Direction finding method and system based on unmanned aerial vehicle cluster under multi-factor error condition

Abstract

The invention discloses a direction finding method based on an unmanned aerial vehicle cluster under a multi-factor error condition, and the method comprises the following steps: S10, in a laboratory, respectively considering a position error, a radio frequency channel amplitude-phase error, a nonlinear processing error and a coupling mutual interference error through a control variable method, and then constructing a network model considering the four errors; s20, deploying an unmanned aerial vehicle cluster carrying an electronic countermeasure reconnaissance load, sensing and receiving signals from each target radiation source, and recording data; s30, performing error correction on the data received in the step S20 by using the network model in the step S10; s40, after the error correction data in the step S30 are received, performing virtual-real separation and then entering a neural network; s50, sequentially processing the data in the neural network for six times to obtain an output function, combining virtual and real parts of the output function to obtain a complex function, multiplying the complex function with the steering vector to obtain a spatial spectrum, and searching peaks according to the spatial spectrum to carry out direction finding on each radiation source; the invention further discloses a corresponding system. The accuracy, reliability and flexibility of electronic countermeasure reconnaissance are improved, and the direction finding process is simplified.

Description

Technical Field

[0001] The present invention relates to the technical field of electronic countermeasures, and more specifically, to a direction finding method and system based on an unmanned aerial vehicle (UAV) cluster under multi-factor error conditions. Background Art

[0002] With the rapid development of modern electronic information technology, the complexity of the electromagnetic environment is increasing day by day. Electromagnetic situation awareness has become one of the key technologies to ensure information security and implement electronic countermeasures. Among them, accurate direction finding and positioning of target radiation sources are not only an important part of electromagnetic situation awareness but also a prerequisite for military operations such as precise strikes, electronic reconnaissance, and anti-reconnaissance. Traditional ground electronic countermeasure reconnaissance means are restricted by factors such as terrain occlusion and line-of-sight distance limitations and are difficult to meet the requirements in complex and changing battlefield environments.

[0003] To solve these problems, some existing technologies have proposed a UAV-borne direction finding and ambiguity resolution method based on an interferometer. By optimizing the interferometer direction finding system, the direction finding accuracy and stability have been effectively improved. However, it cannot achieve simultaneous direction finding of multiple radiation sources. There are also some existing technologies that implement direction finding based on UAV clusters and intelligent metasurfaces. Their system is compressive sensing sparse reconstruction, innovating the atomic norm direction finding method. But the influence of actual existing errors is not considered. There are also some existing technologies that combine neural networks with direction finding methods, consider the situation of existing errors, fit functions by network training methods, and eliminate the influence of errors. However, the error model is relatively single and there is a gap from the actual situation; the input and output of the neural network are the measured value and the actual value respectively, and the characteristic information of the original data is not fully utilized. Summary of the Invention

[0004] In view of the above defects or improvement requirements of the existing technology, the present invention provides a direction finding method and system based on a UAV cluster under multi-factor error conditions. By comprehensively considering the UAV position error, radio frequency channel amplitude-phase inconsistency error, non-linear processing error, and coupling and mutual interference error between the reconnaissance payloads of each UAV, a more accurate and practical network model is constructed; the accuracy and reliability of UAV clusters in electronic countermeasure reconnaissance are significantly improved.

[0005] To achieve the above object, according to the first aspect of the present invention, there is provided a direction finding method based on a UAV cluster under multi-factor error conditions, including the following steps: S10, network training. Under the ideal state in the laboratory, by the method of controlling variables, respectively consider the UAV position error, radio frequency channel amplitude-phase inconsistency error, non-linear processing error, and coupling and mutual interference error between the reconnaissance payloads of each UAV, and then construct a network model considering the four errors. S20, Data acquisition: Deploy a drone swarm and equip it with electronic countermeasure reconnaissance payloads. When the drones are flying in the air, they sense and receive signals from various target radiation sources, and at the same time record the data received by the electronic countermeasure reconnaissance payloads of each drone. S30, Data correction: Use the network model in S10 to correct the errors in the data received in S20. S40, Virtual-real separation processing: After the ground station receives the data with errors corrected in S30, it performs virtual-real separation and then sends this data into the neural network. S50, Neural network processing: In the neural network, the data sequentially passes through a fully connected layer, tensor deformation, convolutional layer, batch normalization, activation function, and fully connected layer to obtain an output function. Combine the imaginary part and the real part of the output function to obtain a complex function, and then multiply this complex function by the steering vector to obtain a spatial spectrum. Finally, peak search is performed based on the spatial spectrum to achieve direction finding for each radiation source.

[0006] Furthermore, the method for constructing the network model considering the four types of errors in S10 is as follows: S11, Conduct network training in the laboratory without considering the position error of the drones, amplitude-phase inconsistency error of the RF channels, nonlinear processing error, and coupling and mutual interference error between the reconnaissance payloads of each drone, and receive data in an ideal state. S12, Conduct training considering only the position error of the drones in S11. Assume that the position error follows a Gaussian distribution with a mean of 0, and the standard deviation of the position error follows a uniform distribution, and the maximum value is ; S13, Conduct training considering only the amplitude-phase inconsistency error of the RF channels in S11. Assume that the amplitude-phase inconsistency error follows a Gaussian distribution with a mean of 0, the standard deviation of the RF channel amplitude and the standard deviation of the RF channel phase follow a uniform distribution, and the maximum values of the standard deviation of the RF channel amplitude and the standard deviation of the RF channel phase are and ; S14, Conduct training considering only the nonlinear processing error in S11. The nonlinear error is described by , where is the nonlinear error coefficient; S15, Conduct training considering only the coupling and mutual interference error in S11. The coupling coefficient follows a uniform distribution, and the maximum value of the coupling coefficient is ; At S16, perform training by comprehensively considering the position error in S12, the amplitude-phase inconsistency error in S13, the non-linear processing error in S14, and the coupling and mutual interference error in S15, and construct a network model considering the four errors.

[0007] Further, the method for recording the data received by each UAV electronic countermeasure reconnaissance payload in S20 is as follows: There are radiation sources, and the incoming wave direction of each radiation source is , where , is the radiation source number, and there are UAVs, with the UAV at the outermost edge of the cluster as the reference UAV; The th UAV is at a distance of from the reference UAV, where ; At moment, the total data received by the reference UAV's electronic countermeasure reconnaissance payload from radiation sources: , In the formula, is the data received by the reference UAV's electronic countermeasure reconnaissance payload from the th radiation source at moment. Then, without considering errors, the data received by the th UAV's electronic countermeasure reconnaissance payload is : , In the above formula, is the radiation source signal wavelength, is the noise, j is the complex unit.

[0008] Further, the method for correcting the errors of the received data in S30 is as follows: Considering the position error of the UAV, the distance between the th UAV and the reference UAV will change directly; Considering the amplitude-phase inconsistency error of the RF channel, different amplitude gains will be generated in the RF channel of the UAV electronic countermeasure reconnaissance payload. The amplitude gain corresponding to the th UAV is denoted as ; and different UAV reconnaissance payload RF channels have different phase shifts. The phase shift corresponding to the th UAV is denoted as ; Considering the non - linear processing error, the received data is processed by the function g to compensate for the non - linear processing error, where the function g does not have an additional impact on the noise part of the data during the processing; After considering the above three errors, the received data is : .

[0009] Furthermore, considering that there is also coupling interference between the reconnaissance payloads of each UAV, the coupling coefficient between the electronic counter - reconnaissance payloads of the th UAV and the th UAV is , then the received data changes from to : ; In the formula, is the amplitude gain of the th UAV, is the phase shift of the th UAV, is the distance between the th UAV and the reference aircraft; then the total data received by N UAVs is: .

[0010] Furthermore, the method for separating real and imaginary parts in S40 is as follows: Separate the real part and the imaginary part of the total data received by the N UAVs to obtain the neural input value : , In the formula, and respectively represent 's real part and imaginary part.

[0011] Furthermore, the method for obtaining the output function in S50 is as follows: The input value Y of the neural network is the batch - processed data obtained after multiple snapshots accumulation of the neural input function , then: , In the formula, is the batch - processing quantity, 's dimension is , N is the total number of UAVs, and it specifically includes the following steps: S41, First, it passes through a fully connected layer. The output dimension of the fully connected layer is , where is the number of filters in the subsequent convolutional layer; is the value required for dimensional expansion; S42, after output from the fully connected layer, tensor deformation is performed, and the output dimension is

[0012] S43, then it enters the convolutional layer, and the output is processed by batch normalization and activation function. The number of convolutional layers is , and the final output dimension is ; S44, passing through the fully connected layer again, transforms the dimension into to obtain the output function.

[0013] Furthermore, the method for obtaining the spatial spectrum in S50 is as follows: Combining the imaginary part and the real part of the output function to obtain a complex function z, then the spatial spectrum is: , In the formula, is the steering vector.

[0014] According to the second aspect of the present invention, a direction finding system based on an unmanned aerial vehicle cluster under multi-factor error conditions is provided, including: A network training module, under the ideal state of the laboratory, by the method of controlling variables, respectively considering the position error of the unmanned aerial vehicle, the amplitude-phase inconsistency error of the radio frequency channel, the non-linear processing error, and the coupling and mutual interference error between the reconnaissance payloads of each unmanned aerial vehicle, and then constructing a network model considering the four errors; A data acquisition module, deploying an unmanned aerial vehicle cluster and equipping it with an electronic countermeasure reconnaissance payload; when the unmanned aerial vehicle is flying in the air, it senses and receives signals from each target radiation source, and at the same time records the data received by the electronic countermeasure reconnaissance payload of each unmanned aerial vehicle; A data correction module, using the network model in S10 to correct the errors of the data received in S20; A virtual-real separation processing module, after the ground station receives the data with errors corrected in S30, performs virtual-real separation and then sends these data into the neural network; A neural network processing module, in the neural network, the data sequentially passes through a fully connected layer, tensor deformation, a convolutional layer, batch normalization, an activation function, and a fully connected layer to obtain an output function, combines the imaginary part and the real part of the output function to obtain a complex function, multiplies the complex function by the steering vector to obtain a spatial spectrum, and finally realizes direction finding for each radiation source according to the peak search of the spatial spectrum.

[0015] Generally speaking, compared with the prior art, the above technical solution conceived by the present invention can achieve the following beneficial effects: 1. In the direction finding method of the present invention, by comprehensively considering the UAV position error, the amplitude-phase inconsistency error of the RF channels, the non-linear processing error, and the coupling and mutual interference error between the reconnaissance payloads of each UAV, a more accurate and practical network model is constructed, improving the accuracy and reliability of the UAV cluster in electronic countermeasure reconnaissance.

[0016] 2. The direction finding method of the present invention adopts a unique scheme with the original signal data as the input, retains the original features of the signal, and extracts features through the convolutional layer, which can capture signal information more comprehensively, avoiding information loss and feature distortion. At the same time, through the tensor deformation technology, the complexity of data processing is reduced and the calculation efficiency is improved. In addition, the multiplication operation of the complex function at the output end and the steering vector directly obtains the spatial spectrum without estimating the number of signal sources, further simplifying the direction finding process and enhancing the adaptability and flexibility of the system.

[0017] 3. In the network training process of the direction finding method of the present invention, by gradually increasing the consideration of error factors, the generalization ability of the model is improved, and the stability and accuracy of the model under different error conditions are ensured. In addition, direction finding of multiple radiation sources is realized through spatial spectrum peak searching, providing an efficient and accurate means for electronic reconnaissance and intelligence collection. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 Schematic diagram of the measurement method provided by a preferred embodiment of the present invention; Figure 2 Information processing flow provided by a preferred embodiment of the present invention; Figure 3 Neural network architecture provided by a preferred embodiment of the present invention; Figure 4 Network training process provided by a preferred embodiment of the present invention; Figure 5 Test results provided by a preferred embodiment of the present invention; Figure 6 Schematic diagram of the structure of a direction finding system based on a UAV cluster under multi-factor error conditions provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0019] In order to make the objectives, technical solutions, and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0020] Based on the above problems, the present invention proposes a direction finding method based on an unmanned aerial vehicle (UAV) cluster under multi-factor error conditions. This method combines laboratory research and practical applications, and constructs a network model to address various error sources such as UAV position error, radio frequency channel error, non-linear processing error, and interference between UAVs.

[0021] In actual operation, the UAV cluster carries reconnaissance equipment to collect signal data from different radiation sources in the air. These data are then corrected for errors through a specially designed algorithm to improve the accuracy of the data. Next, the data is separated into real and imaginary parts and input into a neural network for further analysis.

[0022] Through multi-layer processing, including steps such as fully connected, convolutional, normalization, and activation functions, the neural network finally outputs a function. This function is combined with the steering vector to generate a spatial spectrum for determining the direction of the radiation source.

[0023] Specifically, Please refer to Figure 1 and Figure 4 , the present invention relates to a direction finding method based on an unmanned aerial vehicle (UAV) cluster under multi-factor error conditions, including the following steps: S10, network training. Under ideal laboratory conditions, by using the method of controlling variables, the position error of the UAV, the amplitude-phase inconsistency error of the radio frequency channel, the non-linear processing error, and the coupling and mutual interference error between the reconnaissance payloads of each UAV are considered separately, and then a network model considering the four errors is constructed.

[0024] The method for constructing a network model considering the four errors in S10 is as follows: S11, conduct network training in the laboratory without considering the position error of the UAV, the amplitude-phase inconsistency error of the radio frequency channel, the non-linear processing error, and the coupling and mutual interference error between the reconnaissance payloads of each UAV, and receive data in an ideal state; S12, conduct training considering only the position error of the UAV in S11. Assume that the position error follows a Gaussian distribution with an average value of 0, and the standard deviation of the position error follows a uniform distribution with a maximum value of ; S13. Conduct training considering only the amplitude-phase inconsistency error in S11. Assume that the amplitude-phase inconsistency error follows a Gaussian distribution with a mean of 0, the standard deviation of the radio frequency channel amplitude and the standard deviation of the radio frequency channel phase follow a uniform distribution. The maximum values of the standard deviation of the radio frequency channel amplitude and the standard deviation of the radio frequency channel phase are and respectively; S14. Conduct training considering only the non-linear processing error in S11. The non-linear error is described by where is the non-linear error coefficient; S15. Conduct training considering only the coupling and mutual interference error in S11. The coupling coefficient follows a uniform distribution, and the maximum value of the coupling coefficient is ; S16. Conduct training considering comprehensively the position error in S12, the amplitude-phase inconsistency error in S13, the non-linear processing error in S14, and the coupling and mutual interference error in S15, and construct a network model considering the four types of errors.

[0025] S20. Data acquisition. Deploy a drone swarm and equip it with an electronic countermeasure reconnaissance payload. When the drones are flying in the air, they sense and receive signals from various target radiation sources, and at the same time record the data received by the electronic countermeasure reconnaissance payloads of each drone.

[0026] The method for recording the data received by the electronic countermeasure reconnaissance payloads of each drone in S20 is as follows: There are radiation sources, and the incoming wave direction of each radiation source is where , is the radiation source number. There are drones, and the outermost drone in the swarm is used as the reference drone; The distance between the th drone and the reference drone is where ; At time, the total data received by the electronic countermeasure reconnaissance payload of the reference drone from radiation sources: , In the formula, is the data received by the electronic countermeasure reconnaissance payload of the reference drone from the rd radiation source at time. Then, without considering errors, the data received by the electronic countermeasure reconnaissance payload of the th drone is : , In the above formula, is the wavelength of the radiation source signal, is the noise, j is the imaginary unit.

[0027] S30, data correction, using the network model in S10, correct the error of the data received in S20; The method for correcting the error of the received data in S30 is: Considering the position error of the UAV, the distance between the th UAV and the reference aircraft is which will change directly; Considering the amplitude-phase inconsistency error of the RF channel, different amplitude gains will be generated in the RF channel of the UAV electronic countermeasure reconnaissance payload. Among them, the amplitude gain corresponding to the th UAV is denoted as ; and different UAV reconnaissance payload RF channels have different phase shifts. Among them, the phase shift corresponding to the th UAV is denoted as ; Considering the non-linear processing error, process the received data through the function g to compensate for the non-linear processing error. Among them, the function g does not have an additional impact on the noise part of the data during the processing; After considering the above three errors, the received data is : .

[0028] Considering that there is also coupling interference between the reconnaissance payloads of each UAV, the coupling coefficient between the reconnaissance payloads of the th UAV and the th UAV is , then the received data changes from to : ; In the formula, is the amplitude gain of the th UAV, is the phase shift of the th UAV, is the distance between the th UAV and the reference aircraft; then the total data received by N UAVs is: .

[0029] Please refer to Figure 2 and Figure 3 For S40, real - imaginary separation processing: After the ground station receives the error - corrected data in S30, perform real - imaginary separation and then send this data into the neural network.

[0030] The method for real - imaginary separation in S40 is as follows: Separate the real part and the imaginary part of the total data received by the N UAVs to obtain the neural input value : , In the formula, and respectively represent the real part and the imaginary part.

[0031] For S50, neural network processing: In the neural network, the data sequentially passes through a fully - connected layer, tensor deformation, convolutional layer, batch normalization, activation function, and fully - connected layer to obtain an output function. Combine the imaginary part and the real part of the output function to obtain a complex - valued function, then multiply this complex - valued function by the steering vector to obtain a spatial spectrum, and finally perform peak searching based on the spatial spectrum to achieve direction finding for each radiation source.

[0032] The method for obtaining the output function in S50 is as follows: The input value Y of the neural network is the batch - processed data obtained after multiple snapshots accumulation of the neural input function , then: , In the formula, is the batch - processing quantity, The dimension of is S41, First, pass through the fully - connected layer. The output dimension of the fully - connected layer is , where is the number of filters in the subsequent convolutional layer; is the value required for dimension expansion; S42, After output from the fully - connected layer, perform tensor deformation, and the output dimension is

[0033] S43, Then enter the convolutional layer, and the output is processed by batch normalization and activation function. The number of convolutional layers is , and the final output dimension is ; S44, Pass through the fully - connected layer again, and transform the dimension to , the output function is obtained.

[0034] The method for obtaining the spatial spectrum in S50 is as follows: Combining the imaginary part and the real part of the output function to obtain a complex function z, then the spatial spectrum is: , In the formula, is the steering vector.

[0035] As Figure 6 shown, as another aspect of the present invention, it also relates to a direction finding system based on an unmanned aerial vehicle (UAV) cluster under multi-factor error conditions, including: A network training module, under ideal laboratory conditions, by the method of controlling variables, respectively considering the position error of the UAVs, the amplitude-phase inconsistency error of the radio frequency channels, the non-linear processing error, and the coupling and mutual interference error between the reconnaissance payloads of each UAV, and then constructing a network model considering the four errors; A data acquisition module, deploying a UAV cluster and equipping it with electronic countermeasure reconnaissance payloads; when the UAVs fly in the air, they sense and receive signals from each target radiation source, and at the same time record the data received by the electronic countermeasure reconnaissance payloads of each UAV; A data correction module, using the network model in S10 to correct the errors of the data received in S20; A virtual-real separation processing module, after the ground station receives the data with errors corrected in S30, performs virtual-real separation and then sends these data into a neural network; A neural network processing module, in the neural network, the data sequentially passes through a fully connected layer, tensor deformation, a convolutional layer, batch normalization, an activation function, and a fully connected layer to obtain an output function, combines the imaginary part and the real part of the output function to obtain a complex function, then multiplies the complex function by the steering vector to obtain a spatial spectrum, and finally realizes direction finding for each radiation source according to the peak search of the spatial spectrum.

[0036] Embodiment 1 The test conditions are set as follows. The number of UAVs , the number of radiation sources . Neural network hyperparameters: batch size , the number of filters in the convolutional layer , the number of convolutional layers . Parameters related to the error situation: the maximum standard deviation of the UAV position error , the maximum standard deviation of the radio frequency channel amplitude inconsistency , the maximum standard deviation of the radio frequency channel phase inconsistency , the non-linear error coefficient , the maximum coupling coefficient .

[0037] The test results are as follows Figure 5 shown. The abscissa is the direction finding angle and the ordinate is the spatial spectrum. It can be seen that the actual incoming wave directions of each radiation source , , , and the direction finding results of the present invention are , , , which are relatively close and can achieve accurate direction finding under various error factors.

[0038] It is easy for those skilled in the art to understand that the above are only preferred embodiments of the present invention and are not used to limit the present invention. Any modifications, equivalent replacements, and improvements made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A direction finding method based on drone clusters under multi-factor error conditions, characterized in that: The following steps are involved: S10, network training, under ideal laboratory conditions, through the control variable method, the position error of the UAV, the amplitude and phase inconsistency error of the RF channel, the nonlinear processing error and the coupling and mutual interference error between the reconnaissance payloads of each UAV are considered respectively, and then a network model considering the four errors is constructed; S20, data acquisition, deploying a drone cluster and making it carry an electronic countermeasure reconnaissance payload; when the drones are flying in the air, they sense and receive signals from each target radiation source, and at the same time record the data received by the electronic countermeasure reconnaissance payload of each drone; S30, data correction, using the network model in S10 to perform error correction on the data received in S20; S40, virtual-real separation processing, after receiving the error-corrected data in S30, the ground station performs virtual-real separation and then sends the data to the neural network; S50, neural network processing, in which the data passes through a fully connected layer, a tensor deformation, a convolutional layer, a batch normalization, an activation function and a fully connected layer in sequence to obtain an output function, the imaginary part and the real part of the output function are combined to obtain a complex function, and then the complex function is multiplied by the steering vector to obtain a spatial spectrum, and finally the direction finding of each radiation source is achieved according to the peak finding of the spatial spectrum.

2. The direction finding method based on drone clusters under multi-factor error conditions according to claim 1 is characterized in that: The method for constructing the network model considering the four errors in S10 is: S11, network training is performed in the laboratory without considering the position error of the UAV, the amplitude and phase inconsistency error of the RF channel, the nonlinear processing error, and the coupling and mutual interference error between the reconnaissance payloads of each UAV, and the data of the ideal state is received; S12, perform training considering only the position error of the drone in S11, assuming that the position error follows a Gaussian distribution, the average value is 0, and the standard deviation of the position error is It follows a uniform distribution, with a maximum value of ; S13, training is performed considering only the amplitude-phase inconsistency error in S11, assuming that the amplitude-phase inconsistency error obeys a Gaussian distribution, the average value is 0, and the RF channel amplitude standard deviation is And the RF channel phase standard deviation It follows a uniform distribution, and the maximum values ​​of the standard deviations of the RF channel amplitude and RF channel phase are and ; S14, training is performed considering only the nonlinear processing error in S11, and the nonlinear error is expressed as Description, where Nonlinear error coefficient; S15, training is performed considering only the coupling mutual interference error in S11, and the coupling coefficient It obeys uniform distribution, and the maximum coupling coefficient is ; S16, conduct training that comprehensively considers the position error in S12, the amplitude-phase inconsistency error in S13, the nonlinear processing error in S14, and the coupling mutual interference error in S15, and constructs a network model that considers the four errors.

3. The direction finding method based on drone clusters under multi-factor error conditions according to claim 2 is characterized in that: The method for recording the data received by each UAV electronic countermeasure reconnaissance payload in S20 is: have radiation sources, the direction of the wave from each radiation source is ,in , Number the radiation source, the drone has The drone at the edge of the cluster is used as the reference drone. No. The distance between the UAV and the reference aircraft is ,in ; exist At this moment, the reference aircraft's electronic countermeasure reconnaissance payload changes from Total data received from radiation sources: ， In the formula, The electronic countermeasure reconnaissance payload of the reference aircraft is Moment If the data received from the radiation source is not considered, the The data received by the UAV electronic countermeasure reconnaissance payload is : ， In the above formula, is the wavelength of the radiation source signal, For noise, j A plural unit.

4. The direction finding method based on drone clusters under multi-factor error conditions according to claim 3 is characterized in that: The method for performing error correction on the received data in S30 is: Considering the position error of the UAV, The distance between the UAV and the reference aircraft is will change directly; Considering the inconsistency of amplitude and phase of the RF channel, the RF channel of the UAV electronic countermeasure reconnaissance payload will produce different amplitude gains, among which the first The amplitude gain corresponding to the UAV is recorded as ; Different UAV reconnaissance payload RF channels have different phase shifts, among which The phase shift corresponding to the UAV is recorded as ; Considering the nonlinear processing error, processing the received data by a function g to compensate for the nonlinear processing error, wherein the function g does not generate an additional effect on the noise part in the data during the processing; After considering the above three errors, the received data is : 。 5. The direction finding method based on drone clusters under multi-factor error conditions according to claim 4 is characterized in that: Considering the coupling and mutual interference between the reconnaissance payloads of each UAV, The first drone and The coupling coefficient of the UAV electronic countermeasure reconnaissance load is , The received data is then becomes : ； In the formula, For the The amplitude gain of the UAV, For the The phase shift of the drone, For the The distance between the UAV and the reference aircraft; The total data received by N drones is for: 。 6. The direction finding method based on drone clusters under multi-factor error conditions according to claim 5 is characterized in that: The method for performing virtual-real separation in S40 is: The total data received by the N drones Separate the real and imaginary parts of the neural input value : ， In the formula, and Respectively The real and imaginary parts of .

7. The direction finding method based on drone clusters under multi-factor error conditions according to claim 6 is characterized in that: The method of obtaining the output function in S50 is: The input value Y of the neural network is the neural input function After accumulating multiple snapshots, the batch data is: ， In the formula, is the batch size, The dimension is , N is the total number of drones, specifically including the following steps: S41, First, it passes through the fully connected layer. The dimension of the output of the fully connected layer is ,in, is the number of filters in the subsequent convolutional layer; The value required for dimension expansion; S42, after output from the fully connected layer, performs tensor deformation, and the output dimension is ; S43, then enters the convolution layer, the output is batch normalized and activated, the number of convolution layers is , the final output dimension is ; S44, after passing through the fully connected layer again, the dimension Transformed to , and obtain the output function.

8. The direction finding method based on drone clusters under multi-factor error conditions according to claim 7 is characterized in that: The method for obtaining the spatial spectrum in S50 is: The imaginary part and the real part of the output function are combined to obtain the complex function z, then the spatial spectrum for: ， In the formula, is the guiding vector.

9. A direction finding system based on drone clusters under multi-factor error conditions, characterized in that: include: The network training module, under ideal laboratory conditions, uses the control variable method to consider the position error of the UAV, the amplitude and phase inconsistency error of the RF channel, the nonlinear processing error, and the coupling and mutual interference error between the reconnaissance payloads of each UAV, and then constructs a network model that considers the four errors; The data acquisition module deploys a cluster of drones and enables them to carry electronic countermeasure reconnaissance payloads; when the drones are flying in the air, they sense and receive signals from each target radiation source and record the data received by the electronic countermeasure reconnaissance payload of each drone; A data correction module, using the network model in S10 to perform error correction on the data received in S20; A virtual-real separation processing module, after receiving the error-corrected data in S30, the ground station performs virtual-real separation and then sends the data into the neural network; In the neural network processing module, data in the neural network passes through the fully connected layer, tensor deformation, convolution layer, batch normalization, activation function and fully connected layer in sequence to obtain the output function, the imaginary part and the real part of the output function are combined to obtain the complex function, and then the complex function is multiplied by the steering vector to obtain the spatial spectrum, and finally the direction finding of each radiation source is realized according to the peak finding of the spatial spectrum.

Images