CN118590107B - Terminal direct connection and non-cellular heterogeneous network access mode selection method - Google Patents

Abstract

The present invention provides a method for selecting a terminal direct and cellular-free heterogeneous network access mode, comprising the following steps: adopting a time division duplex operation mode, utilizing uplink and downlink channel reciprocity, and performing heterogeneous network system channel estimation based on terminal direct and cellular-free heterogeneous networks through pilot information sent in an uplink; precoding transmission symbols in a downlink data transmission phase through a maximum ratio precoding scheme based on a channel estimation value, and then using a conjugate beamforming method to send signals to users; modeling the problem of maximizing downlink communication and rate, and using a user communication mode selection algorithm based on a graph neural network to solve it, thereby obtaining a user communication mode with the highest communication rate, realizing terminal direct and cellular-free heterogeneous network access mode selection, and realizing an improvement in the downlink communication rate.

Description

A method for selecting terminal direct and non-cellular heterogeneous network access modes

Technical Field

The invention relates to the technical field of wireless communications and proposes a method for selecting a terminal direct-through and non-cellular heterogeneous network access mode.

Background Art

In the traditional Multiple-Input Multiple-Output (MIMO) system, a large number of antennas are concentrated in the base station, and the user terminals are distributed around the base station, which has the advantages of low data sharing overhead and front-end transmission requirements. However, the distributed MIMO system can provide uniform and good service to all user terminals by taking advantage of the independent fading of the signal, thereby achieving the purpose of obtaining high diversity gain against shadow fading. The cellless network is a deconstruction of the traditional MIMO system. A large number of antennas are distributed in different locations over a wide area, and the users are also distributed over this wide area. These antennas are called access points. In theory, each user can communicate with each access point. By relying on time division duplex operation, with the help of the fronthaul network and central processing unit operating in the same time-frequency resources, a large number of geographically dispersed antennas serve a small number of user terminals together. The central processing unit sends downlink data and power control coefficients to the access point, and the access point feeds back the data received from the user terminal in the uplink to the central processing unit through the fronthaul link. All access points are connected to the central processor through the backhaul link for phase-coherent collaboration, serving all users simultaneously on the same time-frequency resources. At the same time, D2D communication, as a way of direct communication between devices, has attracted widespread attention. Through D2D communication, devices can exchange information directly without going through base stations, thereby improving communication efficiency, reducing network congestion, and providing possibilities for new application scenarios. Combining D2D with non-cellular networks can further optimize the utilization of network resources and improve communication quality. However, the simultaneous operation of non-cellular networks and D2D links on the same carrier frequency will cause mutual interference and performance degradation, so it is necessary to select an appropriate user communication mode to increase the communication rate of the system.

Deep learning-based methods are widely used to solve problems in the field of mobile communications and always achieve the desired results. Currently, optimization methods are usually used to solve the user mode selection problem, which will increase the demand for computing resources when processing large-scale data. Deep learning models may be more suitable for processing large-scale data due to their automatic feature learning and end-to-end learning characteristics. Mobile communication networks usually have complex spatial structures and topological relationships, such as connections between base stations and relationships between users. Graph neural networks can effectively capture these spatial structures and topological information, and better understand the interactive relationships between nodes in the network, making graph neural networks perform well in solving resource allocation problems in mobile communications.

In summary, utilizing graph neural networks for the selection of user communication patterns will produce the desired results.

Summary of the invention

According to the technical problems raised above, a method for selecting terminal direct and non-cellular heterogeneous network access modes is provided.

The technical means adopted by the present invention are as follows: a method for selecting a terminal direct and non-cellular heterogeneous network access mode, comprising the following steps:

S1. Using the time division duplex operation mode and utilizing the reciprocity of uplink and downlink channels, the channel estimation of the heterogeneous network system based on terminal direct access and non-cellular heterogeneous network is performed through the pilot information sent in the uplink;

S2, precoding the transmission symbols in the downlink data transmission phase through a maximum ratio precoding scheme based on the channel estimation value, and then sending the signal to the user using a conjugate beamforming method;

S3. Model the problem of maximizing downlink communication and rate, and use a user communication mode selection algorithm based on graph neural network to solve it, obtain the user communication mode with the highest communication rate, and realize terminal direct and cellular-free heterogeneous network access mode selection.

Furthermore, the specific process of adopting the time division duplex operation mode, utilizing the reciprocity of uplink and downlink channels, and performing channel estimation of a heterogeneous network system based on terminal direct access and non-cellular heterogeneous networks through pilot information sent in the uplink is as follows:

S11. A heterogeneous network system based on terminal direct communication and non-cellular heterogeneous networks, with a total of M multi-antenna access points, K single-antenna users and I transmitters that can establish terminal direct communication with the users, each access point is connected to the central processor via a backhaul link, and the M access points and I transmitters serve the K users under the same time and frequency resources;

In the terminal direct and non-cellular heterogeneous network system, the time division duplex operation mode is adopted, and the channel reciprocity is used to perform channel estimation to all users at each access point and transmitter. The obtained channel state information is used for uplink data transmission decoding and downlink data transmission encoding;

The channel coefficient between the kth user and the mth access point is expressed as express:

The channel coefficient between the kth user and the i-th access point is expressed as express:

Where, m = 1,…,M, k = 1,…,K, i = 1 ,…, I, is the large-scale fading coefficient between access point m and user k, The large-scale fading coefficient between transmitter i and user k mainly reflects the impact of path loss and shadow fading on the channel. and They represent the small-scale fading coefficients between access point m and user k and between transmitter i and user k, respectively. Each small-scale fading coefficient and are all independent and identically distributed, represents a complex Gaussian random variable with mean 0 and variance 1, are independent and identically distributed random variables, whose distribution conforms to , N is the number of antennas at the access point; the large-scale fading coefficient is calculated by path loss and uncorrelated log-normal shadowing To model:

represents the path loss, With standard deviation and The shadow fading of It is represented by:

in: , is the carrier frequency, is the antenna height of the access point, is the user's antenna height, is the distance from the mth access point to the kth user, and is the reference distance; a two-component model is used to calculate the shadow fading coefficient :

in, , , are two independent random variables, is a parameter, , and The covariance function of is:

in, It is access point and The distance between access points, It is Users and The distance between users, is the correlation distance, The modeling method and same;

By channel conditions and , we get the pilot information sent by user k received by the mth access point in the uplink and the i-th transmitter receives the pilot information sent by user k in the uplink :

Different orthogonal pilot information is assigned to the access point and the transmitter to eliminate coupling, where is the uplink pilot transmission duration, and is the pilot sequence used by the kth user to send to the access point and transmitter, where and is the random variable of user k, , , Representation in the complex domain dimensional vector, is the Euclidean norm, and are the normalized signal-to-noise ratios when the user sends the pilot signal to the access point and the transmitter, respectively. and is the additional noise at the mth access point and the i-th transmitter;

Based on the received pilot sequence, the mth access point and the i-th transmitter perform channel estimation. exist Projection on and exist Projection for:

in: and They are and The conjugate transpose of Indicates users, where k and are all included in the user set K, For users A random variable, , ;

S12, according to the minimum mean square error criterion, the channel coefficient and Estimated to be:

in, Indicates the distance from the mth access point to the The large-scale fading coefficient between users, Indicates the transmission from the i-th transmitter to the The large-scale fading coefficient between users, represents the mean, Indicates conjugation.

Furthermore, the specific process of precoding the transmission symbols in the downlink data transmission phase by using the maximum ratio precoding scheme based on the channel estimation value and then sending the signal to the user by using the conjugate beamforming method is as follows:

S21. In the downlink data transmission phase, the access point m and the transmitter i use beamforming to precode the data to be transmitted to the user k according to the channel estimation result. The signal sent by the mth access point is:

The signal sent by the i-th transmitter is:

in, is a binary variable for establishing a terminal-to-terminal D2D communication link. , is the symbol sent to the kth user, and , , and is the normalized downlink signal-to-noise ratio, is the power control coefficient of the downlink from the mth access point to the kth user, is the power control coefficient of the downlink from the ith transmitter to the kth user; and The conjugate beam technology is used. In the signal transmission part, and represents the conjugate form of the channel estimate;

The selection of the power control coefficient needs to satisfy the power constraint of each access point:

Also expressed as: , , , Expressed as the channel coefficient estimate and The mean square of only considers the user access mode selection, so the transmitter uses the maximum transmission power, the access point uses full power transmission, and for the mth access point, the transmission power for each user is the same;

Signal compression is required when transmitting signals through the fronthaul link, which in turn causes compression distortion noise accompanying the transmitted signal. Defined as:

in: and It is independent of the noise, indicating additive distortion noise, which obeys a complex Gaussian distribution with a mean of 0 and a variance of Q, and a capacity of The amount of information to be transmitted in the forward link is expressed as a function of differential entropy:

,

in: express and The mutual information between Represents the differential entropy function, from which the quantization noise power is obtained :

,

In the downlink data transmission phase in the heterogeneous system of cellular-free and D2D communication, all access points send data signals to users on the same time-frequency resources at the same time;

S22, the signal received by the user without cellular network is :

The signal received by the D2D user is:

in, and are the additive noise of the kth non-cellular network user and the kth D2D user respectively;

Assuming that each user knows the channel statistics, the received signal Written as:

here,

;

in, represents the strength of the desired signal of the kth user, represents the uncertainty in the beamforming gain, Indicates from No interference from cellular network users, represents the interference from i transmitters, represents the quantization noise distortion power;

Downlink communication rate of the kth user without cellular network The expression is:

in, refers to the achievable rate of the kth user in the downlink, Expressed as the channel coefficient estimate The mean square of

Similarly, the signal-to-interference-to-noise ratio expression of the kth D2D user downlink is derived as:

Further, the downlink communication rate of the kth D2D user is obtained .

Furthermore, the problem of maximizing downlink communication and rate is modeled, and a user communication mode selection algorithm based on a graph neural network is used to solve it, so as to obtain the user communication mode with the highest communication rate, and realize terminal direct connection and non-cellular heterogeneous network access mode selection as follows:

S31. Select the user's communication mode to maximize the user's communication and rate, as follows:

The constrained optimization problem of maximizing the communication rate is modeled as follows:

Among them, C1 and C2 are user association constraints, that is, each user can only form D2D communication with at most one transmitter;

The user, access point and transmitter are regarded as three types of nodes in the heterogeneous graph, where the user and the access point are fully connected, and the user and the transmitter are fully connected. The graph of the graph optimization model is Indicates that, Represents a collection of nodes. , and denote users, access points and transmitter nodes respectively; Represents the set of adjacent nodes that form the edge, and x and y represent the set The nodes in Represents node characteristics, , Represents edge features, ,in, Represented as a complex field, and Represented as the dimension size of node features and edge features;

Define node feature matrix ,in , represented as the i-th row of the node feature matrix Z, that is, the node feature of the i-th node, Represents the i-th node, and sets the node features to a tensor of all 1s. Adjacent feature tensor ,in , represented as the feature of the edge between node i and node j, where , Represented as the edge between node i and node j, using the large-scale fading coefficient and As a feature of the edge;

S32. Define the loss function as the negative value of the objective function :

Graph neural networks extend convolutional neural networks to graphs. In a graph neural network layer, each node updates its hidden state based on aggregated information from neighboring nodes.

S33, heterogeneous neural networks are used in non-cellular network systems. Each layer of the neural network contains two types of message transmission, namely, messages transmitted from the access point to the user, and messages transmitted from the user to the access point. Therefore, different weight matrices are used to parameterize different message transmission processes; the feature of node m is Initialize to an empty vector ,in is an empty vector in the real field;

For a T-layer graph neural network, the update of node k in layer t as follows:

in, represents the aggregation information of node k in the t-th layer of the neural network, represents the neighboring nodes of node k, is the pooling function, is the aggregate neural network at layer t, are the learnable weights in the t-th layer of the neural network, is the node feature of node k in the Tth layer of neurons, is the pooling function. The pooling function here chooses the average, that is, the average of the aggregated features. is a combinatorial function, is a combination of neural networks, is the node feature of node k in the t-1th layer of the graph neural network, is a multi-layer perceptron that maps the final hidden state to the transmit power.

Compared with the prior art, the present invention has the following advantages:

The present invention provides a terminal direct and cellular-free heterogeneous network access mode selection method, which can precode downlink data transmission according to uplink channel estimation, optimize downlink transmission power through a graph neural network, and effectively improve the communication of the system.

Based on the above reasons, the present invention can be widely promoted in fields such as wireless communication.

BRIEF DESCRIPTION OF THE DRAWINGS

In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

Fig. 1 is a flow chart of the method of the present invention;

FIG2 is a scene diagram used in an embodiment of the present invention;

FIG3 is a graph optimization model diagram provided by an embodiment of the present invention;

FIG4 is a flowchart of a neural network provided by an embodiment of the present invention;

FIG5 is a graph showing the results of a test set of a trained neural network provided by an embodiment of the present invention;

Figure 6 is a comparison chart between the user access mode selection method based on graph neural network and the multi-layer perceptron method.

DETAILED DESCRIPTION

In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

FIG. 1 is a flow chart of the method of the present invention.

As shown in FIG1 , an embodiment of the present invention provides a method for selecting a terminal direct access mode and a non-cellular heterogeneous network access mode. The transmitter searches for devices around it through a device search to find devices with which D2D communication can be established. The user sends pilot information to the access point and the transmitter at the same time, performs channel estimation at the access point and the transmitter, and adopts a graph representation learning method to select a user communication mode that makes the system communication rate the highest, that is, the user chooses to access a non-cellular network or D2D communication, including the following steps:

S1. Using the time division duplex operation mode and utilizing the reciprocity of uplink and downlink channels, the channel estimation of the heterogeneous network system based on terminal direct access and non-cellular heterogeneous network is performed through the pilot information sent in the uplink;

S2, precoding the transmission symbols in the downlink data transmission phase through a maximum ratio precoding scheme based on the channel estimation value, and then sending the signal to the user using a conjugate beamforming method;

S3. Model the problem of maximizing downlink communication and rate, and use a user communication mode selection algorithm based on graph neural network to solve it, obtain the user communication mode with the highest communication rate, and realize terminal direct and cellular-free heterogeneous network access mode selection.

The time division duplex operation mode is adopted, the reciprocity of uplink and downlink channels is utilized, and the channel estimation of the heterogeneous network system based on terminal direct connection and non-cellular heterogeneous network is performed through the pilot information sent by the uplink. Specifically:

S11, FIG2 is a scene diagram used in an embodiment of the present invention, and FIG3 is a graph optimization model diagram provided in an embodiment of the present invention. In a heterogeneous network system based on terminal direct communication and non-cellular heterogeneous networks, there are M multi-antenna access points, K single-antenna users and I transmitters that can establish D2D communication with the users. Each access point is connected to the central processor through a backhaul link. The M access points and I transmitters serve K users under the same time and frequency resources.

The transmitter first sends discovery information to the surrounding user equipment. After receiving the information, the user sends a pilot sequence to the transmitter. At the same time, the user sends a pilot sequence to all access points for uplink pilot training. In D2D and non-cellular heterogeneous network systems, time division duplex operation mode is adopted. Channel reciprocity is used to estimate the channel to all users at each access point and transmitter. The acquired channel state information is used for uplink data transmission decoding and downlink data transmission encoding:

The channel coefficient between the kth user and the mth access point is expressed as express:

The channel coefficient between the kth user and the i-th access point is expressed as express:

Where, m = 1, …, M, k = 1, …, K, i = 1, …, I, is the large-scale fading coefficient between access point m and user k, The large-scale fading coefficient between transmitter i and user k mainly reflects the impact of path loss and shadow fading on the channel. and They represent the small-scale fading coefficients between access point m and user k and between transmitter i and user k, respectively. Each small-scale fading coefficient and are all independent and identically distributed, represents a complex Gaussian random variable with mean 0 and variance 1, are independent and identically distributed random variables, whose distribution conforms to , N is the number of antennas at the access point; the large-scale fading coefficient is calculated by path loss and uncorrelated log-normal shadowing To model:

represents the path loss, With standard deviation and The shadow fading of It can be expressed as follows:

in: , is the carrier frequency, is the antenna height of the access point, is the user's antenna height, is the distance from the mth access point to the kth user, and For reference distance, in the real world, adjacent transmitters and receivers may be surrounded by common obstacles. Therefore, shadow fading is interrelated. We use a model with two components to calculate the shadow fading coefficient:

in, , , are two independent random variables, , is a parameter. and The covariance function of is:

in, It is access point and The distance between access points, It is Users and The distance between users, is the relevant distance, which depends on the environment and is generally between 20m and 200m. The modeling method and same.

By channel conditions and , we get the pilot information sent by user k received by the mth access point in the uplink and the i-th transmitter receives the pilot information sent by user k in the uplink :

In order to improve the quality of channel estimation, different orthogonal pilot information is allocated to the access point and the transmitter to eliminate coupling, where: is the uplink pilot transmission duration, and is the pilot sequence used by the kth user to send to the access point and transmitter, where and is the random variable of user k, , , Representation in the complex domain dimensional vector, is the Euclidean norm, and are the normalized signal-to-noise ratios when the user sends the pilot signal to the access point and the transmitter, respectively. and is the additional noise at the mth access point and the i-th transmitter;

Based on the received pilot sequence, the mth access point and the i-th transmitter perform channel estimation. exist Projection on and exist Projection for:

in: and They are and The conjugate transpose of Indicates users, where k and are all included in the user set K, For users A random variable, , .

S12, according to the minimum mean square error criterion, the channel coefficient and Estimated to be:

in, Indicates the distance from the mth access point to the The large-scale fading coefficient between users, Indicates the transmission from the i-th transmitter to the The large-scale fading coefficient between users, represents the mean, Indicates conjugation.

The specific process of precoding the transmission symbols in the downlink data transmission phase by using the maximum ratio precoding scheme based on the channel estimation value and then sending the signal to the user by using the conjugate beamforming method is as follows:

S21. In the downlink data transmission phase, the access point m and the transmitter i use beamforming to precode the data to be transmitted to the user k according to the channel estimation result. The signal sent by the mth access point is:

The signal sent by the i-th transmitter is:

in, is a binary variable for establishing a D2D communication link, , is the symbol sent to the kth user, and , , and is the normalized downlink signal-to-noise ratio, is the power control coefficient for the downlink from the mth access point to the kth user, is the power control coefficient for the downlink from the ith transmitter to the kth user ; and The conjugate beam technology is used. In the signal transmission part, and represents the conjugate form of the channel estimate;

The selection of the power control coefficient needs to satisfy the power constraint of each access point:

Also expressed as: , , , Expressed as the channel coefficient estimate and The mean square of only considers the user access mode selection, so the transmitter uses the maximum transmission power and the access point uses full power transmission. For the mth access point, the transmission power for each user is the same.

For non-cellular networks, the front-end capacity limitation needs to be considered, and the number of bits for transmitting signals between the access point and the central processor is limited. Therefore, signal compression is required when transmitting signals through the fronthaul link, which in turn leads to compression distortion noise accompanying the transmitted signal. The compressed signal can be defined as:

in and It is independent of the noise, indicating additive distortion noise, which obeys a complex Gaussian distribution with a mean of 0 and a variance of Q, and a capacity of The amount of information to be transmitted in the forward link can be expressed as a function of differential entropy:

.

in express and The mutual information between represents the differential entropy function. From this, we can get the quantization noise power :

.

In the downlink data transmission phase in the heterogeneous system of cellular-free and D2D communication, all access points send data signals to users on the same time-frequency resources at the same time.

S22. The signal received by users without cellular network is:

The signal received by the D2D user is:

in, and are the additive noise of the kth non-cellular network user and the kth D2D user respectively.

Assuming that each user knows the channel statistics, the received signal Written as:

here,

.

in, represents the strength of the desired signal of the kth user, represents the uncertainty in the beamforming gain, Indicates from No interference from cellular network users, represents the interference from i transmitters, represents the quantization noise distortion power;

The downlink communication rate expression of the kth user without cellular network is:

in, refers to the achievable rate of the kth user in the downlink, Expressed as the channel coefficient estimate The mean square of .

Similarly, the signal-to-interference-to-noise ratio expression of the kth D2D user downlink can be derived as:

The downlink communication rate of the kth D2D user can be further obtained:

The specific process of modeling the problem of maximizing downlink communication and rate, solving it by adopting a user communication mode selection algorithm based on graph neural network, obtaining the user communication mode with the highest communication rate, and realizing the selection of terminal direct connection and non-cellular heterogeneous network access mode is as follows:

S31. Select the user's communication mode to maximize the user's communication and rate, as follows:

The constrained optimization problem of maximizing the communication rate is modeled as follows:

Among them, C1 and C2 are user association constraints, that is, each user can only form D2D communication with at most one transmitter;

The heterogeneous network system model based on terminal direct access and non-cellular heterogeneous network is constructed as a heterogeneous graph, and users, access points and transmitters are regarded as three types of nodes in the heterogeneous graph, in which users and access points are fully connected, and users and transmitters are fully connected. The graph of the graph optimization model is composed of Indicates that, Represents a collection of nodes. , and denote users, access points and transmitter nodes respectively. Represents the set of adjacent nodes that form the edge, and x and y represent the set Nodes in . Represents node characteristics, , Represents edge features, ,in, Represented as a complex field, and Represented as the dimension size of node features and edge features;

Define node feature matrix ,in , represented as the i-th row of the node feature matrix Z, that is, the node feature of the i-th node, Represents the i-th node, and sets the node features to a tensor of all 1s. Adjacent feature tensor ,in , represented as the feature of the edge between node i and node j, where , Represented as the edge between node i and node j, using the large-scale fading coefficient and As a feature of the edge;

S32. Define the loss function as the negative value of the objective function:

Graph neural networks extend convolutional neural networks to graphs. In a graph neural network layer, each node updates its hidden state based on aggregated information from neighboring nodes.

S33, heterogeneous neural networks are used in non-cellular network systems. Each layer of the neural network contains two types of message transmission, namely, messages transmitted from the access point to the user, and messages transmitted from the user to the access point. Therefore, different weight matrices are used to parameterize different message transmission processes; the feature of node m is Initialize to an empty vector ,in is an empty vector in the real field;

FIG4 is a flowchart of a neural network provided by an embodiment of the present invention;

For a T-layer graph neural network, the update of node k in layer t as follows:

in, represents the aggregation information of node k in the t-th layer of the neural network, represents the neighboring nodes of node k, is the pooling function, is the aggregate neural network at layer t, are the learnable weights in the t-th layer of the neural network, is the node feature of node k in the Tth layer of neurons, is the pooling function. The pooling function here chooses the average, that is, the average of the aggregated features. is a combinatorial function, is a combination of neural networks, is the node feature of node k in the t-1th layer of the graph neural network, is a multi-layer perceptron that maps the final hidden state to the transmit power.

Simulation conditions

In the simulation scenario, the access points and users are randomly distributed in a rectangular area of 1km*1km, and the normalized signal-to-noise ratio of each pilot is , normalized downlink signal-to-noise ratio , , the number of access point antennas N = 2, the front-end capacity .

Simulation content and result analysis

Simulation 1: The graph neural network is trained with the training set and verified with test set data of different sizes to reflect the generalization performance of the graph neural network model.

As shown in Figure 5, the graph data constructed by the system model shown in Figure 2 is input into the neural network for training. The number of access points, the number of users, and the number of transmitters selected in the training set are 20, 20, and 10. The number of users and transmitters in the test set are consistent with the training set, and the number of access points is selected from 20 to 80. After the test set is passed through the network trained by the training set, it can still show good performance, which reflects the generalization performance of the graph neural network.

Simulation 2: The method proposed in this application is compared with the multi-layer perceptron method. As shown in Figure 6, the proposed user communication mode selection method based on graph neural network performs better than the multi-layer perceptron, verifying the effectiveness of the method of the present invention.

Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein by equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims (1)

1. A method for selecting a terminal direct access and non-cellular heterogeneous network access mode, characterized in that it comprises the following steps: S1. Using the time division duplex operation mode and utilizing the reciprocity of uplink and downlink channels, the channel estimation of the heterogeneous network system based on terminal direct access and non-cellular heterogeneous network is performed through the pilot information sent in the uplink; S2, precoding the transmission symbols in the downlink data transmission phase through a maximum ratio precoding scheme based on the channel estimation value, and then sending the signal to the user using a conjugate beamforming method; S3. Model the problem of maximizing downlink communication and rate, and use a user communication mode selection algorithm based on graph neural network to solve it, obtain the user communication mode with the highest communication rate, and realize terminal direct and non-cellular heterogeneous network access mode selection; The specific process of precoding the transmission symbols in the downlink data transmission phase by using the maximum ratio precoding scheme based on the channel estimation value and then sending the signal to the user by using the conjugate beamforming method is as follows: S21. In the downlink data transmission phase, the access point M and the transmitter I use beamforming to precode the data to be transmitted to the user K according to the channel estimation result. The signal sent by the Mth access point is: The signal sent by the first transmitter is: Wherein, α _i,k is a binary variable for establishing a terminal-to-terminal D2D communication link, α _i,k = {0,1}, s _mk is a symbol sent to the Kth user, and ρ ^c and ρ ^d are the normalized downlink signal-to-noise ratios, η _m,k is the downlink power control coefficient from the Mth access point to the Kth user, ξ _i,k is the downlink power control coefficient from the Ith transmitter to the Kth user; and The conjugate beam technology is used. In the signal transmission part, and represents the conjugate form of the channel estimate; The selection of the power control coefficient needs to satisfy the power constraint of each access point: Also expressed as: Expressed as the channel coefficient estimate and The mean square of only considers the user access mode selection, so the transmitter uses the maximum transmission power, the access point uses full power transmission, and for the Mth access point, the transmission power for each user is the same; Signal compression is required when transmitting signals through the fronthaul link, which in turn causes compression distortion noise accompanying the transmitted signal. Defined as: in: Independent of _smk , it represents additive distortion noise, obeys a complex Gaussian distribution with a mean of 0 and a variance of Q. The amount of information to be transmitted in the fronthaul link with a capacity of _Cm is expressed as a function of differential entropy: in: express The mutual information between and s _mk , h(·) represents the differential entropy function, from which the quantization noise power Q _m is obtained: In the downlink data transmission phase in the heterogeneous system of cellular-free and D2D communication, all access points send data signals to users on the same time-frequency resources at the same time; S22, the signal received by the user without cellular network is The signal received by the D2D user is: in, and are the additive noise of the Kth non-cellular network user and the Kth D2D user respectively; Assuming that each user knows the channel statistics, the received signal Written as: here, Where DS _k represents the strength of the desired signal of the Kth user, BU _k represents the uncertainty of the beamforming gain, CUI _k' represents the interference from k' non-cellular network users, DUI _i represents the interference from I transmitters, and Q _m,k represents the quantization noise distortion power; Downlink communication rate of the Kth user without cellular network The expression is: Similarly, the signal-to-interference-to-noise ratio expression of the Kth D2D user downlink is derived as: Further, the downlink communication rate of the Kth D2D user is obtained The specific process of adopting the time division duplex operation mode, utilizing the reciprocity of uplink and downlink channels, and performing channel estimation of a heterogeneous network system based on terminal direct access and non-cellular heterogeneous networks through pilot information sent in the uplink is as follows: S11. A heterogeneous network system based on terminal direct communication and non-cellular heterogeneous networks, with a total of M multi-antenna access points, K single-antenna users and I transmitters that can establish terminal direct communication with the users, each access point is connected to the central processor via a backhaul link, and the M access points and I transmitters serve the K users under the same time and frequency resources; In the terminal direct and non-cellular heterogeneous network system, the time division duplex operation mode is adopted, and the channel reciprocity is used to perform channel estimation to all users at each access point and transmitter. The obtained channel state information is used for uplink data transmission decoding and downlink data transmission encoding; The channel coefficient between the Kth user and the Mth access point is represented by gm _,k : The channel coefficient between the Kth user and the Ith access point is represented by g _i,k : Among them, M＝1,…,M，K＝1,…,K，I＝1,…,I，um _,k is the large-scale fading coefficient between access point M and user K，ui _,k is the large-scale fading coefficient between transmitter I and user K, which mainly reflects the impact of path loss and shadow fading on the channel. and They represent the small-scale fading coefficients between access point M and user K and between transmitter I and user K, respectively. Each small-scale fading coefficient h _m,k and h _i,k is independent and identically distributed. represents a complex Gaussian random variable with mean 0 and variance 1, are independent and identically distributed random variables, whose distribution conforms to N is the number of antennas at the access point; the large-scale fading coefficient _um,k is modeled by path loss and uncorrelated log-normal shadowing: PL _m,k represents the path loss, is a function with standard deviation _σsh and The shadow fading of , where the path loss PL _m,k is expressed as follows: in: L = 46.3 + 33.9log ₁₀ (f) - 13.82log ₁₀ (h _AP ) - (1.1log ₁₀ (f) - 0.7) h _u + (1.56log ₁₀ (f) - 0.8), f is the carrier frequency, h _AP is the antenna height of the access point, h _u is the antenna height of the user, d _mk is the distance from the Mth access point to the Kth user, d ₀ and d ₁ are the reference distances; a two-component model is used to calculate the shadow fading coefficient z _m,k : in, are two independent random variables, δ is a parameter, 0≤δ≤1, and the covariance function of a _m and b _k is: Where d _AP(m,m′) is the distance between the mth access point and the m'th access point, d _UE(k,k') is the distance between the kth user and the k'th user, d _decorr is the correlation distance, and u _i,k is modeled in the same way as u _m,k ; Through the channel conditions g _m,k and g _i,k , we can get the pilot information sent by user K received by the Mth access point in the uplink: The pilot information sent by user K is received by the Ith transmitter in the uplink Different orthogonal pilot information is allocated to the access point and the transmitter to eliminate coupling, where τ _p is the uplink pilot transmission duration, and is the pilot sequence used by the Kth user to send to the access point and the transmitter, where φ _k and ψ _k are random variables of user K, ||φ _k ||＝1, ||ψ _k ||＝1, represents a vector of τ _p ×1 dimension in the complex field, ||·|| is the Euclidean norm, and are the normalized signal-to-noise ratios when the user sends the pilot signal to the access point and the transmitter, respectively. and is the additional noise at the Mth access point and the Ith transmitter; Based on the received pilot sequence, the Mth access point and the Ith transmitter perform channel estimation. exist Projection on and exist Projection for: in: and are the conjugate transposes of φ _k and ψ _k respectively, k' represents the k'th user, where K and k' are both included in the user set K, φ _k' is the random variable of user k', ||φ _k' ||＝1, ||ψ _k' ||＝1; S12. According to the minimum mean square error criterion, the channel coefficients g _m,k and g _i,k are estimated as: Where _um,k' represents the large-scale fading coefficient between the Mth access point and the k'th user, u _i,k' represents the large-scale fading coefficient from the Ith transmitter to the k'th user, represents the mean, () ^* represents the conjugate; The problem of maximizing downlink communication and rate is modeled, and a user communication mode selection algorithm based on graph neural network is used to solve it, so as to obtain the user communication mode with the highest communication rate, and realize terminal direct connection and non-cellular heterogeneous network access mode selection as follows: S31. Select the user's communication mode to maximize the user's communication and rate, as follows: Model the constrained optimization problem of maximizing the communication rate, as follows: Among them, C1 and C2 are user association constraints, that is, each user can only form D2D communication with at most one transmitter; The user, access point and transmitter are regarded as three types of nodes in the heterogeneous graph, where the user and the access point are fully connected, and the user and the transmitter are fully connected. The graph of the graph optimization model is represented by G = (V, E, s, t), where V = V _UE ∪ V _AP ∪ V _TX represents the node set, V _UE , V _AP and V _TX represent the user, access point and transmitter nodes respectively; represents the set of adjacent nodes that form the edge, x and y represent the nodes in the set V, s represents the node feature, s: t represents the edge feature, t: in, It is represented as a complex field, l ₁ and l ₂ represent the dimension size of node features and edge features; Define node feature matrix Where Z _(i,:) = s(ν _i ), which represents the i-th row of the node feature matrix Z, i.e., the node feature of the i-th node, ν _i represents the i-th node, and the node feature is set to a tensor of all 1s, and the adjacent feature tensor in It is represented as the feature of the edge between node i and node j, where i,j∈V, {i,j} represents the edge between node i and node j, and the large-scale fading coefficients um _,k and ui _,k are used as the features of the edge; S32. Define the loss function as the negative value Loss of the objective function: Graph neural networks extend convolutional neural networks to graphs. In a graph neural network layer, each node updates its hidden state based on aggregated information from neighboring nodes. S33, heterogeneous neural networks are used in non-cellular network systems. Each layer of the neural network contains two types of message transmission, namely, messages transmitted from the access point to the user, and messages transmitted from the user to the access point. Therefore, different weight matrices are used to parameterize different message transmission processes; the feature of node m is Initialize to an empty vector in is an empty vector in the real field; For a T-layer graph neural network, the update of node k in layer t as follows: in, represents the aggregated information of node k in the t-th layer of the neural network, N(k) represents the neighboring nodes of node k, PL(·) is the pooling function, and the pooling function here is the average, that is, the average of the aggregated features. is the aggregate neural network of layer t, W ₁ ^(t) , are the learnable weights in the t-th layer of the neural network, is the node feature of node k in the Tth layer of neurons, CB(·) is the combination function, is a combination of neural networks, is the node feature of node k in the t-1th layer of the graph neural network, and MLP(·) is a multi-layer perceptron that maps the final hidden state to the transmission power.