Multi-node cooperative interference frequency utilization situation reasoning cognitive method and system with limited perception capability

Abstract

The invention discloses a perception-limited multi-node cooperative interference frequency situation inference cognitive method and system, and provides an interference inference cognitive method based on node cooperation for solving the problems that the perception ability of a single node is limited and the overall interference rule is difficult to obtain, so as to meet the requirement of wireless ad hoc network cooperative interference resistance in a complex electromagnetic environment. The problems of spectrum situation reasoning and interference mode cognition under objective limiting conditions such as incomplete spectrum situation perception and limited samples are solved. The method can be extensively applied to the fields needing electromagnetic spectrum analysis support, such as spectrum planning and network electric countermeasure at the same time.

Description

TECHNICAL FIELD

[0001] The application relates to a method and system for reasoning and cognition of a frequency situation of multi-node cooperative interference with limited sensing capability, and belongs to the technical field of wireless communication. BACKGROUND

[0002] Spectrum situation reasoning and cognition in a complex electromagnetic environment is one of important prerequisites for implementing intelligent frequency use, anti-interference frequency decision, and realizing dynamic spectrum access by fully utilizing spectrum resources. Traditionally, spectrum situation reasoning and cognition depend on complete sensing of electromagnetic environment spectrum data, that is, analysis is performed on the basis of obtaining overall spectrum data. However, this requires large and special spectrum sensing equipment, and needs high-density deployment and strong information interaction capability to realize, which is not realistic in many actual situations based on the spectrum environment sensing of communication equipment itself. Especially, for sensing and reasoning the frequency use law of interference, since the working frequency band of the interference radiation source is very wide and exceeds the sensing range of general communication equipment, the existing spectrum situation reasoning and cognition method based on a deep neural network lacks practicality when applied to self-organizing network anti-interference and the like due to the lack of sensing capability. SUMMARY

[0003] The application provides a method and system for reasoning and cognition of a frequency situation of multi-node cooperative interference with limited sensing capability, and solves the problems disclosed in the background.

[0004] In order to solve the above technical problems, the technical scheme adopted by the application is as follows: The method for reasoning and cognition of a frequency situation of multi-node cooperative interference with limited sensing capability comprises the following steps: A plurality of nodes in an arbitrary area are grouped into a wireless self-organizing network and communication is established; For any node n, a certain frequency range is selected, spectrum data monitoring and recording are performed, and the spectrum monitoring data is interacted between nodes through the self-organizing network; Any node n obtains the spectrum monitoring data of other nodes through the self-organizing network, stores the data in the local node according to a time index, and forms a spectrum monitoring matrix; Any node n performs neural network training according to the local spectrum monitoring matrix, and performs sensing prediction according to the trained neural network to improve the sensing capability of the node.

[0005] Further, for any node n, the method for performing spectrum data monitoring and recording in a certain frequency range and interacting the spectrum monitoring data between nodes through the self-organizing network is as follows: For any node n, according to the working frequency band of its own business communication, a communication working center frequency point F nThe nearby frequency range 'a' is designated as the monitoring range for interference frequency sensing, i.e., the monitoring range G. n =[F n -a, F n +a], the bandwidth of each monitoring channel is B j Then the number of channels is M=G n / B j For any node n, starting from the initial frequency, monitor G one by one. n Within the range of channels, obtain the signal power value of the channel and record it to obtain a 1*M spectrum monitoring signal strength vector.

[0006] Furthermore, any node n obtains spectrum monitoring data from other nodes through the ad hoc network, stores it locally according to time index, and forms a spectrum monitoring matrix as follows: Assume that the spectrum monitoring signal strength vector shared by node m through the ad hoc network is 1*M*(N-1), and the spectrum monitoring signal strength vector shared by node h through the ad hoc network is 1*M. Combine them to obtain a 1*M*N spectrum monitoring signal strength vector; after T monitoring cycles, the spectrum monitoring matrix is ​​obtained.

[0007] Furthermore, methods for training neural networks based on local spectrum monitoring matrices include: The M×N dimensional spectrum matrices collected over T monitoring periods were time-aligned, and outliers caused by network latency or packet loss were removed. 200 samples were collected for each type of interference, resulting in 2400 data sets. A stratified sampling method was used to divide the data into training and validation sets, with 70% used for training and 30% for validation. The neural network architecture model was implemented in Python and consisted of convolutional blocks, thresholding denoising blocks, flattening layers, and fully connected layers. The convolutional blocks extracted frequency domain spatial features through convolution, and used ReLU activation and max pooling to achieve feature dimensionality reduction. The thresholding denoising block filtered out noise by adaptively setting thresholds and introduced a Dropout layer to prevent overfitting. The flattening layer converted the feature map into a one-dimensional vector. The fully connected layer completed the final classification decision through a hidden layer.

[0008] Furthermore, based on the local spectrum monitoring matrix, the method for training the neural network also includes: the neural network structure uses three convolutional blocks to extract high-level features of the data layer by layer, and a threshold denoising block is introduced after the convolutional blocks. The model is trained using the Adam optimizer with a learning rate of 0.001, 150 iterations, and 128 sets of data per iteration.

[0009] A second aspect of the present invention provides a multi-node cooperative interference frequency situational reasoning cognitive system with limited perception capabilities, comprising: The self-organizing network module consists of multiple nodes forming a wireless self-organizing network and establishing communication. Data interaction module, for any one node n, select a certain frequency range, spectrum data monitoring and recording, through the ad hoc network, the interaction of the spectrum monitoring data between the nodes; Spectrum monitoring matrix module, for any node n, through the ad hoc network, obtain the spectrum monitoring data of other nodes, store in local according to time index, form the spectrum monitoring matrix; Neural network module, for any one node n, according to the local spectrum monitoring matrix, neural network training, according to the training results to improve the sensing ability of the node.

[0010] The third aspect of the present application provides a computer readable storage medium storing one or more programs, characterized in that: the one or more programs include instructions, which, when executed by a computing device, cause the computing device to perform any of the above methods.

[0011] The fourth aspect of the present application provides a computing device, comprising: One or more processors, one or more memories and one or more programs, wherein the one or more programs are stored in the one or more memories and configured to be executed by the one or more processors, and the one or more programs include instructions for executing any of the above methods.

[0012] The present application has the beneficial effects: the present application is aimed at the problem that the sensing ability of a single node is limited and it is difficult to obtain the overall interference law, proposes an interference reasoning and cognition method based on node cooperation, to adapt to the need of wireless ad hoc network cooperative anti-interference in complex electromagnetic environment, solve the problems of incomplete spectrum situation awareness and sample limitation, and solve the problems of spectrum situation reasoning and interference mode cognition under the objective limitation conditions such as incomplete spectrum situation awareness and sample limitation. The present application can be simultaneously applied to the fields such as spectrum planning and network electric countermeasure which need electromagnetic spectrum analysis support. BRIEF DESCRIPTION OF DRAWINGS

[0013] Figure 1 is the flowchart of the embodiment of the present application; Figure 2 is the neural network structure diagram of the embodiment of the present application; Figure 3 is the real picture of the interference frequency rule of the embodiment of the present application; Figure 4 is the local matrix diagram of the spectrum monitoring of the embodiment of the present application; Figure 5 is the confusion matrix provided by the embodiment of the present application; Figure 6 is the interference rule identification result diagram provided by the embodiment of the present application. DETAILED DESCRIPTION

[0014] The application will be further described below with reference to the drawings. The following examples are only used to more clearly illustrate the technical solutions of the application, and cannot be used to limit the protection scope of the application.

[0015] As Figure 1 shown, the application provides a cognitive method for interference frequency state situation reasoning based on limited perception ability of multiple nodes. The core idea of the application is a multiple node data cooperation and fusion mechanism. Through cooperative perception and data fusion between nodes, and by using a deep learning model to dynamically adjust network parameters, efficient feature extraction and classification of interference patterns are realized, and the limitations of single node perception and identification are avoided.

[0016] Specifically, the following steps are included: Step 1, multiple nodes form a wireless ad hoc network and establish communication. While performing business communication, the nodes are also used to interact with spectrum sensing data. Let the frequency range of interest of the system be [Fs, Fe], and let the number of nodes in the system be N.

[0017] Step 2, for any node n, a certain frequency range is selected to monitor and record the signal power value, and then the monitoring record is interacted between nodes through the ad hoc network. Specifically, for any node n, according to the working frequency band of its own business communication, the center frequency point F n of its communication is selected, and a frequency range a around the center frequency point is selected as the monitoring range for interference frequency sensing, that is, the monitoring range G n =[F n -a, F n +a], and the channel bandwidth is B j , then the number of channels is M=G n / B j . Any node n starts from the starting frequency and monitors the channels in the range G n one by one to obtain the signal power value of the channel and record it to obtain a 1*M spectrum monitoring signal strength vector. Step 3, the nodes obtain a spectrum monitoring matrix based on spectrum monitoring data sharing. Any node n in the system obtains the spectrum monitoring signal strength vector of other nodes through the ad hoc network and stores it locally according to the time index to form a spectrum monitoring matrix. Specifically, assuming that the spectrum monitoring signal strength vector shared by node m through the ad hoc network is 1*M*(N-1), and the spectrum monitoring signal strength vector shared by node h through the ad hoc network is 1*M, the combination obtains a 1*M*N spectrum monitoring signal strength vector; after T monitoring periods, a spectrum monitoring matrix is obtained.

[0018] Step 4: For any node n, neural network training is performed based on the local spectrum monitoring matrix. Specifically, the M×N dimensional spectrum matrices collected over T monitoring periods are time-aligned, outliers caused by network latency or packet loss are removed, and 200 samples are collected for each type of interference, totaling 2400 sets of data. A stratified sampling method is used to divide the data into training and validation sets, with 70% used for training and 30% for validation. The neural network structure model is implemented in Python and consists of convolutional blocks, thresholding denoising blocks, flattening layers, and fully connected layers. The convolutional blocks extract frequency domain spatial features through convolution, and use ReLU activation and max pooling to achieve feature dimensionality reduction. The thresholding denoising block filters noise by adaptively setting a threshold and introduces a Dropout layer to prevent overfitting. The flattening layer converts the feature map into a one-dimensional vector. The fully connected layer completes the final classification decision through a 3D hidden layer. This network structure uses three convolutional blocks to extract high-level features of the data layer by layer, and a thresholding denoising block is introduced after the convolutional blocks to further improve the robustness and noise resistance of the model. The model was trained using the Adam optimizer with a learning rate of 0.001, for 150 iterations, with 128 sets of data trained each time.

[0019] The accuracy of a trained neural network recognition model is measured by the F1 score, defined as follows: , Precision rate is used to measure the proportion of true positives among samples predicted as positive by the model. ; Recall rate is used to measure the proportion of samples that are actually positive but are correctly predicted by the model. .in, This represents the number of true samples. This represents the number of false positives. This represents the number of false negatives.

[0020] The perception capabilities of this node are improved by performing perception predictions based on the trained neural network. Example

[0021] In this embodiment, spectral data for interference types based on 12 different interference-to-noise ratio (IRR) conditions were constructed. The frequency range of the interference signals was set between 1500 MHz and 2500 MHz, covering wide-band interference characteristics. To accurately reflect the dynamic changes of actual noise, noise intensity was described by noise power spectral density, whose value fluctuated randomly, thereby simulating changes in noise intensity under different environments.

[0022]

[0023] Table 1 Table 1 shows the parameter settings for the spectrum datasets of 12 interference types.Figure 2 Fig. 1 shows a schematic diagram of a neural network structure, Figure 3 Fig. 2 shows a real graph of interference frequency rules, Figure 4 Fig. 3 shows a local matrix graph of spectrum monitoring.

[0024] In processing the interference monitoring spectrum data in the 1 GHz frequency range, since the frequency band coverage capability of the spectrum monitoring sensing node is limited to 400 MHz, in order to achieve comprehensive interference monitoring, the 1 GHz frequency band is divided into three sub-frequency bands for monitoring. The specific division scheme is: node 1 covers the [0, 400] MHz frequency band, node 2 covers the [300, 700] MHz frequency band, and node 3 covers the [600, 1000] MHz frequency band.

[0025] Under this division, Figure 5 The identification results of the interference types detected by each node are respectively shown. Due to the limited frequency band coverage range of the spectrum monitoring sensing node, the identification accuracy of each node in the interference type is reduced because the data is not complete. Specifically, when the interference signal spans multiple sub-frequency bands, a single node can only obtain part of the spectrum information of the interference, and cannot completely capture all the characteristics of the interference signal, thereby affecting the accurate judgment of the model on the interference type. For example, for an interference signal with a center frequency of 500 MHz, node 1 cannot detect the high frequency part (> 400 MHz) of the signal, and node 3 cannot obtain the low frequency component (< 600 MHz), and although node 2 can cover the main frequency band of the interference, it still may lead to insufficient feature extraction due to the lack of complete spectrum context information. The identification accuracy of each type of interference after node cooperation is higher than that of single node, which further verifies the effectiveness of the node cooperation method.

[0026] Figure 6 The interference rule identification result graph is shown, it can be seen that with the change of the jamming-to-noise ratio, the proposed model is better than other models, especially when the jamming-to-noise ratio is -4 dB, the F1-score of the proposed model is improved by about 20.6% compared with the single node 1 model, and compared with the single node 2 model and the single node 3 model, the F1-score of the proposed model is also improved by 6% and 20.4% respectively. The experimental results show that the proposed method can effectively improve the accuracy of interference identification, and further prove that the proposed method has achieved good performance in the task of processing interference identification.

[0027] The perception ability limited multi-node cooperative interference frequency situation reasoning cognition method provided by the application effectively solves the problems of limited single node perception ability and incomplete spectrum information by constructing a neural network architecture of a distributed spectrum sensing network. The method adopts a multi-node data cooperation and fusion mechanism, and the identification accuracy of each type of interference is more than 90%, which is higher than the identification accuracy of each type of interference of a single node. The F1 score of a single node is 0.96, and the F1 score of interference types after node cooperation is 0.985, which further verifies the effectiveness of the method. The method not only significantly improves the overall identification rate, but also ensures that the identification accuracy of all interference types is better than that of a single node mode, which provides reliable technical support for solving the spectrum situation perception problem in a complex electromagnetic environment.

[0028] The above only describes the preferred embodiments of the application, and it should be noted that those skilled in the art can make some improvements and modifications without departing from the technical principles of the application, and these improvements and modifications should also be considered as the protection scope of the application.

[0029] A computer-readable storage medium storing one or more programs, the one or more programs comprising instructions which, when executed by a computing device, cause the computing device to perform the perception ability limited multi-node cooperative interference frequency situation reasoning cognition method.

[0030] A computing device comprising one or more processors, one or more memories, and one or more programs, wherein the one or more programs are stored in the one or more memories and configured to be executed by the one or more processors, and the one or more programs comprise a program for performing the perception ability limited multi-node cooperative interference frequency situation reasoning cognition method.

[0031] Those skilled in the art should understand that the embodiments of the application can be provided as a method, a system, or a computer program product. Therefore, the application can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the application can take the form of a computer program product implemented on one or more computer usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer usable program code.

[0032] The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart or flows and / or block diagram block or blocks. Figure 1 one or more flowcharts and / or blocks Figure 1 means for functionally implementing the steps in one or more flowcharts and / or blocks

[0033] The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart or flows and / or block diagram block or blocks. Figure 1 one or more flowcharts and / or blocks Figure 1 means for functionally implementing the steps in one or more flowcharts and / or blocks

[0034] The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart or flows and / or block diagram block or blocks. Figure 1 one or more flowcharts and / or blocks Figure 1 means for functionally implementing the steps in one or more flowcharts and / or blocks

[0035] The above merely provides an embodiment of the present application, but is not intended to limit the present application, and any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall fall within the scope of protection of the present application.

Claims

1. A cognitive method for reasoning about the situation using frequency-based interference with limited perception capabilities across multiple nodes, characterized by: To form a wireless ad hoc network and establish communication between multiple nodes in any area; For any node n, a certain frequency range is selected to monitor and record spectrum data, and the spectrum monitoring data is exchanged between nodes through a self-organizing network; Any node n can obtain spectrum monitoring data from other nodes through the self-organizing network, store it locally by time index, and form a spectrum monitoring matrix; For any node n, a neural network is trained based on the local spectrum monitoring matrix, and the perception prediction is performed based on the trained neural network to improve the perception capability of this node.

2. The multi-node collaborative interference frequency situation reasoning and cognition method with limited perception capability according to claim 1, characterized in that, For any node n, a certain frequency range is selected for spectrum data monitoring and recording. The method for exchanging this spectrum monitoring data among the nodes via a self-organizing network is as follows: For any node n, select its communication center frequency F based on the operating frequency band of its own business communication. n The nearby frequency range 'a' is designated as the monitoring range for interference frequency sensing, i.e., the monitoring range G. n =[F n -a, F n +a], the bandwidth of each monitoring channel is B j Then the number of channels is M=G n / B j For any node n, starting from the initial frequency, monitor G one by one. n Within the range of channels, obtain the signal power value of the channel and record it to obtain a 1*M spectrum monitoring signal strength vector.

3. The multi-node collaborative interference frequency situation reasoning and cognition method with limited perception capability according to claim 2, characterized in that, Any node n obtains spectrum monitoring data from other nodes through the ad hoc network, stores it locally according to time index, and forms a spectrum monitoring matrix as follows: Assume that the spectrum monitoring signal strength vector shared by node m through the ad hoc network is 1*M*(N-1), and the spectrum monitoring signal strength vector shared by node h through the ad hoc network is 1*M. Combine them to obtain a 1*M*N spectrum monitoring signal strength vector; after T monitoring cycles, the spectrum monitoring matrix is ​​obtained.

4. The multi-node cooperative interference frequency situation reasoning and cognition method with limited perception capability according to claim 3, characterized in that, Methods for training neural networks based on local spectrum monitoring matrices include: The M×N dimensional spectrum matrices collected over T monitoring periods were time-aligned, and outliers caused by network latency or packet loss were removed. 200 samples were collected for each type of interference, resulting in 2400 data sets. A stratified sampling method was used to divide the data into training and validation sets, with 70% used for training and 30% for validation. The neural network architecture model was implemented in Python and consisted of convolutional blocks, thresholding denoising blocks, flattening layers, and fully connected layers. The convolutional blocks extracted frequency domain spatial features through convolution, and used ReLU activation and max pooling to achieve feature dimensionality reduction. The thresholding denoising block filtered out noise by adaptively setting thresholds and introduced a Dropout layer to prevent overfitting. The flattening layer converted the feature map into a one-dimensional vector. The fully connected layer completed the final classification decision through a hidden layer.

5. The multi-node cooperative interference frequency situation reasoning and cognition method with limited perception capability according to claim 4, characterized in that, The method for training the neural network based on the local spectrum monitoring matrix also includes: the neural network structure uses three convolutional blocks to extract high-level features of the data layer by layer, and a threshold denoising block is introduced after the convolutional blocks. The model is trained using the Adam optimizer with a learning rate of 0.001, 150 iterations, and 128 sets of data per training iteration.

6. A multi-node collaborative interference frequency situational reasoning cognitive system with limited perception capabilities, characterized in that, include: The self-organizing network module consists of multiple nodes forming a wireless self-organizing network and establishing communication. The data interaction module is used to monitor and record spectrum data for any node n within a certain frequency range, and to exchange the spectrum monitoring data between nodes through an ad hoc network. The spectrum monitoring matrix module is used to obtain spectrum monitoring data of other nodes through the self-organizing network for any node n, store it locally by time index, and form a spectrum monitoring matrix. The neural network module is used to train a neural network for any node n based on the local spectrum monitoring matrix, and to improve the perception capability of the node based on the training results.

7. A computer-readable storage medium for storing one or more programs, characterized in that: The one or more programs include instructions that, when executed by a computing device, cause the computing device to perform any of the methods according to claims 1 to 5.

8. A computing device, characterized in that, include: One or more processors, one or more memories, and one or more programs, wherein the one or more programs are stored in the one or more memories and configured to be executed by the one or more processors, the one or more programs including instructions for performing any of the methods according to claims 1 to 5.

Images