PRECODING METHOD AND APPARATUS

Abstract

This application provides a precoding method and an apparatus for improving a power utilization of antennas. The method includes: determining a generalized inverse matrix Hm+ of a channel value matrix Hm of K users in an mth iteration in J iteration operations, where m=1, 2, . . . , and J, 1≤J≤M−K+1, and M is a quantity of antennas; selecting, from a set S1m-1 based on Hm+, an antenna index nm and an update coefficient αm corresponding to nm, where S1m-1 is a set of n unselected in antenna indexes n of the M antennas until an (m−1)th iteration ends; determining a weight matrix Wm based on Hm+ and αm; and assigning a row in Wm and corresponding to nm to a row in a final weight matrix Wopt and corresponding to nm, where Wopt is to adjust transmit powers of at least J of the M antennas to a preset maximum transmit power P after the J iteration operations end.

Description

TECHNICAL FIELD

This application relates to the field of communication technologies, and in particular, to a precoding method and an apparatus.

BACKGROUND

In a multi-antenna multi-user transmission system, to avoid impact on orthogonality between a plurality of users caused when a transmit power of an antenna exceeds a maximum value, a base station generally performs maximum transmit power normalization processing on transmit powers of a plurality of antennas by using a power-limited eigenvector beamforming (PEBF) algorithm. That is, initial transmit powers of the plurality of antennas are divided by a transmit power of an antenna with a largest transmit power in the plurality of antennas. In this way, transmit powers of all the antennas except the antenna with the largest transmit power are reduced, which reduces a power utilization of the antennas, and results in a problem of power waste.

SUMMARY

This application provides a precoding method and an apparatus, to improve a power utilization of antennas.

According to a first aspect, an embodiment of this application provides a precoding method, including: determining a generalized inverse matrix H_m⁺ of a channel value matrix H_m of K users in an m^th iteration in J iteration operations, where m=1, 2, . . . , and J, 1≤J≤M−K+1, J is an integer, and M is a quantity of antennas; selecting, from a set S1_m-1 based on H_m, an antenna index n_m and an update coefficient α_m corresponding to n_m, where S1_m-1 is a set of n unselected in antenna indexes n of the M antennas until an (m−1)^th iteration ends; determining a weight matrix W_m based on H_m⁺ and α_m; and assigning a row that is in W_m and that corresponds to n_m to a row that is in a final weight matrix W_opt and that corresponds to n_m, where W_opt is used to adjust transmit powers of at least J of the M antennas to a preset maximum transmit power P after the J iteration operations end.

According to the precoding method provided in this application, W_opt obtained through the J iteration operations can adjust the transmit powers of the at least J of the M antennas to the preset maximum transmit power P, thereby improving a power utilization of the M antennas.

In an embodiment, a product of H_m and W_m is a diagonal matrix or approximates to a diagonal matrix.

In an embodiment, the obtained W_opt can ensure orthogonality between the plurality of users.

In an embodiment, the determining a weight matrix W_m based on H_m⁺ and α_m includes: calculating W_m by using a formula W_m=W_m-1+α_mH_m⁺, where W_m-1 is a weight matrix used in the (m−1)^th iteration.

In an embodiment, when m is greater than 2, the determining H_m⁺ of H_m of K users includes: calculating H_m⁺ by using a formula

$H_m^{+}={\left(H_{m-1}+{{\beta }_{n_{m-1}}}^{*}{{\beta }_{n_{m-1}}}^H\right)}^{-1}=H_{m-1}^{+}-\frac{H_{m-1}^{+}{{\beta }_{n_{m-1}}}^{*}{{\beta }_{n_{m-1}}}^H{H}_{m-1}^{+}}{1+{{\beta }_{n_{m-1}}}^H{H}_{m-1}^{+}{\beta }_{n_{m1}}},$

where H_m-1⁺ represents a generalized inverse matrix of a channel value matrix H_m-1 of the K users in the (m−1)^th iteration, and β_nm-1 represents a column that is in H_m-1 and that corresponds to an antenna index n_m-1 selected in the (m−1)^th iteration.

In an embodiment, complexity of a plurality of large matrix inversions can be reduced.

In an embodiment, the determining H_m⁺ of a channel value matrix H_m includes: setting, to zero, a column that is in an initial channel value matrix H of the K users and that corresponds to an antenna index n selected until the (m−1)^th iteration ends, to obtain H_m; and calculating H_m⁺ based on H_m.

In an embodiment, the selecting, from a set S1_m-1 based on H_m, n_m and an update coefficient α_m corresponding to n_m includes: substituting M−m+1 n in S1_m-1 into a formula one by one:

$\alpha =\frac{\begin{array}{c}-\mathrm{real}\left({r\left(H_m^{+},n\right)}^{*}{r\left(W_{m-1},n\right)}^H\right)+\\ \sqrt{{\mathrm{real}\left({r\left(H_m^{+},n\right)}^{*}{r\left(W_{m-1},n\right)}^H\right)}^2-{r\left(H_m^{+},n\right)}^2\left({r\left(W_{m-1},n\right)}^2-P\right)}\end{array}}{{r\left(H_m^{+},n\right)}^2}$ $\mathrm{and}\text{/}\mathrm{or}$ $\alpha =\frac{\begin{array}{c}-\mathrm{real}\left({r\left(H_m^{+},n\right)}^{*}{r\left(W_{m-1},n\right)}^H\right)-\\ \sqrt{{\mathrm{real}\left(r\left(H_m^{+},n\right)*{r\left(W_{m-1},n\right)}^H\right)}^2-{r\left(H_m^{+},n\right)}^2\left({r\left(W_{m-1},n\right)}^2-P\right)}\end{array}}{{r\left(H_m^{+},n\right)}^2};$

and

determining n making a smallest |α| as an antenna index n_m selected in the m^th iteration process, and determining the smallest |α| as α_m.

In an embodiment, when m=J, the assigning a row that is in W_m and that corresponds to n_m to a row that is in a final weight matrix W_opt and that corresponds to n_m includes: r(W_opt,S1_M-J+1)=r(W_J,S1_M-J+1).

In an embodiment, after the row that is in W_m and that corresponds to n_m is assigned to the (n_m)^th row in the final weight matrix W_opt, the method further includes: deleting rows that are in W_m and H_m⁺ and that correspond to n_m; and deleting a column that is in H_m and that corresponds to n_m.

In an embodiment, a set S2_m-1 is set, M−m+1 numbers a in S2_m-1 are in a one-to-one correspondence with M−m+1 n in S1_m-1, and the selecting, from a set S1_m-1 based on H_m, n_m and an update coefficient α_m corresponding to n_m includes: substituting the M−m+1 numbers a in the set S2_m-1 into a formula one by one:

$\alpha =\frac{\begin{array}{c}-\mathrm{real}\left({r\left(H_m^{+},a\right)}^{*}{r\left(W_{m-1},a\right)}^H\right)+\\ \sqrt{{\mathrm{real}\left({r\left(H_m^{+},a\right)}^{*}{r\left(W_{m-1},a\right)}^H\right)}^2-{r\left(H_m^{+},a\right)}^2\left({r\left(W_{m-1},a\right)}^2-P\right)}\end{array}}{{r\left(H_m^{+},a\right)}^2}$ $\mathrm{and}\text{/}\mathrm{or}$ $\alpha =\frac{\begin{array}{c}-\mathrm{real}\left({r\left(H_m^{+},a\right)}^{*}{r\left(W_{m-1},a\right)}^H\right)-\\ \sqrt{{\mathrm{real}\left({r\left(H_m^{+},a\right)}^{*}{r\left(W_{m-1},a\right)}^H\right)}^2-{r\left(H_m^{+},a\right)}^2\left({r\left(W_{m-1},a\right)}^2-P\right)}\end{array}}{{r\left(H_m^{+},a\right)}^2};$

and

determining n that is in S1_m-1 and that corresponding to a making a smallest |α| as an antenna index n_m selected in the m^th iteration process, and determining the smallest |α| as α_m.

In an embodiment, when m=J, the assigning a row that is in W_m and that corresponds to n_m to a row that is in a final weight matrix W_opt and that corresponds to n_m includes: r(W_opt, S1_M-J+1)=r(W_J, S2_M-J+1).

In an embodiment, before the J iteration operations are performed, the method further includes: performing normalization processing on transmit powers of the M antennas to obtain an original precoding weight matrix; calculating a power utilization of the transmit powers of the M antennas based on the original precoding weight matrix; and when the power utilization is less than a preset threshold, performing the J iteration operations.

According to a second aspect, this application provides an apparatus. The apparatus may be a network device, or may be a chip in the network device. The apparatus has a function of implementing the method according to the first aspect. The function may be implemented by hardware, or may be implemented by hardware by executing corresponding software. The hardware or the software includes one or more modules corresponding to the function. The apparatus may include a first weight calculation module, a data weighting module, a power decision module, a second weight calculation module, and the like.

In an embodiment, when the apparatus is a network device, the first weight calculation module, the data weighting module, the power decision module, and the second weight calculation module may be integrated into a processor of the network device. In an embodiment, the network device further includes a memory. When the network device includes a memory, the memory is configured to store computer execution instructions, the processor is connected to the memory, and the processor executes the computer execution instructions stored in the memory, to enable the network device to perform the method in the first aspect.

In an embodiment, when the apparatus is a chip in the network device, the first weight calculation module, the data weighting module, the power decision module, and the second weight calculation module may be integrated into a processor of the chip. In an embodiment, the chip further includes a memory. When the chip includes a memory, the memory is configured to store computer execution instructions, the processor is connected to the memory, and the processor executes the computer execution instructions stored in the memory, to enable the chip in the network device to perform the method in the first aspect. In an embodiment, the memory is a memory inside the chip, for example, a register or a buffer, or the memory may be a memory located inside the terminal but outside the chip, for example, a read-only memory (ROM) or another type of static storage device that can store static information and instructions, or a random access memory (RAM).

Specifically, the first weight calculation module is configured to determine a generalized inverse matrix H_m⁺ of a channel value matrix H_m of K users in an m^th iteration in J iteration operations, where m=1, 2, . . . , and J, 1≤J≤M−K+1, J is an integer, and M is a quantity of antennas; select, from a set S1_m-1 based on H_m⁺, an antenna index n_m and an update coefficient α_m corresponding to n_m, where S1_m-1 is a set of n unselected in antenna indexes n of the M antennas until an (m−1)^th iteration ends; determine a weight matrix W_m based on H_m⁺ and α_m; and assign a row that is in W_m and that corresponds to n_m to a row that is in a final weight matrix W_opt and that corresponds to n_m, where W_opt is used by the data weighting module to adjust transmit powers of at least J of the M antennas to a preset maximum transmit power P after the first weight calculation module completes the J iteration operations.

In an embodiment, a product of H_m and W_m is a diagonal matrix or approximates to a diagonal matrix.

In an embodiment, that the first weight calculation module determines a weight matrix W_m based on H and α_m includes: calculating W_m by using a formula W_m=W_m-1+α_mH_m⁺, where W_m-1 is a weight matrix used in the (m−1)^th iteration.

In an embodiment, when m is greater than 2, that the first weight calculation module determines H_m⁺ of H_m of K users includes: calculating H_m⁺ by using a formula

$H_m^{+}={\left(H_{m-1}+{{\beta }_{n_{m-1}}}^{*}{{\beta }_{n_{m-1}}}^H\right)}^{-1}=H_{m-1}^{+}-\frac{H_{m-1}^{+}{{\beta }_{n_{m-1}}}^{*}{{\beta }_{n_{m-1}}}^H{H}_{m-1}^{+}}{1+{{\beta }_{n_{m-1}}}^H{H}_{m-1}^{+}{\beta }_{n_{m-1}}},$

where H_m-1⁺ represents generalized inverse matrix of a channel value matrix H_m-1 of the K users in the (m−1)^th iteration, and β_nm-1 represents a column that is in H_m-1 and that corresponds to an antenna index n_m-1 selected in the (m−1)^th iteration.

In an embodiment, that the first weight calculation module determines H_m⁺ of a channel value matrix H_m includes: setting, to zero, a column that is in an initial channel value matrix H of the K users and that corresponds to an antenna index n selected until the (m−1)^th iteration ends, to obtain H_m; and calculating H_m⁺ based on H_m.

In an embodiment, that the first weight calculation module selects, from a set S1_m-1 based on H_m, n_m and an update coefficient α_m corresponding to n_m includes: substituting M−m+1 n in S1_m-1 into a formula one by one:

$\alpha =\frac{\begin{array}{c}-\mathrm{real}\left({r\left(H_m^{+},n\right)}^{*}{r\left(W_{m-1},n\right)}^H\right)+\\ \sqrt{{\mathrm{real}\left({r\left(H_m^{+},n\right)}^{*}{r\left(W_{m-1},n\right)}^H\right)}^2-{r\left(H_m^{+},n\right)}^2\left({r\left(W_{m-1},n\right)}^2-P\right)}\end{array}}{{r\left(H_m^{+},n\right)}^2}$ $\mathrm{and}\text{/}\mathrm{or}$ $\alpha =\frac{\begin{array}{c}-\mathrm{real}\left({r\left(H_m^{+},n\right)}^{*}{r\left(W_{m-1},n\right)}^H\right)-\\ \sqrt{{\mathrm{real}\left({r\left(H_m^{+},n\right)}^{*}{r\left(W_{m-1},n\right)}^H\right)}^2-{r\left(H_m^{+},n\right)}^2\left({r\left(W_{m-1},n\right)}^2-P\right)}\end{array}}{{r\left(H_m^{+},n\right)}^2};$

and

determining n making a smallest |α| as an antenna index n_m selected in the m^th iteration process, and determining the smallest |α| as α_m.

In an embodiment, when m=J, that the first weight calculation module assigns a row that is in W_m and that corresponds to n_m to a row that is in a final weight matrix W_opt and that corresponds to n_m includes: r(W_opt,S1_M-J+1)=r(W_J,S1_M-J+1).

In an embodiment, the first weight calculation module is further configured to: after the row that is in W_m and that corresponds to n_m is assigned to the (n_m)^th row in the final weight matrix W_opt, delete rows that are in W_m and H_m⁺ and that correspond to n_m; and delete a column that is in H_m and that corresponds to n_m.

In an embodiment, a set S2_m-1 is set, M−m+1 numbers a in S2_m-1 are in a one-to-one correspondence with M−m+1 n in S1_m-1, and that the first weight calculation module selects, from a set S1_m-1 based on H_m, n_m and an update coefficient α_m corresponding to n_m includes: substituting the M−m+1 numbers a in the set S2_m-1 into a formula one by one:

$\alpha =\frac{\begin{array}{c}-\mathrm{real}\left({r\left(H_m^{+},a\right)}^{*}{r\left(W_{m-1},a\right)}^H\right)+\\ \sqrt{{\mathrm{real}\left({r\left(H_m^{+},a\right)}^{*}{r\left(W_{m-1},a\right)}^H\right)}^2-{r\left(H_m^{+},a\right)}^2\left({r\left(W_{m-1},a\right)}^2-P\right)}\end{array}}{{r\left(H_m^{+},a\right)}^2}$ $\mathrm{and}\text{/}\mathrm{or}$ $\alpha =\frac{\begin{array}{c}-\mathrm{real}\left({r\left(H_m^{+},a\right)}^{*}{r\left(W_{m-1},a\right)}^H\right)-\\ \sqrt{{\mathrm{real}\left({r\left(H_m^{+},a\right)}^{*}{r\left(W_{m-1},a\right)}^H\right)}^2-{r\left(H_m^{+},a\right)}^2\left({r\left(W_{m-1},a\right)}^2-P\right)}\end{array}}{{r\left(H_m^{+},a\right)}^2};$

and

determining n that is in S1_m-1 and that corresponding to a making a smallest |α| as an antenna index n_m selected in the m^th iteration process, and determining the smallest |α| as α_m.

In an embodiment, when m=J, that the first weight calculation module assigns a row that is in W_m and that corresponds to n_m to a row that is in a final weight matrix W_opt and that corresponds to n_m includes: r(W_opt,S1_M-J+1)=r(W_J,S2_M-J+1).

In an embodiment, the apparatus further includes the second weight calculation module and the power decision module. The second weight calculation module is configured to: before the first weight calculation module performs the J iteration operations, perform normalization processing on transmit powers of the M antennas to obtain an original precoding weight matrix. The power decision module is configured to calculate a power utilization of the transmit powers of the M antennas based on the original precoding weight matrix calculated by the second weight calculation module. The first weight calculation module is configured to: when the power utilization calculated by the power decision module is less than a preset threshold, perform the J iteration operations.

For technical effects of the apparatus provided in this application, refer to technical effects of the first aspect or the implementations of the first aspect. Details are not described herein again.

The processor mentioned anywhere above may be a general-purpose central processing unit (CPU), a microprocessor, an application-specific integrated circuit (ASIC), or one or more integrated circuits configured to control program execution of the method in the second aspect.

According to a third aspect, an embodiment of this application provides a computer storage medium. The computer storage medium stores a program used to implement the method in the first aspect. When the program runs in a network device, the network device is enabled to perform the method in the first aspect.

According to a fourth aspect, an embodiment of this application provides a computer program product. The program product includes a program, and when the program is run, the method in the first aspect is performed.

BRIEF DESCRIPTION OF DRAWINGS

FIG. 1 is a schematic diagram of a structure of a network device according to this application;

FIG. 2 is a flowchart 1 of an embodiment of a precoding method according to this application;

FIG. 3 is a schematic diagram of functions of antennas according to this application;

FIG. 4 is a flowchart 2 of another embodiment of a precoding method according to this application;

FIG. 5 is a schematic diagram of algorithm modules of a network device according to this application;

FIG. 6 is a schematic diagram of comparison between simulation effects of power utilizations according to this application;

FIG. 7 is a schematic diagram of comparison between simulation effects of average cell throughputs according to this application;

FIG. 8 is a schematic diagram of comparison between simulation effects of spectral efficiency gains according to this application;

FIG. 9 is a schematic diagram 1 of a structure of an apparatus according to this application; and

FIG. 10 is a schematic diagram 2 of a structure of an apparatus according to this application.

DESCRIPTION OF EMBODIMENTS

First, words such as “example” or “for example” below are used to represent an example, an illustration, or a description. Any embodiment or design solution described as “example” or “for example” in embodiments of this application should not be interpreted as being more preferable or more advantageous than another implementation or design solution. Exactly, use of the terms such as “example” or “for example” is intended to present a related concept in a specific manner.

Unless otherwise stated, in this specification, “/” usually represents that associated objects before and after “/” are in an “or” relationship. For example, A/B may represent A or B. The term “and/or” describes only an association relationship for describing associated objects and represents that three relationships may exist. For example, A and/or B may represent the following three cases: Only A exists, both A and B exist, and only B exists. In addition, in descriptions of this application, “a plurality of” refers to two or more.

The precoding method provided in this application is applied to a network device having a multi-antenna transmission module. The network device can support a fourth generation (4G) access technology, such as a long term evolution (LTE) access technology, or support a fifth generation (5G) access technology, such as a new radio (NR) access technology, or support a plurality of wireless technologies, such as an LTE technology and an NR technology. In addition, the network device may also be applied to a future-oriented communication technology.

The network device may be a base station, for example, an evolved nodeB (eNB) in a communication system based on the 4G access technology, a next generation nodeB (gNB) in a communication system based on the 5G access technology, a transmission reception point (TRP), a relay node or an access point (AP).

FIG. 1 is a schematic diagram of a structure of a network device according to this application. The network device includes a processor 101, a memory 102, a transceiver 103, and a communication bus 104. The processor 101, the memory 102, and the transceiver 103 are connected via a communication bus 104.

The processor 101 in this embodiment of this application may include at least one of the following types: a general-purpose central processing unit (CPU), a digital signal processor (DSP), a microprocessor, an application-specific integrated circuit (ASIC), a microcontroller unit (MCU), a field programmable gate array (FPGA), or an integrated circuit configured to implement a logic operation. For example, the processor 101 may be a single-core (single-CPU) processor or a multi-core (multi-CPU) processor. The at least one processor 101 may be integrated into one chip or located on a plurality of different chips.

The memory 102 in this embodiment of this application may include at least one of the following types: a read-only memory (ROM) or another type of static storage device that can store static information and instructions, a random access memory (RAM) or another type of dynamic storage device that can store information and instructions, or an electrically erasable programmable read-only memory (EEPROM). In some scenarios, the memory may alternatively be a compact disc read-only memory (CD-ROM) or another compact disc storage medium, an optical disc storage medium (including a compact disc, a laser disc, an optical disc, a digital versatile disc, a Blu-ray disc, and the like), a magnetic disk storage medium or another magnetic storage device, or any other medium that can be configured to carry or store expected program code in a form of an instruction or a data structure and that can be accessed by a computer. However, the memory is not limited thereto.

The memory 102 may exist independently, and is connected to the processor 101. In an embodiment, the memory 102 and the processor 101 may be integrated, for example, integrated into one chip. The memory 102 can store program code for executing the technical solutions in the embodiments of this application, and the processor 101 controls execution of the program code. Various types of computer program code to be executed may also be considered as drivers of the processor 101. For example, the processor 101 is configured to execute the computer program code stored in the memory 102, to implement the technical solutions in the embodiments of this application.

The transceiver 103 may be configured to support receiving or sending of an air interface signal between network devices, and the transceiver 103 may be connected to a plurality of antennas 105. The transceiver 103 includes a transmitter Tx and a receiver Rx. Specifically, one or more antennas 105 may receive a radio frequency signal. The receiver Rx of the transceiver 103 is configured to: receive the air interface signal from the antenna, convert the air interface signal into a digital baseband signal or a digital intermediate frequency signal, and provide the digital baseband signal or the digital intermediate frequency signal to the processor 101, so that the processor 101 further processes the digital baseband signal or the digital intermediate frequency signal, for example, performs demodulating processing and decoding processing. In addition, the transmitter Tx in the transceiver 103 is further configured to: receive a modulated digital baseband signal or digital intermediate frequency signal from the processor 101, convert the modulated digital baseband signal or digital intermediate frequency signal into an air interface signal, and send the air interface signal through the one or more antennas 105.

In the precoding method provided in this application, for M (M≥1) antennas of a network device, transmit powers of J selected antennas can be adjusted to a maximum transmit power through J (1≤J≤M−K+1, where J is an integer) iteration operations, and it can be ensured that transmit powers of M-J unselected antennas do not exceed the maximum transmit power. In other words, according to the precoding method provided in this application, powers of at least J of the M antennas can reach a full power, thereby improving a power utilization.

The precoding method and the apparatus provided in this application are illustrated below with reference to specific embodiments.

It is assumed that a network device has M (M≥1, where M is an integer) antennas, and antenna indexes n are respectively configured for the M antennas. For example, the antenna indexes of the M antennas are respectively 0 to M−1 (n=0, 1, . . . , and M−1), 1 to M (n=1, 2, . . . , and M), or other continuous/discontinuous values. When a quantity of users for whom the M antennas are spatially multiplexed is K (K≥1, where K is an integer), a maximum of M−K+1 iteration operations may be performed based on the precoding method provided in this application.

FIG. 2 is a flowchart of an embodiment of a precoding method according to this application, and mainly describes an m^th (m=1, 2, . . . , and J) iteration process in J iteration processes. The method includes the following operations.

Step 201: The network device determines a generalized inverse matrix H_m⁺ of a channel value matrix H_m of the K users.

In this application, the channel value matrix H_m required in the m^th iteration process may be first calculated, and then H_m⁺ is calculated based on H_m.

In an embodiment, a column that is in an initial channel value matrix H of the K users and that corresponds to an antenna index n selected until an (m−1)^th iteration ends is set to zero, to obtain H_m, and then H_m is inverted to calculate H_m⁺.

In this application, for each iteration operation, the network device selects an antenna, and calculates a corresponding weight, to adjust a transmit power of the antenna to a maximum transmit power. Therefore, to avoid impact of a weight calculated in a current iteration process on a transmit power of an antenna that has been selected, a channel value corresponding to the antenna that has been selected may be set to zero.

For example, assuming that the antenna indexes n of the M antennas are respectively 1 to M, a set S1 may be set to represent a set of unselected n. Before the J iteration operations are performed, S1={1, 2, . . . , M}. Each time an iteration is performed, a selected n is deleted from S1. Then, S1 until the (m−1)^th iteration ends may be represented by S1_m-1, to record n unselected in the antenna indexes n of the M antennas until the (m−1)^th iteration ends.

Then, H_m may be calculated by using the following formula (1):

H _m =H diag{x₁,x₂, . . . ,x_M}  (1)

where x_n represents a coefficient corresponding to an antenna index n. In the m^th iteration process, when n∉S1_m-1, x_n=0; when n∈S1_m-1, x_n=1. The formula (1) indicates that columns corresponding to m−1 antennas selected until the (m−1)^th iteration ends are set to zero.

After H_m is calculated by using the formula (1) H_m⁺ may be calculated based on H_m.

In an embodiment, the network device may perform a column deletion operation when each iteration ends. For example, after the (m−1)^th iteration operation ends, a column that is in H_m-1 and that corresponds to an antenna index n_m-1 selected in the (m−1)^th iteration is deleted to obtain H_m. Then, H_m⁺ is calculated based on H_m.

In an embodiment, the network device may alternatively use a matrix inversion lemma, and obtain H_m based on a generalized inverse matrix H_m-1⁺ of a channel value matrix of the K users that is obtained in the (m−1)^th iteration. In this way, complexity of a plurality of large matrix inversions is reduced.

For example, based on a matrix inversion lemma shown in the following formula (2):

$\begin{array}{cc}{\left(A+a{b}^H\right)}^{-1}=A^{-1}-\frac{A^{-1}a{b}^H{A}^{-1}}{1+b^H{A}^{-1}a}& \left(2\right)\end{array}$

the following formula (3) may be obtained:

$\begin{array}{cc}{H}_m^{+}={\left(H_{m-1}+{{\beta }_{n_{m-1}}}^{*}{{\beta }_{n_{m-1}}}^H\right)}^{-1}=H_{m-1}^{+}-\frac{H_{m-1}^{+}{{\beta }_{n_{m-1}}}^{*}{{\beta }_{n_{m-1}}}^H{H}_{m-1}^{+}}{1+{{\beta }_{n_{m-1}}}^H{H}_{m-1}^{+}{\beta }_{n_{m-1}}}& \left(3\right)\end{array}$

where H_m-1⁺ represents the generalized inverse matrix of the channel value matrix H_m-1 of the K users in the (m−1)^th iteration, and β_nm-1 represents the column that is in H_m-1 and that corresponds to the antenna index n_m-1 selected in the (m−1)^th iteration.

In the first iteration process, that is, when m=1, the network device may directly obtain a corresponding generalized inverse matrix by using the initial channel value matrix H. When m is greater than 2, the network device may directly substitute H_m-1⁺ and β_nm-1 into the formula (3) to obtain H_m⁺.

Step 202: The network device selects, from a set S1_m-1 based on H_m⁺, n_m and an update coefficient α_m corresponding to n_m.

For example, when H_m⁺ is calculated in a zero setting manner, the network device may substitute M−m+1 n in S_m-1 into the following formula (4) and/or formula (5) one by one:

$\begin{array}{cc}\alpha =\frac{\begin{array}{c}-\mathrm{real}\left({r\left(H_m^{+},n\right)}^{*}{r\left(W_{m-1},n\right)}^H\right)+\\ \sqrt{{\mathrm{real}\left({r\left(H_m^{+},n\right)}^{*}{r\left(W_{m-1},n\right)}^H\right)}^2-{r\left(H_m^{+},n\right)}^2\left({r\left(W_{m-1},n\right)}^2-P\right)}\end{array}}{{r\left(H_m^{+},n\right)}^2}\mathrm{and}& \left(4\right)\\ \alpha =\frac{\begin{array}{c}-\mathrm{real}\left({r\left(H_m^{+},n\right)}^{*}{r\left(W_{m-1},n\right)}^H\right)-\\ \sqrt{{\mathrm{real}\left({r\left(H_m^{+},n\right)}^{*}{r\left(W_{m-1},n\right)}^H\right)}^2-{r\left(H_m^{+},n\right)}^2\left({r\left(W_{m-1},n\right)}^2-P\right)}\end{array}}{{r\left(H_m^{+},n\right)}^2};& \left(5\right)\end{array}$

and

determine n making a smallest |α| as an antenna index n_m selected in the m^th iteration process, and determine the smallest |α| as α_m, where W_m-1 is a weight matrix used in the (m−1)^th iteration.

In this application, the selected antenna index indicates that an antenna indicated by the antenna index is selected.

Alternatively, when H_m⁺ is calculated in a row deletion manner, a set S2_m-1 may be set, and M−m+1 numbers a in S2_m-1 are in a one-to-one correspondence with M−m+1 n in S1_m-1 The network device may substitute the M−m+1 numbers a in the set S2_m-1 into the following formula (6) and/or formula (7) one by one:

$\begin{array}{cc}\alpha =\frac{\begin{array}{c}-\mathrm{real}\left({r\left(H_m^{+},a\right)}^{*}{r\left(W_{m-1},a\right)}^H\right)+\\ \sqrt{{\mathrm{real}\left({r\left(H_m^{+},a\right)}^{*}{r\left(W_{m-1},a\right)}^H\right)}^2-{r\left(H_m^{+},a\right)}^2\left({r\left(W_{m-1},a\right)}^2-P\right)}\end{array}}{{r\left(H_m^{+},a\right)}^2}& \left(6\right)\\ \alpha =\frac{\begin{array}{c}-\mathrm{real}\left({r\left(H_m^{+},a\right)}^{*}{r\left(W_{m-1},a\right)}^H\right)-\\ \sqrt{{\mathrm{real}\left({r\left(H_m^{+},a\right)}^{*}{r\left(W_{m-1},a\right)}^H\right)}^2-{r\left(H_m^{+},a\right)}^2\left({r\left(W_{m-1},a\right)}^2-P\right)}\end{array}}{{r\left(H_m^{+},a\right)}^2};\mathrm{and}& \left(7\right)\end{array}$

and

determine n that is in S1_m-1 and that corresponds to a making a smallest |α| as an antenna index n_m selected in the m^th iteration process, and determine the smallest |α| as α_m.

Step 203: Determine a weight matrix W_m based on H_m⁺ and α_m.

For example, the weight matrix used in the m^th iteration may be calculated by using the weight matrix W_m-1 used in the (m−1)^th iteration and H_m⁺. For example, W_m may be calculated by using the following formula (8):

W _m =W _m-1+α_mH_m⁺  (8)

In this application, on one hand, W_m needs to ensure orthogonality between the plurality of users.

Assuming that an initial weight matrix W₀ is a zero matrix, in the first iteration, W₁=W₀+α₁H⁺=α₁H⁺, in the second iteration, W₂=W₁+α₂H₂⁺, and so on. It can be learned that the weight matrix is related to the generalized inverse matrix of the channel value matrix, and is a row-full rank matrix.

Such a theorem exists: If there is a matrix A and A is a row-full rank L×Q(L≤Q), there is a matrix B that is a row-full rank L×Q (where L-Q columns are zero vectors), a non-zero column vector of the matrix B is a submatrix of A. In other words, a non-zero column vector of B is a subset of columns of A. Then, AB⁺=I_L.

The following proof (9) may be obtained based on the theorem:

AB ⁺=[A]_c,d[B⁺]_d,e=[B]_c,f[B⁺]_f,e=BB⁺=I_L  (9)

W_m-1 and H_m⁺ meet the condition of the foregoing theorem, and therefore, based on the theorem and the proof (9), it can be learned that:

$\begin{array}{cc}\{\begin{array}{c}{H}_{m}r\left(W_{m-1},n\right)=0\\ H_{m}r\left(H_m^{+},n\right)=0\end{array}& \left(10A\right)\end{array}$

where r(H_m⁺, n) represents an n^th row of H_m⁺, and r(W_m-1, n) represents an n^th row of W_m-1.

Then, based on the formula (10A) and the formula (8), it can be learned that a product of H_m and W_m is a diagonal matrix or approximates to a diagonal matrix. That is, when i=j, [H_mW_m]_i,j≠0; when i≠j, [H_mW_m]_i,j is equal to 0 or approximates to 0. Therefore, the weight matrix W_m calculated in the m^th iteration process can ensure the orthogonality between the plurality of users.

On the other hand, W_m needs to ensure that a transmit power of an antenna indicated by n_m selected in the m^th iteration process can be adjusted to a maximum transmit power P, and transmit powers of antennas indicated by remaining antenna indexes in S1_m-1 other than n_m are less than or equal to P.

That is, n_m in S1_m-1 makes ∥r(W_m-1, n_m)∥²=P, and a remaining n in S1_m-1 other than n_m makes ∥r(W_m-1, n)∥²≤P.

Then, an equation (10B) may be obtained:

f _m,n(α)=∥r(W_m-1,n)+αr(H_m⁺,n)+α∥²−P  (10B)

For the equation (10B), n_m making f_m,n(α)=0 may be found, and a remaining n in S1_m-1 other than n_m makes f_m,n(α)≤0.

Based on the following solving process:

${r\left(W_{m-1},n\right)+\alpha r\left(H_m^{+},n\right)+\alpha }^2-P={\left(r\left(W_{m-1},n\right)+\alpha r\left(H_m^{+},n\right)\right)}^H\left(r\left(W_{m-1},n\right)+\alpha r\left(H_m^{+},n\right)\right)-P={r\left(W_{m-1},n\right)}^{H}r(W_{m-1},\mathrm{n\_}+\alpha {r\left(W_{m-1},n\right)}^{H}r\left(H_m^{+},n\right)+\alpha {r\left(H_m^{+},n\right)}^{H}r\left(W_{m-1},n\right)+{\alpha }^2{r\left(H_m^{+},n\right)}^{H}r\left(H_m^{+},n\right)-P={r\left(W_{m-1},n\right)}^2=P+2\alpha \mathrm{real}\left({r\left(W_{m-1},n\right)}^{H}r\left(H_m^{+},n\right)\right)+{\alpha }^2{r\left(H_m^{+},n\right)}^2=0$

it can be learned that the equation (10B) is a convex function and has one or two solutions. Then, the equation 8 is converted into a binomial solution formula, to obtain a formula (11) (that is, the foregoing formula (4) and formula (5)):

$\begin{array}{cc}\alpha =\frac{\begin{array}{c}-\mathrm{real}\left({r\left(H_m^{+},n\right)}^{*}{r\left(W_{m-1},n\right)}^H\right)\pm \\ \sqrt{{\mathrm{real}\left({r\left(H_m^{+},n\right)}^{*}{r\left(W_{m-1},n\right)}^H\right)}^2-{r\left(H_m^{+},n\right)}^2\left({r\left(W_{m-1},n\right)}^2-P\right)}\end{array}}{\left({r\left(H_m^{+},n\right)}^2\right)}& \left(11\right)\end{array}$

In the formula (11), an absolute value of the second term of the numerator is always greater than the first term, and therefore, α has a positive solution and a negative solution.

Each n in S1_m-1 is substituted into the formula (4), the formula (5), or the formula (4) and the formula (5) one by one, to obtain corresponding α. After each n and the corresponding α are substituted into the equation (10B), a function curve of an antenna corresponding to each n in S1_m-1 may be obtained. For example, it is assumed that S1_m-1 includes four n: n1, n2, n3, and n4. After corresponding α is obtained based on the formula (4) and/or the formula (5), n1, n2, n3, n4, and corresponding α are substituted into the equation (10B). Functions of antennas indicated by n1, n2, n3, and n4 are shown in FIG. 3.

Based on FIG. 3, it can be learned that a smallest |α| (α_m) obtained based on n1 can make f_m,n1(α_m)=0, and f_m,n(α_m)≤0, where n=n2, n3, n4.

Therefore, based on the formula (8) and the methods for calculating α_m and determining n_m that are provided in this application, the calculated W_m can ensure that the transmit power of the antenna indicated by n_m can be adjusted to the maximum power P, and the transmit powers of the antennas indicated by the remaining antenna indexes in S1_m-1 other than n_m are less than or equal to P.

Step 204: The network device assigns a row that is in W_m and that corresponds to n_m to a row that is in a final weight matrix W_opt and that corresponds to n_m.

W_opt is used to adjust the transmit powers of the M antennas after the J iteration operations end. In this application, the row that is in W_m and that corresponds to n_m can adjust the transmit power of the antenna corresponding to n_m to P. Therefore, the row that is in W_m and that corresponds to n_m may be assigned to the row that is in W_opt and that corresponds to n_m, to ensure that W_opt can adjust the transmit power of the antenna corresponding to n_m to P.

Then, when the J iterations end, after the transmit powers of the M antennas are adjusted by using W_opt, transmit powers of antennas selected in the J iteration processes can be adjusted to P.

It may be understood that if H_m is calculated based on the formula (1), each time an iteration is performed, a column corresponding to a selected antenna in a channel value matrix is set to zero. Therefore, a corresponding row in a weight matrix calculated in a subsequent iteration process is also set to zero. Therefore, the weight matrix calculated in the subsequent iteration process does not affect a transmit power of the antenna that has been selected. In other words, the antenna selected in the (m−1)^th iteration process before the m^th iteration is still maintained at P.

In an embodiment, after each iteration ends, the network device may alternatively delete rows corresponding to a selected antenna in a weight matrix obtained in the current iterative calculation and a generalized inverse matrix of a channel value matrix, and delete a column corresponding to the selected antenna in the channel value matrix.

For example, after the m^th iteration ends, a column that is in H_m and that corresponds to n_m is deleted, and rows that are in W_m and H_m⁺ and that corresponds to n_m are deleted, to avoid impact of a subsequent iteration process on the transmit power of the antenna indicated by n_m. In addition, calculation complexity of the subsequent iteration process can be reduced.

W_J in a J^th iterative calculation can ensure that a transmit power of a selected antenna is adjusted to P, and transmit powers of unselected antennas are less than or equal to P. Therefore, the network device may directly assign a row that is in W_J and that corresponds to n_J selected in the J^th iterative calculation and rows corresponding to M-J n unselected until the J iterations to rows in W_opt one by one. In this way, after the transmit powers of the M antennas are adjusted by using W_opt, the transmit powers of the antennas selected in the J iteration processes can be adjusted to P, and transmit powers of M-J remaining unselected antennas are less than or equal to P.

Iteration operations of the precoding method provided in this application are illustrated below respectively with reference to the zero setting and row deletion manners by using J=3 as an example.

When the zero setting manner is used, three iterations are as follows:

S301. Initialize parameters:

S1₀=[1, . . . ,M],W₀=[0,0, . . . ,0,0]∈C^M×K, and H=[h₁^H;h₂^H; . . . ;h_K^H]

where h₁ represents a channel or a channel feature vector of a first user, h₂ represents a channel or a channel feature vector of a second user, . . . , and h_K represents a channel or a channel feature vector of a K^th user. The antenna indexes n of the M antennas are respectively 1, 2, . . . , and M, respectively corresponding to a first row, a second row, . . . , and an M^th row of the initial weight matrix W₀.

S302. Start the first iteration, and calculate H₁⁺ of H₁:

In the first iteration process, the initial channel value matrix H may be used as the channel value matrix H₁ used in the first iteration, and then H₁⁺ is directly calculated based on H₁. For example, an inverse matrix U₁=inv(H₁*H₁^H) of H₁ is calculated, and then H₁⁺ is calculated by using H₁⁺=H₁^H*U₁.

S303. Search for n₁ that is in S1₀ and that makes a smallest |α|, and α₁:

A for loop may be used to substitute the M antenna indexes n in S1₀ into a formula

$\alpha =\frac{\sqrt{P}}{r\left(H^{+},n\right)}$

one by one, to obtain [α₁, n₁]=argmin(|α|_min).

S304. Update a weight matrix.

A weight matrix W₁=W₀+α₁H₁⁺=α₁H₁⁺ in the first iteration is calculated, where W₁∈C^M×K.

A row that is in W₁ and that corresponds to n₁ (that is, an (n₁)^th row in W₁) is assigned to a row that is in W_opt (W_opt∈C^M×K) and that corresponds to n₁ (that is, an (n₁)^th row in W_opt). That is, r(W_opt,n₁)=r(W₁,n₁).

S305. Delete n₁ from S1₀ to obtain S1₁, that is, S1₁ includes M−1 remaining antenna indexes other than n₁.

S306. Start the second iteration, and calculate H₂⁺ of H₂:

H₂ is calculated based on the formula (1). Because n₁∉S1₁ and x_n1=0, an (n₁)th column in H is set to zero, to obtain H₂.

Then, H₂⁺ is calculated based on H₂. For example, an inverse matrix U₂=inv(H₂*H₂^H) of H₂ is calculated, and then H₂⁺ is calculated by using H₂⁺=H₂^H*U₂.

S307. Search for n₂ that is in S1₁ and that makes a smallest |α|, and α₂:

The for loop may be used to substitute the M−1 n in S1₁ into the formula (4) and/or the formula (5) one by one, to obtain [α₂, n₂]=argmin(|α|_min).

S308. Update the weight matrix.

A weight matrix W₂=W₁+α₂H₂⁺ in the second iteration is calculated, where W₂∈C^M×K.

A row that is in W₂ and that corresponds to n₂ (that is, an (n₂)^th row in W₂) is assigned to a row that is in W_opt and that corresponds to n₂ (that is, an (n₂)^th row in W_opt). That is, r(W_opt,n₂)=r(W₂,n₂).

S309. Delete n₂ from S1₁ to obtain S1₂, that is, S1₂ includes M−2 remaining antenna indexes other than n₁ and n₂.

S310. Start the third iteration, and calculate H₃⁺ of H₃:

H₃ is calculated based on the formula (1). Because n₁∉S1₁, x_n1=0, n₂∉S1₂, and x_n2=0, (n₁)th and (n₂)th columns in H are set to zero, to obtain H₃.

Then, H₃⁺ is calculated based on H₃. For example, an inverse matrix U₃=inv(H₃*H₃^H) of H₃ is calculated, and then H₃⁺ is calculated by using H₃⁺=H₃^H*U₃.

S311. Search for n₃ that is in S1₂ and that makes a smallest |α|, and α₃:

The for loop may be used to substitute the M−2 n in S1₂ into the formula (4) and/or the formula (5) one by one, to obtain [α₃, n₃]=argmin(|α|_min).

S312. Update the weight matrix.

A weight matrix W₃=W₂+α₃H₃⁺ in the first iteration is calculated, where W₃∈C^M×K.

A row that is in W₃ and that corresponds to each n in S1₂ is assigned to a row that is in W_opt and that corresponds to each n in S1₂. That is, r(W_opt, S1₂)=r(W₃, S1₂).

S313. The iteration ends.

So far, the output W_opt can adjust transmit powers of antennas corresponding to the antenna indexes n₁, n₂, and n₃ in the M antennas of the network device to the maximum transmit power, and ensure that transmit powers of the M−3 remaining antennas are less than or equal to the maximum transmit power.

When the row deletion manner is used, three iterations are as follows:

S401. Initialize parameters:

S1₀=[1, . . . , M],

S2₀=[1, . . . , M],

W₀=[0,0, . . . ,0,0]∈C^M×K, and

H=[h₁^H; h₂^H; . . . ; h_K^H]

where h₁ represents a channel or a channel feature vector of a first user, h₂ represents a channel or a channel feature vector of a second user, . . . , and h_K represents a channel or a channel feature vector of a K^th user. The antenna indexes n of the M antennas are respectively 1, 2, . . . , and M, respectively corresponding to a first row, a second row, . . . , and an M^th row of the initial weight matrix W₀. The set S2₀ of numbers a corresponding to the antenna indexes (or the rows of the weight matrix) is set. Initially, M numbers in S2₀ are in a one-to-one correspondence with the M antenna indexes in S1₀.

S402. Start the first iteration, and calculate H₁⁺ of H₁:

In the first iteration process, the initial channel value matrix H may be used as the channel value matrix H₁ used in the first iteration, and then H₁⁺ is directly calculated based on H₁. For example, an inverse matrix U₁=inv(H₁*H₁^H) of H₁ is calculated, and then H₁⁺ is calculated by using H₁⁺=H₁^H*U₁.

S403. Determine n₁ and α₁.

A for loop may be used to substitute the M numbers a in S2₀ into a formula

$\left|\alpha \right|=\frac{\sqrt{P}}{r\left(H^{+},a\right)}$

one by one, to obtain [α₁,a₁]=argmin(|α|_min). n₁ that is in S1₀ and that corresponds to a₁ in S2₀ is determined as a selected antenna index. For example, a₁ is 3. Then, corresponding n₁ is 3.

S404. Update a weight matrix.

A weight matrix W₁=W₀+α₁H₁⁺=α₁H₁⁺ in the first iteration is calculated, where W₁∈C^M×K.

An (a₁)^th row in W₁ is assigned to a row that is in W_opt (W_opt∈C^M×K) and that corresponds to n₁ (that is, an (n₁)^th row in W_opt). That is, r(W_opt, n₁)=r(W₁, a₁).

S405. Obtain a column that is in H₁ and that corresponds to n₁ (that is, an (a₁)^th column in H₁), that is, g₁=c(H₁,a₁); delete a row that is in W₁ and that corresponds to n₁ (that is, an (a₁)^th row in W₁), delete a row that is in H₁⁺ and that corresponds to n₁ (that is, an (a₁)^th row in H₁⁺), and delete a column that is in H₁ and that corresponds to n₁ (that is, an (a₁)^th column in H₁, to obtain H₂; delete n₁ from S1₀ to obtain S1₁, that is, S1₁ includes M−1 remaining antenna indexes other than n₁; and update S2₀ to S2₁[1, . . . , M−1]. 1, 2, . . . , and M−1 in S2₁ are in a one-to-one correspondence with the M−1 antenna indexes in S1₁.

For example, when M=5, S1₀=[1,2,3,4,5] and S2₀=[1,2,3,4,5]. Initially, a number 1 in S2₀ corresponds to an antenna index 1 in S1₀, a number 2 in S2₀ corresponds to an antenna index 2 in S1₀, . . . , and a number 5 in S2₀ corresponds to an antenna index 5 in S1₀.

Assuming that a₁=3 is obtained through calculation in S403, it corresponds to the antenna index 3 in S1₀. That is, n₁=3 is determined. Then, in S404, r(W_opt, 3)=r(W₁, 3) is performed. In S405, g₁=c (H₁, 3), a row that is in W₁ and that corresponds to the antenna index 3 (that is, a third row in W₁) is deleted, a row that is in H₁⁺ and that corresponds to the antenna index 3 (that is, a third row in H₁⁺) is deleted, and a column that is in H₁ and that corresponds to the antenna index 3 (that is, a third column in H₁) is deleted, to obtain H₂. 3 is deleted from S1₀ to obtain S1₁, that is, S1₁ includes four remaining antenna indexes other than 3. That is, S1₁=[1,2,4,5]. S2₀ is updated to S2₁=[1,2,3,4] In this case, the number 1 in S2₁ corresponds to the antenna index 1 in S1₁, the number 2 in S2₁ corresponds to the antenna index 2 in S1₁, the number 3 in S2₁ corresponds to the antenna index 4 in S1₂, and the number 4 in S2₁ corresponds to the antenna index 5 in S1₁.

It should be noted that, if the generalized inverse matrix of the channel value matrix is calculated by using the formula (3), the column that is in H₁ and that corresponds to n₁ needs to be obtained for preparation, and the column that is in H₁ and that corresponds to n₁ is deleted to obtain H₂, as described in S405. If the channel value matrix is first calculated and then the generalized inverse matrix of the channel value matrix is calculated, the column that is in H₁ and that corresponds to n₁ does not need to be obtained, and the column that is in H₁ and that corresponds to n₁ is directly deleted to obtain H₂.

S406. Start the second iteration, and calculate H₂⁺ of H₂:

H₂⁺ is calculated based on the formula (3) by using H₁⁺. For example, after H₁⁺ and β_n1 (that is, g₁) are substituted into the formula (3), a simplified derivation process of the formula (3) is as follows:

$D1=g_1^{H*}{U}_1$ $G1=\frac{D1}{1-D{1}^{*}{g}_1}$ $H_2^{+}=H_1^{+}+{\left(H_1^{+*}{g}_1\right)}^{*}G1$

For ease of the third iterative calculation, an intermediate parameter U₂=U₁+(U₁*g₁)*G1 may be calculated herein for preparation.

S407. Determine n₂ and α₂.

The for loop may be used to substitute the M−1 numbers a in S2₁ into the formula

$\left|\alpha \right|=\frac{\sqrt{P}}{r\left(H^{+},a\right)}$

one by one, to obtain [α₂, a₂]=argmin(|α|_min). n₂ that is in S1₁ and that corresponds to a₂ in S2₁ is determined as a selected antenna index.

S408. Update the weight matrix.

A weight matrix W₂=W₁+α₂H₂⁺ in the second iteration is calculated, where W₂∈C^M-1×K.

An (a₂)^th row (a row corresponding to n₂) in W₂ is assigned to a row that is in W_opt (W_opt∈C^M×K) and that corresponds to n₂ (that is, an (n₁)^th row in W_opt). That is, r(W_opt, n₂)=r(W₂, a₂).

S409. Obtain a column that is in H₂ and that corresponds to n₂ (that is, an (a₂)^th column in H₂), that is, g₂=c(H₂, a₂); delete a row that is in W₂ and that corresponds to n₂ (that is, an (a₂)^th row in W₂), delete a row that is in H₂⁺ and that corresponds to n₂ (that is, an (a₂)^th row in H₂⁺), and delete a column that is in H₂ and that corresponds to n₂ (that is, an (a₂)^th column in H₂, to obtain H₃; delete n₂ from S1₁ to obtain S1₂, that is, S1₂ includes M−2 remaining antenna indexes other than n₁ and n₂; and update S2₁ to S2₂=[1, . . . , M−2]. 1, 2, . . . , and M−1 in S2₁ are in a one-to-one correspondence with the M−1 antenna indexes in S1₁.

For example, based on the foregoing example, a₂=3 is obtained through calculation in S407. Then, it corresponds to the antenna index 4 in S1₁. That is, n₂=4 is determined. Then, in S408, r(W_opt, 4)=r(W₂, 3) is performed. In S409, g₂=c(H₂, 3) a row that is in W₂ and that corresponds to the antenna index 4 (that is, a third row in W₂) is deleted, a row that is in H₂⁺ and that corresponds to the antenna index 4 (that is, a third row in H₂⁺) is deleted, and a column that is in H₂ and that corresponds to the antenna index 4 (that is, a third column in H₂) is deleted, to obtain H₃. 4 is deleted from S1₁ to obtain S1₂, that is, S1₂ includes three remaining antenna indexes other than 3 and 4. That is, S1₂=[1,2,5]. S2₁ is updated to S2₂=[1,2,3]. In this case, the number 1 in S2₂ corresponds to the antenna index 1 in S1₂, the number 2 in S2₂ corresponds to the antenna index 2 in S1₂, and the number 3 in S2₂ corresponds to the antenna index 5 in S1₂.

S410. Start the third iteration, and calculate H₃⁺ of H₃:

H₃⁺ is Calculated based on the formula (3) by using H₂⁺. For example, after H₂⁺ and β_n2 (that is, g₂) are substituted into the formula (3), a simplified derivation process of the formula (3) is as follows:

$D2=g_2^{H*}{U}_{2}G2=\frac{D2}{1-D{2}^{*}{g}_2}{H}_3^{+}=H_2^{+}+{\left(H_2^{+*}{g}_2\right)}^{*}G2$

S411. Determine n₃ and α₃.

The for loop may be used to substitute the M−2 numbers a in S2₂ into the formula

$\left|\alpha \right|=\frac{\sqrt{P}}{r\left(H^{+},a\right)}$

one by one, to obtain [α₃, a₃]=argmin(|α|_min). n₃ that is in S1₂ and that corresponds to a₃ in S2₂ is determined as a selected antenna index.

S412. Update the weight matrix.

A weight matrix W₃=W₂+α₃H₃⁺ in the third iteration is calculated, where W₃∈C^M-2×K.

A row that is in W₃ and that corresponds to each n in S1₂ is assigned to a row that is in W_opt (W_opt∈C^M×K) and that corresponds to each n in S1₂. That is, r(W_opt,S1₂)=r(W₃,S2₂).

S413. The iteration ends.

Based on the foregoing example, a₃=2 is obtained through calculation in S411. Then, it corresponds to the antenna index 2 in S1₂. That is, n₃=2 is determined. Then, in S412, r(W_opt,S1₂)=r(W₃,S2₂) is performed, and based on a correspondence between S2₂ and S1₂, a first row in W₃ is assigned to a first row in W_opt, a second row in W₃ is assigned to a second row in W_opt, and a third row in W₃ is assigned to a fifth row in W_opt.

So far, the output W_opt can adjust transmit powers of antennas with the antenna indexes 2, 3, and 4 in the M antennas of the network device to the maximum transmit power, and ensure that transmit powers of remaining antennas with the antenna indexes 1 and 5 are less than or equal to the maximum transmit power.

FIG. 4 is a flowchart of another embodiment of a precoding method according to this application. The method includes the following operations.

Step 401: The network device performs normalization processing on transmit powers of the M antennas to obtain an original precoding weight matrix.

For example, if the network device performs normalization processing on the transmit powers of the M antennas based on a conventional PEBF algorithm, the obtained original precoding weight matrix is a PEBF weight matrix.

Step 402: The network device calculates a power utilization of the transmit powers of the M antennas based on the original precoding weight matrix.

The network device calculates, based on the PEBF weight matrix, a sum of transmit powers finally used by the M antennas, and divides the sum by a sum (MP) of maximum transmit powers allowed to be used by the M antennas, to obtain the power utilization of the transmit powers of the M antennas.

For example, the power utilization η may be calculated by using the following formula (12):

$\begin{array}{cc}\eta =\frac{\sum _{i=0}^{M-1}\sum _{j=0}^{N_{\mathrm{RB}}-1}\sum _{k=0}^{N_{j,L}-1}{W_{i,j,k,\mathrm{PEBF}}}^2}{\mathrm{MP}}& \left(12\right)\end{array}$

where W_i,j,k,PEBF represents a complex value of an i^th antenna at a k^th layer of PEBF weights of a j^th RB, N_RB represents a quantity of RBs, and N_j,L represents a layer quantity of the j^th RB.

Step 403: The network device determines whether the calculated power utilization is greater than a preset threshold THr.

Step 404: When η<THr, perform J iteration operations.

Step 405: When η≥THr, perform data weighting on the transmit powers of the M antennas by using the original weight matrix.

Based on the example, as shown in FIG. 5, algorithm modules of the network device may include a PEBF module, a power utilization judging module, a sum rate maximum algorithm (SRMA) module, and a data weighting module. The network device may first perform power normalization processing on the M antennas of the network device based on the PEBF module, to obtain a PEBF weight matrix. The PEBF module outputs the PEBF weight matrix to the power utilization judging module, and the power utilization judging module calculates and judges a power utilization. When the power utilization judging module determines that the calculated η is less than THr, a signal is output to the SRMA module, so that the SRMA module performs J iteration operations to obtain W_opt. Then, W_opt is output to the data weighting module, so that the data weighting module performs data weighting on the transmit powers of the M antennas. When the power utilization judging module determines the calculated η is greater than or equal to THr, the PEBF weight matrix is directly output to the data weighting module, so that the data weighting module performs data weighting on the transmit powers of the M antennas.

It should be noted that a power utilization of antennas, a throughput of a cell, and a spectral efficiency gain of a system can be effectively improved by using the precoding provided in this application.

For example, FIG. 6 to FIG. 8 are schematic diagrams of comparison between simulation effects of precoding performed by using the conventional PEBF algorithm and the precoding method (referred to as the SRMA algorithm below) provided in this application. FIG. 6 is a schematic diagram of comparison between simulation effects of power utilizations. It can be learned that when the conventional PEBF algorithm is used for precoding, a power utilization of the antennas is 71.6%. However, when the SRMA algorithm provided in this application is used for precoding, a power utilization of the antennas is 99.2%.

FIG. 7 is a schematic diagram of comparison between simulation effects of average cell throughputs obtained by using four user equipments (UEs), six UEs, and eight UEs as examples. It can be learned that, an average cell throughput obtained by performing precoding by using the SRMA algorithm provided in this application may be at least 20% higher than an average cell throughput obtained by performing precoding by using the conventional PEBF algorithm.

FIG. 8 is a schematic diagram of comparison between simulation effects of spectral efficiency gains. It can be learned that a spectral efficiency gain of the system after the SRMA algorithm provided in this application is used for precoding may be at least 7% higher than a spectral efficiency gain of the system after the conventional PEBF algorithm is used for precoding.

Referring to FIG. 9, an apparatus provided in this application includes a first weight calculation module 901 and a data weighting module 902.

The first weight calculation module 901 is configured to determine a generalized inverse matrix H_m⁺ of a channel value matrix H_m of K users in an m^th iteration in J iteration operations, where m=1, 2, . . . , and J, 1≤J≤M−K+1, J is an integer, and M is a quantity of antennas; select, from a set S1_m-1 based on H_m⁺, an antenna index n_m and an update coefficient α_m corresponding to n_m, where S1_m-1 is a set of n unselected in antenna indexes n of the M antennas until an (m−1)^th iteration ends; determine a weight matrix W_m based on H_m⁺ and α_m; and assign a row that is in W_m and that corresponds to n_m to a row that is in a final weight matrix W_opt and that corresponds to n_m, where W_opt is used by the data weighting module 902 to adjust transmit powers of at least J of the M antennas to a preset maximum transmit power P after the first weight calculation module 901 completes the J iteration operations.

In an embodiment, a product of H_m and W_m is a diagonal matrix or approximates to a diagonal matrix.

In an embodiment, that the first weight calculation module 901 determines a weight matrix W_m based on H_m⁺ and α_m includes: calculating W_m by using a formula W_m=W_m-1+α_mH_m⁺, where W_m-1 is a weight matrix used in the (m−1)^th iteration.

In an embodiment, when m is greater than 2, that the first weight calculation module 901 determines H_m⁺ of H_m of K users includes: calculating H_m⁺ by using a formula

$H_m^{+}={\left(H_{m-1}+{\beta }_{n_{m-1}}*{\beta }_{n_{m-1}}^H\right)}^{-1}=H_{m-1}^{+}-\frac{H_{m-1}^{+}{\beta }_{n_{m-1}}*{\beta }_{n_{m-1}}^H{H}_{m-1}^{+}}{1+{\beta }_{n_{m-1}}^H{H}_{m-1}^{+}{\beta }_{n_{m-1}}},$

where H_m-1⁺ represents a generalized inverse matrix of a channel value matrix H_m-1 of the K users in the (m−1)^th iteration, and β_nm-1 represents a column that is in H_m-1 and that corresponds to an antenna index n_m-1 selected in the (m−1)^th iteration.

In an embodiment, that the first weight calculation module 901 determines H_m⁺ of a channel value matrix H_m includes: setting, to zero, a column that is in an initial channel value matrix H of the K users and that corresponds to an antenna index n selected until the (m−1)^th iteration ends, to obtain H_m; calculating H_m⁺ based on H_m.

In an embodiment, that the first weight calculation module 901 selects, from a set S1_m-1 based on H_m, n_m and an update coefficient α_m corresponding to n_m includes: substituting M−m+1 n in S1_m-1 into a formula one by one:

$\alpha =\frac{\begin{array}{c}-\mathrm{real}\left(r\left(H_m^{+},n\right)*{r\left(W_{m-1},n\right)}^H\right)+\\ \sqrt{{\mathrm{real}\left(r\left(H_m^{+},n\right)*{r\left(W_{m-1},n\right)}^H\right)}^2-{r\left(H_m^{+},n\right)}^2\left({r\left(W_{m-1},n\right)}^2-\right)P)}\end{array}}{{r\left(H_m^{+},n\right)}^2}$ $\mathrm{and}\text{/}\mathrm{or}$ $\alpha =\frac{\begin{array}{c}-\mathrm{real}(r\left(H_m^{+},n\right)*{r\left(W_{m-1},n\right)}^H-\\ \sqrt{{\mathrm{real}\left(r\left(H_m^{+},n\right)*{r\left(W_{m-1},n\right)}^H\right)}^2-{r\left(H_m^{+},n\right)}^2\left({r\left(W_{m-1},n\right)}^2-\right)P)}\end{array}}{{r\left(H_m^{+},n\right)}^2};$

and

determining n making a smallest |α| as an antenna index n_m selected in the m^th iteration process, and determining the smallest |α| as α_m.

In an embodiment, when m=J, that the first weight calculation module 901 assigns a row that is in W_m and that corresponds to n_m to a row that is in a final weight matrix W_opt and that corresponds to n_m includes: r(W_opt,S1_M-J+1)=r(W_J,S1_M-J+1).

In an embodiment, the first weight calculation module 901 is further configured to: after the row that is in W_m and that corresponds to n_m is assigned to the (n_m)^th row in the final weight matrix W_opt, delete rows that are in W_m and H_m⁺ and that correspond to n_m; and delete a column that is in H_m and that corresponds to n_m.

In an embodiment, a set S2_m-1 is set, M−m+1 numbers a in S2_m-1 are in a one-to-one correspondence with M−m+1 n in S1_m-1, and that the first weight calculation module 901 selects, from a set S1_m-1 based on H_m, n_m and an update coefficient α_m corresponding to n_m includes:

substituting the M−m+1 numbers a in the set S2_m-1 into a formula one by one:

$\alpha =\frac{\begin{array}{c}-\mathrm{real}\left(r\left(H_m^{+},a\right)*{r\left(W_{m-1},a\right)}^H\right)+\\ \sqrt{{\mathrm{real}\left(r\left(H_m^{+},a\right)*{r\left(W_{m-1},a\right)}^H\right)}^2-{r\left(H_m^{+},a\right)}^2\left({r\left(W_{m-1},a\right)}^2-\right)P)}\end{array}}{{r\left(H_m^{+},a\right)}^2}$ $\mathrm{and}\text{/}\mathrm{or}$ $\alpha =\frac{\begin{array}{c}-\mathrm{real}\left(r\left(H_m^{+},a\right)*{r\left(W_{m-1},a\right)}^H\right)-\\ \sqrt{{\mathrm{real}\left(r\left(H_m^{+},a\right)*{r\left(W_{m-1},a\right)}^H\right)}^2-{r\left(H_m^{+},a\right)}^2\left({r\left(W_{m-1},a\right)}^2-\right)P)}\end{array}}{{r\left(H_m^{+},a\right)}^2};$

and

determining n that is in S1_m-1 and that corresponding to a making a smallest |α| as an antenna index n_m selected in the m^th iteration process, and determining the smallest |α| as α_m.

In an embodiment, when m=J, that the first weight calculation module 901 assigns a row that is in W_m and that corresponds to n_m to a row that is in a final weight matrix W_opt and that corresponds to n_m includes: r(W_opt,S1_M-J+1)=r(W_J,S2_M-J+1.)

In an embodiment, as shown in FIG. 10, the apparatus further includes a second weight calculation module 903 and a power decision module 904. The second weight calculation module 903 is configured to: before the first weight calculation module 901 performs the J iteration operations, perform normalization processing on transmit powers of the M antennas to obtain an original precoding weight matrix. The power decision module 904 is configured to calculate a power utilization of the transmit powers of the M antennas based on the original precoding weight matrix calculated by the second weight calculation module 903. The first weight calculation module 901 is configured to: when the power utilization calculated by the power decision module 904 is less than a preset threshold, perform the J iteration operations.

The apparatus may be the network device in the embodiments of this application, for example, may be the network device shown in FIG. 1. The first weight calculation module 901, the data weighting module 902, the second weight calculation module 903, and the power decision module 904 may be integrated into the processor 101 shown in FIG. 1. The memory 102 shown in FIG. 1 may store computer execution instructions of the precoding method provided in this application. The processor 101 completes the J iteration operations by executing the computer execution instructions, and uses the obtained W_opt to adjust transmit powers of the M antennas of the network device. The transceiver 103 may then be controlled to transmit a signal by using an adjusted transmit power.

Alternatively, the apparatus may be a chip in the network device in the embodiments of this application. The first weight calculation module 901, the data weighting module 902, the second weight calculation module 903, and the power decision module 904 may be integrated into a processor in the chip.

An embodiment of this application further provides a computer-readable storage medium. The methods described in the foregoing embodiments may be all or partially implemented by using software, hardware, firmware, or any combination thereof. If the methods are implemented in the software, functions used as one or more instructions or code may be stored in the computer-readable medium or transmitted on the computer-readable medium. The computer-readable medium may include a computer storage medium and a communication medium, and may further include any medium that can transfer a computer program from one place to another. The storage medium may be any available medium accessible to a computer.

In an embodiment, the computer-readable medium may include a RAM, a ROM, an EEPROM, a CD-ROM or another optical disc storage, a magnetic disk storage or another magnetic storage device, or any other medium that can be configured to carry or store required program code in a form of an instruction or a data structure and that may be accessed by the computer. In addition, any connection is appropriately referred to as a computer-readable medium. For example, if a coaxial cable, an optical fiber cable, a twisted pair, a digital subscriber line (DSL), or wireless technologies (such as infrared, radio, and microwave) are used to transmit software from a website, a server, or another remote source, the coaxial cable, the optical fiber cable, the twisted pair, the DSL, or the wireless technologies such as infrared, radio, and microwave are included in a definition of the medium. Magnetic disks and optical discs used in this specification include a compact disk (CD), a laser disk, an optical disc, a digital versatile disc (DVD), a floppy disk, and a Blu-ray disc. The magnetic disks usually magnetically reproduce data, and the optical discs optically reproduce data by using laser light. The foregoing combinations should also be included within the scope of the computer-readable medium.

An embodiment of this application further provides a computer program product. The methods described in the foregoing embodiments may be all or partially implemented by using software, hardware, firmware, or any combination thereof. When the methods are implemented in the software, the methods may be all or partially implemented in a form of the computer program product. The computer program product includes one or more computer instructions. When the foregoing computer program instructions are loaded and executed on a computer, the procedures or functions described in the foregoing method embodiments are all or partially generated. The computer may be a general-purpose computer, a special-purpose computer, a computer network, a network device, user equipment, or another programmable apparatus.

The objectives, technical solutions, and benefits of the present invention are further described in detail in the foregoing specific embodiments. It should be understood that the foregoing descriptions are merely specific embodiments of the present invention, but are not intended to limit the protection scope of the present invention. Any modification, equivalent replacement, or improvement made based on the technical solutions of the present invention shall fall within the protection scope of the present invention.

Claims

1. A precoding method, comprising: determining a generalized inverse matrix Hm+ of a channel value matrix Hm of K users in an mth iteration in J iteration operations, wherein m=1, 2, . . . , and J, 1≤J≤M−K+1, J is an integer, and M is a quantity of antennas;selecting, from a set S1m-1 based on Hm+, an antenna index nm and an update coefficient αm corresponding to nm, wherein S1m-1 is a set of columns that are unselected in antenna indexes of the M antennas until an (m−1)th iteration ends;determining a weight matrix Wm based on Hm+ and αm; andassigning a row that is in Wm and that corresponds to nm to a row that is in a final weight matrix Wopt and that corresponds to nm, wherein Wopt is to adjust transmit powers of at least J of the M antennas to a preset transmit power P after the J iteration operations end.

2. The method according to claim 1, wherein a product of Hm and Wm is a diagonal matrix.

3. The method according to claim 1, wherein the determining of the weight matrix Wm based on Hm+ and αm comprises: calculating Wm by using a formula Wm=Wm-1+αmHm+, wherein Wm-1 is a weight matrix used in the (m−1)th iteration.

4. The method according to claim 1, wherein when m is greater than 2, the determining of the generalized inverse matrix Hm+ of the channel value matrix Hm of the K users comprises: calculating Hm+ by using a formula

5. The method according to claim 1, wherein the determining of the generalized inverse matrix Hm+ of the channel value matrix Hm comprises: setting, to zero, a column that is in an initial channel value matrix H of the K users and that corresponds to an antenna index column selected until the (m−1)th iteration ends, to obtain Hm; andcalculating Hm+ based on Hm.

6. The method according to claim 1, wherein the selecting of nm and the update coefficient αm corresponding to nm comprises: substituting M−m+1 n in S1m-1 into a formula one by one:

7. The method according to claim 1, wherein when m=J, the assigning of the row that is in Wm and that corresponds to nm to the row that is in the final weight matrix Wopt and that corresponds to nm comprises: r(Wopt,S1M-J+1)=r(WJ,S1M-J+1).

8. The method according to claim 1, wherein after the row that is in Wm and that corresponds to nm is assigned to the (nm)th row in the final weight matrix Wopt, the method further comprises: deleting rows that are in Wm and Hm+ and that correspond to nm; anddeleting a column that is in Hm and that corresponds to nm.

9. The method according to claim 8, further comprising: setting a set S2m-1, wherein the set S2m-1 includes M−m+1 numbers a that are in a one-to-one correspondence with M−m+1 n in S1m-1; andwherein the selecting, from the set S1m-1 based on Hm, of nm and the update coefficient αm corresponding to nm comprises:substituting the M−m+1 numbers a in the set S2m-1 into a formula one by one:

10. The method according to claim 9, wherein when m=J, the assigning of the row that is in Wm and that corresponds to nm to the row that is in a final weight matrix Wopt and that corresponds to nm comprises: r(Wopt,S1M-J+1)=r(WJ,S2M-J+1).

11. The method according to claim 1, wherein before the J iteration operations are performed, the method further comprises: performing normalization processing on transmit powers of the M antennas to obtain an original precoding weight matrix;calculating a power utilization of the transmit powers of the M antennas based on the original precoding weight matrix; andwhen the power utilization is less than a preset threshold, performing the J iteration operations.

12. An apparatus, comprising: a processor;a memory coupled to the processor, and storing program instructions, which, when executed, cause the processor to perform operations comprising:determining a generalized inverse matrix Hm+ of a channel value matrix Hm of K users in an mth iteration in J iteration operations, wherein m=1, 2, . . . , and J, 1≤J≤M−K+1, J is an integer, and M is a quantity of antennas;selecting, from a set S1m-1 based on Hm+, an antenna index nm and an update coefficient αm corresponding to nm, wherein S1m-1 is a set of columns that are unselected in antenna indexes of the M antennas until an (m−1)th iteration ends;determining a weight matrix Wm based on Hm+ and αm; andassigning a row that is in Wm and that corresponds to nm to a row that is in a final weight matrix Wopt and that corresponds to nm, wherein Wopt is used by the data weighting module to adjust transmit powers of at least J of the M antennas to a preset transmit power P after the first weight calculation module completes the J iteration operations.

13. The apparatus according to claim 12, wherein a product of Hm and Wm is a diagonal matrix.

14. The apparatus according to claim 12, wherein the determining of the weight matrix Wm based on Hm+ and αm comprises: calculating Wm by using a formula Wm=Wm-1+αmHm+, wherein Wm-1 is a weight matrix used in the (m−1)th iteration.

15. The apparatus according to claim 12, wherein when m is greater than 2, the determining of Hm+ of Hm of the K users comprises: calculating Hm+ by using a formula

16. The apparatus according to claim 12, wherein the determining of Hm+ of a channel value matrix Hm comprises: setting, to zero, a column that is in an initial channel value matrix H of the K users and that corresponds to an antenna index n selected until the (m−1)th iteration ends, to obtain Hm; andcalculating Hm+ based on Hm.

17. The apparatus according to claim 12, wherein the selecting, of nm and the update coefficient αm corresponding to nm comprises: substituting M−m+1 n in S1m-1 into a formula one by one:

18. The apparatus according to claim 12, wherein when m=J, the assigning of the row that is in Wm and that corresponds to nm to the row that is in a final weight matrix Wopt and that corresponds to nm comprises: r(Wopt,S1M-J+1)=r(WJ,S1M-J+1).

19. The apparatus according to claim 12, the operations further comprising: after the row that is in Wm and that corresponds to nm is assigned to the (nm)th row in the final weight matrix Wopt, deleting rows that are in Wm and Hm+ and that correspond to nm; and delete a column that is in Hm and that corresponds to nm.

20. The apparatus according to claim 19, the operations further comprising: setting a set S2m-1, wherein the set S2m-1 includes M−m+1 numbers a that are in a one-to-one correspondence with M−m+1 n in S1m-1;wherein the selecting, from a set S1m-1 based on Hm, of nm and the update coefficient αm corresponding to nm comprises:substituting the M−m+1 numbers a in the set S2m-1 into a formula one by one: