Intelligent anti-interference communication implementation method and system based on dual-network structure

Abstract

The invention relates to an intelligent anti-interference communication implementation method and system based on a dual-network structure. The method comprises the following steps: S1, scanning a frequency spectrum; s2, data acquisition; s3, data preprocessing; s4, interference type identification; and S5, generating a communication strategy. A spectrum sensing system of a dual-network structure is responsible for rapid sensing of a spatial spectrum, and a frequency use behavior analysis network of a communication decision system analyzes a spectrum use condition of a working space of the system; the multi-agent communication decision-making network of the communication decision-making system focuses on communication strategy generation, efficient transmission of communication data is achieved, the communication rate and reliability are improved, the decision-making accuracy is improved, the decision-making time is shortened, the congestion degree and delay of the network are reduced, and the communication efficiency is improved. And the transmission efficiency of communication signals is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of wireless communication, and particularly relates to a method and system for realizing intelligent anti-interference communication based on a dual-network structure. Background Art

[0002] Communication anti-interference technology plays a crucial role in ensuring the efficient operation of equipment and the reliable transmission of information. Traditional anti-interference technologies such as spread-spectrum communication and adaptive power adjustment adopt a passive anti-interference method, which is difficult to adapt to increasingly complex and changeable scenarios of targeted malicious interference and intelligent interference.

[0003] Currently, the anti-interference used in the prior art is all of a single-network structure. Through the retrieval of public patent documents, the following similar public patent document was found:

[0004] The publication number is: CN 118338317 A, and the patent name is: An invention patent for an intelligent anti-interference method based on joint decision-making of power and modulation mode. By constructing a communication model under an artificial interference environment, two key dimensions of intelligent decision-making, namely power allocation and modulation mode selection, are integrated into a comprehensive decision-making framework, and MDP is used as the basic framework of the system. The anti-interference communication method adopted by this invention patent is a single-network one, and it has the following disadvantages and deficiencies:

[0005] (1) By using the detection ability of the transmitter itself, only the normalized interference power information of the channel can be sensed; (2) The communication decisions of network decision-making are the power and modulation mode of each channel. The magnitude of the interference power of each sensed channel cannot avoid interference by switching frequency points, and can only increase the probability of successful communication by adjusting power and adjusting the modulation mode; (3) A reasonable reward function is not designed; (4) The communication strategy decision is completed in a single-network structure. The single-network structure requires the neural network to sense the spatial spectrum and make communication decisions at the same time, which will lead to difficult convergence during network training.

[0006] Therefore, the present invention introduces artificial intelligence deep learning technology and reinforcement learning technology into the anti-interference communication system to realize an intelligent anti-interference communication system. In the intelligent anti-interference communication system, communication devices use spectrum scanning and spatial spectrum sensing technology based on deep learning to achieve rapid sensing of the spatial spectrum, and then combine reinforcement learning technology to achieve intelligent cooperative anti-interference networking communication among multiple communication devices, thereby solving the problem of highly reliable communication in scenarios of targeted malicious interference and intelligent interference in complex electromagnetic environments. Summary of the Invention

[0007] The object of the present invention is to overcome the deficiencies of the prior art and provide a method and system for realizing intelligent anti-interference communication based on a dual-network structure, including a spectrum sensing system and a communication decision-making system. The communication decision-making system is implemented by combining a frequency usage behavior analysis network and a multi-agent communication decision-making network. The frequency usage behavior analysis network is used to analyze the spectrum usage situation in space, and the multi-agent communication decision-making network is used to realize intelligent collaborative anti-interference networking communication between multiple communication devices.

[0008] The present invention solves its technical problems through the following technical solutions:

[0009] A method for realizing intelligent anti-interference communication based on a dual-network structure, characterized in that the steps of the method are as follows:

[0010] S1. Spectrum scanning: Divide the spectrum space within the entire communication bandwidth of the communication device into N sub-band spaces. The spectrum scanning device sequentially scans each sub-band to obtain spatial spectrum data and transmits it to the data acquisition module of the communication decision-making system.

[0011] S2. Data acquisition: The data acquisition module of the communication decision-making system acquires the spatial spectrum data of the spectrum scanning device and the communication quality data of the communication device, forms a communication quality matrix C ij of the communication quality matrix C ij and transmits it to the communication policy generation module;

[0012] S3. Data preprocessing: The spectrum cognition module receives the spatial spectrum data acquired by the data acquisition module, completes the preprocessing of the spatial spectrum data, and transmits the preprocessed data to the frequency usage behavior analysis network;

[0013] S4. Interference category identification: The frequency usage behavior analysis network performs interference identification and perception on the preprocessed spatial spectrum data to obtain a frequency usage behavior matrix and the frequency behavior matrix and transmits it to the communication policy generation module;

[0014] S5. Communication policy generation: The knowledge base stores the communication quality matrix C ij and the frequency usage behavior matrix and obtains a communication policy matrix A under the same communication quality matrix C ij and the frequency usage behavior matrix . The communication policy generation module retrieves from the knowledge base whether there is a pre-set policy. If it exists, it directly outputs the corresponding configuration information from the knowledge base and configures it to the communication device; if it does not exist, it generates an optimal communication policy matrix A through the multi-agent communication decision-making network and configures the corresponding configuration information to the communication device.

[0015] Moreover, the communication quality data of S2 includes signal-to-noise ratio snr, communication quality sqt and bit error rate P e The data acquisition module forms a communication quality matrix C with the communication quality data of each communication device. ij , the communication quality matrix C ij Transmitted to the communication strategy generation module,

[0016]

[0017] Where: i, j are the communication device numbers i, j∈1,2,3...U, and U is the number of communication devices in the intelligent communication system;

[0018] C ij A communication quality matrix representing when a communication device with device number i communicates with a communication device with device number j;

[0019] snr ij It represents the signal-to-noise ratio when the communication device with device number i communicates with the communication device with device number j, and when i=j, snr j =0;

[0020] sqt ij Indicates the signal quality when the communication device with device number i communicates with the communication device with device number j, and when i=j, sqt j =0;

[0021] represents the bit error rate when the communication device with device number i communicates with the communication device with device number j, and when i=j, P ej =0.

[0022] Moreover, the S3 is specifically:

[0023] (1) The spectrum recognition module segments the received N sub-band spatial spectrum data according to the set fixed time window t;

[0024] (2) performing time-frequency conversion on the N sub-band spatial spectrum data intercepted according to the fixed time window by means of fast Fourier transform, and converting the time domain data into the frequency domain;

[0025] (3) The spectrum recognition module sequentially converts the processed time window t m N sub-band spatial frequency domain data within Passed to the frequency behavior analysis network, where t represents the size of the time window t>0; m represents the number of time windows m∈1,2,...m; N represents the number of sub-bands, N∈1,2,...N.

[0026] Moreover, the S4 is specifically:

[0027] (1) The input of the frequency usage behavior analysis network is t m N sub - frequency bands in a time period, and the output is the interference category of the sub - frequency bands For the current t m The frequency - domain data of the N sub - frequency bands in the time period are sequentially analyzed by the frequency usage behavior analysis network to analyze the spatial spectrum distribution, and finally a frequency usage behavior matrix of m * N dimensions covering the entire communication bandwidth within the entire time period T (1 to m time windows) is formed where t m is the current time period, N is the frequency band number, representing the interference types on each sub - frequency band N within m time windows t; where t represents the size of the time window t > 0; m represents the number of time windows m ∈ 1, 2,... m; N represents the number of sub - frequency bands, N ∈ 1, 2,... N;

[0028] (2) Transfer the frequency usage behavior matrix to the communication strategy generation module for communication decision - making

[0029] Moreover, the specific content of S5 is as follows:

[0030] (1) The communication strategy generation module retrieves the communication quality matrix C ij and the frequency usage behavior matrix If there is a communication strategy, the communication strategy matrix A is transferred to the communication device. The communication device configures parameters according to the communication strategy matrix A, re - configures the device, and conducts communication; if not, it makes a decision through the multi - agent communication decision network, generates the optimal communication strategy matrix A and configures the corresponding configuration information to the communication device to achieve closed - loop control;

[0031] (2) The knowledge base receives the communication strategy matrix A generated by the communication strategy generation module, and stores the communication quality matrix C ij and the frequency usage behavior matrix and the communication strategy matrix A in the knowledge base to form a communication strategy knowledge base

[0032] Moreover, the multi - agent communication decision network is implemented by means of reinforcement learning. The Markov decision - making process is as follows:

[0033] (1) The state space of the multi - agent communication decision network is S

[0034] where: represents the spatial spectrum distribution perceived by the frequency usage behavior spectrum cognition module; C ij represents the communication quality situation during communication between each communication device. These two matrices include the external environment and internal communication quality situation of the entire communication system during the communication process

[0035] (2) The decision action space of the multi-agent communication decision network is the communication policy matrix A = [a f , a pow , a bw , a cm , where a f , a pow , a bw , a cm are configurable parameters of the communication device, and the specific meanings are as follows: a f is the frequency band, a pow is the power, a bw is the communication bandwidth, a cm is the carrier modulation method;

[0036] (3) The reward function of the multi-agent communication decision network is R = w 1 * r i + w 2 * r IN , where r i is the external reward, r IN is the internal reward, and w 1 , w 2 are weight parameters;

[0037] External reward:

[0038] Among them: U is the number of communication devices in the intelligent communication system;

[0039]

[0040] represents the bit error rate when the communication device with device number i communicates with the communication device with device number j, and when i = j, P ej = 0;

[0041] Internal reward r IN_i :

[0042] Among them: a i is the decision action of the multi-agent communication decision network, is the optimal decision action of the multi-agent communication decision network in the current state S.

[0043] An intelligent anti-jamming communication implementation system based on a dual-network structure, characterized in that it is used to execute the intelligent anti-jamming communication implementation method based on the dual-network structure, including a spectrum sensing system and a communication decision system. The frequency usage behavior analysis network of the communication decision system is used to analyze the spectrum usage situation in space, and the multi-agent communication decision network of the communication decision system is used for communication decision-making;

[0044] The spectrum sensing system includes a number of spectrum scanning devices and communication devices connected through channels; the spectrum scanning devices extract features from the spectrum in a wide frequency domain and transmit the spectrum data to the communication decision-making network; the communication devices communicate according to the communication decision output by the communication decision-making system.

[0045] The communication decision-making system includes a data acquisition module, a spectrum cognition module, a communication strategy generation module, a knowledge base, a frequency usage behavior analysis network, and a multi-agent communication decision-making network.

[0046] The data acquisition module receives the spectrum data of the spectrum scanning devices and the communication quality data of the communication devices, and forms a communication quality matrix C with the communication quality data. ij ,and transmits the communication quality matrix C ij to the communication strategy generation module.

[0047] The spectrum cognition module receives the spectrum data of the data acquisition module and performs data preprocessing, and transmits the preprocessed data to the frequency usage behavior analysis network.

[0048] The frequency usage behavior analysis network identifies and senses the interference category of the preprocessed data to obtain a frequency usage behavior matrix.

[0049] The knowledge base receives and stores the communication strategy matrix A under the same frequency usage behavior matrix communication quality matrix C ij .

[0050] The multi-agent communication decision-making network generates an optimal communication strategy matrix A in real time according to the frequency usage behavior matrix communication quality matrix C ij and configures the corresponding configuration information to the communication devices.

[0051] The positive effects that the present invention can produce are as follows:

[0052] 1. The spectrum sensing system with a dual-network structure of the present invention is responsible for the rapid sensing of spatial spectrum, and the frequency usage behavior analysis network of the communication decision-making system analyzes the spectrum usage situation in the working space of the system; the multi-agent communication decision-making network of the communication decision-making system focuses on the generation of communication strategies, realizes the efficient transmission of communication data, improves the communication rate and reliability. The way of division of labor and cooperation of the dual-network structure not only improves the accuracy of decision-making, but also reduces the decision-making time, thereby reducing the network congestion degree and delay, and improving the transmission efficiency of communication signals.

[0053] 2. The dual-network structure of the present invention has good scalability and flexibility, can be dynamically adjusted and optimized according to actual application requirements, provides a solid technical foundation for the application of multi-agent systems in complex environments, and performs particularly well when dealing with emergencies and high-concurrency communication requirements. It has broad application prospects and important research value in future development.

[0054] 3. The system of the present invention uses a frequency usage behavior analysis network (artificial intelligence method) to analyze the spatial interference information of each channel and each time period, and can perceive richer spatial spectrum information. At the same time, combined with the multi-agent communication decision network, a better communication strategy can be made.

[0055] 4. The present invention can achieve cooperative communication between multiple communication devices through the multi-agent communication decision network and the MADDPG training framework. It can not only adjust the power and debugging method, but also change the communication frequency point to achieve intelligent anti-interference avoidance.

[0056] 5. The present invention designs a reward function suitable for application in an actual communication system by adopting a joint reward function (including internal reward and external reward), and the implementation process is more complete.

[0057] 6. The present invention adopts a dual-network structure, and the training process is easier to converge; at the same time, the decision-making process combines the knowledge base with the multi-agent communication decision network. The existing strategies are directly generated from the knowledge base, and the decision-making time is shorter. The new strategy is generated by the multi-agent communication decision network and can be applied to more complex intelligent interference scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] Figure 1 is the system framework diagram of the present invention;

[0059] Figure 2 is the functional block diagram of the dual-network structure of the present invention;

[0060] Figure 3 is the system block diagram for constructing a random interference environment of the present invention;

[0061] Figure 4 is the reward function curve diagram for training the multi-agent communication decision network of the present invention in a random interference environment;

[0062] Figure 5 is the reward function curve diagram for the decision-making of the multi-agent communication decision network of the present invention in a random interference environment;

[0063] Figure 6 is the time curve diagram for the decision-making of the multi-agent communication decision network of the present invention in a random interference environment. DETAILED DESCRIPTION OF THE INVENTION

[0064] The present invention will be further described in detail through specific embodiments below. The following embodiments are only descriptive and not restrictive, and the protection scope of the present invention cannot be limited thereby.

[0065] As Figure 1 shown, a method for realizing intelligent anti-interference communication based on a dual-network structure is innovative in that the steps of the method are as follows:

[0066] S1. Spectrum scanning: Divide the spectrum space within the entire communication bandwidth of the communication device into N sub-band spaces, and the spectrum scanning device sequentially scans each sub-band to obtain spatial spectrum data and transmits it to the data acquisition module of the communication decision system;

[0067] S2. Data acquisition: The data acquisition module of the communication decision system acquires the spatial spectrum data of the spectrum scanning device and the communication quality data of the communication device, forms a communication quality matrix C ij for the communication quality data, and transmits the communication quality matrix C ij to the communication policy generation module;

[0068] The communication quality data includes signal-to-noise ratio snr, communication quality sqt, and bit error rate P e . The data acquisition module forms a communication quality matrix C from the communication quality data of each communication device ij ,

[0069]

[0070] where: i and j are communication device numbers, i, j ∈ 1, 2, 3... U, and U is the number of communication devices in the intelligent communication system;

[0071] C ij represents the communication quality matrix when the communication device with device number i communicates with the communication device with device number j;

[0072] snr ij represents the signal-to-noise ratio when the communication device with device number i communicates with the communication device with device number j, and when i = j, snr j = 0;

[0073] sqt ij represents the signal quality when the communication device with device number i communicates with the communication device with device number j, and when i = j, sqt j = 0;

[0074] represents the bit error rate when the communication device with device number i communicates with the communication device with device number j, and when i = j, P ej = 0.

[0075] S3. Data preprocessing: The spectrum cognition module receives the spatial spectrum data collected by the data acquisition module, completes the preprocessing of the spatial spectrum data, and transmits the preprocessed data to the frequency usage behavior analysis network;

[0076] (1) The spectrum cognition module segments and intercepts the received N sub-band spatial spectrum data according to the set fixed time window t;

[0077] (2) The N sub-band spatial spectrum data intercepted according to the fixed time window is subjected to time-frequency conversion by means of fast Fourier transform, and the time-domain data is converted to the frequency domain;

[0078] (3) The spectrum cognition module sequentially transmits the processed N sub-band spatial frequency domain data within the time window t m to the frequency usage behavior analysis network.

[0079] S4. Interference category identification: The frequency usage behavior analysis network performs interference identification and perception on the preprocessed spatial spectrum data to obtain the frequency usage behavior matrix

[0080] (1) The input of the frequency usage behavior analysis network is N sub-bands in the time period t m , and the output is the interference category of the sub-band; the frequency domain data of the N sub-bands in the current time period t m are sequentially analyzed by the frequency usage analysis module to analyze the spatial spectrum distribution, and finally form an m*N-dimensional frequency usage behavior matrix covering the entire communication bandwidth within the entire time period T (0 to m time windows) where t m is the current time period, N is the frequency band number, representing the interference types on each sub-band N within m time windows t. Where t represents the size of the time window t>0; m represents the number of time windows m∈1, 2,...m; N represents the number of sub-bands, N∈1, 2,...N;

[0081] (2) Transmit the frequency usage behavior matrix to the communication strategy generation module for communication decision-making.

[0082] S5. Communication strategy generation: The knowledge base stores the communication quality matrix C ij and the frequency usage behavior matrix and obtains the same communication quality matrix C ij and the frequency usage behavior matrix ​For the communication policy matrix A below, the communication policy generation module retrieves from the knowledge base whether there is a preset policy. If it exists, the corresponding configuration information is directly output from the knowledge base and configured to the communication device. If it does not exist, the optimal communication policy matrix A is generated through a multi-agent communication decision network, and the corresponding configuration information is configured to the communication device.

[0083] (1) First, the communication policy generation module retrieves the communication quality matrix C in the knowledge base ij , the frequency usage behavior matrix If there is a communication policy, the communication policy matrix A is passed to the communication device. The communication device configures parameters according to the communication policy matrix A, reconfigures the device, and conducts communication. If not, a decision is made through a multi-agent communication decision network to generate the optimal communication policy matrix and configure the corresponding configuration information to the communication device, thus realizing closed-loop control;

[0084] (2) The knowledge base receives the communication policy matrix A generated by the communication policy generation module and stores the communication quality matrix C ij , the frequency usage behavior matrix The communication policy matrix A in the knowledge base to form a communication policy knowledge base.

[0085] The multi-agent communication decision network is implemented by means of reinforcement learning. The Markov decision-making process is as follows:

[0086] (1) The state space S of the multi-agent communication decision network is

[0087] (2) The decision action space of the multi-agent communication decision network is the communication policy matrix A = [a f , a pow , a bw , a cm , where a f is the frequency band, a pow is the power, a bw is the communication bandwidth, a cm is the carrier modulation method;

[0088] (3) The reward function R of the multi-agent communication decision network is R = w 1 * r i + w 2 * r IN , where r i is the external reward, r IN is the internal reward, w 1 , w 2 are weight parameters;

[0089] External reward:

[0090] Where: U is the number of communication devices in the intelligent communication system, represents the bit error rate when the communication device with device number i communicates with the communication device with device number j, and when i = j, P ej = 0;

[0091] Internal reward r IN_i :

[0092] Where: a i is the decision-making action of the multi-agent communication decision-making network, is the optimal decision-making action of the multi-agent communication decision-making network in the current state S.

[0093] The multi-agent communication decision-making network is implemented using the MADDPG framework.

[0094] 1) The Actor network uses an LSTM network: the first layer is an LSTM network, the middle layer is a fully connected layer, using the Relu activation function, and the last layer uses a fully connected layer with an output quantity of n for outputting discrete action values, where n is the number of action spaces.

[0095] 2) The Critic network uses an LSTM network model: the first layer is an LSTM network, the middle layer is a fully connected layer, using the Relu activation function, and the last layer uses a fully connected layer with an output quantity of 1 to output the state-action value.

[0096] As Figure 2 shown, an intelligent anti-jamming communication implementation system based on a dual-network structure, its innovation lies in: including a spectrum sensing system and a communication decision system, the frequency usage behavior analysis network of the communication decision system is used to analyze the spectrum usage situation of the space, and the multi-agent communication decision-making network of the communication decision system is used for communication decision-making;

[0097] The spectrum sensing system includes a number of spectrum scanning devices and communication devices connected through channels; the spectrum scanning devices extract features from the wide-frequency spectrum and transmit the spectrum data to the communication decision-making network; the communication devices communicate according to the communication decisions output by the communication decision-making network;

[0098] The communication decision system includes a data acquisition module, a spectrum cognition module, a communication strategy generation module, a knowledge base, a frequency usage behavior analysis network, and a multi-agent communication decision-making network;

[0099] The data acquisition module receives the spectrum data of the spectrum scanning devices and the communication quality data of the communication devices, and forms the communication quality matrix C ij and transmits the communication quality matrix Cij Transmitted to the communication policy generation module;

[0100] The spectrum cognition module receives the spectrum data of the data acquisition module and performs data preprocessing, and transmits the preprocessed data to the frequency usage behavior analysis network;

[0101] The frequency usage behavior analysis network identifies and perceives the interference category of the preprocessed data to obtain the frequency usage behavior matrix

[0102] The knowledge base receives and stores the same frequency usage behavior matrix Communication quality matrix C ij The communication policy matrix A under;

[0103] The multi-agent communication decision-making network generates the optimal communication policy matrix A in real time according to the frequency usage behavior matrix Communication quality matrix C ij , and configures the corresponding configuration information to the communication device.

[0104] Such as Figure 3 , Figure 4 The random interference environment constructed for this system and the training results in this environment are as follows: The specific implementation steps are as follows:

[0105] Such as Figure 3 As shown, the communication bandwidth covered by the entire communication device is divided into 5 sub-bands. The interference devices on each sub-band can generate 5 types of interference, such as linear frequency sweep interference, random number interference, broadband interference, comb interference, and fixed frequency interference, through the random interference strategy control, and verify the performance of the communication decision-making system implemented by the present invention in this environment. The frequency behavior analysis network and the multi-agent communication decision-making network in the communication decision-making system constructed by the present invention need to be trained before actual application; after the training is completed, the frequency behavior analysis network and the multi-agent communication decision-making network are directly used for spatial spectrum analysis and communication decision-making. The specific implementation process is as follows:

[0106] In this environment, a distributed training method is adopted to train the frequency usage behavior analysis network and the multi-agent communication decision-making network respectively:

[0107] Train the frequency usage behavior analysis network. First, collect the spatial spectrum data of each sub-band in the space and the interference types sent through the constructed random interference environment to form the training label data, and then train the frequency usage behavior analysis network;

[0108] Train the multi-agent communication decision-making network. When training the multi-agent communication decision-making network, directly form the frequency usage behavior matrix according to the output result of the random interference strategy This ensures that the accurate spatial spectrum situation is input into the multi-agent communication decision-making network. Under this condition, the multi-agent communication decision-making network is trained, and the reward function value (rewards) curves of 5 multi-agent communication decision-making networks are as Figure 4 shown. In a random interference environment, when the multi-agent communication decision-making network is trained to 1200 epochs, the reward value reaches the maximum value of 1, and at this time, the communication success rate between 5 communication devices reaches 100%, that is, 5 communication devices have achieved intelligent collaborative anti-interference networking communication.

[0109] As Figure 5 , Figure 6 show the decision accuracy and decision time of the system in the randomly interfered environment constructed for this system. The specific implementation steps when the system is actually applied are as follows:

[0110] In the actual application process, the frequency behavior matrix generated by the frequency behavior analysis network When and the communication quality matrix C ij are input into the communication strategy generation module as Figure 2 shown for communication decision-making;

[0111] 100 experiments were carried out in a random interference environment, and the reward values of 5 communication devices are as Figure 5 shown. It can be seen that the reward function values of 5 multi-agent communication decision-making networks all reach about 0.985, that is, the decision accuracy is 98.5%; as Figure 6 shown, the decision time of the multi-agent communication decision-making network is between 0.9 ms and 1.2 ms. Therefore, the system can quickly complete spatial spectrum analysis and intelligent collaborative anti-interference communication decision-making.

[0112] Although the embodiments and drawings of the present invention are disclosed for illustrative purposes, those skilled in the art can understand that: various substitutions, changes and modifications are possible without departing from the spirit and scope of the present invention and the appended claims. Therefore, the scope of the present invention is not limited to the content disclosed in the embodiments and drawings.

Claims

1. A method for realizing intelligent anti-interference communication based on a dual network structure, characterized in that: The steps of the method are: S1. Spectrum scanning: Divide the spectrum space within the entire communication bandwidth of the communication device into N sub-band spaces. The spectrum scanning device scans each sub-band in turn to obtain spatial spectrum data and transmits it to the data acquisition module of the communication decision system; S2. Data acquisition: The data acquisition module of the communication decision system collects the spatial spectrum data of the spectrum scanning device and the communication quality data of the communication device, and converts the communication quality data into a communication quality matrix C ij , the communication quality matrix C ij transmitting to a communication strategy generation module; S3, data preprocessing: the spectrum recognition module receives the spatial spectrum data collected by the data acquisition module, completes the spatial spectrum data preprocessing, and transmits the preprocessed data to the frequency usage behavior analysis network; S4. Interference category identification: The frequency behavior analysis network performs interference identification and perception on the pre-processed spatial spectrum data to obtain the frequency behavior matrix And transform the frequency behavior into a matrix transmitting to a communication strategy generation module; S5. Communication strategy generation: the knowledge base stores the communication quality matrix C ij and frequency behavior matrix And obtain the same communication quality matrix C ij and frequency behavior matrix The communication strategy matrix A under the communication strategy generation module retrieves whether there is a preset strategy from the knowledge base, and if so, directly outputs the corresponding configuration information from the knowledge base and configures it to the communication device; If it does not exist, the optimal communication strategy matrix A is generated through the multi-agent communication decision network, and the corresponding configuration information is configured to the communication device.

2. The method for realizing intelligent anti-interference communication based on a dual network structure according to claim 1, characterized in that: The communication quality data of S2 includes signal-to-noise ratio snr, communication quality sqt and bit error rate P e The data acquisition module forms a communication quality matrix C with the communication quality data of each communication device. ij , the communication quality matrix C ij Transmitted to the communication strategy generation module, Where: i, j are the communication device numbers i, j∈1,2,3...U, and U is the number of communication devices in the intelligent communication system; C ij A communication quality matrix representing when a communication device with device number i communicates with a communication device with device number j; snr ij It represents the signal-to-noise ratio when the communication device with device number i communicates with the communication device with device number j, and when i=j, snr j =0; sqt ij Indicates the signal quality when the communication device with device number i communicates with the communication device with device number j, and when i=j, sqt j =0; represents the bit error rate when the communication device with device number i communicates with the communication device with device number j, and when i=j, P ej =0.

3. The method for realizing intelligent anti-interference communication based on a dual network structure according to claim 1, characterized in that: The S3 is specifically: (1) The spectrum recognition module segments the received N sub-band spatial spectrum data according to the set fixed time window t; (2) performing time-frequency conversion on the N sub-band spatial spectrum data intercepted according to the fixed time window by means of fast Fourier transform, and converting the time domain data into the frequency domain; (3) The spectrum recognition module sequentially converts the processed time window t m N sub-band spatial frequency domain data within Passed to the frequency behavior analysis network, where t represents the size of the time window t>0; m represents the number of time windows m∈1,2,...m; N represents the number of sub-bands, N∈1,2,...N.

4. The method for realizing intelligent anti-interference communication based on a dual network structure according to claim 1, characterized in that: The S4 is specifically: (1) The input of the frequency behavior analysis network is t m N sub-bands in the time period, the output is the interference category of the sub-band Set the current t m The frequency domain data of the N sub-bands in the time period are analyzed in turn through the frequency behavior analysis network to analyze the spatial spectrum distribution, and finally form an m*N-dimensional frequency behavior matrix covering the entire communication bandwidth in the entire time period T (1 to m time windows). where t m is the current time period, N is the frequency band number, Table 1 shows the interference type on each sub-band N in m time windows t; where t represents the size of the time window t>0; m represents the number of time windows m∈1,2,...m; N represents the number of sub-bands, N∈1,2,...N; (2) The frequency behavior matrix Passed to the communication strategy generation module for communication decision.

5. The method for realizing intelligent anti-interference communication based on a dual network structure according to claim 1, characterized in that: The S5 is specifically: (1) The communication strategy generation module retrieves the communication quality matrix C in the knowledge base ij , frequency behavior matrix If a communication strategy exists, the communication strategy matrix A is passed to the communication device, and the communication device configures parameters according to the communication strategy matrix A, reconfigures the device, and communicates; if it does not exist, a decision is made through the multi-agent communication decision network, the optimal communication strategy matrix A is generated, and the corresponding configuration information is configured to the communication device to achieve closed-loop control; (2) The knowledge base receives the communication strategy matrix A generated by the communication strategy generation module and converts the communication quality matrix C ij , frequency behavior matrix The communication strategy matrix A is stored in the knowledge base to form a communication strategy knowledge base.

6. The method for realizing intelligent anti-interference communication based on a dual network structure according to claim 5, characterized in that: The multi-agent communication decision network is implemented by reinforcement learning, and the Markov decision establishment process is: (1) The state space of the multi-agent communication decision network is S. in: represents the spatial spectrum distribution perceived by the spectrum recognition module of the frequency usage behavior; C ij Indicates the communication quality between various communication devices. These two matrices include the external environment and internal communication quality of the entire communication system during the communication process. (2) The decision action space of the multi-agent communication decision network is the communication strategy matrix A = [a f ,a pow ,a bw ,a cm ], where a f ,a pow ,a bw ,a cm It is a configurable parameter of the communication device, and its specific meaning is as follows: a f is the frequency band, a pow is power, a bw is the communication bandwidth, a cm is the carrier modulation method; (3) Reward function of the multi-agent communication decision network R = w1*r i +w2*r IN , where r i For external rewards, r IN Internal reward, w1 and w2 are weight parameters; External Rewards: Where: U is the number of communication devices in the intelligent communication system; represents the bit error rate when the communication device with device number i communicates with the communication device with device number j, and when i=j, P ej =0; Internal Rewards IN_i : Among them: a i is the decision action of the multi-agent communication decision network, It is the optimal decision action of the multi-agent communication decision network in the current state S.

7. An intelligent anti-interference communication implementation system based on a dual network structure, characterized in that: Used to execute the intelligent anti-interference communication implementation method based on the dual network structure as described in any one of claims 1 to 6, comprising a spectrum sensing system and a communication decision system, wherein the frequency usage behavior analysis network of the communication decision system is used to analyze the spectrum usage of the space, and the multi-agent communication decision network of the communication decision system is used for communication decision; The spectrum sensing system includes a plurality of spectrum scanning devices and communication devices connected through a channel; the spectrum scanning device extracts features of a spectrum in a wide frequency domain and transmits spectrum data to a communication decision network; the communication device communicates according to a communication decision output by the communication decision system; The communication decision system includes a data acquisition module, a spectrum recognition module, a communication strategy generation module, a knowledge base, a frequency usage behavior analysis network and a multi-agent communication decision network; The data acquisition module receives the spectrum data of the spectrum scanning device and the communication quality data of the communication device, and forms the communication quality matrix C ij , the communication quality matrix C ij transmitting to a communication strategy generation module; The spectrum recognition module receives spectrum data from the data acquisition module and performs data preprocessing, and transmits the preprocessed data to the frequency usage behavior analysis network; The frequency behavior analysis network identifies and perceives the interference category of the preprocessed data to obtain the frequency behavior matrix The knowledge base receives and stores the same frequency usage behavior matrix Communication quality matrix C ij The communication strategy matrix A below; The multi-agent communication decision network is based on the frequency behavior matrix Communication quality matrix C ij , generate the optimal communication strategy matrix A in real time, and configure the corresponding configuration information to the communication device.

Images