A control method for a microwave power amplifier of an unmanned aerial vehicle

Abstract

The present invention relates to a control method for a microwave power amplifier of an unmanned aerial vehicle, wherein signals of different types and intensities are sequentially transmitted to the unmanned aerial vehicle, and the signal detection module of the unmanned aerial vehicle identifies the type and intensity of the received signal and assigns different gains; the microwave power amplifier performs signal amplification, and the execution system performs corresponding actions according to the received signal; various data of the unmanned aerial vehicle test flight time period are obtained, and the gain data corresponding to the unmanned aerial vehicle execution action state meeting the specified requirements are extracted; training data samples are constructed and input into a gain prediction model for training; received signals are collected in real time and input into a gain prediction model for gain prediction; corresponding gain amplification processing is performed on the received signals; the state of the unmanned aerial vehicle execution action is monitored in real time, and fed back to the gain prediction model, and the parameters of the gain prediction model are optimized. Different amplification processing of signals of different types and intensities is realized, signal quality is enhanced, and signal stability and reliability are improved.

Description

Technical Field

[0001] The invention belongs to the technical field of microwave power amplifiers, and in particular relates to a control method for a microwave power amplifier of an unmanned aerial vehicle. Background Art

[0002] ‌ A microwave power amplifier is a device that amplifies the power of a microwave signal to a certain level. ‌ It is usually composed of one or more microwave transistors, field effect transistors or other power amplifier devices. The microwave signal is input into the amplifier, amplified by the electronic devices inside the amplifier, and then output to the load or other circuits. It is mainly used to increase the power of microwave signals, enhance signal quality, and improve signal stability and reliability. ‌ It can increase the power of microwave signals several times or even dozens of times, thereby enhancing the penetration and coverage area of ​​the signal and meeting the needs of long-distance transmission.

[0003] Microwave power amplifiers play a vital role in the control of drones. Since the communication signal of a drone during flight will gradually attenuate during transmission, if it is not amplified, the signal strength will be very weak and cannot be interpreted at the receiving end. Microwave power amplifiers can increase the strength and stability of the signal, thereby improving the performance of the entire drone communication system, so that the drone can achieve stable communication even in complex environments.

[0004] However, in the existing UAV communication process, microwave power amplifiers will receive various types of communication signals, including radio frequency signals, GPS signals, control signals, remote sensing signals, etc. Due to the different types of communication signals and the different strengths of the signals received by the UAV, the existing UAV power amplifiers will make different signals show different effects according to the same amplification method.

[0005] Therefore, how to improve the control process of the existing microwave power amplifier of the UAV so as to achieve the best amplification effect for signals of different types and intensities, so as to enhance the signal quality, improve the stability and reliability of the signal, and realize the safe, stable and efficient operation of the UAV. Summary of the invention

[0006] The purpose of the present invention is to provide a control method for a microwave power amplifier of an unmanned aerial vehicle, so as to improve the control process of the microwave power amplifier of the existing unmanned aerial vehicle, so as to achieve the best amplification effect when amplifying signals of different types and intensities, so as to enhance the signal quality, improve the stability and reliability of the signal, and realize the safe, stable and efficient operation of the unmanned aerial vehicle.

[0007] In order to solve the above technical problems, the technical solution adopted by the present invention is as follows:

[0008] A method for controlling a microwave power amplifier of an unmanned aerial vehicle comprises the following steps:

[0009] S1: Test-fly the UAV, transmit different types of signals to the UAV in turn, adjust the signal strength of various signals, the UAV's signal receiving module receives the signals in turn and transmits them to the microwave power amplifier, and the microwave power amplifier receives the signals in turn;

[0010] S2: The signal detection module inside the microwave power amplifier identifies the type of the received signal and detects the strength of the received signal, and assigns different gains according to the type and strength of the received signal in accordance with preset rules;

[0011] S3: The microwave power amplifier amplifies the received signal according to the different gains assigned, and transmits the amplified signal to the execution system of the UAV. The execution system of the UAV executes the corresponding action on the flight of the UAV according to the received signal;

[0012] S4: Obtain the received signal data and gain data of the microwave power amplifier during the UAV test flight period and the state data of the UAV execution action, and extract the gain data corresponding to the UAV execution action state meeting the specified requirements;

[0013] S5: constructing and labeling training data samples based on the gain data of the drone's action state that meets the preset requirements and the corresponding signal data, creating a gain prediction model, and inputting the labeled training data samples into the gain prediction model to train the model;

[0014] S6: collecting the received signal of the signal receiving module of the UAV in real time and inputting it into the gain prediction model, wherein the gain prediction model predicts the corresponding gain based on the acquired received signal and outputs the gain prediction result to the microwave power amplifier;

[0015] S7: the microwave power amplifier performs corresponding gain amplification processing on the received signal based on the gain prediction result, and transmits the amplified signal to the UAV execution system after signal conversion, and the UAV execution system performs corresponding actions;

[0016] S8: Monitor the status of the UAV's action in real time, and identify and analyze the status of the action to determine whether its status meets the preset requirements. If so, the microwave power amplifier continues to amplify according to the predicted gain. If not, it is fed back to the gain prediction model to adjust and optimize the gain prediction model parameters.

[0017] Preferably, the specific process of identifying the type of the received signal and detecting the strength of the received signal in step S2 is as follows:

[0018] S21: The signal recognition module establishes a signal sequence based on the received signal, decomposes the signal into wavelet coefficients at each preset scale according to the signal sequence, performs nonlinear transformation on the wavelet coefficients at each preset scale, and removes the signal that meets the preset noise coefficient.

[0019] S22: extracting features from the signal after noise removal, wherein the extracted features include time domain features, frequency domain features and wavelet features of the signal, and calling a preset signal feature relationship table to match the corresponding signal type according to the time domain features, frequency domain features and wavelet features;

[0020] S23: Detecting the strength of the received signal through a signal detection module;

[0021] The signal feature relationship table is a relationship mapping table between time domain features, frequency domain features, wavelet features and signal types.

[0022] Preferably, in step S3, the microwave power amplifier includes an electron beam emission module, a speed adjustment module, a uniform speed operation module, an electron clustering module, a signal amplification module and a signal output module; the electron beam emission module is connected to the speed adjustment module, the speed adjustment module is connected to the uniform speed operation module, the uniform speed operation module is connected to the electron clustering module, the electron clustering module is connected to the signal amplification module, and the signal amplification module is connected to the signal output module.

[0023] Preferably, the specific process of amplifying the received signal according to the different assigned gains is as follows:

[0024] S31: the electron beam emission module emits an electron beam with a preset initial velocity into the speed adjustment module;

[0025] S32: The speed regulating module accelerates or decelerates the electron beam having a certain initial velocity to a preset speed by receiving a microwave signal, i.e., an alternating voltage;

[0026] S33: after the acceleration or deceleration in step S32, the electron beam enters the uniform speed operation module and performs uniform linear motion;

[0027] S34: The electron beam in uniform linear motion forms a plurality of electron cluster blocks in the electron cluster module at the end of the uniform speed operation module, and the speeds of the plurality of electron cluster blocks are the same;

[0028] S35: The electron clustering block enters the signal amplification module and interacts with the high-frequency electromagnetic field in the signal amplification module, so that the kinetic energy of electrons in the electron clustering block is converted into electromagnetic field energy, thereby amplifying and outputting the received signal.

[0029] Preferably, the gain prediction model comprises an input layer, an encoder, a decoder and an output layer; the input layer is connected to the encoder, the encoder is connected to the decoder, and the decoder is connected to the output layer;

[0030] The input layer inputs the labeled training data samples or the received signals of the signal receiving module of the real-time acquisition drone;

[0031] The encoder includes a plurality of stacked encoder layers, each of which includes two encoder sublayers, one of which is a multi-head self-attention layer and the other is a feedforward fully connected layer;

[0032] The decoder comprises a plurality of stacked decoder layers, each of which comprises three decoder sublayers, one of which is a masked multi-head self-attention layer, one of which is a multi-head self-attention layer, and another of which is a feedforward fully connected layer;

[0033] The output layer includes a linear layer and a normalization layer.

[0034] Preferably, the specific process of inputting the labeled training data samples into the gain prediction model for training is as follows:

[0035] S51: converting the input training data into a query vector, a key vector and a value vector, and mapping the query vector, the key vector and the value vector to different feature subspaces by linear transformation;

[0036] S52: split the query vector, key vector and value vector into multiple heads respectively through the multi-head self-attention layer of the encoder, perform attention calculation on each head, and input them into the feedforward fully connected layer of the encoder;

[0037] S53: performing a nonlinear transformation on the attention calculation result through an activation function through the feedforward fully connected layer of the encoder, mapping the nonlinear transformation result to the feature subspace and outputting it to the decoder;

[0038] S54: the decoder calculates the self-attention weights of the query vector and the key vector using the scaled dot product attention inside each segmented head based on the creation of the mask matrix, and applies the self-attention weights to the value vector to obtain a weighted intermediate output;

[0039] S55: concatenate the weighted outputs of all heads, perform linear transformation and normalization on the output layer to obtain the final output result.

[0040] Preferably, each encoder sublayer is connected to a residual connection and a normalization layer, and each decoder sublayer is connected to a residual connection and a normalization layer.

[0041] Preferably, the signal detection module includes a signal detection circuit, including a first resistor, a second resistor, a third resistor, a first capacitor, a second capacitor, a third capacitor, a fourth capacitor, a fifth capacitor, and a chip, one end of the first capacitor and one end of the second capacitor are both connected to the voltage input terminal, the other end of the second capacitor is respectively connected to the DRI port of the chip and one end of the first resistor, the other end of the first resistor is respectively connected to the VCC port of the chip, the third capacitor and the fourth capacitor, the other end of the fourth capacitor is connected to the FB port of the chip, the SE port and the TC port of the chip are respectively connected to the fifth capacitor, one end of the second resistor is connected to the fifth capacitor, and the other end is connected to the third resistor, the chip, the third capacitor, the fifth capacitor and the third resistor are all grounded, and the other end of the third resistor is connected to the voltage output terminal.

[0042] The beneficial effects of the present invention include:

[0043] The control method of the microwave power amplifier of the drone provided by the present invention includes the following steps: during the drone test flight, signals of different types and intensities are sequentially transmitted to the drone, the type of received signal is identified and the intensity is detected, and different gains are assigned according to the type and intensity of the received signal according to preset rules; the microwave power amplifier performs signal amplification, and the execution system performs corresponding actions according to the received signal; various data of the drone test flight time period are obtained, and the gain data corresponding to the drone execution action state meeting the specified requirements is extracted; training data samples are constructed and input into the gain prediction model for training; the received signal is collected in real time and input into the gain prediction model for gain prediction; the received signal is subjected to corresponding gain amplification processing; the state of the drone execution action is monitored in real time, and is fed back to the gain prediction model, and the gain prediction model parameters are iteratively adjusted and optimized. The best amplification effect can be achieved when signals of different types and intensities are amplified, the signal quality is enhanced, the stability and reliability of the signal are improved, and the safe, stable and efficient operation of the drone is achieved.

[0044] First, a signal sequence is established based on the received signal through the signal recognition module, and the signal is decomposed into wavelet coefficients on each preset scale according to the signal sequence. The wavelet coefficients on each preset scale are nonlinearly transformed, and the signals corresponding to the preset noise coefficients are eliminated, thereby realizing signal denoising processing and avoiding noise data from interfering with the execution process of the drone. By extracting features from the signal after noise removal, the extracted features include the time domain features, frequency domain features and wavelet features of the signal. According to the time domain features, frequency domain features and wavelet features, the preset signal feature relationship table is called to match the corresponding signal type and the strength of the received signal is detected through the signal detection module. The accurate identification of the signal type and the precise detection of the strength provide an effective data basis for the subsequent matching of the corresponding gain, ensuring the stable operation of the drone.

[0045] Secondly, by setting up an electron beam emission module, a speed adjustment module, a uniform speed operation module, an electron clustering module, a signal amplification module and a signal output module through a microwave power amplifier, precise amplification of the signal according to the corresponding gain is achieved.

[0046] Thirdly, an electron beam with a preset initial velocity is emitted through the electron beam emission module and enters the speed regulation module; the speed regulation module accelerates or decelerates the electron beam with a certain initial velocity to a preset speed by receiving a microwave signal, and the electron beam after acceleration or deceleration enters the uniform speed operation module and performs uniform linear motion; the electron beam in uniform linear motion forms a number of electron clustering blocks in the electron clustering module at the end of the uniform speed operation module, and the electron clustering blocks enter the signal amplification module and interact with the high-frequency electromagnetic field in the signal amplification module, so that the process of converting the kinetic energy of electrons in the electron clustering blocks into electromagnetic field energy realizes the precise amplification of the received signal.

[0047] Furthermore, the gain prediction model sets an input layer, an encoder, a decoder and an output layer; the encoder includes multiple stacked encoder layers, each encoder layer includes two encoder sub-layers, one is a multi-head self-attention layer, and the other is a feedforward fully connected layer; the decoder includes multiple stacked decoder layers, each decoder layer includes three decoder sub-layers, one is a masked multi-head self-attention layer, one is a multi-head self-attention layer, and the other is a feedforward fully connected layer; the output layer includes a structure of a linear layer and a normalization processing layer, which can efficiently and accurately extract the characteristic relationship between signals of different types and intensities and gains, and realize accurate prediction of gains.

[0048] Furthermore, by converting the input training data into a query vector, a key vector and a value vector, and mapping the query vector, key vector and value vector to different feature subspaces by means of linear transformation; the multi-head self-attention layer of the encoder divides the query vector, key vector and value vector into multiple heads respectively, and performs attention calculation on each head, and inputs them into the feedforward fully connected layer of the encoder; the feedforward fully connected layer of the encoder performs a nonlinear transformation on the attention calculation result through an activation function, maps the nonlinear transformation result to the feature subspace and outputs it to the decoder; the decoder, based on creating a mask matrix, uses scaled dot product attention to calculate the self-attention weights of the query vector and key vector within each divided head, and applies the self-attention weights to the value vector to obtain a weighted intermediate output; the weighted outputs of all heads are spliced, and the linear transformation and normalization of the output layer are performed to obtain the final output result, so that accurate gain prediction results can be obtained efficiently.

[0049] Finally, one end of the first capacitor and one end of the second capacitor of the signal detection circuit are connected to the voltage input terminal, the other end of the second capacitor is respectively connected to the DRI port of the chip and one end of the first resistor, the other end of the first resistor is respectively connected to the VCC port of the chip, the third capacitor and the fourth capacitor, the other end of the fourth capacitor is connected to the FB port of the chip, the SE port and the TC port of the chip are respectively connected to the fifth capacitor, one end of the second resistor is connected to the fifth capacitor, and the other end is connected to the third resistor, the chip, the third capacitor, the fifth capacitor and the third resistor are all grounded, and the other end of the third resistor is connected to the voltage output terminal, so that the signal strength can be accurately tested. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 The figure is a flow chart of a method for controlling a microwave power amplifier of a UAV according to the present invention.

[0051] Figure 2 It is a schematic diagram of the circuit structure of the signal detection module of the present invention.

[0052] Figure 3 A schematic diagram of a flow chart for training the gain prediction model of the present invention.

[0053] Description of the drawings: R1 is the first resistor, R2 is the second resistor, R3 is the third resistor, C1 is the first capacitor, C2 is the second capacitor, C3 is the third capacitor, C4 is the fourth capacitor, C5 is the fifth capacitor, U is the chip, Vin is the voltage output terminal, and Vout is the voltage output terminal. DETAILED DESCRIPTION

[0054] The following is combined with Figure 1~Figure 3 The present invention is further described in detail:

[0055] Example 1

[0056] See attached Figure 1 As shown, a control method for a microwave power amplifier of an unmanned aerial vehicle comprises the following steps:

[0057] S1: Test-fly the UAV, transmit different types of signals to the UAV in turn, adjust the signal strength of various signals, the UAV's signal receiving module receives the signals in turn and transmits them to the microwave power amplifier, and the microwave power amplifier receives the signals in turn. The test-flight stage of the UAV is to obtain sufficient signal data and operation data, so that the characteristic relationship between signal data, gain and operation data can be captured through the gain prediction model in the future, providing an accurate data basis for the subsequent gain prediction.

[0058] S2: The signal detection module inside the microwave power amplifier identifies the type of received signal and detects the strength of the received signal, and assigns different gains according to the type and strength of the received signal according to the preset rules. Since different types of signal data need to be amplified with different gains, and different strength signal data of the same type also need to be amplified with different gains, the final execution command for the drone execution system can be provided more accurately, so as to achieve stable flight of the drone.

[0059] S3: The microwave power amplifier amplifies the received signal according to the different gains assigned, and transmits the amplified signal to the execution system of the UAV. The execution system of the UAV executes the corresponding action on the flight of the UAV according to the received signal;

[0060] S4: Obtain the received signal data, gain data and state data of the UAV execution action of the microwave power amplifier during the UAV test flight period, and extract the gain data corresponding to the state of the UAV execution action meeting the specified requirements. Therefore, by extracting the gain data corresponding to the state of the UAV execution action meeting the specified requirements, it can provide a data basis for the subsequent training of the gain prediction model, which is conducive to the gain prediction model capturing the characteristic relationship between the UAV received signal and the optimal gain.

[0061] S5: Based on the gain data and corresponding signal data that meet the preset requirements of the drone's action state, training data samples are constructed and labeled, a gain prediction model is created, and the labeled training data samples are input into the gain prediction model to train the model. The parameters of the created gain prediction model are adjusted through the training process so that the model can capture the characteristic relationship between the received signal and the optimal gain from the input data, so as to accurately predict the corresponding gain of the subsequent real-time collected signal data.

[0062] S6: The received signal of the signal receiving module of the UAV is collected in real time and input into the gain prediction model. The gain prediction model predicts the corresponding gain based on the acquired received signal and outputs the gain prediction result to the microwave power amplifier.

[0063] S7: The microwave power amplifier performs corresponding gain amplification processing on the received signal based on the gain prediction result, and transmits the amplified signal to the UAV execution system after signal conversion, and the UAV execution system performs corresponding actions.

[0064] S8: Monitor the state of the drone's action in real time, and identify and analyze the state of the action to determine whether its state meets the preset requirements. If so, the microwave power amplifier continues to amplify according to the predicted gain. If not, it feeds back to the gain prediction model to adjust and optimize the gain prediction model parameters. By monitoring the operating state of the drone's action in real time and feeding back the monitoring results to the gain prediction model, the gain prediction model parameters are adjusted through the feedback process, and then the gain prediction model is iteratively optimized to ensure that the model can adapt to new input data and has better robustness.

[0065] Example 2

[0066] On the basis of Example 1, the specific process of identifying the type of the received signal and detecting the strength of the received signal in step S2 is as follows:

[0067] S21: The signal recognition module establishes a signal sequence based on the received signal, decomposes the signal into wavelet coefficients at each preset scale according to the signal sequence, performs nonlinear transformation on the wavelet coefficients at each preset scale, and removes the signal that meets the preset noise coefficient.

[0068] S22: extracting features from the signal after noise removal, wherein the extracted features include time domain features, frequency domain features and wavelet features of the signal, and calling a preset signal feature relationship table to match the corresponding signal type according to the time domain features, frequency domain features and wavelet features;

[0069] S23: Detecting the strength of the received signal through a signal detection module;

[0070] The signal feature relationship table is a relationship mapping table between time domain features, frequency domain features, wavelet features and signal types.

[0071] In this embodiment, a signal sequence is established based on the received signal through the signal recognition module, and the signal is decomposed into wavelet coefficients on each preset scale according to the signal sequence, and the wavelet coefficients on each preset scale are nonlinearly transformed, and the signal corresponding to the preset noise coefficient is eliminated, thereby realizing the denoising of the signal and avoiding the interference of noise data on the execution process of the drone. By extracting features from the signal after noise removal, the extracted features include the time domain features, frequency domain features and wavelet features of the signal, calling the preset signal feature relationship table according to the time domain features, frequency domain features and wavelet features to match the corresponding signal type and detecting the strength of the received signal through the signal detection module, the accurate identification of the signal type and the precise detection of the strength provide an effective data basis for the subsequent matching of the corresponding gain, ensuring the stable operation of the drone.

[0072] The signal detection module includes a signal detection circuit, including a first resistor R1, a second resistor R2, a third resistor R3, a first capacitor C1, a second capacitor C2, a third capacitor C3, a fourth capacitor C4, a fifth capacitor C5, and a chip U. One end of the first capacitor C1 and one end of the second capacitor C2 are both connected to a voltage input terminal Vin, the other end of the second capacitor C2 is respectively connected to a DRI port of the chip U and one end of the first resistor R1, the other end of the first resistor R1 is respectively connected to a VCC port of the chip U, the third capacitor C3 and the fourth capacitor C4, the other end of the fourth capacitor C4 is connected to an FB port of the chip U, the SE port and the TC port of the chip U are respectively connected to the fifth capacitor C5, one end of the second resistor R2 is connected to the fifth capacitor C5, and the other end is connected to the third resistor R3, the chip U, the third capacitor C3, the fifth capacitor C5 and the third resistor R3 are all grounded, and the other end of the third resistor R3 is connected to a voltage output terminal Vout.

[0073] Example 3

[0074] On the basis of Example 1 or Example 2, the microwave power amplifier in step S3 includes an electron beam emission module, a speed adjustment module, a uniform speed operation module, an electron clustering module, a signal amplification module and a signal output module; the electron beam emission module is connected to the speed adjustment module, the speed adjustment module is connected to the uniform speed operation module, the uniform speed operation module is connected to the electron clustering module, the electron clustering module is connected to the signal amplification module, and the signal amplification module is connected to the signal output module.

[0075] In this embodiment, the specific process of amplifying the received signal according to the different assigned gains is as follows:

[0076] S31: the electron beam emission module emits an electron beam with a preset initial velocity into the speed adjustment module;

[0077] S32: The speed regulating module accelerates or decelerates the electron beam having a certain initial velocity to a preset speed by receiving a microwave signal, i.e., an alternating voltage;

[0078] S33: after the acceleration or deceleration in step S32, the electron beam enters the uniform speed operation module and performs uniform linear motion;

[0079] S34: The electron beam in uniform linear motion forms a plurality of electron cluster blocks in the electron cluster module at the end of the uniform speed operation module, and the speeds of the plurality of electron cluster blocks are the same;

[0080] S35: The electron clustering block enters the signal amplification module and interacts with the high-frequency electromagnetic field in the signal amplification module, so that the kinetic energy of electrons in the electron clustering block is converted into electromagnetic field energy, thereby amplifying and outputting the received signal.

[0081] The electron beam emitting module emits an electron beam with a preset initial velocity and enters the speed regulating module; the speed regulating module accelerates or decelerates the electron beam with a certain initial velocity to a preset speed by receiving a microwave signal, and the electron beam after acceleration or deceleration enters the uniform speed operation module and performs uniform linear motion; the electron beam in uniform linear motion forms a plurality of electron clustering blocks in the electron clustering module at the end of the uniform speed operation module, and the electron clustering blocks enter the signal amplification module and interact with the high-frequency electromagnetic field in the signal amplification module, so that the process of converting the kinetic energy of electrons in the electron clustering blocks into electromagnetic field energy realizes the precise amplification of the received signal.

[0082] Example 4

[0083] On the basis of Example 1 or Example 2 or Example 3, the gain prediction model includes an input layer, an encoder, a decoder and an output layer; the input layer is connected to the encoder, the encoder is connected to the decoder, and the decoder is connected to the output layer. The input layer inputs the labeled training data sample or the received signal of the signal receiving module of the real-time acquisition drone. The encoder includes a plurality of stacked encoder layers, each of which includes two encoder sublayers, one of which is a multi-head self-attention layer, and the other is a feedforward fully connected layer. The decoder includes a plurality of stacked decoder layers, each of which includes three decoder sublayers, one of which is a masked multi-head self-attention layer, one of which is a multi-head self-attention layer, and the other is a feedforward fully connected layer. The output layer includes a linear layer and a normalization layer. Each encoder sublayer is connected to a residual connection and a normalization layer, and each decoder sublayer is connected to a residual connection and a normalization layer.

[0084] In this embodiment, the specific process of inputting the labeled training data samples into the gain prediction model for training is as follows:

[0085] S51: converting the input training data into a query vector, a key vector and a value vector, and mapping the query vector, the key vector and the value vector to different feature subspaces by linear transformation;

[0086] S52: split the query vector, key vector and value vector into multiple heads respectively through the multi-head self-attention layer of the encoder, perform attention calculation on each head, and input them into the feedforward fully connected layer of the encoder;

[0087] S53: performing a nonlinear transformation on the attention calculation result through an activation function through the feedforward fully connected layer of the encoder, mapping the nonlinear transformation result to the feature subspace and outputting it to the decoder;

[0088] S54: the decoder calculates the self-attention weights of the query vector and the key vector using the scaled dot product attention inside each segmented head based on the creation of the mask matrix, and applies the self-attention weights to the value vector to obtain a weighted intermediate output;

[0089] S55: concatenate the weighted outputs of all heads, perform linear transformation and normalization on the output layer to obtain the final output result.

[0090] The input training data is converted into a query vector, a key vector and a value vector, and the query vector, key vector and value vector are mapped to different feature subspaces by linear transformation; the multi-head self-attention layer of the encoder divides the query vector, key vector and value vector into multiple heads respectively, and performs attention calculation on each head, and inputs them into the feedforward fully connected layer of the encoder; the feedforward fully connected layer of the encoder performs a nonlinear transformation on the attention calculation result through an activation function, maps the nonlinear transformation result to the feature subspace and outputs it to the decoder; the decoder creates a mask matrix, uses scaled dot product attention to calculate the self-attention weights of the query vector and key vector within each divided head, and applies the self-attention weights to the value vector to obtain a weighted intermediate output; the weighted outputs of all heads are spliced, and the linear transformation and normalization of the output layer are performed to obtain the final output result, so that accurate gain prediction results can be obtained efficiently.

[0091] In summary, the control method of the microwave power amplifier of the drone provided by the present invention is that the drone is tested, signals of different types and intensities are transmitted to the drone in turn, the type of received signal is identified and the intensity is detected, and different gains are assigned according to the type and intensity of the received signal according to the preset rules; the microwave power amplifier amplifies the signal, and the execution system performs corresponding actions according to the received signal; various data of the drone test flight time period are obtained, and the gain data corresponding to the drone execution action state that meets the specified requirements is extracted; training data samples are constructed and input into the gain prediction model for training; the received signal is collected in real time and input into the gain prediction model for gain prediction; the received signal is subjected to corresponding gain amplification processing; the state of the drone execution action is monitored in real time, and fed back to the gain prediction model, and the gain prediction model parameters are iteratively adjusted and optimized. The best amplification effect can be achieved when amplifying signals of different types and intensities, the signal quality is enhanced, the stability and reliability of the signal are improved, and the safe, stable and efficient operation of the drone is realized.

Claims

1. A method for controlling a microwave power amplifier of an unmanned aerial vehicle, characterized in that: The following steps are involved: S1: Test-fly the UAV, transmit different types of signals to the UAV in turn, adjust the signal strength of various signals, the UAV's signal receiving module receives the signals in turn and transmits them to the microwave power amplifier, and the microwave power amplifier receives the signals in turn; S2: The signal detection module inside the microwave power amplifier identifies the type of the received signal and detects the strength of the received signal, and assigns different gains according to the type and strength of the received signal in accordance with preset rules; S3: The microwave power amplifier amplifies the received signal according to the different gains assigned, and transmits the amplified signal to the execution system of the UAV. The execution system of the UAV executes the corresponding action on the flight of the UAV according to the received signal; S4: Obtain the receiving signal data and gain data of the microwave power amplifier during the UAV test flight period and the state data of the UAV execution action, and extract the gain data corresponding to the UAV execution action state that meets the specified requirements; S5: constructing and labeling training data samples based on the gain data of the drone's action state that meets the preset requirements and the corresponding signal data, creating a gain prediction model, and inputting the labeled training data samples into the gain prediction model to train the model; S6: collecting the received signal of the signal receiving module of the drone in real time and inputting it into the gain prediction model, wherein the gain prediction model predicts the corresponding gain based on the acquired received signal and outputs the gain prediction result to the microwave power amplifier; S7: the microwave power amplifier performs corresponding gain amplification processing on the received signal based on the gain prediction result, and transmits the amplified signal to the UAV execution system after signal conversion, and the UAV execution system performs corresponding actions; S8: Monitor the status of the UAV's action in real time, and identify and analyze the status of the action to determine whether its status meets the preset requirements. If so, the microwave power amplifier continues to amplify according to the predicted gain. If not, it is fed back to the gain prediction model to adjust and optimize the gain prediction model parameters.

2. The control method of a microwave power amplifier of an unmanned aerial vehicle according to claim 1, characterized in that: The specific process of identifying the type of received signal and detecting the strength of received signal in step S2 is as follows: S21: The signal recognition module establishes a signal sequence based on the received signal, decomposes the signal into wavelet coefficients at each preset scale according to the signal sequence, performs nonlinear transformation on the wavelet coefficients at each preset scale, and removes the signal that meets the preset noise coefficient. S22: extracting features from the signal after noise removal, wherein the extracted features include time domain features, frequency domain features and wavelet features of the signal, and calling a preset signal feature relationship table to match the corresponding signal type according to the time domain features, frequency domain features and wavelet features; S23: Detecting the strength of the received signal through a signal detection module; The signal feature relationship table is a relationship mapping table between time domain features, frequency domain features, wavelet features and signal types.

3. The control method of a microwave power amplifier of an unmanned aerial vehicle according to claim 1, characterized in that: In step S3, the microwave power amplifier includes an electron beam emission module, a speed adjustment module, a uniform speed operation module, an electron clustering module, a signal amplification module and a signal output module; the electron beam emission module is connected to the speed adjustment module, the speed adjustment module is connected to the uniform speed operation module, the uniform speed operation module is connected to the electron clustering module, the electron clustering module is connected to the signal amplification module, and the signal amplification module is connected to the signal output module.

4. The control method of a microwave power amplifier of an unmanned aerial vehicle according to claim 3, characterized in that: The specific process of amplifying the received signal according to the different gains assigned is as follows: S31: the electron beam emission module emits an electron beam with a preset initial velocity into the speed adjustment module; S32: The speed regulating module accelerates or decelerates the electron beam having a certain initial velocity to a preset speed by receiving a microwave signal, i.e., an alternating voltage; S33: after the acceleration or deceleration in step S32, the electron beam enters the uniform speed operation module and performs uniform linear motion; S34: The electron beam in uniform linear motion forms a plurality of electron cluster blocks in the electron cluster module at the end of the uniform speed operation module, and the speeds of the plurality of electron cluster blocks are the same; S35: The electron clustering block enters the signal amplification module and interacts with the high-frequency electromagnetic field in the signal amplification module, so that the kinetic energy of electrons in the electron clustering block is converted into electromagnetic field energy, thereby amplifying and outputting the received signal.

5. The control method of a microwave power amplifier of an unmanned aerial vehicle according to claim 1, characterized in that: The gain prediction model comprises an input layer, an encoder, a decoder and an output layer; the input layer is connected to the encoder, the encoder is connected to the decoder, and the decoder is connected to the output layer; The input layer inputs the labeled training data samples or the received signals of the signal receiving module of the real-time acquisition drone; The encoder includes a plurality of stacked encoder layers, each of which includes two encoder sublayers, one of which is a multi-head self-attention layer and the other is a feedforward fully connected layer; The decoder comprises a plurality of stacked decoder layers, each of which comprises three decoder sublayers, one of which is a masked multi-head self-attention layer, one of which is a multi-head self-attention layer, and another of which is a feedforward fully connected layer; The output layer includes a linear layer and a normalization layer.

6. The control method of a microwave power amplifier of an unmanned aerial vehicle according to claim 5, characterized in that: The specific process of inputting the labeled training data samples into the gain prediction model for training is as follows: S51: converting the input training data into a query vector, a key vector and a value vector, and mapping the query vector, the key vector and the value vector to different feature subspaces by linear transformation; S52: split the query vector, key vector and value vector into multiple heads respectively through the multi-head self-attention layer of the encoder, perform attention calculation on each head, and input them into the feedforward fully connected layer of the encoder; S53: performing a nonlinear transformation on the attention calculation result through an activation function through the feedforward fully connected layer of the encoder, mapping the nonlinear transformation result to the feature subspace and outputting it to the decoder; S54: the decoder calculates the self-attention weights of the query vector and the key vector using the scaled dot product attention inside each segmented head based on the creation of the mask matrix, and applies the self-attention weights to the value vector to obtain a weighted intermediate output; S55: concatenate the weighted outputs of all heads, perform linear transformation and normalization on the output layer to obtain the final output result.

7. The control method of a microwave power amplifier of an unmanned aerial vehicle according to claim 5, characterized in that: Each encoder sublayer is connected to a residual connection and a normalization layer, and each decoder sublayer is connected to a residual connection and a normalization layer.

8. The control method of a microwave power amplifier of an unmanned aerial vehicle according to claim 2, characterized in that: The signal detection module comprises a signal detection circuit, comprising a first resistor (R1), a second resistor (R2), a third resistor (R3), a first capacitor (C1), a second capacitor (C2), a third capacitor (C3), a fourth capacitor (C4), a fifth capacitor (C5), and a chip (U), one end of the first capacitor (C1) and one end of the second capacitor (C2) are both connected to a voltage input terminal (Vin), the other end of the second capacitor (C2) is respectively connected to a DRI port of the chip (U) and one end of the first resistor (R1), and the first resistor (R1) is connected to a voltage input terminal (Vin). The other end is respectively connected to the VCC port of the chip (U), the third capacitor (C3) and the fourth capacitor (C4); the other end of the fourth capacitor (C4) is connected to the FB port of the chip (U); the SE port and the TC port of the chip (U) are respectively connected to the fifth capacitor (C5); one end of the second resistor (R2) is connected to the fifth capacitor (C5) and the other end is connected to the third resistor (R3); the chip (U), the third capacitor (C3), the fifth capacitor (C5) and the third resistor (R3) are all grounded; the other end of the third resistor (R3) is connected to the voltage output terminal (Vout).

Images