CN117527017A

CN117527017A - Information transmission method, information transmission device, electronic device, and storage medium

Info

Publication number: CN117527017A
Application number: CN202311322814.2A
Authority: CN
Inventors: 李洋; 林庆丰; 刘亚锋
Original assignee: Shenzhen Research Institute of Big Data SRIBD
Current assignee: Shenzhen Research Institute of Big Data SRIBD
Priority date: 2023-10-12
Filing date: 2023-10-12
Publication date: 2024-02-06

Abstract

The application provides an information sending method, an information sending device, electronic equipment and a storage medium, which belong to the technical field of wireless communication, and the information sending method, the information sending device, the electronic equipment and the storage medium can solve the beam forming problem in a multi-group multicast scene by obtaining channel gain state information of a base station to a multicast group, performing first mapping on the channel gain state information to obtain a reference parameter for beam forming, performing second mapping on the channel gain state information according to the reference parameter to obtain a target beam forming vector of the base station to the multicast group, and sending the target information from the base station to the multicast group according to the channel gain state information, the target beam forming vector and preset noise data.

Description

Information transmission method, information transmission device, electronic device, and storage medium

Technical Field

The present invention relates to the field of wireless communications technologies, and in particular, to an information transmission method, an information transmission device, an electronic device, and a storage medium.

Background

In the related art, the performance of a wireless communication system is improved by solving the communication problem in a Multi-user, multi-packet scenario through a Multi-group multicast beamforming (Multi-Group Multicast Beamforming) technique. In a multi-user, multi-packet scenario, a transmitter needs to transmit information to multiple receivers. The receivers may be divided into different groups, with the receivers of each group only receiving information for the group in which they are located and not receiving information for other groups. This makes it necessary for the beamforming technique to effectively suppress signal interference while ensuring the quality of information transmission. How to perform beam forming in a multi-group multicast scenario becomes a problem to be solved.

Disclosure of Invention

The embodiment of the application mainly aims to provide an information sending method, an information sending device, electronic equipment and a storage medium, and aims to solve the problem of beam forming in a multi-group multicast scene, ensure the transmission quality of multicast group information and simultaneously effectively inhibit noise.

To achieve the above object, a first aspect of an embodiment of the present application provides an information sending method, including:

acquiring channel gain state information of a base station to a multicast group;

performing first mapping on the channel gain state information to obtain a reference parameter for beam forming;

performing second mapping on the channel gain state information according to the reference parameters to obtain a target beam forming vector from the base station to the multicast group;

and transmitting target information from the base station to the multicast group according to the channel gain state information, the target beam forming vector and the preset noise data.

In some embodiments, the first mapping the channel gain status information to obtain a reference parameter for beamforming includes:

performing linear mapping on the channel gain state information to obtain an initial channel embedded vector;

Performing first self-attention transformation on the initial channel embedded vector to obtain a first channel embedded vector;

performing second self-attention transformation on the first channel embedded vector to obtain a second channel embedded vector;

and linearly mapping the second channel embedded vector to obtain the reference parameter.

In some embodiments, the performing a first self-attention transformation on the initial channel embedded vector to obtain a first channel embedded vector includes:

performing multi-head self-attention calculation on the initial channel embedded vector to obtain a multi-head attention vector;

vector fusion is carried out on the multi-head attention vector and the initial channel embedded vector to obtain a first fusion vector;

normalizing the first fusion vector to obtain a normalized vector, and performing forward calculation on the normalized vector to obtain a forward vector;

vector fusion is carried out on the normalized vector and the forward vector to obtain a second fusion vector;

and normalizing the second fusion vector to obtain the first channel embedded vector.

In some embodiments, the performing second mapping on the channel gain status information according to the reference parameter to obtain a target beamforming vector from the base station to the multicast group includes:

Carrying out nonlinear mapping on the channel gain state information according to the reference parameters to obtain candidate beamforming vectors;

and carrying out gradient descent updating on the candidate beam forming vector to obtain the target beam forming vector.

In some embodiments, the reference parameters include a first reference parameter and a second reference parameter, and the nonlinear mapping is performed on the channel gain state information according to the reference parameters to obtain candidate beamforming vectors, including:

performing linear mapping on the channel gain state information according to the first reference parameter to obtain a first beam forming vector;

summing the preset identity matrix and the first beam forming vector to obtain a second beam forming vector;

performing inversion processing on the second beam forming vector to obtain a third beam forming vector;

and carrying out linear mapping on the channel gain state information according to the third beam forming vector and the second reference parameter to obtain the candidate beam forming vector.

In some embodiments, the gradient descent update of the candidate beamforming vector to obtain the target beamforming vector includes:

Gradient determination is carried out according to preset constraint violation data, so that gradient data are obtained;

and performing gradient descent update on the candidate beam forming vector according to the gradient data and a preset gradient descent step length to obtain the target beam forming vector.

In some embodiments, before the gradient determining is performed according to the preset target loss data to obtain gradient data, the information sending method further includes:

calculating the signal-to-interference-and-noise ratio of the multicast group according to the candidate beamforming vector, the channel gain state information and preset noise data;

and performing constraint violation punishment according to the signal-to-interference-and-noise ratio to obtain the preset constraint violation data.

To achieve the above object, a second aspect of the embodiments of the present application proposes an information transmission apparatus, including:

the acquisition module is used for acquiring channel gain state information of the base station to the multicast group;

the first mapping module is used for carrying out first mapping on the channel gain state information to obtain a reference parameter for beam forming;

a second mapping module, configured to perform second mapping on the channel gain status information according to the reference parameter, to obtain a target beamforming vector from the base station to the multicast group;

And the information sending module is used for sending target information from the base station to the multicast group according to the channel gain state information, the target beam forming vector and the preset noise data.

To achieve the above object, a third aspect of the embodiments of the present application proposes an electronic device, including a memory and a processor, where the memory stores a computer program, and the processor implements the information sending method according to the first aspect when executing the computer program.

To achieve the above object, a fourth aspect of the embodiments of the present application proposes a computer-readable storage medium storing a computer program that, when executed by a processor, implements the information transmission method described in the first aspect.

According to the information sending method, the information sending device, the electronic equipment and the computer readable storage medium, the channel gain state information of the base station to the multicast group is obtained through first mapping, the reference parameters for beam forming are obtained, and the calculation complexity of multi-group multicast beam forming can be reduced through the low-dimensional structure of the reference parameters due to the fact that the reference parameters are low-dimensional relative to the channel gain state information. And performing second mapping on the channel gain state information according to the reference parameters to obtain a target beam forming vector of the base station to the multicast group, so that the target beam forming vector can meet the requirements of actual communication environments, the signal phase and amplitude of each antenna of the base station can be controlled by the beam forming vector to strengthen the signal in the expected direction, and interference signals in useless directions are restrained, thereby improving the communication quality and communication efficiency. And transmitting target information from the base station to the multicast group according to the channel gain state information, the target beam forming vector and the preset noise data, so that the multicast group can receive effective information, and the quality of information transmission is ensured.

Drawings

Fig. 1 is a flowchart of an information sending method provided in an embodiment of the present application;

fig. 2 is a flowchart of step S120 in fig. 1;

fig. 3 is a schematic diagram of mapping relationship from channels to beamforming vectors according to an embodiment of the present application;

fig. 4 is a schematic diagram of mapping relationship from a channel to an intermediate parameter according to an embodiment of the present application;

fig. 5 is a flowchart of step S220 in fig. 2;

FIG. 6 is a schematic diagram of the structure of an attention module provided in an embodiment of the present application;

fig. 7 is a flowchart of step S130 in fig. 1;

fig. 8 is a flowchart of step S710 in fig. 7;

fig. 9 is a flowchart of step S720 in fig. 7;

fig. 10 is another flowchart of an information sending method provided in an embodiment of the present application;

FIG. 11 is a graph of transmit power test results for different numbers of users according to an embodiment of the present application;

FIG. 12 is a graph of constraint violation degree test results for different numbers of users according to an embodiment of the present application;

FIG. 13 is a graph of calculated time test results for different numbers of users according to an embodiment of the present application;

fig. 14 is a graph of transmit power test results for different multicast group numbers according to an embodiment of the present application;

FIG. 15 is a graph of constraint violation degree test results for different multicast group numbers according to an embodiment of the present application;

FIG. 16 is a graph of calculated time test results for different multicast group numbers according to an embodiment of the present application;

fig. 17 is a schematic structural diagram of an information transmitting apparatus provided in an embodiment of the present application;

fig. 18 is a schematic hardware structure of an electronic device according to an embodiment of the present application.

Detailed Description

In order to make the objects, technical solutions and advantages of the present application more apparent, the present application will be further described in detail with reference to the accompanying drawings and examples. It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the present application.

It should be noted that although functional block division is performed in a device diagram and a logic sequence is shown in a flowchart, in some cases, the steps shown or described may be performed in a different order than the block division in the device, or in the flowchart. The terms first, second and the like in the description and in the claims and in the above-described figures, are used for distinguishing between similar elements and not necessarily for describing a particular sequential or chronological order.

Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of the present application only and is not intended to be limiting of the present application.

With the widespread use of mobile devices, wireless communication technology plays an increasingly important role in meeting the user's demands for high quality and efficient information transmission. In the related art, the performance of the wireless communication system is improved by solving the communication problem in a multi-user and multi-packet scenario through a multi-group multicast beam forming technology. In a multi-user, multi-packet scenario, a transmitter needs to transmit information to multiple receivers. The receivers may be divided into different groups, with the receivers of each group only receiving information for the group in which they are located and not receiving information for other groups. This makes it necessary for the beamforming technique to effectively suppress signal interference while ensuring the quality of information transmission. How to perform beam forming in a multi-group multicast scenario becomes a problem to be solved.

Based on this, the embodiment of the application provides an information sending method, an information sending device, an electronic device and a computer readable storage medium, which aim to solve the problem of beam forming in a multi-group multicast scene, and can effectively inhibit noise while guaranteeing the transmission quality of multicast group information.

The information transmission method, the information transmission device, the electronic apparatus, and the computer-readable storage medium provided in the embodiments of the present application are specifically described by the following embodiments, and the information transmission method in the embodiments of the present application is first described.

The embodiment of the application provides an information sending method, which relates to the technical field of wireless communication. The information sending method provided by the embodiment of the application can be applied to the terminal, the server side and software running in the terminal or the server side. In some embodiments, the terminal may be a smart phone, tablet, notebook, desktop, etc.; the server side can be configured as an independent physical server, a server cluster or a distributed system formed by a plurality of physical servers, and a cloud server for providing cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDNs, basic cloud computing services such as big data and artificial intelligent platforms and the like; the software may be an application or the like that implements the information transmission method, but is not limited to the above form.

The subject application is operational with numerous general purpose or special purpose computer system environments or configurations. For example: personal computers, server computers, hand-held or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, distributed computing environments that include any of the above systems or devices, and the like. The application may be described in the general context of computer-executable instructions, such as program modules, being executed by a computer. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform particular tasks or implement particular abstract data types. The application may also be practiced in distributed computing environments where tasks are performed by remote processing devices that are linked through a communications network. In a distributed computing environment, program modules may be located in both local and remote computer storage media including memory storage devices.

Fig. 1 is an optional flowchart of an information sending method provided in an embodiment of the present application, where the method in fig. 1 may include, but is not limited to, steps S110 to S140.

Step S110, obtaining channel gain state information of a base station to a multicast group;

step S120, carrying out first mapping on channel gain state information to obtain a reference parameter for beam forming;

step S130, performing second mapping on the channel gain state information according to the reference parameters to obtain a target beam forming vector of the base station to the multicast group;

step S140, transmitting the target information from the base station to the multicast group according to the channel gain status information, the target beamforming vector and the preset noise data.

In the steps S110 to S140 illustrated in the embodiments of the present application, the channel gain status information of the base station to the multicast group is obtained, and the channel gain status information is mapped for the first time to obtain the reference parameter for beamforming. And performing second mapping on the channel gain state information according to the reference parameters to obtain a target beam forming vector of the base station to the multicast group, so that the signal phase and amplitude of each antenna of the base station can be controlled by the beam forming vector to strengthen the signal in the expected direction and inhibit interference signals in useless directions, thereby improving the communication quality and communication efficiency. And transmitting target information from the base station to the multicast group according to the channel gain state information, the target beam forming vector and the preset noise data, so that the multicast group can receive effective information, and the quality of information transmission is ensured.

In step S110 of some embodiments, a core goal of the multi-group multicast beamforming technique is to achieve efficient antenna resource sharing and optimization among different multicast groups to improve the performance of the overall wireless communication system. Consider a downlink multi-user system in which a base station equipped with N antennas serves M multicast groups, where the number of antennas serving each multicast group is N, where N may be greater than or equal to M, or N may be less than M. Each multicast group is composed of K _m A single antenna user, which needs to receive common messages from the base station. A single antenna user refers to a user's receiving antenna being a single antenna. Different onesThe multicast groups are disjoint, meaning that the total number of users isAcquiring channel gain state information H from base station to mth multicast group _m Channel gain state information H _m Expressed as a channel matrix, i.eh _m,k H is the channel vector from the base station to the kth user in the mth multicast group _m,k ∈C ^N C represents the complex domain and N represents the N antennas of the base station. It will be appreciated that for N transmit antennas and one receive antenna, the dimension of the channel vector is N, with the nth dimension representing the channel gain between the nth transmit antenna and the receive antenna.

Referring to fig. 2, in some embodiments, step S120 may include, but is not limited to, steps S210 to S240:

step S210, carrying out linear mapping on channel gain state information to obtain an initial channel embedding vector;

step S220, performing a first self-attention transformation on the initial channel embedded vector to obtain a first channel embedded vector;

step S230, performing a second self-attention transformation on the first channel embedded vector to obtain a second channel embedded vector;

step S240, the second channel embedded vector is linearly mapped to obtain the reference parameter.

In step S210 of some embodiments, some key issues need to be addressed when implementing a multi-group multicast beamforming technique. First, it is necessary to consider how to optimize the beams so that the signal energy is concentrated at the target receiver, reducing signal interference to other multicast groups. Secondly, it is necessary to balance the performance among the multicast groups in terms of data transmission rate, signal to noise ratio, etc., to avoid that the performance of some multicast groups is too high while the performance of other multicast groups is too low. In addition, factors such as power limitation and hardware constraint in practical application are also required to be considered, so that the multi-group multicast beam forming technology can adapt to the requirements of practical communication environments.

In the related art, the forming of multiple groups of multicast beams is performed by methods of minimum mean square error, signal-to-noise ratio maximization, signal-to-interference-and-noise ratio maximization and the like, and the methods improve the forming performance of the multiple groups of multicast beams to a certain extent. In order to further improve the performance of beam forming and solve the non-convex optimization problem in the multi-group multicast beam forming problem, approximation algorithms such as semi-positive relaxation, continuous convex approximation and the like are introduced, but the complexity of the optimization algorithms is larger. When facing a large-scale antenna system and a large-scale user group, the method can lead to large calculation amount and poor real-time performance of beam forming.

In order to solve the problems of large calculation amount and poor real-time performance of beam forming, as shown in fig. 3, the channel gain state information H is first mapped by a first mapping function f (·) to obtain a reference parameter, where the first mapping function isRepresenting the mapping from channel H to intermediate parameters (alpha, lambda), R ₊ Representing the positive real number field and C representing the complex number field. The reference parameter is an intermediate parameter variable on which the multi-group multicast beamforming depends, denoted by (α, λ). I.e., (α, λ) =f (H). The reference parameters include a first reference parameter and a second reference parameter, the first reference parameter being λ, the second reference parameter being α. Wherein, the channel gain status information h= [ H ] ₁ ,H ₂ ,...,H _M ]First reference parameter->The first reference parameter comprises real vectors of a plurality of multicast groups, and the real vector of the mth multicast group is expressed asλ _m,k Greater than 0, indicating the lagrangian multiplier for the kth user in the mth multicast group. Second reference parameter->α _m Representing the repetition of the mth multicast groupWeight vector-> α _m,k Representing the real number corresponding to the kth user in the mth multicast group. T denotes a transpose operation.

It should be noted that the mapping from channel H to the intermediate parameters (α, λ) should be independent of the index of the users and multicast groups. In particular, if the index of any user within the same multicast group is swapped, or the index of any multicast group is swapped, f (·) should produce a corresponding permutation of the original output. In other words, f (·) essentially has a permutation such as a hierarchical permutation that is a permutation of the user indices within each multicast group and a permutation of the multicast group indices.

The first mapping function consists of one embedding layer, L layering layers and one de-embedding layer. And carrying out real part and imaginary part separation on the channel gain state information H through the embedding layer to obtain a real matrix, and carrying out linear mapping on the real matrix to obtain an initial channel embedding vector. If the real matrix is expressed as H ^em The initial channel embedding vector is denoted as X ⁽⁰⁾ ，H ^em ＝[R(H) ^T ,Im(H) ^T ] ^T R (H) represents the real part of H, and Im (H) represents the imaginary part of H. H has dimensions of NxM, [ R (H) ^T ,Im(H) ^T ]Is M x 2n, h ^em Is 2N x M. Initial channel embedding vector X ⁽⁰⁾ Can be unfolded into Is the channel vector h _m,k ∈C ^N Is embedded in the memory.

The method of linear mapping is shown in formula (1).

Wherein W is ^em For weight parameters of the embedded layer, W ^em ∈R ^d×2N ；b ^em B is the bias parameter of the embedded layer ^em ∈R ^d ；W ^em And b ^em Are training parameters;representing a dot product operation; />Representing an all 1 vector in the M dimension.

In step S220 of some embodiments, each hierarchical layer includes a first attention module and a second attention module, and the network structures of the first attention module and the second attention module of each hierarchical layer are the same. As shown in FIG. 4, the first attention module of the first hierarchical layer is represented asThe second attention module of the first hierarchical layer is represented asEmbedding vector X into an initial channel by a first attention module of a first hierarchical layer ⁽⁰⁾ Performing a first self-attention transformation to obtain a first channel embedded vector Z ⁽¹⁾ 。Z ⁽¹⁾ ∈R ^d×K ，/>

In step S230 of some embodiments, the first channel embedded vector is subjected to a second self-attention transformation by a second attention module of the first hierarchical layer, the first hierarchical layer outputting a second channel embedded vector X ⁽¹⁾ 。X ⁽¹⁾ ∈R ^d×K ，

In step S240 of some embodiments, the second channel output by the last hierarchical layer is embedded into vector X ^(l-1) A first attention module input to the first hierarchical layerMake a first attention shift, will +.>Conversion to->I.e. < ->Obtaining a first channel embedded vector Z of a first layering layer ^(l) A second attention module via the first hierarchical layer->Embedding vector Z into first channel ^(l) Performing a second attention transformation, and outputting a second channel embedded vector X by the first layering layer ^(l) 。X ^(l) ∈R ^d×K ，Inputting the second channel embedded vector output by the last layering layer to the next layering layer until reaching the L-th layering layer, so that the L-th layering layer outputs the second channel embedded vector X ^(L) . Second channel embedded vector X output by de-embedding layer to the L-th hierarchical layer ^(L) And (5) performing linear mapping to obtain a reference parameter (alpha, lambda).

De-embedding layer X by linear projection ^(L) ∈R ^d×K Conversion to X ^de ∈R ^3×K The specific method of linear projection is shown in formula (2).

Wherein W is ^de For de-embedding weight parameters of the layer, W ^de ∈R ^3×d ；b ^de B for de-embedding bias parameters of layer ^de ∈R ³ ；W ^de And b ^de Are training parameters;representing a dot product operation; />Representing an all 1 vector in the M dimension.

Comprises (alpha) _m,k ,λ _m,k ) As shown in the formulas (3) to (5).

Wherein R represents a real part; im represents the imaginary part; lambda (lambda) _m,k Representing a first reference parameter corresponding to a kth user in an mth multicast group; alpha _m,k Representing a second reference parameter corresponding to a kth user in an mth multicast group;representation->Data in a first dimension; />Representation->Data in a second dimension; />Representation->Data in a third dimension; reLU is a ReLU activation function.

If in the same multicast groupAny element in->The index is exchanged, or any multicast group +.>The indices of f (·) should be swapped to produce the corresponding permutation of the original output. To guarantee the permutation etc. variability properties of the users within each multicast group +.>It needs to be shared between different multicast groups. Due to each->All are treated independently and identically, so the hierarchical structure has permutation etc. variability in multicast groups. In order to allow interaction between different multicast groups, a second attention module is used +.>To treatI.e. < ->

The above steps S210 to S240, the reference parameter is low-dimensional with respect to the channel gain status information, and by this low-dimensional structure, the computational complexity of the multi-group multicast beamforming when N is large can be reduced, and effective guidance is provided for constructing the mapping function from the wireless channel to the multi-group multicast beamforming.

Referring to fig. 5, in some embodiments, step S220 may include, but is not limited to, steps S510 to S550:

step S510, multi-head self-attention calculation is carried out on the initial channel embedded vector to obtain a multi-head attention vector;

step S520, vector fusion is carried out on the multi-head attention vector and the initial channel embedded vector to obtain a first fusion vector;

step S530, carrying out normalization processing on the first fusion vector to obtain a normalized vector, and carrying out forward computation on the normalized vector to obtain a forward vector;

step S540, carrying out vector fusion on the normalized vector and the forward vector to obtain a second fusion vector;

step S550, the second fusion vector is normalized to obtain a first channel embedded vector.

In step S510 of some embodiments, a self-attention module in a transducer is employed to processUse->Representing the mapping function of the self-attention module. As shown in fig. 6, the first attention module +.>Comprising a plurality ofA head attention module, a normalization module and a forward calculation module. And performing multi-head self-attention calculation on the initial channel embedded vector through a multi-head attention module to obtain a multi-head attention vector. The multi-head attention module comprises a first linear layer, a self-attention layer and a second linear layer, wherein the first linear layer and the second linear layer are all fully connected layers. After the initial channel embedded vector is linearly mapped through the first linear layer, multi-head self-attention calculation is carried out through the self-attention layer, and attention vectors of a plurality of heads are obtained. Vector stitching is performed on the attention vectors of the plurality of heads to obtain a stitched vector. And linearly mapping the spliced vector through a second linear layer to obtain the multi-head attention vector.

In step S520 of some embodiments, vector alignment addition is performed on the multi-headed attention vector and the initial channel embedded vector to obtain a first fusion vector.

In step S530 of some embodiments, the normalization module normalizes the first fusion vector to obtain a normalized vector. The normalization module is used for smoothing the first fusion vector to enable the first fusion vector to have relatively stable distribution. If the first fusion vector is denoted as X, the normalization process is calculated as follows: the mean μ and variance σσ of the first fusion vector X input to the normalization module are calculated, i.e., μ=mean (X), σσ=var (X). The first fusion vector X is normalized according to the mean μ and the variance σ to obtain a normalized vector x_norm= (X- μ)/sqrt (σσ+ε), where ε is a small constant for preventing errors due to division by zero.

And carrying out forward calculation on the normalized vector through a forward calculation module to obtain a forward vector. The forward computation module in the transducer includes two fully connected layers (linear layers) and an activation function. And after the normalized vectors are subjected to linear mapping through the two full-connection layers in sequence, performing activation processing on the normalized vectors subjected to linear mapping through an activation function to obtain forward vectors. Unlike traditional optimization-based methods, deep learning-based methods make multi-group multicast beamforming techniques a more practical solution to real-time implementation due to the extremely short computation time of fast feed-forward operations.

In step S540 of some embodiments, vector alignment addition is performed on the normalized vector and the forward vector to obtain a second fusion vector.

In step S550 of some embodiments, the normalization module normalizes the second fusion vector to obtain a first channel embedded vector. The normalization processing method may refer to step S530, and will not be described herein.

The network structures of the first attention module and the second attention module are the same, and the second self-attention transforming method can refer to steps S510 to S550, which are not repeated here.

Through the above steps S510 to S550, a first channel embedding vector can be obtained to establish a mapping of channels to intermediate parameters based on the first channel embedding vector.

As shown in fig. 3, the reference parameter (α, λ) and the channel gain state information H are input to a second mapping function g (·, ·, ·), and the second mapping function g (·, ·) is used to perform second mapping on the reference parameter (α, λ) and the channel gain state information H, so as to obtain an initial beamforming vector W of the base station to the multicast group. Represents the mapping from (H, α, λ) to beamforming W, i.e. w=g (H, α, λ). The second mapping function g (·, ·, ·) is the same as the first mapping function f (·), and also has a hierarchical permutation and other variability.

Referring to fig. 7, in some embodiments, step S130 may include, but is not limited to, steps S710 to S720:

step S710, carrying out nonlinear mapping on the channel gain state information according to the reference parameters to obtain candidate beam forming vectors;

step S720, gradient descent updating is performed on the candidate beamforming vector to obtain a target beamforming vector.

In step S710 of some embodiments, the second mapping function is composed of a build layer and R approximationsBeam enhancement layer composition. The architecture layer computes multiple sets of multicast beamforming by nonlinear operations. And the construction layer carries out nonlinear mapping on the channel gain state information according to the reference parameters to obtain candidate beamforming vectors. The candidate beamforming vector is denoted as W ⁽⁰⁾ ，The dimension of the beamforming vector is also N for N transmit antennas, with amplitude and phase control in each dimension of the transmit signal such that the target signal energy is concentrated in the corresponding multicast group.

In step S720 of some embodiments, the base station transmits information to different multicast groups, and the transmit power of the base station transmit antenna is minimized by using beamforming technology when the received signal-to-noise ratio of each multicast group meets the specified requirement. To minimize the total transmit power while meeting per-user signal-to-interference-and-noise ratio constraints, the multi-group multicast beamforming problem can be described as:

Wherein w is _m A beamforming vector representing the mth multicast group,is the variance of additive white gaussian noise at the kth user in the mth multicast group. />The total transmit power of the base station, which is independent of the channel, is the energy of the beamforming vector. />Signalling for the kth user in the mth multicast groupThe noise ratio, which can be expressed as SINR _m,k The method comprises the steps of carrying out a first treatment on the surface of the The superscript H denotes a conjugate transpose; gamma ray _m Representing the signal-to-interference-and-noise ratio targets of the users in the mth multicast group.

Because the signal-to-interference-and-noise ratio constraint shown in equation (7) is a non-convex constraint, the resulting multi-group multicast beamforming problem is a non-convex problem. Existing approaches to solving the non-convex problem include semi-positive relaxation and continuous convex approximation, however, these approaches require a large number of iterations to solve the problem and are iterated for fixed wireless channels. Once the wireless channel changes, these optimization-based iterative algorithms must be re-executed, resulting in significant computational complexity and delay.

W ⁽⁰⁾ The initial solution of the multi-group multicast beam forming problem at the corresponding H can be shown as formula (6) and formula (7), but it is difficult to ensure that the initial solution satisfies the non-convex constraint of formula (7). The constraint condition processing module is added in the embodiment of the application due to the lack of effective processing of the constraint condition in the traditional neural network architecture design, namely, a constraint enhancement layer based on gradient descent is added, so that the constraint condition can be better met. To better satisfy the non-convex constraint, the candidate beamforming vectors are gradient-descent updated by R constraint enhancement layers, and multiple groups of multicast beamforming matrixes are distributed from W ⁽⁰⁾ Update to W ^(R) A target beamforming vector is obtained.

The above steps S710 to S720 provide effective guidance for establishing a mapping function from (H, α, λ) to beamforming W through the construction layer, and enable the target beamforming vector to satisfy the non-convex inequality constraint through the constraint enhancement layer, thereby improving the efficiency and quality of wireless communication.

Referring to fig. 8, in some embodiments, the reference parameters include a first reference parameter and a second reference parameter, and step S710 may include, but is not limited to, steps S810 to S840:

step S810, carrying out linear mapping on channel gain state information according to a first reference parameter to obtain a first beam forming vector;

step S820, summing the preset identity matrix and the first beam forming vector to obtain a second beam forming vector;

step S830, performing inversion processing on the second beam forming vector to obtain a third beam forming vector;

in step S840, the channel gain status information is mapped linearly according to the third beamforming vector and the second reference parameter, so as to obtain a candidate beamforming vector.

In step S810 of some embodiments, unlike conventional neural network designs, which construct a single mapping from channels to beamforming, the embodiments of the present application propose a method to solve the problem of multi-group multicast beamforming design constrained by signal-to-interference-and-noise ratio, which incorporates the mathematical structure of optimal beamforming into the neural network architecture design, giving multi-group multicast beamforming structure at the construction layer. Based on the multi-group multicast beam forming structure, the mapping function from the wireless channel to the multi-group multicast beam forming is decomposed into two independent mapping functions, and the two mapping functions have the property of modification such as layering replacement and the like, so that key guidance is provided for the design of the neural network architecture. The multi-group multicast beamforming structure is shown in equation (8).

Wherein w is _m Candidate beamforming vectors for the mth multicast group; lambda (lambda) _j,k A corresponding Lagrangian multiplier is constrained for the signal-to-interference-and-noise ratio of the kth user in the jth multicast group; gamma ray _j The target signal-to-interference-and-noise ratio for the users in the j-th multicast group is a scalar, and the target signal-to-interference-and-noise ratios for all users in the same group are the same. I _N Is an N x N unit array; h is a _j,k Channel vector for kth user in jth multicast group; alpha _m Is the complex weight vector of the mth multicast group.

It should be noted that λ and α are two super parameters on which optimal beamforming depends, and therefore, a mapping relationship between the channel and the two parameters needs to be learned. Since the multi-group multicast beamforming structure is explicitly given by equation (8), the build layer does not necessarily involve any trainable parameters.

As shown in formula (8), the channel gain status information H is mapped linearly according to the first reference parameter lambda to obtain a first beamforming vector as

In step S820 of some embodiments, for the preset identity matrix I _N And a first beamforming vectorAdding to obtain a second beamforming vector of +.>

In step S830 of some embodiments, a second beamforming vector is usedPerforming inversion processing to obtain a third beam forming vector as

In step S840 of some embodiments, the third beamforming vector is multiplied by the channel gain state information, and the result of the multiplication is multiplied by the second reference parameter to obtain a candidate beamforming vector.

The above steps S810 to S840 introduce an optimal beamforming mathematical structure, and the inherent low-dimensional structure in the mathematical structure reduces the computational complexity of obtaining multiple groups of multicast beamforming when N is larger, and provides effective guidance for constructing the mapping function from wireless channels to multiple groups of multicast beamforming.

Referring to fig. 9, in some embodiments, step S720 may include, but is not limited to, steps S910 to S920:

step S910, performing gradient determination according to preset constraint violation data to obtain gradient data;

step S920, gradient descent updating is performed on the candidate beam forming vector according to the gradient data and the preset gradient descent step length, so as to obtain the target beam forming vector.

In step S910 of some embodiments, an inverse constraint violation number, i.e., a constraint violation function, is preset. To solve the non-convex inequality constraint, a constraint violation function V (H, W) is defined, the definition of which is shown in equation (9).

To guarantee the original constraint, V (H, W) is minimized over W such that V (H, W) reaches a minimum of 0. To this end, the function value of V (H, W) is reduced by performing a number of gradient descent steps by R constraint enforcement layers, each constraint enforcement layer performing one gradient descent update on the multicast beamforming matrix. In the training phase, r=r ^train Set to a relatively small value, e.g. R ^train 5 to facilitate back propagation of the gradient during training. In the test phase, r=r ^test Set to a relatively large value, and R ^test ≥R ^train To ensure feasibility of the test procedure constraints.

Specifically, for constraint violation functions, W is biased to obtain gradient data. Gradient data for the r-th constrained enhancement layer is represented asRepresenting V (H, W) at W ^(r-1) A gradient thereat.

In step S920 of some embodiments, for the constraint violation function V (H, W), the candidate beamforming vector W output for the last constraint enhancement layer ^(r-1) Obtaining the bias guide to obtain gradient data of the current constraint enhancement layerGradient data +.>And under a preset gradientAnd multiplying the step-down steps to obtain the subtraction parameter of the current constraint enhancement layer. Candidate beamforming vector W outputting the last constrained enhancement layer ^(r-1) Subtracting the subtraction parameters to obtain candidate beamforming vector W ^(r-1) And performing gradient descent update until reaching the R constraint enhancement layer, and taking the output of the R constraint enhancement layer as an initial beam forming vector. The gradient descent update method is shown in formula (10).

Wherein eta represents the gradient descent step size,representing V (H, W) at W ^(r-1) A gradient thereat.

Through the steps S910 to S920, the initial beamforming vector can better satisfy the non-convex constraint condition, and the accuracy of beamforming vector calculation is improved.

Referring to fig. 10, in some embodiments, before step S910, the information transmission method may include, but is not limited to, steps S1010 to S1020:

step S1010, calculating the signal-to-interference-and-noise ratio of the multicast group according to the candidate beam forming vector, the channel gain state information and the preset noise data;

and step S1020, performing constraint violation punishment according to the signal-to-interference-and-noise ratio to obtain preset constraint violation data.

In step S1010 of some embodiments, the candidate beamforming vector w, the channel gain state information h, and the preset noise data are usedThe signal-to-interference-plus-noise ratio of the multicast group is calculated, and the preset noise data refers to the variance of the additive white gaussian noise. The calculation method of the signal-to-interference-and-noise ratio is shown in the formula (11).

In step S1020 of some embodiments, a constraint violation penalty is performed on the signal-to-interference-and-noise ratio to obtain the preset constraint violation data as Σ _(m,k) (ReLU(γ _m -SINR _m,k )) ² 。SINR _m,k For the signal-to-interference-and-noise ratio at the kth user in the mth multicast group, gamma _m Representing the signal-to-interference-and-noise ratio targets of the users in the mth multicast group.

Through the steps S1010 to S1020, preset constraint violation data can be obtained, so as to gradient down the candidate beamforming vector based on the preset constraint violation data, and obtain an optimal beamforming vector.

In step S140 of some embodiments, additive white gaussian noise is generated according to a preset mean value, and a variance represented by the preset noise data, where the preset mean value is 0. And transmitting target information from the base station to the multicast group according to the channel gain state information, the target beamforming vector and the additive white gaussian noise, wherein the information received by the multicast group is shown in a formula (12).

Wherein y is _m,k Representing information received at a kth user in an mth multicast group; s is(s) _m Target information indicating an mth multicast group; n is n _m,k Mean zero and variance for the kth user in the mth multicast groupAdditive white gaussian noise of (c).

In some embodiments, consider a downlink multi-user system, with system parameters set as follows: the (x, y, z) coordinates (in meters) of the base station are (0,0,20), rectangular areas [85,95 ] of the user in the (x, y) plane]×[85,115]Uniformly distributed, z=0. x represents the abscissa, y represents the ordinate, and z represents the height. The large scale fading coefficient is 32.6+36.7log according to the path loss model ₁₀ D (unit: decibel) generation, wherein D is base station and userThe distance between them (unit: m), the small-scale rayleigh fading coefficients follow a complex gaussian distribution with a mean of 0 and a variance of 1. Background noise powerTarget gamma of signal to noise ratio of-100 dBm for each user _m 10dB.

The neural network parameters were set as follows: the embedding dimension d is set to 128, the layer number L of the layering layers is set to 2, the attention head number is 4, and a forward neural network in the self-attention module adopts a ReLU activation function and a double-layer fully-connected neural network with 512 hidden neurons. The forward neural network is formed by connecting four layers of a full connection layer, an active layer, a full connection layer and an active layer in series. The input to the neural network is channel H, and performance, including total transmit power and average constraint violation, is evaluated over 1280 different channel input samples.

The definition of the average constraint violation degree is shown in formula (13).

The variable optimized by the embodiment of the application is multi-group multicast beam forming, and the beam forming vector is adjusted according to the signal transmitting power and the signal-to-interference-and-noise ratio to obtain the target beam forming vector. The signal transmitting power of the transmitting antenna is minimized under the condition that the received signal-to-noise ratio of each multicast group meets the specified requirement through the beam forming technology. The penalty term for the constraint violation is added to the penalty function during training to better balance the trade-off between total transmit power and constraint violation. Specifically, energy of the beam forming vector is calculated, and the energy is used as signal transmitting power of the base station. If the beamforming vector is represented as w= [ W ] ₁ ,w ₂ ,...,w _m ,...,w _M ]The signal transmission power is expressed as

To signal transmission powerAnd carrying out weighted summation on the preset constraint violation data to obtain target loss data. The calculation method of the target loss data is shown in formula (14).

Where ρ >0 represents a penalty factor.

Minimizing target loss data constantly adjusts network parameters of the neural network.

Unlike the permutation etc. denaturation of classical convertors, both mapping functions of the embodiments of the present application essentially have the permutation etc. denaturation of layering, i.e. between users within each group, and also have the permutation etc. denaturation between different groups, and neural network architectures are designed that can guarantee this property, i.e. the permutation etc. convertors of layering proposed by the embodiments of the present application. Unlike classical transformers, hierarchical substitution and like variable transformers have more structured hierarchical substitution and like variability, which is highly desirable in multi-group multicast beamforming designs. The hierarchical permutation and the like transformation transformers provided by the embodiment of the application show quite good generalization performance under different user numbers and different multicast group numbers.

The generalization performance of the hierarchical permutation and the like transformation proposed in the examples of the present application was tested at different numbers of users (from 4 to 28) and compared with other benchmark methods. Fig. 11 to 13 are simulation results of performing the user generalization test in the group. In fig. 11 and 12, the layered replacement or the like transformation was trained with k=4, and the test was performed at different K (from 4 to 28). Since the architecture of a fully connected neural network depends on the number of users, it can be retrained at a different K. As can be seen from fig. 11, as K increases, the transmit power of the hierarchical permutation and the like variable transformers of the embodiments of the present application is always much lower than that of the fully connected neural network, but very close to the continuous convex approximation algorithm. As seen from fig. 12, the classical transducer exhibits a greater degree of constraint violation when K >4, and therefore the demonstration of classical transducer transmit power is omitted from fig. 11. In addition, fig. 12 demonstrates that hierarchical permutation alike transformation transformers are the smaller transmit power advantage obtained without sacrificing the signal-to-interference-and-noise ratio constraints. In contrast, other deep learning-based methods (fully connected neural networks and classical transducers) do not meet the signal-to-interference-and-noise ratio constraints well. To demonstrate the advantages of hierarchical permutation and the like transformation in real-time implementation, its computation time on the CPU is further illustrated in fig. 13. Since the transmitting power of the fully connected neural network point is higher, the display of the calculation time is omitted. As can be seen from fig. 13, the calculation time of the hierarchical substitution alike transformation transformers of the embodiments of the present application is much less than the calculation time of the optimization-based continuous convex approximation algorithm.

In addition to testing the generalization performance of the variant transformers such as hierarchical permutation over different numbers of users within the multicast group, the generalization performance of the variant transformers such as hierarchical permutation over different numbers of multicast groups was further tested and compared to other baseline methods. During training, the number of multicast groups is fixed to m=4. In the test procedure, the generalization performance of the variant transformers such as hierarchical permutation under different multicast group numbers M (from 1 to 6) is tested. The number of base station antennas is set to n=16, and the number of users in each multicast group is set to K _m =2. As can be seen from fig. 14, the transmit power of the hierarchical permutation alike transformation transformers in the present invention at different M is very close to that of the optimization-based continuous convex approximation. As shown in fig. 14 and 15, the classical transducer performs much worse than the layered replacement or the like in the present invention, especially when M is different from M in the training process. In fig. 15, since a large degree of constraint violation occurs in the classical transformers at m=1 and m=2, a display of the transmission power thereof is omitted in fig. 14. Furthermore, fig. 16 shows that at different M, the calculation speed of hierarchical permutation etc. transformation transformers of the embodiments of the present application is much faster than continuous convex approximation based on optimization.

Referring to fig. 17, an embodiment of the present application further provides an information sending apparatus, which may implement the above information sending method, where the apparatus includes:

an obtaining module 1710, configured to obtain channel gain status information of a base station to a multicast group;

a first mapping module 1720, configured to perform first mapping on the channel gain status information to obtain a reference parameter for beamforming;

a second mapping module 1730, configured to perform second mapping on the channel gain status information according to the reference parameter, to obtain a target beamforming vector of the base station to the multicast group;

the information sending module 1740 is configured to send target information from the base station to the multicast group according to the channel gain status information, the target beamforming vector and the preset noise data.

The specific implementation of the information sending device is basically the same as the specific embodiment of the information sending method, and will not be described herein.

The embodiment of the application also provides electronic equipment, which comprises a memory and a processor, wherein the memory stores a computer program, and the processor realizes the information sending method when executing the computer program. The electronic equipment can be any intelligent terminal including a tablet personal computer, a vehicle-mounted computer and the like.

Referring to fig. 18, fig. 18 illustrates a hardware structure of an electronic device according to another embodiment, the electronic device includes:

the processor 1810 may be implemented by a general-purpose CPU (central processing unit), a microprocessor, an application-specific integrated circuit (ApplicationSpecificIntegratedCircuit, ASIC), or one or more integrated circuits, etc. for executing related programs to implement the technical solutions provided by the embodiments of the present application;

memory 1820 may be implemented in the form of read-only memory (ReadOnlyMemory, ROM), static storage, dynamic storage, or random access memory (RandomAccessMemory, RAM), among others. The memory 1820 may store an operating system and other application programs, and when the technical solutions provided in the embodiments of the present disclosure are implemented in software or firmware, relevant program codes are stored in the memory 1820 and called by the processor 1810 to perform the information transmission method of the embodiments of the present disclosure;

an input/output interface 1830 for implementing information input and output;

the communication interface 1840 is configured to implement communication interaction between the device and other devices, and may implement communication in a wired manner (e.g., USB, network cable, etc.), or may implement communication in a wireless manner (e.g., mobile network, WIFI, bluetooth, etc.);

A bus 1850 that transfers information between the various components of the device (e.g., the processor 1810, the memory 1820, the input/output interface 1830, and the communication interface 1840);

wherein the processor 1810, the memory 1820, the input/output interface 1830, and the communication interface 1840 enable communication connection therebetween within the device via the bus 1850.

The embodiment of the application also provides a computer readable storage medium, wherein the computer readable storage medium stores a computer program, and the computer program realizes the information sending method when being executed by a processor.

The memory, as a non-transitory computer readable storage medium, may be used to store non-transitory software programs as well as non-transitory computer executable programs. In addition, the memory may include high-speed random access memory, and may also include non-transitory memory, such as at least one magnetic disk storage device, flash memory device, or other non-transitory solid state storage device. In some embodiments, the memory optionally includes memory remotely located relative to the processor, the remote memory being connectable to the processor through a network. Examples of such networks include, but are not limited to, the internet, intranets, local area networks, mobile communication networks, and combinations thereof.

According to the information sending method, the information sending device, the electronic equipment and the computer readable storage medium, the channel gain state information of the base station to the multicast group is obtained, the first mapping is carried out on the channel gain state information to obtain the reference parameters for beam forming, and the calculation complexity of multi-group multicast beam forming can be reduced through the low-dimensional structure of the reference parameters due to the fact that the reference parameters are low-dimensional relative to the channel gain state information. And performing second mapping on the channel gain state information according to the reference parameters to obtain a target beam forming vector of the base station to the multicast group, so that the signal phase and amplitude of each antenna of the base station can be controlled by the beam forming vector to strengthen the signal in the expected direction and inhibit interference signals in useless directions, thereby improving the communication quality and communication efficiency. And transmitting target information from the base station to the multicast group according to the channel gain state information, the target beam forming vector and the preset noise data, so that the multicast group can receive effective information, and the quality of information transmission is ensured.

The embodiments described in the embodiments of the present application are for more clearly describing the technical solutions of the embodiments of the present application, and do not constitute a limitation on the technical solutions provided by the embodiments of the present application, and as those skilled in the art can know that, with the evolution of technology and the appearance of new application scenarios, the technical solutions provided by the embodiments of the present application are equally applicable to similar technical problems.

It will be appreciated by those skilled in the art that the technical solutions shown in the figures do not constitute limitations of the embodiments of the present application, and may include more or fewer steps than shown, or may combine certain steps, or different steps.

The above described apparatus embodiments are merely illustrative, wherein the units illustrated as separate components may or may not be physically separate, i.e. may be located in one place, or may be distributed over a plurality of network elements. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

Those of ordinary skill in the art will appreciate that all or some of the steps of the methods, systems, functional modules/units in the devices disclosed above may be implemented as software, firmware, hardware, and suitable combinations thereof.

The terms "first," "second," "third," "fourth," and the like in the description of the present application and in the above-described figures, if any, are used for distinguishing between similar objects and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used may be interchanged where appropriate such that embodiments of the present application described herein may be implemented in sequences other than those illustrated or otherwise described herein. Furthermore, the terms "comprises," "comprising," and "having," and any variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed but may include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus.

It should be understood that in this application, "at least one" means one or more, and "a plurality" means two or more. "and/or" for describing the association relationship of the association object, the representation may have three relationships, for example, "a and/or B" may represent: only a, only B and both a and B are present, wherein a, B may be singular or plural. The character "/" generally indicates that the context-dependent object is an "or" relationship. "at least one of" or the like means any combination of these items, including any combination of single item(s) or plural items(s). For example, at least one (one) of a, b or c may represent: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", wherein a, b, c may be single or plural.

In the several embodiments provided in this application, it should be understood that the disclosed apparatus and method may be implemented in other ways. For example, the above-described apparatus embodiments are merely illustrative, and for example, the above-described division of units is merely a logical function division, and there may be another division manner in actual implementation, for example, a plurality of units or components may be combined or may be integrated into another system, or some features may be omitted, or not performed. Alternatively, the coupling or direct coupling or communication connection shown or discussed with each other may be an indirect coupling or communication connection via some interfaces, devices or units, which may be in electrical, mechanical or other form.

The units described above as separate components may or may not be physically separate, and components shown as units may or may not be physical units, may be located in one place, or may be distributed over a plurality of network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

In addition, each functional unit in each embodiment of the present application may be integrated in one processing unit, or each unit may exist alone physically, or two or more units may be integrated in one unit. The integrated units may be implemented in hardware or in software functional units.

The integrated units, if implemented in the form of software functional units and sold or used as stand-alone products, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present application may be embodied essentially or in part or all of the technical solution or in part in the form of a software product stored in a storage medium, including multiple instructions to cause a computer device (which may be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods of the various embodiments of the present application. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a random access Memory (Random Access Memory, RAM), a magnetic disk, or an optical disk, or other various media capable of storing a program.

Preferred embodiments of the present application are described above with reference to the accompanying drawings, and thus do not limit the scope of the claims of the embodiments of the present application. Any modifications, equivalent substitutions and improvements made by those skilled in the art without departing from the scope and spirit of the embodiments of the present application shall fall within the scope of the claims of the embodiments of the present application.

Claims

1. An information transmission method, characterized in that the method comprises:

2. The information transmission method according to claim 1, wherein the first mapping the channel gain status information to obtain a reference parameter for beamforming includes:

3. The information transmission method according to claim 2, wherein the performing a first self-attention transformation on the initial channel embedded vector to obtain a first channel embedded vector includes:

4. The method for transmitting information according to any one of claims 1 to 3, wherein said performing second mapping on the channel gain status information according to the reference parameter to obtain a target beamforming vector from the base station to the multicast group includes:

5. The information transmission method according to claim 4, wherein the reference parameters include a first reference parameter and a second reference parameter, the nonlinear mapping is performed on the channel gain state information according to the reference parameters to obtain candidate beamforming vectors, and the method includes:

6. The method of information transmission according to claim 4, wherein the performing gradient descent update on the candidate beamforming vector to obtain the target beamforming vector includes:

7. The information transmission method according to claim 6, wherein before the gradient determination is performed based on the preset constraint violation data to obtain gradient data, the information transmission method further comprises:

8. An information transmission apparatus, characterized in that the apparatus comprises:

9. An electronic device, characterized in that the electronic device comprises a memory storing a computer program and a processor implementing the information transmission method according to any one of claims 1 to 7 when executing the computer program.

10. A computer-readable storage medium storing a computer program, characterized in that the computer program, when executed by a processor, implements the information transmission method of any one of claims 1 to 7.