GAT-based air-ground network anti-interference transmission and deployment method and system

Abstract

The invention discloses an air-ground network anti-interference transmission and deployment method and system based on a GAT, and belongs to the technical field of communication signal processing. Firstly, a communication scene is constructed, each unmanned aerial vehicle base station performs legal communication with a legal user served by the unmanned aerial vehicle base station, and meanwhile, a plurality of interference sources are randomly distributed in an area to interfere with legal communication; obtaining the instantaneous receiving rate of each legal receiver under interference, and forming a problem model which takes maximization of the anti-interference receiving rate of each legal user of the system as a target and takes beamforming transmitted by each unmanned aerial vehicle base station and position deployment of each unmanned aerial vehicle as variables; then, a framework based on a graph attention network is developed, a beam forming vector is solved through neural network training, and unmanned aerial vehicle base station position deployment is solved by training a multi-agent depth deterministic strategy gradient network nested with the graph attention network; neural network parameters are optimized by adopting a stochastic gradient descent method, so that the purposes of effectively suppressing interference and improving the communication performance of a physical layer are achieved.

Description

Background Art

[0002] With the rapid popularization of wireless communication technology and the continuous enrichment of application scenarios, anti-interference ability has become an important indicator for measuring the performance of communication systems. The openness of wireless channels and limited spectrum resources lead to the inevitability of interference. Especially in complex air-ground communication environments, problems such as multi-user interference, intentional interference, and the ever-changing UAV communication channels are becoming increasingly serious. These interferences will reduce the signal-to-noise ratio of the received signal, thus seriously affecting the stability of the system and the accuracy of data transmission. Therefore, anti-interference technology has gradually become a key means to improve communication quality. By using advanced signal processing technologies, adaptive filtering algorithms, intelligent spectrum management, and UAV location deployment methods, anti-interference technology can effectively weaken the impact of interference and optimize the reliability and robustness of wireless communication systems. At the same time, by means of strategies such as multi-antenna technology, beamforming, and cooperative communication, the interference suppression ability of the system can be further enhanced, thus providing a solid technical guarantee for achieving efficient and secure communication.

[0003] Physical layer anti-interference technology provides a more effective solution for dynamic channel conditions and computational complexity challenges. Conventional physical layer anti-interference solutions often rely on the successive convex approximation algorithm. By expanding the non-convex problem into a convex form, the sub-optimal solution is approximated through iterative loops. However, the computational amount required by traditional optimization algorithms is relatively large, and the scenario parameters they target are rigid, unable to fully meet the growing dynamic requirements of air-ground communication physical layer security. In this context, new research has proposed using more efficient deep learning methods to improve the anti-interference performance of the air-ground communication physical layer. Its basic model is the multi-layer perceptron MLP. The MLP can counter the non-linearity and non-convexity of air-ground communication physical layer security problems and minimize the demand for computing resources by utilizing parallel processing and avoiding the iterative process in the air-ground communication physical layer security solution process. Nguyen et al. used a double-Q reinforcement learning network based on MLP to learn effective communication strategies, adjust channel access and transmission power to cope with different interference scenarios. Liu et al. used deep reinforcement learning to adjust the reflection element parameters of reconfigurable intelligent surfaces, significantly improving the anti-interference signal-to-noise ratio of legitimate receivers.

[0004] Although reinforcement learning has powerful performance in dealing with anti-interference at the physical layer, it relies on empirical exploration of the environment and is limited by the fixed parameters of the environment. The mapping ability of the action-state setting to the anti-interference physical scenario is limited. However, due to factors such as channel condition changes and user mobility, the wireless anti-interference communication environment at the physical layer has inherent complexity and dynamics. Traditional reinforcement learning representations are insufficient to capture and solve complex variables and are also difficult to accurately model non-Euclidean complex physical relationships. In contrast, graph neural networks are specifically designed to directly process graph-structured data in non-Euclidean spaces. In particular, heterogeneous graph neural networks allow the coexistence of multiple types of nodes and connection relationships, providing a more flexible and adaptable framework for capturing the irregularities and dynamics of wireless communication scenarios. Graph attention networks can efficiently process heterogeneous graph data with spatial structures composed of multiple nodes and edges, adapt to scenario expansion, and introduce an attention weight mechanism to enhance the network's processing ability for features. Compared with traditional multi-layer perceptron schemes, graph attention networks have advantages in dealing with the anti-interference scenario of the physical layer of air-ground communication with diverse communication members and complex communication relationships. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to provide a method and system for anti-interference transmission and deployment of an air-ground network based on GAT in view of the above deficiencies in the prior art. The potential of graph attention networks in improving the anti-interference performance at the physical layer is studied, and graph attention networks are used to achieve highly scalable and efficient strategy design, so as to solve the technical problems of long operation time, poor generalization performance, and difficult position deployment in solving the anti-interference problem at the physical layer in wireless communication scenarios.

[0006] The present invention adopts the following technical solutions:

[0007] An anti-interference transmission and deployment method for an air-ground network based on GAT includes the following steps:

[0008] Construct a communication scenario with multiple UAV base stations, legitimate communication users, and multiple interference sources;

[0009] Based on the given direction of the UAV base station transmitter signal arrival and the interference power of the interference source, obtain the channel gain between each UAV base station transmitter and the corresponding legitimate user, the channel gain of other UAV base stations interfering with the legitimate user, and the channel gain of the interference source to each legitimate user, and determine the received signal expression of the legitimate user under the influence of the interference source;

[0010] According to the obtained expression of the received signal at the legitimate user, construct a problem model with the transmit beamforming vectors of each UAV base station and the deployment positions of each UAV base station as variables and the anti-interference rate of maximizing this anti-interference system as the objective function;

[0011] Based on the constructed communication scenario and the obtained problem model, define the nodes and connection edges of the graph mapped by the scenario and the problem model;

[0012] Based on the problem model, design a graph attention network encoder, and define the methods for message generation and message aggregation in the graph attention network;

[0013] Based on the constructed communication scenario and the problem model, define the exploration agents in the scenario and the problem model, and design the actions and states of the multi-agent deep deterministic policy gradient network.

[0014] Design the loss function of the graph attention network and the rewards of the multi-agent deep deterministic policy gradient network, define the neural network learning method and parameters, train the neural network, and obtain a problem model with generalization ability to adapt to the dynamic topology and communication conditions of the air-ground network, and realize the anti-interference communication decision output based on the graph neural network in the interference environment.

[0015] Preferably, in the communication scenario, the UAV base station has N antennas, the legitimate user has a single antenna, and K different UAV base stations and legitimate user pairs are distributed in a fixed area, denoted as The UAV base stations and users in the same group are represented by the same serial number. At the same time, there are J interference sources in the whole area, denoted as Interfere with the communication between the UAV base station and the user.

[0016] Preferably, the received signal expression of the legitimate user under the influence of the interference source is:

[0017]

[0018] Among them, L represents the geographical locations of all UAV base stations in the area, l k represents the geographical coordinates of the UAV base station k, P is the transmission power of each interference source, f k represents the beamforming vector of the base station k to the corresponding user, h l,k (L) represents the communication channel vector from the base station l to the user k when the geographical locations of all UAV base stations in the area are L, g j,k (L) represents the communication channel vector from the interference source j to the user k when the geographical locations of all UAV base stations in the area are L, The transmission power of each interference source is represented as P, n k represents the additive white Gaussian noise at the receiving end of the user k, s k is the signal transmitted by the base station to the user k, K is the number of UAV base stations in the area, and H is the symbol for conjugate transpose.

[0019] Preferably, the problem model is described as follows:

[0020]

[0021] Among them, R(L,F,P) is the anti-interference rate of the system, and γ k is the signal-to-interference-plus-noise ratio of the legitimate link signal received by user k. L represents the geographical locations of all UAV base stations in the area, and l k represents the geographical coordinates of UAV base station k. P is the transmission power of each interference source, and f k represents the beamforming vector of base station k for the corresponding user. h l,k (L) represents the communication channel vector from base station l to user k when the geographical locations of all UAV base stations in the area are L. g j,k (L) represents the communication channel vector from interference source j to user k when the geographical locations of all UAV base stations in the area are L. The transmission power of each interference source is expressed as P, and n k represents the additive white noise that follows a complex Gaussian distribution at the receiving end of user k.

[0022] Preferably, the characteristics of interference source node j are set as its corresponding power p j , and the corresponding dimensions are supplemented, and the ~ symbol is used to distinguish the characteristics of interference source nodes, which is described as follows:

[0023]

[0024] The characteristics of the connected edges are designed as the channels corresponding to the interference relationships, and the ~ symbol is used to label a type of connected edges connected to the interference source nodes:

[0025]

[0026] Among them, represents the connection edge from UAV base station-legitimate user node l to UAV base station-legitimate user node k. represents the connection edge from interference source node j to UAV base station-legitimate user node k. represents taking the real part of the interference channel gain from the base station in UAV base station-legitimate user l to the receiver in UAV base station-legitimate user k. represents taking the imaginary part of the interference channel from the base station in UAV base station-legitimate user l to the receiver in UAV base station-legitimate user k. represents taking the real part of the interference channel gain from interference source j to the receiver in UAV base station-legitimate user k. represents taking the imaginary part of the interference channel from interference source j to the receiver in UAV base station-legitimate user k.

[0027] Preferably, in the graph attention network encoder, let be the feature of the interfering source node j in the d-th message passing. The message to be transmitted generated by the interfering source node j in the d-th message passing is Let be the feature of the UAV base station-legitimate user node k in the d-th message passing. The message to be transmitted generated by the UAV base station-legitimate user node k in the d-th message passing is According to the attention coefficient and the information generated by the neighbor nodes, the aggregated information of the central node k is described as follows:

[0028]

[0029] Among them, is regarded as the contribution of the connection edge between the neighbor node u and the central node k to constructing the feature of the central node k;

[0030] When both the node u and the node k are UAV base station-legitimate user nodes, the contribution information of the neighbor node u to the central node k and when the node q is a UAV base station-legitimate user node and the node k is an interfering source node, the contribution information in the d-th message passing is expressed as follows:

[0031]

[0032] Among them, is the feature aggregation neural network, is the neighbor node feature, is the edge feature, is the legitimate user feature.

[0033] Preferably, for the cell UAV base station agent, the relationship between the action and the state transformation is described as follows:

[0034] s′ k =[s k [0]+Δx k ,s k [1]+Δy k ,s k [2],s k [3]]=[x k +Δx k ,y k +Δy k ,x k,u ,y k,u

[0035] Among them, s′ is the updated state, s k [0] is the first dimension of the state, Δx​k is the position update in the x direction, Δy k is the position update in the y direction, x k is the position in the x direction, y k is the position in the x direction, x k,u is the x-direction position of the served user, y k,u is the y-direction position of the served user;

[0036] The state of the interference source j at the nth step is set to:

[0037] s′ j = [x j,J , y j,J , 0, p j

[0038] where, x j,J is the abscissa of the interference source j, y j,J is the abscissa of the interference source j, p j is the interference power of the interference source j;

[0039] The action of the interference source agent is set to the change of power:

[0040] a j = [s j , Δp j

[0041] where, s j represents the symbolic term, Δp j represents the interference source power change value term;

[0042] When the interference source agent j takes the action a j in the current situation, the power state of the next-step interference source agent j is described as follows:

[0043]

[0044] where, sign(·) represents the sign function, the function outputs 1 when the input is positive, and the function outputs -1 when the input is negative. sign(·) is used to extract the sign, Δp j is the minimum power change, is the interference source set, P is the maximum power threshold, Δp i is the minimum power change of any interference source, p i is the power of any interference source.

[0045] Preferably, the loss function loss of the graph attention neural network is:

[0046]

[0047] ​​where \(R(L,F,P)\) is the sum rate of anti - interference communication between each UAV base station and legitimate users in the region, and \(\gamma\) k is the anti - interference received signal - to - noise ratio of the UAV base station - legitimate user \(k\).

[0048] Preferably, the reward of the UAV agent in the multi - agent deep deterministic policy gradient network is set as the communication rate of the UAV base station to the legitimate user in the state after completing the current action:

[0049] \(r\) k \(=\log_2(1 + \gamma\) k (L, gnn(L,P),P))

[0050] where \(L\) is the position variable of the UAV base station, \(gnn(L,P)\) is the output of the inner - layer graph neural network in the current network state, and \(P\) is the power variable of the interference source;

[0051] The reward of the interference - source agent in the multi - agent deep deterministic policy gradient network is set according to the purpose of the game model, and the reward is set to a negative value:

[0052]

[0053] where \(F\) is generated by a scheme for anti - interference beamforming design based on a graph attention network, and \(\beta\) is the reward coefficient.

[0054] In a second aspect, an embodiment of the present invention provides an anti - interference transmission and deployment system for an air - ground network based on GAT, including:

[0055] A scenario module that constructs a communication scenario with multiple UAV base stations, legitimate communication users, and multiple interference sources; where each UAV base station conducts legitimate communication with its served legitimate users, and at the same time, multiple interference sources are randomly distributed in the region to interfere with legitimate communication;

[0056] A signal module that, based on the given direction of the UAV base station transmitter signal arrival and the interference power of the interference source, obtains the channel gain between each UAV base station transmitter and the corresponding legitimate user, the channel gain of other UAV base stations interfering with the legitimate user, and the channel gain of the interference source to each legitimate user, and determines the received signal expression of the legitimate user under the influence of the interference source;

[0057] A function module that, according to the obtained received signal expression at the legitimate user, constructs a problem model with the transmit beamforming vectors of each UAV base station and the deployment positions of each UAV base station as variables and the anti - interference rate of maximizing this anti - interference system as the objective function;

[0058] A network module, based on the constructed communication scenario and problem model, defines the nodes and connection edges of the graph mapped by the scenario and problem model; based on the problem model, designs a graph attention network encoder and defines the methods for message generation and message aggregation in the graph attention network;

[0059] An output module, based on the constructed communication scenario and problem model, defines the exploration agents in the scenario and problem model, designs the actions and states of a multi-agent deep deterministic policy gradient network; designs the loss function of the graph attention network and the rewards of the multi-agent deep deterministic policy gradient network, defines the neural network learning method and parameters, trains the neural network, and obtains a problem model with generalization ability to adapt to the dynamic topology and communication conditions of the air-ground network, and realizes the anti-jamming communication decision output based on the graph neural network under interference environments.

[0060] In a third aspect, a computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of the above-mentioned anti-jamming transmission and deployment method for air-ground networks based on GAT are implemented.

[0061] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium including a computer program. When the computer program is executed by a processor, the steps of the above-mentioned anti-jamming transmission and deployment method for air-ground networks based on GAT are implemented.

[0062] In a fifth aspect, a chip includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of the above-mentioned anti-jamming transmission and deployment method for air-ground networks based on GAT are implemented.

[0063] In a sixth aspect, an embodiment of the present invention provides an electronic device including a computer program. When the computer program is executed by the electronic device, the steps of the above-mentioned anti-jamming transmission and deployment method for air-ground networks based on GAT are implemented.

[0064] Compared with the prior art, the present invention has at least the following beneficial effects:

[0065] An anti-jamming transmission and deployment method for air-ground networks based on GAT considers an anti-jamming transmission and deployment system for air-ground networks based on a graph attention network, and proposes to use a graph attention network to generate multi-base station beamforming strategies for anti-jamming and assist in generating multi-UAV deployment strategies in multi-UAV base station control, thereby improving the physical layer anti-jamming performance. From an optimization perspective, this problem is a non-convex problem and is difficult to solve effectively. Although continuous convex approximation can approximately solve convex problems, it is difficult to ensure the effectiveness of its solutions and consumes a large amount of computing resources.

[0066] In addition, in wireless communication scenarios, the communication relationships and interference relationships are complex. The high mobility of drones also makes the air-ground communication characterized by frequent changes and complex situations. It is difficult to use a general method to solve the problems of different numbers of drone base stations-legitimate users and interference sources. Based on this, a framework based on graph attention network is developed. The drone base stations, legitimate user pairs, interference sources, legitimate communication relationships, and the communication interference relationships of other drone base stations and interference sources on the legitimate user receivers in the wireless communication scenario are modeled as various types of nodes and edges of a graph. The beamforming vectors of each drone base station and the deployment positions of each drone are used as variables to improve the anti-interference communication performance of the physical layer in the wireless communication scenario. The neural network parameters are optimized by using the stochastic gradient descent method to achieve the purpose of effectively improving the anti-interference communication performance of the physical layer of wireless communication. The numerical results show that this method can significantly improve the decision-making efficiency and obtain good generalization performance.

[0067] Furthermore, a communication scenario with multiple drone base stations-legitimate users and multiple interference sources is designed. Among them, each drone base station conducts legitimate communication with the legitimate users it serves, while multiple interference sources are randomly distributed in the area to interfere with the legitimate communication. This scenario conforms to the general scenario characteristics in multi-drone air-ground wireless communication in reality and is easy to be extended to real communication scenarios. The design of multiple drone base stations-legitimate users and multiple interference sources fully reflects the complexity and randomness of the communication scenario, which is beneficial to improving the adaptability of the scheme to various situations and ensuring the smooth underlying logic of the scalability of the scheme.

[0068] Furthermore, based on the current positions of each drone base station, the given directions of the transmitter signals arriving, and the interference powers of the interference sources, the channel gains between the transmitters of each drone base station and the corresponding legitimate users, the channel gains of the interference from other drone base stations to the legitimate users, and the channel gains of the interference sources to the corresponding legitimate users of each drone base station are obtained. The expression of the received signal of the legitimate user under the influence of the interference source is determined, laying the necessary conditions for calculating the anti-interference communication rate of the system to measure the anti-interference performance of the system, and providing a theoretical basis for the subsequent problem model representation.

[0069] Furthermore, according to the expression of the received signal at the legitimate user, a problem model is constructed with the beamforming vectors of the transmitters of each drone base station and the deployment positions of the drone base stations as variables and the maximization of the anti-interference rate of this anti-interference system as the objective function, clearly defining the problem-solving objective in a concise form, and clarifying the optimization variables and constraints, providing favorable conditions for the subsequent problem-solving.

[0070] Furthermore, based on the constructed scenario and the problem model, the nodes and connecting edges of the graph mapped by the scenario and the problem are defined, serving the design of the subsequent graph attention network encoder and providing support for using the graph attention network encoder to process the problems in the scenario.

[0071] Furthermore, based on the problem model, a graph attention network encoder is designed, and methods for message generation and message aggregation in the graph attention network are defined, providing a solution for processing the features of nodes and edges of the heterogeneous graph mapped by the scenario and laying a foundation for the training of the neural network.

[0072] Furthermore, based on the constructed scenario and problem models, exploration agents in the scenario and problem are defined, and the actions and states of the multi-agent deep deterministic policy gradient network are designed, laying a foundation for the graph attention network to combine multi-agent reinforcement learning to achieve the deployment of UAV base station positions.

[0073] Furthermore, the loss function of the graph attention network and the rewards of the multi-agent deep deterministic policy gradient network are designed, the neural network learning method and parameters are defined, the neural network is trained, a problem model with generalization ability is obtained, and the anti-interference performance of the physical layer is improved.

[0074] It can be understood that the beneficial effects of the above second aspect to sixth aspect can be referred to the relevant descriptions in the above first aspect, and will not be elaborated here.

[0075] In summary, the present invention applies graph neural network technology to the UAV air-ground wireless communication scenario with intentional and unintentional interference, uses the beamforming vectors of each UAV base station as variables, and solves the complex multi-cell anti-interference communication problem through the training of the neural network. In this regard, the neural network can approximate the internal relationship between the system input and output in various complex mathematical forms to achieve the best anti-interference communication performance of the physical layer.

[0076] Next, through the drawings and embodiments, the technical solutions of the present invention will be further described in detail. Description of the Drawings

[0077] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required to be used in the embodiments of the present application will be briefly introduced below. Obviously, the following described drawings are only some embodiments of the present application, and those of ordinary skill in the art can obtain other drawings according to these drawings without creative efforts.

[0078] Figure 1 It is a model diagram of the wireless transmission system constructed by the present invention;

[0079] Figure 2 It is a flowchart of the present invention;

[0080] Figure 3 It is a schematic diagram of the relationship between the physical layer anti-interference communication rate of the system of the present invention and the number of cells composed of UAV-legitimate users under different schemes;

[0081] Figure 4 It is a schematic diagram showing the relationship between time and the number of cells composed of drones - legitimate users under the processing of different schemes of the system of the present invention;

[0082] Figure 5 It is a schematic diagram of a computer device provided by an embodiment of the present invention;

[0083] Figure 6 It is a block diagram of an electronic device provided by the present invention according to an embodiment.

[0084] Among them, 60. Computer device; 61. Processor; 62. Memory; 63. Computer program; 600. Electronic device; 610. Processing unit; 620. Storage unit; 6201. Random access storage unit; 6202. Cache storage unit; 6203. Read-only storage unit; 6204. Program / utility; 6205. Program module; 630. Bus; 640. Display unit; 650. Input / output interface; 660. Network adapter; 700. External device. Detailed implementation manners

[0085] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0086] In the description of the present invention, it should be understood that the terms "include" and "comprise" indicate the presence of the described features, wholes, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components and / or their combinations.

[0087] It should also be understood that the terms used in the specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. As used in the specification of the present invention and the appended claims, unless the context clearly indicates otherwise, the singular forms "a", "an" and "the" are intended to include the plural forms.

[0088] It should also be further understood that the term " / and" used in the specification of the present invention and the appended claims refers to any combination and all possible combinations of one or more of the related listed items, and includes these combinations. For example, A and / or B can represent: A exists alone, A and B exist simultaneously, and B exists alone. In addition, the character " / " in the present invention generally represents an "or" relationship between the preceding and following related objects.

[0089] It should be understood that although terms such as first, second, and third may be used in the embodiments of the present invention to describe preset ranges, etc., these preset ranges should not be limited to these terms. These terms are only used to distinguish the preset ranges from each other. For example, without departing from the scope of the embodiments of the present invention, the first preset range may also be referred to as the second preset range, and similarly, the second preset range may also be referred to as the first preset range.

[0090] Depending on the context, the word "if" as used herein can be interpreted as "when" or "while" or "in response to determining" or "in response to detecting". Similarly, depending on the context, the phrase "if determined" or "if detected (stated condition or event)" can be interpreted as "when determined" or "in response to determining" or "when detected (stated condition or event)" or "in response to detecting (stated condition or event)".

[0091] Various structural schematic diagrams according to the disclosed embodiments of the present invention are shown in the drawings. These figures are not drawn to scale, where for the purpose of clear expression, some details are enlarged and some details may be omitted. The shapes of various regions and layers shown in the figures, as well as their relative sizes and positional relationships, are only exemplary. In practice, there may be deviations due to manufacturing tolerances or technical limitations, and those skilled in the art can design regions / layers with different shapes, sizes, and relative positions according to actual needs.

[0092] The present invention provides an anti-interference transmission and deployment method for air-ground networks based on GAT, revealing the potential of graph attention networks in anti-interference for UAV air-ground communication. A scenario with multiple UAV base station-legitimate user pairs and multiple interference sources is considered, and it is proposed to use graph attention networks on each UAV base station to help generate beamforming strategies and UAV base station deployment strategies against interference. A scenario with multiple UAV base station-legitimate user pairs and multiple interference sources is constructed, and the instantaneous reception rate of each legitimate user receiver receiving the signal from the corresponding UAV base station under the action of intentional interference from interference sources and unintentional interference from other UAV base stations is analyzed, and a problem model is formed with the goal of maximizing the anti-interference reception rate of this system, where the transmit beamforming vectors of each UAV base station and the deployment positions of each UAV are variables; then a framework based on graph attention networks is developed, where the beamforming vectors are solved by training a neural network, and the UAV deployment positions are solved by training a multi-agent deep deterministic policy gradient network with a nested graph attention network; the neural network parameters are optimized using the stochastic gradient descent method to achieve the purpose of effectively suppressing interference and improving the anti-interference performance of the physical layer.

[0093] Embodiment 1

[0094] Please refer to Figure 2 , a method for anti - interference transmission and deployment of an air - ground network based on GAT according to the present invention includes the following steps:

[0095] S1. Construct a communication scenario with multiple unmanned aerial vehicle (UAV) base stations, legitimate communication users, and multiple interference sources; wherein, each UAV base station conducts legitimate communication with the legitimate users it serves, and at the same time, multiple interference sources are randomly distributed in the area to interfere with the legitimate communication.

[0096] Please refer to Figure 1 , in the air - ground wireless communication system model diagram constructed by the present invention, in a scenario with multiple UAV base stations, multiple legitimate users, and multiple interference sources, the UAV base station has N antennas while the legitimate user has a single antenna. K different UAV base stations and legitimate user pairs are distributed in a fixed area, denoted as The UAV base stations and users can be represented by the same serial number. At the same time, there are J interference sources in the whole area, denoted as Interfere with the communication between the UAV base station and the user.

[0097] S2. Based on the given direction of the signal arrival of the UAV base station transmitter and the interference power of the interference source, obtain the channel gain between each UAV base station transmitter and the corresponding legitimate user, the channel gain of the interference from other UAV base stations to the legitimate user, and the channel gain from the interference source to each legitimate user, and determine the received signal expression of the legitimate user under the influence of the interference source.

[0098] The received signal y k (F) is:

[0099]

[0100] Wherein, represents the geographical locations of all UAV base stations in the area, l k =(x k , y k ) represents the geographical coordinates of UAV base station k, and the transmission power of each interference source is denoted as represents the beamforming vector of base station to the corresponding user, F = [f1, f2,..., f K , represents the communication channel vector from base station to user when the geographical locations of all UAV base stations in the area are L, represents the communication channel vector from interference source to user when the geographical locations of all UAV base stations in the area are L, and the transmission power of each interference source is denoted as wherein in the formula represents the user additive white noise at the receiving end that follows a complex Gaussian distribution, and s k is the base station transmitted to the user signal, K is the number of unmanned aerial vehicle (UAV) base stations in the area, and H is the symbol for conjugate transpose.

[0101] S3. According to the expression of the received signal at the legitimate user obtained in step S2, construct a problem model with the transmit beamforming vectors of each UAV base station and the deployment positions of each UAV base station as variables and the maximization of the anti-interference rate of this anti-interference system as the objective function;

[0102] The problem model is described as follows:

[0103]

[0104] where R(L,F,P) is the anti-interference rate of the system, and γ k is the signal-to-interference-plus-noise ratio (SINR) of the received legitimate link signal of the user , where represents the geographical locations of all UAV base stations in the area, l k =(x k ,y k ) represents the geographical coordinates of UAV base station k, the transmit power of each interference source is expressed as represents the beamforming vector of the base station for the corresponding user, F = [f1, f2,..., f K , represents the communication channel vector from the base station to the user when the geographical locations of all UAV base stations in the area are L, represents the communication channel vector from the interference source to the user when the geographical locations of all UAV base stations in the area are L, the transmit power of each interference source is expressed as wherein in the formula represents the user additive white noise at the receiving end that follows a complex Gaussian distribution. The problem is solved by dividing it into inner and outer layer problems. The inner layer problem is to design a more suitable beamforming vector group F for the system to improve the anti-interference transmission performance of the system when the positions of each UAV base station are fixed, and the outer layer problem is to maximize the legitimate communication rate of the system by deploying the positions of UAV base stations and simulating the anti-strategy of interference source power.

[0105] S4. Based on the scenario constructed in step S1 and the problem model in step S3, define the nodes and connecting edges of the graph mapped by the scenario and the problem.

[0106] The mutual interference of the signals transmitted by different UAV base stations and the interference of the interference source on the communication between each UAV base station and the legitimate user are similar. Thus, different members and communication relationships in the communication scenario are mapped into the graph structure. The communication pair of each UAV base station and the corresponding user is mapped as a type of node, and the interference source is mapped as another type of node. The two types of connecting edges are: the connecting edge between the interference source node and the UAV base station-legitimate user node, and the connecting edge between different UAV base station-legitimate user nodes.

[0107] In the graph attention network, the initial feature of the UAV base station-legitimate user node k is expressed as:

[0108]

[0109] where is a function to obtain the real part, is a function to obtain the imaginary part, represents the real part of the communication channel gain of the UAV base station to the legitimate user and represents the imaginary part of the communication channel gain of the UAV base station to the legitimate user

[0110] Similarly, set the feature of the interference source node j as its corresponding power p j , complement the corresponding dimension, and use the ~ symbol to distinguish the features of the interference source node, which is described as follows:

[0111]

[0112] Similarly, design the feature of the connecting edge as the channel corresponding to the interference relationship, and use the ~ symbol to label a type of connecting edge connected to the interference source node:

[0113]

[0114] where represents the connecting edge from the UAV base station-legitimate user node l to the UAV base station-legitimate user node k, represents the connecting edge from the interference source node j to the UAV base station-legitimate user node k, represents the real part of the interference channel gain of the base station in the UAV base station-legitimate user l to the receiver in the UAV base station-legitimate user k, ​It represents taking the imaginary part of the interference channel from the base station in the UAV base station-legitimate user l to the receiver in the UAV base station-legitimate user k. It represents taking the real part of the interference channel gain from the interference source j to the receiver in the UAV base station-legitimate user k. It represents taking the imaginary part of the interference channel from the interference source j to the receiver in the UAV base station-legitimate user k.

[0115] S5. Based on the problem model in step S3, design a graph attention network encoder and define the methods for message generation and message aggregation in the graph attention network;

[0116] In the graph attention network encoder, let be the feature of the interference source node j in the d-th message passing. Then, the message to be transmitted generated by the interference source node j in the d-th message passing is Let be the feature of the UAV base station-legitimate user node k in the d-th message passing. Then, the message to be transmitted generated by the UAV base station-legitimate user node k in the d-th message passing is According to the attention coefficient and the information generated by the neighbor nodes, the aggregated information of the central node k is described as follows:

[0117]

[0118] Among them, is regarded as the contribution of the connection edge between the neighbor node u and the central node k to constructing the feature of the central node k, that is, the attention coefficient, and is described as follows:

[0119]

[0120] Among them, the softmax operation is equivalent to calculating the normalized weights of the contribution information of all types of connection edges of the central node, indicating that the softmax operation is equivalent to calculating the normalized weights of the contribution information of all types of connection edges of the central node. represents the contribution information of the neighbor node u to the central node k when both the node u and the node k are UAV base station-legitimate user nodes. represents the contribution information in the d-th message passing when the node q is a UAV base station-legitimate user node and the node k is an interference source node. and are described as follows:

[0121]

[0122]

[0123] S6. Based on the scenario constructed in step S1 and the problem model in step S3, define the exploration agents in the scenario and the problem, and design the actions and states of the multi-agent deep deterministic policy gradient network.

[0124] The outer problem is to maximize the legal communication rate of the system by deploying the positions of the UAV base stations and simulating the interference source power countermeasure strategy. Therefore, two types of agents are set in the scenario. One is the UAV base station agent, and its state is defined as the deployment position of the UAV base station and the positions of its corresponding legitimate users, described as follows:

[0125] s k =[x k ,y k ,x k,u ,y k,u

[0126] where x k is the abscissa of the position of UAV k of the UAV base station, and y k represents the ordinate of UAV base station k, x k,u represents the abscissa of legitimate user k, and y k,u represents the ordinate of legitimate user k.

[0127] The action a k of the UAV base station agent at the k-th step is defined as the displacement of the UAV, described as follows:

[0128] a k =[Δx k ,Δy k

[0129] where Δx k represents the displacement of the UAV base station in the x direction, and Δy k represents the displacement of the UAV base station in the y direction.

[0130] Therefore, for the cell UAV base station agent, the relationship between the action and the state transformation can be described as follows:

[0131] s′ k =[s k [0]+Δx k ,s k [1]+Δy k ,s k [2],s k [3]]=[x k +Δx k ,y k +Δy k ,x k,u ,y k,u

[0132] ​​​For the strategy selection of interference sources, only the interference power variable is involved. Therefore, the state of interference source j at the nth step is set as:

[0133] s j ′=[x j,J ,y j,J ,0,p j

[0134] where x j,J is the abscissa of interference source j, y j,J is the abscissa of interference source j, and p j is the interference power of interference source j. During the exploration of the environment by interference source agent j in one round, x j,J and y j,J remain unchanged, corresponding to the immovable characteristic of the interference source. In the state, p j changes with the action selected by the agent.

[0135] The action of the interference source agent is set as the change of power:

[0136] a j =[s j ,Δp j

[0137] where s j represents the symbol term, and Δp j represents the interference source power change value term. At the same time, considering the power constraint of the interference source, when interference source agent j takes action a j in the current situation, the power state of interference source agent j in the next step is described as follows:

[0138]

[0139] where sign(·) represents the sign function. When the input is positive, the function outputs 1; when the input is negative, the function outputs -1. sign(·) is used to extract the sign.

[0140] S7. Define the loss function of the graph attention network and the reward of the multi-agent deep deterministic policy gradient network, define the neural network learning method and parameters, train the neural network, and obtain a problem model with generalization ability to achieve the improvement of the anti-interference performance of the physical layer.

[0141] The loss function loss of the graph attention neural network is:

[0142]

[0143] where R(L,F,P) is the sum of the rates of anti-interference communication between each UAV base station and legal users in the region, and γ k ​​is the anti-interference received signal-to-noise ratio of the UAV base station - legitimate user k.

[0144] The reward of the UAV agent in the multi-agent deep deterministic policy gradient network is set to the communication rate of the UAV base station to the legitimate user in the state after completing the current action:

[0145] r k = log2(1 + γ k (L, gnn(L, P), P))

[0146] The reward of the interfering source agent in the multi-agent deep deterministic policy gradient network is set according to the purpose of the game model, and the reward is set to a negative value:

[0147]

[0148] where F is generated by the scheme of anti-interference beamforming design based on the graph attention network. Except for constants such as the user's geographical location and channel calculation method that are regarded as fixed values in a single solution, the graph attention network mainly needs to input the position variable L of the UAV base station and the power variable P of the interfering source to obtain the beamforming result vector, and then obtain the communication rate between the UAV and the corresponding legitimate user. These constant information is generally obtained in the form of trained neural network parameters or the state information of the agent during the exploration process of multi-agent reinforcement learning, while the variable L and the variable P are always manifested in the action selection of the agent. Among them, β represents the reward coefficient of the interfering source agent, which is used to ensure the ability of the overall system to keep the overall reward and the reward formula of a single agent positive at the nth step, that is, to balance the ability of the sum of the rewards of all agents to be positive, and to ensure that the goal of reinforcement learning is to provide a strategy for the UAV base station agent to fight against the interfering source agent. Different from the reward of the UAV base station agent, the reward of the interfering source agent reflects the interference degree of the interfering source to all legitimate users, so it is also represented by the negative number of the multiple of the total system communication rate.

[0149] Those skilled in the art of the present technology can understand that various aspects of the present invention can be implemented as a system, a method, or a program product. Therefore, various aspects of the present invention can be specifically implemented in the following forms, namely: a complete hardware implementation, a complete software implementation (including firmware, microcode, etc.), or an implementation combining hardware and software aspects, which can be collectively referred to as "circuit", "module", or "platform" here.

[0150] Embodiment 2

[0151] The present invention provides an anti-jamming transmission and deployment system for air-ground networks based on GAT, which can be used to implement the above-mentioned anti-jamming transmission and deployment method for air-ground networks based on GAT. Specifically, the anti-jamming transmission and deployment system for air-ground networks based on GAT includes a scenario module, a signal module, a function module, a network module, and an output module.

[0152] Among them, the scenario module constructs a communication scenario with multiple unmanned aerial vehicle (UAV) base stations, legitimate communication users, and multiple interference sources. Among them, each UAV base station conducts legitimate communication with the legitimate users it serves, while multiple interference sources are randomly distributed in the area to interfere with the legitimate communication.

[0153] The signal module obtains the channel gain between each UAV base station transmitter and the corresponding legitimate user, the channel gain of the interference from other UAV base stations to the legitimate user, and the channel gain from the interference source to each legitimate user based on the given direction of arrival of the UAV base station transmitter signal and the interference power of the interference source, and determines the received signal expression of the legitimate user under the influence of the interference source.

[0154] The function module constructs a problem model with the transmit beamforming vectors of each UAV base station and the deployment positions of each UAV base station as variables and the anti-jamming rate maximization of this anti-jamming system as the objective function according to the obtained received signal expression at the legitimate user.

[0155] The network module defines the nodes and connection edges of the graph mapped by the constructed communication scenario and problem model based on the constructed communication scenario and problem model. Based on the problem model, it designs a graph attention network encoder and defines the methods of message generation and message aggregation in the graph attention network.

[0156] The output module defines the exploration agents in the scenario and problem model based on the constructed communication scenario and problem model, designs the actions and states of the multi-agent deep deterministic policy gradient network. Designs the loss function of the graph attention network and the rewards of the multi-agent deep deterministic policy gradient network, defines the neural network learning method and parameters, trains the neural network, and obtains a problem model with generalization ability to adapt to the dynamic topology and communication conditions of the air-ground network, and realizes the anti-jamming communication decision output based on the graph neural network in the interference environment.

[0157] Embodiment 3

[0158] The present invention provides a terminal device, which includes a processor and a memory. The memory is used to store a computer program, and the computer program includes program instructions. The processor is used to execute the program instructions stored in the computer storage medium. The processor may be a Central Processing Unit (CPU), or may also be other general-purpose processors, Graphics Processing Unit (GPU), Tensor Processing Unit (TPU), Digital Signal Processor (DSP), Application Specific Integrated Circuit (ASIC), Field-Programmable Gate Array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. It is the computing core and control core of the terminal, and is suitable for implementing one or more instructions. Specifically, it is suitable for loading and executing one or more instructions to implement the corresponding method flow or corresponding function. The processor described in the embodiments of the present invention can be used for the operation of the anti-interference transmission and deployment method for the air-ground network based on GAT, including:

[0159] Construct a communication scenario with multiple UAV base stations, legitimate communication users, and multiple interference sources. Among them, each UAV base station conducts legitimate communication with the legitimate users it serves, while multiple interference sources are randomly distributed in the area to interfere with the legitimate communication. Based on the given direction of the UAV base station transmitter signal arrival and the interference power of the interference sources, obtain the channel gain between each UAV base station transmitter and the corresponding legitimate user, the channel gain of other UAV base stations interfering with the legitimate users, and the channel gain from the interference sources to each legitimate user, and determine the received signal expression of the legitimate user under the influence of the interference sources. According to the obtained received signal expression at the legitimate user, construct a problem model with the transmit beamforming vectors of each UAV base station and the deployment positions of each UAV base station as variables and the maximization of the anti-interference rate of this anti-interference system as the objective function. Based on the constructed communication scenario and the obtained problem model, define the nodes and connection edges of the graph mapped by the scenario and the problem model. Based on the problem model, design a graph attention network encoder and define the methods of message generation and message aggregation in the graph attention network. Based on the constructed communication scenario and the problem model, define the exploration agents in the scenario and the problem model, and design the actions and states of the multi-agent deep deterministic policy gradient network. Design the loss function of the graph attention network and the rewards of the multi-agent deep deterministic policy gradient network, define the neural network learning method and parameters, train the neural network, and obtain a problem model with generalization ability to adapt to the dynamic topology and communication conditions of the air-ground network and achieve anti-interference communication decision output based on the graph neural network in the interference environment.

[0160] Please refer to Figure 5 , the terminal device is a computer device. The computer device 60 in this embodiment includes: a processor 61, a memory 62, and a computer program 63 stored in the memory 62 and executable on the processor 61. When the computer program 63 is executed by the processor 61, it implements the anti-interference transmission and deployment method for the air-ground network based on GAT in the embodiment. To avoid repetition, it will not be elaborated here one by one. Alternatively, when the computer program 63 is executed by the processor 61, it implements the functions of each model / unit in the anti-interference transmission and deployment system for the air-ground network based on GAT in the embodiment. To avoid repetition, it will not be elaborated here one by one.

[0161] The computer device 60 can be a computing device such as a desktop computer, a notebook, a palm computer, and a cloud server. The computer device 60 may include, but is not limited to, a processor 61 and a memory 62. Those skilled in the art can understand that Figure 5 This is only an example of the computer device 60 and does not constitute a limitation on the computer device 60. It may include more or fewer components than shown in the figure, or combine some components, or different components. For example, the computer device may also include input / output devices, network access devices, buses, etc.

[0162] The so-called processor 61 may be a Central Processing Unit (CPU), or may also be other general-purpose processors, Graphics Processing Units (GPUs), Tensor Processing Units (TPUs), Digital Signal Processors (DSPs), Application Specific Integrated Circuits (ASICs), Field-Programmable Gate Arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0163] The memory 62 may be an internal storage unit of the computer device 60, such as the hard disk or memory of the computer device 60. The memory 62 may also be an external storage device of the computer device 60, such as a plug-in hard disk equipped on the computer device 60, a Smart Media Card (SMC), a Secure Digital (SD) card, a Flash Card, etc.

[0164] Furthermore, the memory 62 may also include both the internal storage unit of the computer device 60 and the external storage device. The memory 62 is used to store computer programs and other programs and data required by the computer device. The memory 62 may also be used to temporarily store data that has been output or will be output.

[0165] Please refer to Figure 6 , the terminal device is an electronic device 600, and the electronic device 600 is presented in the form of a general computing device. The components of the electronic device may include, but are not limited to: at least one processing unit 610, at least one storage unit 620, a bus 630 connecting different platform components (including the storage unit 620 and the processing unit 610), a display unit 640, etc.

[0166] Among them, the storage unit stores program code, and the program code can be executed by the processing unit 610, so that the processing unit 610 executes the steps according to various exemplary embodiments of the present invention described in the above method part of this specification. For example, the processing unit 610 can execute the steps as shown in Figure 2 .

[0167] The storage unit 620 may include a readable medium in the form of a volatile storage unit, such as a random access storage unit (RAM) 6201 and / or a cache storage unit 6202, and may further include a read-only storage unit (ROM) 6203.

[0168] The storage unit 620 may also include a program / utilities 6204 having a set (at least one) of program modules 6205. Such program modules 6205 include, but are not limited to: an operating system, one or more application programs, other program modules, and program data. Each or some combination of these examples may include the implementation of a network environment.

[0169] The bus 630 may represent one or more of several types of bus structures, including a memory bus or memory controller, a peripheral bus, an accelerated graphics port, a processing unit, or a local bus using any of a variety of bus structures.

[0170] The electronic device 600 may also communicate with one or more external devices 700 (such as a keyboard, a pointing device, a Bluetooth device, etc.), may also communicate with one or more devices that enable a user to interact with the electronic device 600, and / or may communicate with any device that enables the electronic device 600 to communicate with one or more other computing devices (such as a router, a modem). Such communication may be carried out through the input / output interface 650. Moreover, the electronic device 600 may also communicate with one or more networks (such as a local area network, a wide area network, and / or a public network, such as the Internet) through the network adapter 660. The network adapter 660 may communicate with other modules of the electronic device 600 through the bus 630. It should be understood that, although not shown in the figure, other hardware and / or software modules may be used in conjunction with the electronic device 600, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage platforms, etc.

[0171] Embodiment 4

[0172] The present invention also provides a storage medium, specifically a computer-readable storage medium, which is a memory device in a terminal device and is used to store programs and data. It can be understood that the computer-readable storage medium here can include both the built-in storage medium in the terminal device and, of course, the extended storage medium supported by the terminal device. It can be any tangible medium that contains or stores a program, and this program can be used by or in combination with an instruction execution system, device, or component. The computer-readable storage medium provides a storage space that stores the operating system of the terminal. And, in this storage space, there are also stored one or more instructions suitable for being loaded and executed by a processor, and these instructions can be one or more computer programs (including program codes). It should be noted that more specific examples of the computer-readable storage medium here include: an electrical connection having one or more wires, a portable disk, a hard disk, a random access memory, a read-only memory, an erasable programmable read-only memory, an optical fiber, a portable compact disk read-only memory, an optical storage device, a magnetic storage device, or any suitable combination of the above.

[0173] The computer-readable storage medium also includes data signals propagated in a baseband or as part of a carrier wave, in which the readable program code is carried. Such propagated data signals can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. The readable storage medium can also be any readable medium other than the readable storage medium, and this readable medium can send, propagate, or transmit a program for use by or in combination with an instruction execution system, device, or component. The program code contained on the readable storage medium can be transmitted using any appropriate medium, including but not limited to wireless, wired, optical cable, radio frequency, etc., or any suitable combination of the above.

[0174] The program code for performing the operations of the present invention can be written in any combination of one or more programming languages. The programming languages include object-oriented programming languages - such as Java, C++, etc., and also include conventional procedural programming languages - such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computing device, partially on the user's device, executed as an independent software package, partially on the user's computing device and partially on a remote computing device, or entirely on a remote computing device or server. In the case of a remote computing device, the remote computing device can be connected to the user's computing device through any type of network, including a local area network or a wide area network, or can be connected to an external computing device (for example, by using an Internet service provider to connect through the Internet).

[0175] One or more instructions stored in a computer-readable storage medium can be loaded and executed by a processor to implement the corresponding steps of the method for anti-jamming transmission and deployment of an air-ground network based on GAT in the above embodiments; the one or more instructions in the computer-readable storage medium are loaded and executed by the processor to perform the following steps:

[0176] Construct a communication scenario with multiple unmanned aerial vehicle (UAV) base stations, legitimate communication users, and multiple interference sources; wherein, each UAV base station conducts legitimate communication with the legitimate users it serves, while multiple interference sources are randomly distributed in the area to interfere with the legitimate communication; based on the given direction of the UAV base station transmitter signal arrival and the interference power of the interference sources, obtain the channel gain between each UAV base station transmitter and the corresponding legitimate user, the channel gain of other UAV base stations interfering with the legitimate users, and the channel gain from the interference sources to each legitimate user, and determine the received signal expression of the legitimate users under the influence of the interference sources; according to the obtained received signal expression at the legitimate users, construct a problem model with the transmit beamforming vectors of each UAV base station and the deployment positions of each UAV base station as variables and the anti-jamming rate of maximizing this anti-jamming system as the objective function; based on the constructed communication scenario and the obtained problem model, define the nodes and connection edges of the graph mapped by the scenario and the problem model; design a graph attention network encoder based on the problem model, and define the methods for message generation and message aggregation in the graph attention network; based on the constructed communication scenario and the problem model, define the exploration agents in the scenario and the problem model, and design the actions and states of the multi-agent deep deterministic policy gradient network. Design the loss function of the graph attention network, the rewards of the multi-agent deep deterministic policy gradient network, define the neural network learning method and parameters, train the neural network, and obtain a problem model with generalization ability to adapt to the dynamic topology and communication conditions of the air-ground network, and realize the output of anti-jamming communication decision-making based on the graph neural network in the interference environment.

[0177] The databases involved in the embodiments provided in the present application may include at least one of a relational database and a non-relational database. The non-relational database may include a distributed database based on blockchain, etc., and is not limited thereto. The processors involved in the embodiments provided in the present application may be a general-purpose processor, a central processing unit, a graphics processing unit, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., and is not limited thereto.

[0178] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. Generally, the components described and shown in the accompanying drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the claimed invention, but merely represents selected embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts fall within the scope of protection of the present invention.

[0179] Please refer to Figure 3 , in this solution, the above-mentioned graph neural network RGAT model is directly used in scenarios with other numbers of cells. When the cell distribution area remains unchanged, the cell distribution density changes with the number of users. On different cell distribution densities, the model shows good performance. As can be seen from the figure, compared with the multi-layer perceptron and the successive convex approximation optimization algorithm that are prone to falling into local optimal solutions and unable to capture complex relationships in the data, the graph neural network RGAT shows stable advantages. In addition to numerically showing that the graph neural network RGAT is superior to the successive convex approximation and the multi-layer perceptron algorithm, in actual use, the graph neural network RGAT model can be directly extended to scenarios with other numbers of cells because the addition and deletion of nodes in the graph do not affect the dimensions of the neural network during message passing, which is also an embodiment of the permutation equivariance of the graph neural network. When the unmanned aerial vehicle is used as a base station to communicate with users, its mobility requires the system to have the ability to efficiently generate beamforming vector solutions. Therefore, the graph neural network has obvious advantages over the multi-layer perceptron and the successive convex approximation algorithms in terms of scalability.

[0180] Under different numbers of cells, for the graph neural network RGAT algorithm, the successive convex approximation algorithm, and the multi-layer perceptron algorithm, the processing time for 1000 sets of samples is as Figure 4 shown. Since the execution of the neural network model can be completed quite quickly, the figure shows the time taken for running the neural network 2000 times to adjust the time magnitude for convenient comparison. Among the three algorithms considered, the successive convex approximation algorithm consumes the most time because each operation of the successive convex approximation optimization algorithm requires a large number of loops, while the graph neural network RGAT and the multi-layer perceptron can compress the operation time by setting the number of samples processed in each batch. Compared with the multi-layer perceptron, the operation time of the graph neural network RGAT is not much different, and both consume more operation time as the number of cells and the amount of data increase.

[0181] In summary, for the anti-interference transmission and deployment method and system of the air-ground network based on GAT of the present invention, the multi-UAV anti-interference communication network is mapped into a heterogeneous graph structure, the UAV-users and interference sources are respectively modeled as two types of nodes, and then the secure beamforming problem is converted into a graph learning model. The weights of the connection edges are updated through the attention mechanism, and the different types of connection relationships of the nodes are modeled as the interference between UAVs and the interference from the interference sources through the relational graph message passing mechanism. The anti-interference communication rate of the system is mapped into a loss function to realize the neural network training based on stochastic gradient descent. Further, the relational graph attention network is embedded into the deep reinforcement learning model to optimize the deployment of the multi-UAV base station positions.

[0182] The above content is only to illustrate the technical idea of the present invention, and the protection scope of the present invention cannot be limited thereby. Any changes made on the basis of the technical solution according to the technical idea proposed by the present invention shall fall within the protection scope of the claims of the present invention.

Claims

1. An anti-interference transmission and deployment method for air-ground networks based on GAT, characterized in that, It includes the following steps: Construct a communication scenario with multiple UAV base stations, legitimate communication users, and multiple interference sources; Based on the given direction of the UAV base station transmitter signal arrival and the interference power of the interference sources, obtain the channel gain between each UAV base station transmitter and the corresponding legitimate user, the channel gain of the interference from other UAV base stations to the legitimate user, and the channel gain from the interference sources to each legitimate user, and determine the received signal expression of the legitimate user under the influence of the interference sources; According to the obtained received signal expression at the legitimate user, construct a problem model with the transmit beamforming vectors of each UAV base station and the deployment positions of each UAV base station as variables and the objective function of maximizing the anti-interference rate of this anti-interference system; Based on the constructed communication scenario and the obtained problem model, define the nodes and connection edges of the mapped graph of the scenario and the problem model; Based on the problem model, design a graph attention network encoder, and define the methods of message generation and message aggregation in the graph attention network; Based on the constructed communication scenario and the problem model, define the exploration agents in the scenario and the problem model, and design the actions and states of the multi-agent deep deterministic policy gradient network; Design the loss function of the graph attention network, the reward of the multi-agent deep deterministic policy gradient network, define the neural network learning method and parameters, train the neural network, and obtain a problem model with generalization ability to adapt to the dynamic topology and communication conditions of the air-ground network, and realize the anti-interference communication decision output based on the graph neural network under the interference environment.

2. The anti-interference transmission and deployment method for air-ground network based on GAT according to claim 1, wherein In a communication scenario, the UAV base station has N antennas, and the legitimate user has a single antenna. K different UAV base stations and legitimate user pairs are distributed in a fixed area, denoted as The UAV base stations and users in the same group are represented by the same serial number. At the same time, there are J interference sources in the whole area, denoted as Interfere with the communication between the UAV base station and the user. Each UAV base station conducts legitimate communication with its served legitimate user. At the same time, multiple interference sources are randomly distributed in the area to interfere with the legitimate communication.

3. The anti-interference transmission and deployment method for the air-ground network based on GAT according to claim 1, characterized in that The received signal expression of the legitimate user under the influence of the interference sources is: Among them, \(L\) represents the geographical locations of all UAV base stations in the area, and \(l\) k represents the geographical coordinates of UAV base station \(k\), \(P\) is the transmission power of each interference source, and \(f\) k represents the beamforming vector of base station \(k\) for the corresponding user. \(h\) l,k \((L)\) represents the communication channel vector from base station \(l\) to user \(k\) when the geographical locations of all UAV base stations in the area are \(L\). \(g\) j,k \((L)\) represents the communication channel vector from interference source \(j\) to user \(k\) when the geographical locations of all UAV base stations in the area are \(L\). The transmission power of each interference source is expressed as \(P\), and \(n\) k represents the additive white Gaussian noise at the receiving end of user \(k\), and \(s\) k is the signal transmitted by the base station to user \(k\), \(K\) is the number of UAV base stations in the area, and \(H\) is the symbol for conjugate transpose.

4. The anti-interference transmission and deployment method for air-ground network based on GAT according to claim 1, wherein The problem model is described as follows: Among them, R(L,F,P) is the anti-interference rate of the system, γ k is the signal-to-interference-plus-noise ratio of the legitimate link for user k to receive signals, L represents the geographical locations of all UAV base stations in the area, l k represents the geographical coordinates of UAV base station k, P is the transmission power of each interference source, f k represents the beamforming vector of base station k for the corresponding user, h l,k (L) represents the communication channel vector from base station l to user k when the geographical locations of all UAV base stations in the area are L, g j,k (L) represents the communication channel vector from interference source j to user k when the geographical locations of all UAV base stations in the area are L, The transmission power of each interference source is expressed as P, n k represents the additive white Gaussian noise at the receiving end of user k that follows a complex Gaussian distribution.

5. The anti-interference transmission and deployment method for air-ground network based on GAT according to claim 1, wherein, Set the feature of the interfering source node j to its corresponding power p j , and supplement the corresponding dimension, and use the ~ symbol to distinguish the features of the interfering source nodes, which are described as follows: Design the feature of the connection edge as the channel corresponding to the interference relationship, and use the ~ symbol to label a type of connection edge connected to the interference source node: Among them, represents the connection edge from the UAV base station-legitimate user node l to the UAV base station-legitimate user node k, represents the connection edge from the interference source node j to the UAV base station-legitimate user node k, represents taking the real part of the interference channel gain from the base station in the UAV base station-legitimate user l to the receiver in the UAV base station-legitimate user k, represents taking the imaginary part of the interference channel from the base station in the UAV base station-legitimate user l to the receiver in the UAV base station-legitimate user k, represents taking the real part of the interference channel gain from the interference source j to the receiver in the UAV base station-legitimate user k, represents taking the imaginary part of the interference channel from the interference source j to the receiver in the UAV base station-legitimate user k.

6. The anti-interference transmission and deployment method for the air-ground network based on GAT according to claim 1, wherein In the graph attention network encoder, let be the feature of the interfering source node j in the d-th message passing. The message to be transmitted generated by the interfering source node j in the d-th message passing is Let be the feature of the UAV base station-legitimate user node k in the d-th message passing. The message to be transmitted generated by the UAV base station-legitimate user node k in the d-th message passing is According to the attention coefficient and the information generated by the neighbor nodes, the aggregated information of the central node k is described as follows: Among them, is regarded as the contribution of the connection edge pair between the neighbor node u and the central node k to constructing the feature of the central node k; When both node u and node k are drone base station-legitimate user nodes, the contribution information of neighbor node u to central node k and when node q is a drone base station-legitimate user node and node k is an interference source node, the contribution information in the d-th message passing is expressed as follows: Among them, is a feature aggregation neural network, is the neighbor node feature, is the edge feature, is the legal user feature.

7. The anti-interference transmission and deployment method for air-ground network based on GAT according to claim 1, characterized in that, For the cell UAV base station agent, the relationship between the action and the state transformation is described as follows: s′ k = [s k [0] + Δx k , s k [1] + Δy k , s k [2], s k [3]] = [x k + Δx k , y k + Δy k , x k,u , y k,u ​ where s′ is the updated state, and s k [0] is the first dimension of the state, and Δx k is the position update in the x direction, and Δy k is the position update in the y direction, and x k is the position in the x direction, and y k is the position in the x direction, and x k,u is the x-direction position of the served user, and y k,u is the y-direction position of the served user; The state of the interference source j at the nth step is set as: s′ j = [x j,J , y j,J , 0, p j ​ Among them, x j,J is the abscissa of interference source j, y j,J is the abscissa of interference source j, p j is the interference power of interference source j; The action of the interference source agent is set as the change of power: a j = [s j , Δp j ​ Among them, s j represents the symbol term, and Δp j represents the interference source power change value term; When the interference source agent j takes action a currently j the power state of the interference source agent j in the next step is described as follows: Among them, sign(·) represents the sign function, which outputs 1 when the input is positive and -1 when the input is negative. sign(·) is used to extract the sign, and Δp j is the minimum power change, is the set of interference sources, P is the maximum power threshold, and Δp i is the minimum power change of any interference source, and p i is the power of any interference source.

8. The anti-interference transmission and deployment method for air-ground network based on GAT according to claim 1, characterized in that The loss function loss of the graph attention neural network is: Among them, R(L,F,P) is the sum of the rates of anti-interference communication between each UAV base station and legitimate users in the area, and γ k is the anti-interference received signal-to-noise ratio of the UAV base station - legitimate user k.

9. The anti-interference transmission and deployment method for air-ground network based on GAT according to claim 8, characterized in that, The reward of the UAV agent in the multi-agent deep deterministic policy gradient network is set as the communication rate between the UAV base station and the legitimate user in the state after completing the current action; r k = log2(1 + γ k (L, gnn(L, P), P)) Among them, L is the position variable of the UAV base station, gnn(L, P) is the output of the inner-layer graph neural network in the current network state, and P is the power variable of the interference source; The reward of the interference source agent in the multi-agent deep deterministic policy gradient network is set according to the purpose of the game model, and the reward is set as a negative value: Among them, F is generated by the scheme for realizing anti-interference beamforming design based on the graph attention network, and β is the reward coefficient.

10. An anti-interference transmission and deployment system for air-ground networks based on GAT, characterized in that, It includes: A scenario module that constructs a communication scenario with multiple UAV base stations, legitimate communication users, and multiple interference sources; A signal module that, based on the given direction of the UAV base station transmitter signal arrival and the interference power of the interference sources, obtains the channel gain between each UAV base station transmitter and the corresponding legitimate user, the channel gain of the interference from other UAV base stations to the legitimate user, and the channel gain from the interference sources to each legitimate user, and determines the received signal expression of the legitimate user under the influence of the interference sources; A function module that constructs a problem model with the transmit beamforming vectors of each UAV base station and the deployment positions of each UAV base station as variables and the anti-jamming rate maximization of this anti-jamming system as the objective function according to the obtained expression of the received signal at the legitimate user; A network module that defines the nodes and connection edges of the graph mapped by the constructed communication scenario and problem model based on the constructed communication scenario and problem model; Based on the problem model, design a graph attention network encoder and define the methods of message generation and message aggregation in the graph attention network; An output module that defines the exploration agents in the constructed communication scenario and problem model, designs the actions and states of the multi-agent deep deterministic policy gradient network based on the constructed communication scenario and problem model; designs the loss function of the graph attention network and the rewards of the multi-agent deep deterministic policy gradient network, defines the neural network learning method and parameters, trains the neural network, and obtains a problem model with generalization ability to adapt to the dynamic topology and communication conditions of the air-ground network, and realizes the anti-jamming communication decision output based on the graph neural network in the interference environment.

Images