, the process proceeds unconditionally to the second iteration. . (S103-6)
<Calculation of Second Eigenvalue and Second Eigenvector of Average Transmission Covariance Matrix of Each Subband>
The second eigenvalue calculator 104-2 receives the average transmission covariance matrix of each subband input from the average transmission covariance matrix calculator 102, the first eigenvalue input from the first eigenvalue calculator 104-1 and the first eigenvalue Based on one eigenvector, a second eigenvalue and a second eigenvector are calculated and output to the eigenvalue decomposition result storage unit 105 (S103-7 to S103-13).

A second eigenvalue calculation matrix R ′ _Tx is calculated from the following equation (24) using spectral decomposition of the Hermitian matrix. (S103-8)

As the initial vector, the second vector input from the initial vector calculation unit 103 is used (S103-9).

Hereinafter, in S103-11 to S103-15, the same operation as that of S103-2 to S103-7 is performed on the second eigenvalue calculation matrix R ′ _Tx , and the first eigenvalue of the second eigenvalue calculation matrix is calculated. Find the first eigenvector. Since the first eigenvalue and the first eigenvector of the second eigenvalue calculation matrix are equal to the second eigenvalue and the second eigenvector of the average transmission covariance matrix of each RB, these are output to the eigenvalue decomposition result storage unit 105 ( S103-11 to S103-15).

According to the first embodiment, the first RB is selected from the RBs in the subband, and an eigenvalue decomposition is performed on the eigenvector of the transmission covariance matrix of the average transmission covariance matrix of all RBs in the subband. Value. Since the first RB selects the power dominant in the subband, the first vector that is the eigenvector of the transmission covariance matrix of the first RB is smaller than the randomly generated vector. It is close to the eigenvector of the average transmission covariance matrix of all RBs in the band. Therefore, by using the first vector as the initial vector of the power method, the number of iterations of the power method, and hence the amount of calculation of eigenvalue decomposition can be reduced.

Further, in the calculation of the initial vector, the eigenvector of the transmission covariance matrix of the first RB is obtained by indirectly performing the eigenvalue decomposition of the transmission covariance matrix using the fact that the receiving antenna is significantly smaller than the transmitting antenna. It is possible to calculate.

<Second Exemplary Embodiment>
FIG. 14 shows a conceptual diagram of a second exemplary embodiment.
The calculation method of the first vector of the second embodiment is different from the calculation method of the first embodiment. In the first embodiment, the eigenvector of the 1 RB transmission covariance matrix in the subband is the first vector. On the other hand, in the second embodiment, a first eigenvector and a second eigenvector of a transmission covariance matrix (also referred to as an average transmission covariance matrix between subbands) averaged over a plurality of subbands are set as the first vector, Let it be the second vector.

After the first vector is determined, eigenvalue decomposition is performed by the power method of the average transmission covariance matrix of each subband using the determined first vector as an initial vector, and the result is sent to the eigenvalue decomposition result storage unit 105. Output.

FIG. 7 illustrates the system configuration of the second exemplary embodiment.
The channel matrix storage unit 101 is configured to store a channel matrix for each RB obtained by channel estimation.

The average transmission covariance matrix calculation unit 102A is configured to calculate an average transmission covariance matrix within each subband and between subbands based on the channel matrix for each RB read from the channel matrix calculation unit. The average transmission covariance matrix calculation unit 102A is configured to output an average transmission covariance matrix within each subband and between subbands to the initial vector calculation unit 103A and the eigenvalue decomposition execution unit 104.

The initial vector calculation unit 103A is configured to calculate the first vector based on the average transmission covariance matrix between subbands read from the average transmission covariance matrix calculation unit 102A. The initial vector calculation unit 103A is configured to output the first vector to the eigenvalue decomposition execution unit 104.

The eigenvalue decomposition execution unit 104 performs power multiplication based on the first vector input from the initial vector calculation unit 103A and the average transmission covariance matrix in each subband input from the average transmission covariance matrix calculation unit 102A. Configured to calculate an eigenvector of the average transmission covariance matrix of each subband.

FIG. 2 shows an example of the operation of the second exemplary embodiment.

<Calculation of average transmission covariance matrix>
The average transmission covariance matrix calculation unit 102A calculates an average transmission covariance matrix based on the channel matrix for each RB read from the channel matrix calculation unit, and outputs the average transmission covariance matrix to the initial vector calculation unit 103A and the eigenvalue decomposition execution unit 104 (S101A). ). This operation S101A differs from the operation of S101 in that an average transmission covariance matrix between subbands is obtained. The average transmission covariance matrix within each subband and the average transmission covariance matrix between subbands for generating an initial vector are calculated by the following equations (25) and (26), respectively.

<Calculation of the first vector>
The initial vector calculation unit 103A includes an average transmission covariance matrix between subbands read from the average transmission covariance matrix calculation unit 102A.

Based on, a first vector, which is an initial vector used in the power method, is calculated (S102A).

In S102 in the first embodiment, the eigenvector of the transmission covariance matrix of 1 RB in the subband is the first vector, whereas in S102A of the present embodiment, the first transmission covariance matrix between subbands is calculated. The first eigenvector and the second eigenvector are defined as a first vector and a second vector, respectively.

The number N _{SB of} subbands to be averaged when calculating the average transmission covariance matrix between subbands may be determined by numerical search, and _NSB having a large effect of reducing the amount of computation of eigenvalue decomposition calculated by simulation in advance. .

<Eigenvector calculation>
The eigenvalue decomposition execution unit 104 is based on the average transmission covariance matrix of the subband input from the average transmission covariance matrix calculation unit 102A, and the first vector and the second vector input from the initial vector calculation unit 103A. The eigenvalues and eigenvectors of the average transmission covariance matrix of the subband are calculated and stored in the eigenvalue decomposition result storage unit 105 (S103).

Since this operation S103 is the same as that of the first embodiment, a description thereof will be omitted.
In the present embodiment, by performing eigenvalue decomposition of the average transmission covariance matrix between subbands, a vector close to the eigenvector of the average transmission covariance matrix of subbands can be calculated. By using this as the power vector initial vector, the number of iterations for the average transmission covariance matrix of each subband can be reduced, and the number of computations can be reduced. The calculation of the initial vector requires a certain amount of calculation because eigenvalue decomposition is performed by the power method. However, this initial vector setting improves the number of iterations of the power method of the average transmission covariance matrix in each subband, so that the initial value vector is randomized as the amount of computation of the total eigenvalue decomposition including the initial vector calculation. Compared with the case, the amount of calculation is reduced.

A specific calculation amount reduction effect will be described below using mathematical expressions.
First, the average transmit covariance matrix, and calculated C _rand the average value of the operation quantity when performing the power method the random vector as an initial value.

Next, assuming a certain N _SB , the initial vector generation of this embodiment compared to the case where the initial vector is a random vector during eigenvalue decomposition by the power method of the average transmission covariance matrix in each subband is performed. _Let C _{Imp be} the average value of the amount of improvement in the amount of computation due to use.

Computation of eigenvalue decomposition of N _SB subbands if you did not use the method of determining the initial vector of the present embodiment, the following equation (27).

The first item is the amount of calculation for calculating the initial vector, and the second item is the amount of calculation for eigenvalue decomposition of the average transmission covariance matrix in each subband.
On the other hand, the amount of calculation when the initial vector determination method of the present embodiment is used is expressed by the following equation (28).

Therefore, the calculation amount reduction amount of the initial vector determination method in this embodiment is expressed by the following equation (29).

By determining an appropriate _NSB , the first term of the equation (29) becomes larger than the second term, that is, the amount of calculation by the eigenvalue decomposition in each subband is reduced as the amount of calculation increases by setting the initial vector. And the amount of calculation can be reduced.

In comparison with the first embodiment, the initial vector can be calculated with a low amount of computation in the first embodiment, whereas in this embodiment, the calculation amount of eigenvalue decomposition for initial value calculation is different for each subband. A calculation amount that cannot be ignored for the calculation amount of eigenvalue decomposition is necessary. On the other hand, there is an advantage that the influence of the channel estimation error can be reduced by calculating the initial vector from a wide band.

<Third exemplary embodiment>
FIG. 9 shows the system configuration of the third exemplary embodiment.
In the third embodiment, an initial vector of eigenvalue decomposition using the power method of the average transmission covariance matrix of each subband based on feedback information from the terminal is used.

A system shown in FIG. 9 includes a channel matrix storage unit 101, an average transmission covariance matrix calculation unit 102, an initial vector calculation unit 103B, an eigenvalue decomposition execution unit 104, an eigenvalue decomposition result storage unit 105, and terminal feedback information. And a storage unit 106.

The channel matrix storage unit 101 is connected to the average transmission covariance matrix calculation unit 102. Average transmission covariance matrix calculation section 102 is connected to eigenvalue decomposition execution section 104. The terminal feedback information storage unit 106 is connected to the initial vector calculation unit 103B. The initial vector calculation unit 103B is connected to the eigenvalue decomposition execution unit 104. The eigenvalue decomposition execution unit 104 is connected to the eigenvalue decomposition result storage unit 105.

The channel matrix storage unit 101 is configured to store a channel matrix for each RB obtained by channel estimation.

The terminal feedback information storage unit 106 is configured to store information fed back from the terminal.

The initial vector calculation unit 103B is configured to calculate the first vector and the second vector based on the terminal feedback information read from the terminal feedback information storage unit 106 and output the first vector and the second vector to the eigenvalue decomposition execution unit 104. Yes.

The average transmission covariance matrix calculation unit 102 is configured to calculate an average transmission covariance matrix for each subband based on the channel matrix for each RB read from the channel matrix storage unit 101. Average transmission covariance matrix calculation section 102 outputs the average transmission covariance matrix of each subband to eigenvalue decomposition execution section 104.

The eigenvalue decomposition execution unit 104 converts the first vector and the second vector input from the initial vector calculation unit 103B, and the average transmission covariance matrix of each subband input from the average transmission covariance matrix calculation unit 102. Based on this, the eigenvalues and eigenvectors of the average transmission covariance matrix of each subband are calculated.

<Calculation of average transmission covariance matrix>
The average transmission covariance matrix calculation unit 102 calculates the average transmission covariance matrix of each RB based on the channel matrix for each RB read from the channel matrix storage unit (S101). The calculation method is the same as in the first embodiment.

<Calculation of the first vector>
The initial vector calculation unit 103B reads the terminal feedback information read from the terminal feedback information storage unit 106, and calculates a first vector that is an initial vector used in the power method (S102B).

In LTE, PMI (Precoding Matrix Indicator) is defined as channel information feedback from a terminal. The base station and the terminal maintain a quantized precoding table called a codebook. A suitable precoder determined by the terminal is fed back to the base station as a codebook index.

In this embodiment, the initial vector calculation unit 103B outputs the precoder specified by PMI to the eigenvalue decomposition execution unit 104 as the first vector and the second vector.

Also, in Massive MIMO, for the purpose of increasing array gain, it is possible to increase the number of physical antennas freely without being restricted by the number of antennas determined by the standard. In this case, the number of base station antennas, which is the premise of the codebook precoder, may differ from the actual number of base station antennas. For this reason, a modified precoder obtained by modifying the precoder with respect to the number of transmission antennas may be held. As a setting method of the modified precoder, for example, a steering weight having the same peak direction as the beam pattern of the codec precoder may be used.

If the peak direction of the i-th precoder in the codebook is θ _i , the i-th precoder of the modified code

The n th element of

Is represented by the following equation (30).

<Execution of eigenvalue decomposition>
The eigenvalue decomposition execution unit 104 determines each subband based on the average transmission covariance matrix of each subband input from the average transmission covariance matrix calculation unit 102 and the first vector input from the initial vector calculation unit 103B. Compute the eigenvalues and eigenvectors of the mean transmission covariance matrix of.

Since this operation is the same as that of the first embodiment, a description thereof will be omitted.
[Effect of the embodiment]
In this embodiment, unlike the first embodiment and the second embodiment, the initial vector can be determined simply by referring to the table based on feedback from the terminal. For this reason, there is an advantage that the calculation of the initial vector becomes unnecessary.

<Fourth Exemplary Embodiment>
FIG. 15 shows a wireless station in a fourth exemplary embodiment. The radio station 1000 includes a determination unit 1001 and a calculation unit 1002.

The determining unit 1001 determines an initial vector based on channel information in a propagation path from a radio station having a plurality of antennas to another radio station.

The calculation unit 1002 performs eigenvalue decomposition of the transmission covariance matrix using the iterative method based on the initial vector.

According to this embodiment, the number of iterations of the iterative method in eigenvalue decomposition can be reduced, and the amount of computation can be reduced.

<Other exemplary embodiments>
Although the details have been mainly shown in the above embodiment regarding downlink communication (downlink), it is also applicable to uplink communication (uplink).

In the above-described embodiment, different wireless communication systems (for example, WiFi, WiMAX (Worldwide Interoperability for Microwave Access), IEEE 802.802.) Adopting a TDD (Time Division Duplex) method that uses the same frequency for uplink and downlink at different times. 16m).

Also, the above embodiment may be a wireless communication system that employs an FDD (Frequency Division Duplex) method that uses different frequencies simultaneously on the uplink and downlink.

Also, while the above embodiments have been described with reference to LTE wireless communication systems, at least some of the methods and apparatus of the various embodiments can cover a wide range of communications including many non-LTE and / or non-cellular systems. Applicable to the system. For example, the above embodiment may be a UMTS (Universal Mobile Telecommunications System) system.

The above radio station may be a transmitter, for example. Here, the radio station that receives data from the transmitter may be a receiver.

The above radio station may be a base station, for example. Here, a base station can be used to communicate with one or more wireless terminals, and the functionality of an access point, node, evolved node B (eNB: evolved Node B), or some other network entity May be included in part or in whole. The base station communicates with a UE (User Equipment) via an air interface. This communication can occur through one or more sectors. The base station may act as a router between the UE and the rest of the access network, which may include an Internet Protocol (IP) network, by converting the received air interface frame into an IP packet. The base station may also coordinate management of attributes for the air interface and may be a gateway between the wired network and the wireless network.

The above terminal can also be called a wireless terminal, a mobile terminal, or a user terminal (or user). The terminal is also a system, subscriber unit, subscriber station, mobile station, wireless terminal, mobile device, node, device, remote station, remote terminal, wireless communication device, wireless communication device, wireless communication device or user agent function May include some or all of sex. Terminals include cellular phones, cordless phones, session initiation protocol (SIP) phones, smartphones, wireless local loop (WLL) stations, personal digital assistants (PDAs), laptops, tablets, netbooks, smart books, handheld communication devices, handhelds It may be a computing device, a satellite radio, a wireless modem card and / or another processing device that communicates via a wireless system.

In addition, in the said embodiment, although the example in which the operator of a terminal owns one terminal was shown, it is not restricted to this. One terminal may be shared by a plurality of operators.

Further, the above radio station can be realized by hardware, software, or a combination thereof. The above calculation method can also be realized by hardware, software, or a combination thereof. Here, “realized by software” means realized by a computer reading and executing a program.

The program can be stored and supplied to a computer using various types of non-transitory computer readable media. Non-transitory computer readable media include various types of tangible storage media (tangible storage medium). Examples of non-transitory computer-readable media are magnetic recording media (for example, flexible disks, magnetic tapes, hard disk drives), magneto-optical recording media (for example, magneto-optical disks), CD-ROMs (Compact Disc--Read-Only Memory). , CD-R (Compact-Disc--Recordable), CD-R / W (Compact-Disc--Rewritable), DVD-ROM (Digital-Versatile-Disc--ROM), DVD-R (Digital-Versatile-Disc--Recordable), DVD-R / W (Digital Versatile Disc-Rewritable), semiconductor memory (for example, mask ROM, PROM (Programmable ROM), EPROM (Erasable PROM), flash ROM, RAM (Random Access Memory)).

Also, the program may be supplied to the computer by various types of temporary computer readable media. Examples of transitory computer readable media include electrical signals, optical signals, and electromagnetic waves. The temporary computer-readable medium can supply the program to the computer via a wired communication path such as an electric wire and an optical fiber, or a wireless communication path.

<Appendix>
Part or all of the exemplary embodiments described above can be described as the following supplementary notes. However, the following supplementary notes are merely examples of the present invention, and the present invention is not limited only to such cases.
(Appendix 1)
Determining an initial vector based on channel information in a propagation path from a radio station having a plurality of antennas to another radio station;
Performing eigenvalue decomposition of the transmission covariance matrix using the iterative method based on the initial vector;
Method of calculation.
(Appendix 2)
The channel information is
Related to a part of the radio resource area to be subjected to the eigenvalue decomposition,
The initial vector is
An eigenvector calculated based on the channel information,
The calculation method according to attachment 1.
(Appendix 3)
The transmission covariance matrix is
Calculated based on the eigenvalues of the reception covariance matrix and the eigenvectors of the reception covariance matrix;
The eigenvector is
An eigenvector of the transmission covariance matrix,
The calculation method according to attachment 2.
(Appendix 4)
The partial area is:
It is an area where the power of the channel is highest among the radio resource areas to be subjected to the eigenvalue decomposition.
The calculation method according to attachment 2.
(Appendix 5)
The initial vector is
An eigenvector calculated based on channel information related to a region combining a plurality of radio resource regions to be subjected to eigenvalue decomposition,
The calculation method according to attachment 1.
(Appendix 6)
The eigenvector is
An eigenvector of a transmission covariance matrix averaged in an area obtained by combining a plurality of radio resource areas to be subjected to eigenvalue decomposition;
The calculation method according to attachment 5.
(Appendix 7)
The initial vector is
Based on a precoder selected by the other radio station, determined with reference to a predetermined table,
The calculation method according to attachment 1.
(Appendix 8)
The table is
A precoder obtained by correcting a precoder defined by the system so as to correspond to the number of antennas of the radio station.
The calculation method according to attachment 7.
(Appendix 9)
The channel information is
A channel matrix estimated by the wireless station,
The calculation method according to attachment 1.
(Appendix 10)
The channel information is
A precoder selected by the other radio station;
The calculation method according to attachment 1.
(Appendix 11)
A determination unit that determines an initial vector based on channel information in a propagation path from a wireless station having a plurality of antennas to another wireless station;
A calculation unit that performs eigenvalue decomposition of a transmission covariance matrix using an iterative method based on the initial vector;
Radio station.
(Appendix 12)
The channel information is
Related to a part of the radio resource area to be subjected to the eigenvalue decomposition,
The initial vector is
An eigenvector calculated based on the channel information,
The radio station according to appendix 11.
(Appendix 13)
The transmission covariance matrix is
Calculated based on the eigenvalues of the reception covariance matrix and the eigenvectors of the reception covariance matrix;
The eigenvector is
An eigenvector of the transmission covariance matrix,
The radio station according to attachment 12.
(Appendix 14)
The partial area is:
It is an area where the power of the channel is highest among the radio resource areas to be subjected to the eigenvalue decomposition.
The radio station according to attachment 12.
(Appendix 15)
The initial vector is
An eigenvector calculated based on channel information related to an area obtained by combining a plurality of radio resource areas to be subjected to eigenvalue decomposition;
The radio station according to appendix 11.
(Appendix 16)
The eigenvector is
An eigenvector of a transmission covariance matrix averaged in an area obtained by combining a plurality of radio resource areas to be subjected to eigenvalue decomposition;
The radio station according to attachment 15.
(Appendix 17)
The initial vector is
Based on a precoder selected by the other radio station, determined with reference to a predetermined table,
The radio station according to appendix 11.
(Appendix 18)
The table is
A precoder obtained by correcting a precoder defined by the system so as to correspond to the number of antennas of the radio station.
The radio station according to appendix 17.
(Appendix 19)
The channel information is
A channel matrix estimated by the wireless station,
The radio station according to appendix 11.
(Appendix 20)
The channel information is
A precoder selected by the other radio station;
The radio station according to appendix 11.
(Appendix 21)
Determining an initial vector based on channel information in a propagation path from a radio station having a plurality of antennas to another radio station;
Performing eigenvalue decomposition of the transmission covariance matrix using the iterative method based on the initial vector;
A program that causes a computer to execute.
(Appendix 22)
The channel information is
Related to a part of the radio resource area to be subjected to the eigenvalue decomposition,
The initial vector is
An eigenvector calculated based on the channel information,
The program according to appendix 21.
(Appendix 23)
The transmission covariance matrix is
Calculated based on the eigenvalues of the reception covariance matrix and the eigenvectors of the reception covariance matrix;
The eigenvector is
An eigenvector of the transmission covariance matrix,
The program according to attachment 22.
(Appendix 24)
The partial area is:
It is an area where the power of the channel is highest among the radio resource areas to be subjected to the eigenvalue decomposition.
The program according to attachment 22.
(Appendix 25)
The initial vector is
An eigenvector calculated based on channel information related to an area obtained by combining a plurality of radio resource areas to be subjected to eigenvalue decomposition;
The program according to appendix 21.
(Appendix 26)
The eigenvector is
An eigenvector of a transmission covariance matrix averaged in an area obtained by combining a plurality of radio resource areas to be subjected to eigenvalue decomposition;
The program according to attachment 25.
(Appendix 27)
The initial vector is
Based on a precoder selected by the other radio station, determined with reference to a predetermined table,
The program according to appendix 21.
(Appendix 28)
The table is
A precoder obtained by correcting a precoder defined by the system so as to correspond to the number of antennas of the radio station.
The program according to appendix 27.
(Appendix 29)
The channel information is
A channel matrix estimated by the wireless station,
The program according to appendix 21.
(Appendix 30)
The channel information is
A precoder selected by the other radio station;
The program according to appendix 21.

Furthermore, the present invention is not limited to the above-described embodiments, and various modifications can be made without departing from the gist of the present invention already described. The functions or steps and / or actions according to each embodiment described herein may not be performed in a particular order. Further, although elements of the invention may be described or claimed in the singular, the elements may be in the plural unless expressly stated to be limited to the singular.

This application claims priority based on Japanese Patent Application No. 2016-008497 filed on January 20, 2016, the entire disclosure of which is incorporated herein.

101 Channel matrix storage unit 102 in the first, second, and third embodiments 102 Average transmission covariance matrix calculation unit 102A in the first and third embodiments 102A Average transmission covariance matrix calculation unit 103 in the second embodiment 103 Initial vector calculation unit 103A in the first embodiment 103A Initial vector calculation unit 103B in the second embodiment 103B Initial vector calculation unit 103-1 in the third embodiment 103-1 RB selection unit 103 in the initial vector calculation unit 103 in the first embodiment 103 -2 Reception Covariance Matrix Calculation Unit in Initial Vector Calculation Unit 103 of First Embodiment 103-3 Reception Covariance Matrix Eigenvalue Calculation Unit in Initial Vector Calculation Unit 103 of First Embodiment 103-4 First Embodiment Received covariance matrix eigenvector calculation unit 103-5 in the initial vector calculation unit 103 of the first implementation Transmission covariance matrix eigenvector calculation unit 104 in the initial vector calculation unit 103 of the state 104 eigenvalue decomposition execution unit 104-1 in the first, second, and third embodiments 104-1 first eigenvalue in the eigenvalue decomposition execution unit 104 in the first embodiment Calculation unit 104-2 Second eigenvalue calculation unit in eigenvalue decomposition execution unit 104 in the first embodiment 105 Eigenvalue decomposition result storage unit in the first, second, and third embodiments 106 Terminal feedback information in the third embodiment Storage unit 1000 Radio station 1001 Determination unit 1002 Calculation unit

Claims

Determining an initial vector based on channel information in a propagation path from a radio station having a plurality of antennas to another radio station;
Performing eigenvalue decomposition of the transmission covariance matrix using the iterative method based on the initial vector;
Method of calculation.
The channel information is
Related to a part of the radio resource area to be subjected to the eigenvalue decomposition,
The initial vector is
An eigenvector calculated based on the channel information,
The calculation method according to claim 1.
The transmission covariance matrix is
Calculated based on the eigenvalues of the reception covariance matrix and the eigenvectors of the reception covariance matrix;
The eigenvector is
An eigenvector of the transmission covariance matrix,
The calculation method according to claim 2.
The partial area is:
It is an area where the power of the channel is highest among the radio resource areas to be subjected to the eigenvalue decomposition.
The calculation method according to claim 2.
The initial vector is
An eigenvector calculated based on channel information related to a region combining a plurality of radio resource regions to be subjected to eigenvalue decomposition,
The calculation method according to claim 1.
The eigenvector is
An eigenvector of a transmission covariance matrix averaged in an area obtained by combining a plurality of radio resource areas to be subjected to eigenvalue decomposition;
The calculation method according to claim 5.
The initial vector is
Based on a precoder selected by the other radio station, determined with reference to a predetermined table,
The calculation method according to claim 1.
The table is
A precoder obtained by correcting a precoder defined by the system so as to correspond to the number of antennas of the radio station.
The calculation method according to claim 7.
The channel information is
A channel matrix estimated by the wireless station,
The calculation method according to claim 1.
The channel information is
A precoder selected by the other radio station;
The calculation method according to claim 1.
A determination unit that determines an initial vector based on channel information in a propagation path from a wireless station having a plurality of antennas to another wireless station;
A calculation unit that performs eigenvalue decomposition of a transmission covariance matrix using an iterative method based on the initial vector;
Radio station.
The channel information is
Related to a part of the radio resource area to be subjected to the eigenvalue decomposition,
The initial vector is
An eigenvector calculated based on the channel information,
The radio station according to claim 11.
The transmission covariance matrix is
Calculated based on the eigenvalues of the reception covariance matrix and the eigenvectors of the reception covariance matrix;
The eigenvector is
An eigenvector of the transmission covariance matrix,
The radio station according to claim 12.
The partial area is:
It is an area where the power of the channel is highest among the radio resource areas to be subjected to the eigenvalue decomposition.
The radio station according to claim 12.
The initial vector is
An eigenvector calculated based on channel information related to a region combining a plurality of radio resource regions to be subjected to eigenvalue decomposition,
The radio station according to claim 11.
The eigenvector is
An eigenvector of a transmission covariance matrix averaged in an area obtained by combining a plurality of radio resource areas to be subjected to eigenvalue decomposition;
The radio station according to claim 15.
The initial vector is
Based on a precoder selected by the other radio station, determined with reference to a predetermined table,
The radio station according to claim 11.
The table is
A precoder obtained by correcting a precoder defined by the system so as to correspond to the number of antennas of the radio station.
The radio station according to claim 17.
The channel information is
A channel matrix estimated by the wireless station,
The radio station according to claim 11.
The channel information is
A precoder selected by the other radio station;
The radio station according to claim 11.
Determining an initial vector based on channel information in a propagation path from a radio station having a plurality of antennas to another radio station;
Performing eigenvalue decomposition of the transmission covariance matrix using the iterative method based on the initial vector;
A storage medium for storing a program for causing a computer to execute the above.
The channel information is
Related to a part of the radio resource area to be subjected to the eigenvalue decomposition,
The initial vector is
An eigenvector calculated based on the channel information,
The storage medium according to claim 21.
The transmission covariance matrix is
Calculated based on the eigenvalues of the reception covariance matrix and the eigenvectors of the reception covariance matrix;
The eigenvector is
An eigenvector of the transmission covariance matrix,
The storage medium according to claim 22.
The partial area is:
It is an area where the power of the channel is highest among the radio resource areas to be subjected to the eigenvalue decomposition.
The storage medium according to claim 22.
The initial vector is
An eigenvector calculated based on channel information related to a region combining a plurality of radio resource regions to be subjected to eigenvalue decomposition,
The storage medium according to claim 21.
The eigenvector is
An eigenvector of a transmission covariance matrix averaged in an area obtained by combining a plurality of radio resource areas to be subjected to eigenvalue decomposition;
The storage medium according to claim 25.
The initial vector is
Based on a precoder selected by the other radio station, determined with reference to a predetermined table,
The storage medium according to claim 21.
The table is
The precoder defined in the system is corrected so as to correspond to the number of antennas of the radio station.
The storage medium according to claim 27.
The channel information is
A channel matrix estimated by the wireless station,
The storage medium according to claim 21.
The channel information is
A precoder selected by the other radio station;
The storage medium according to claim 21.