GPS interference phased array monitoring and direction finding antenna system for 1575 frequency band

Abstract

The present invention discloses a GPS interference phased array monitoring and direction-finding antenna system for the 1575 frequency band, relating to the technical field of smart antennas. The system comprises a data integration module for integrating satellite ephemeris, carrier attitude, and airspace coverage data using a weighted averaging method to generate an environmental parameter set; a prediction and calibration module for inputting the environmental parameter set into a bidirectional LSTM network to obtain a phase drift prediction value, and performing nonlinear calibration on the phase drift prediction value using a Kalman-particle hybrid filtering algorithm, while simultaneously monitoring IQ signals using a closed-loop feedback mechanism; and a beamforming module for performing orthogonal modulation and radio frequency up-conversion processing on the IQ signals, and generating a Butler matrix beam using a hybrid topology beam space transformation algorithm. The present invention utilizes a bidirectional LSTM network in combination with a Kalman-particle hybrid filtering algorithm to accurately predict and calibrate phase drift caused by factors such as carrier vibration in real time.

Description

Technical Field

[0001] The present invention relates to the technical field of smart antennas, in particular to a GPS jamming phased array monitoring and direction-finding antenna system for a 1575 frequency band. Background Art

[0002] In recent years, with the widespread application of GPS in various fields, including civilian and military applications, the security and reliability of GPS receivers—phased array antennas—have become a research focus. Phased array antennas operating in the 1575 MHz frequency band, in particular, have attracted increasing attention due to their widespread use in navigation and time synchronization services. In the civilian sector, GPS signals in the 1575 MHz band are ubiquitous, providing high-precision location information and time synchronization services, from personal navigation devices to intelligent transportation, precision agriculture, and disaster monitoring. In military applications, GPS signals in the 1575 MHz band are even more crucial, used for precision-guided weapons, battlefield situational awareness, troop coordination, and communication synchronization. Consequently, researchers have begun exploring more advanced antenna array technologies and signal processing methods. For example, phased array antennas can form beams pointing in specific directions by adjusting the phase relationships of individual array elements, thereby enhancing directional capability and anti-interference performance. Simultaneously, advances in digital signal processing have provided more precise methods for I / Q signal modulation and demodulation, as well as RF upconversion, further improving signal quality and stability. In addition, the introduction of machine learning algorithms, especially deep learning models, has brought new breakthroughs in areas such as satellite ephemeris prediction, phase drift calibration, and multi-source data fusion, making high-precision predictions based on time series analysis possible, and promoting the advancement of space-time filtering technology, laying the foundation for efficient interference suppression within the airspace coverage area.

[0003] However, existing technologies still have some shortcomings. For one thing, in complex dynamic environments, traditional receivers cannot accurately and in real time predict phase drift caused by factors such as carrier vibration, making it difficult to perform timely calibration, which directly affects positioning accuracy and reliability. Furthermore, existing weight allocation strategies typically use a single data processing algorithm for multi-source data fusion and lack adaptive adjustment mechanisms for the characteristics of data from different sources. This results in an inability to effectively integrate data from different sensors or receivers in complex electromagnetic environments, affecting the final positioning results and anti-interference performance. Summary of the Invention

[0004] In view of the above existing problems, the present invention is proposed.

[0005] Therefore, the present invention provides a GPS interference phased array monitoring and direction-finding antenna system for the 1575 frequency band to solve the problems of inaccurate phase drift prediction and insufficient multi-source data fusion.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0007] The present invention provides a GPS interference phased array monitoring and direction-finding antenna system for the 1575 frequency band, which comprises:

[0008] The data integration module is used to integrate the satellite ephemeris, carrier attitude and airspace coverage data through the weighted average method to generate an environmental parameter set;

[0009] The prediction and calibration module is used to input the environmental parameter set into the bidirectional LSTM network to obtain the phase drift prediction value, and then use the Kalman-particle hybrid filter algorithm to perform nonlinear calibration on the phase drift prediction value, while simultaneously using the closed-loop feedback mechanism to monitor the IQ signal;

[0010] The beamforming module is used to perform orthogonal modulation and RF up-conversion on the IQ signals and generate Butler matrix beams using a hybrid topology beam space transformation algorithm;

[0011] The anti-interference positioning module is used to scan within the airspace coverage area using the Butler matrix beam triggering phased array antenna. During the scanning process, space-time filtering is used to suppress GPS interference and obtain preliminary positioning parameters.

[0012] The report generation module is used to perform weighted aggregation and hierarchical optimization on the preliminary positioning parameters using the federated learning mechanism to obtain the optimized positioning parameters, perform ephemeris consistency check and pseudorange residual analysis on the optimized positioning parameters, and generate a phased array antenna monitoring direction finding report.

[0013] As a preferred solution of the GPS interference phased array monitoring and direction-finding antenna system for the 1575 frequency band of the present invention, the specific operation of generating the environmental parameter set is as follows:

[0014] Satellite ephemeris data is cleaned and aligned in time and space using multiple interpolation methods; carrier attitude data is subjected to noise suppression and outlier removal using Kalman filtering and Z-score normalization; and spatial coverage data is interpolated at discrete points using Kriging interpolation.

[0015] Dynamic weights are assigned to the pre-processed satellite ephemeris, carrier attitude and airspace coverage data through weighted averaging method, and fused through Kalman filtering to generate an environmental parameter set.

[0016] As a preferred solution of the GPS interference phased array monitoring and direction-finding antenna system for the 1575 frequency band of the present invention, wherein: the acquisition of the phase drift prediction value is specifically performed as follows:

[0017] In the PyTorch framework, the input layer, hidden layer, and fully connected layer are integrated through the bidirectional parameter method to build a bidirectional LSTM network;

[0018] The environmental parameter set is input into the bidirectional LSTM network through the time series reorganization method, and the input layer performs local feature extraction and dimension upgrading to generate a time three-dimensional tensor;

[0019] The forward LSTM of the hidden layer performs time-series dependency analysis on the three-dimensional time tensor to obtain the hidden state sequence. The backward LSTM performs reverse time-series analysis on the three-dimensional time tensor to generate a reverse hidden state sequence. The hidden state sequence and the reverse hidden state sequence are concatenated to generate a bidirectional feature matrix.

[0020] The fully connected layer uses the ReLU activation function to reduce the dimension of the bidirectional feature matrix and obtain the phase drift prediction value.

[0021] As a preferred solution of the GPS interference phased array monitoring and direction-finding antenna system for the 1575 frequency band of the present invention, wherein: the Kalman-particle hybrid filtering algorithm is used to perform nonlinear calibration on the phase drift prediction value, and the specific operation is as follows:

[0022] The particle filter processes the non-Gaussian noise in the phase drift prediction value through Monte Carlo sampling to obtain the particle state distribution; the particle state distribution is weighted by the second-order moment operation to generate the covariance matrix;

[0023] Kalman filtering is used to remove Gaussian noise in the covariance matrix and generate a filter matrix. The inverse function of the filter matrix is ​​derived by third-order polynomial fitting to generate a dynamic calibration curve. Based on the dynamic calibration curve, the phase drift prediction value is calibrated using the Bayesian weight allocation method.

[0024] As a preferred solution of the GPS interference phased array monitoring and direction-finding antenna system for the 1575 frequency band of the present invention, wherein: the synchronous use of a closed-loop feedback mechanism to monitor IQ signals is specifically performed as follows:

[0025] The phase drift prediction value after nonlinear calibration is dynamically compensated through a closed-loop feedback mechanism, and the IQ signal is sampled using a digital down-conversion method.

[0026] As a preferred solution of the GPS interference phased array monitoring and direction-finding antenna system for the 1575 frequency band described in the present invention, wherein: the IQ signal is subjected to orthogonal modulation and RF up-conversion processing, and a hybrid topology beam space transformation algorithm is used to generate a Butler matrix beam. The specific operations are as follows:

[0027] The IQ signal is image-suppressed using the Bessel interpolation method and quadrature modulated using the QPSK modulation method to obtain a suppressed RF signal. The suppressed RF signal is then up-converted using the polyphase filtering method to generate a multi-beam RF signal.

[0028] A hybrid topology beam space transformation algorithm is used to perform spatial mapping on the multi-beam RF signal to obtain the Butler matrix beam.

[0029] As a preferred solution of the GPS interference phased array monitoring and direction-finding antenna system for the 1575 frequency band of the present invention, wherein: the phased array antenna is triggered by the Butler matrix beam to scan in the airspace coverage range, and the specific operation is as follows:

[0030] The Butler matrix beam is input into the feed channel of the phased array antenna through a directional coupler. In the feed channel, the beam pointing is optimized using a spatial phase gradient compensation algorithm, and a scan trigger signal is obtained through a digital-to-analog converter.

[0031] Based on the scanning trigger signal, the phased array antenna is driven to scan in the spatial coverage range through the time delay scanning mechanism to obtain the array signal.

[0032] As a preferred solution of the GPS interference phased array monitoring and direction-finding antenna system for the 1575 frequency band described in the present invention, wherein: the space-time filtering is used to suppress GPS interference and obtain preliminary positioning parameters. The specific operations are as follows:

[0033] Space-time filtering is used to perform space-time two-dimensional weighting on the array signal to obtain the time domain signal. The GPS interference in the time domain signal is suppressed by the space-time two-dimensional generalized sidelobe cancellation method, and the preliminary positioning parameters are generated using the RAIM integrity monitoring algorithm.

[0034] As a preferred solution of the GPS interference phased array monitoring and direction-finding antenna system for the 1575 frequency band of the present invention, wherein: the acquisition of optimized positioning parameters is specifically performed as follows:

[0035] The federated learning mechanism is used to suppress noise of the preliminary positioning parameters, and the FedAvg algorithm is used for weighted aggregation to obtain multi-source fusion parameters. The hierarchical layer-by-layer alignment method is used to perform hierarchical optimization on the multi-source fusion parameters to obtain the optimized positioning parameters.

[0036] As a preferred solution of the GPS interference phased array monitoring and direction-finding antenna system for the 1575 frequency band of the present invention, the specific operation of generating the phased array antenna monitoring and direction-finding report is as follows:

[0037] The Neville interpolation method is used to perform ephemeris consistency check on the optimized positioning parameters to obtain orbit monitoring data. The double difference analysis method is used to perform pseudorange residual analysis on the optimized positioning parameters to obtain direction monitoring data.

[0038] The track monitoring data and direction monitoring data are integrated using a dynamic phase weighting algorithm to generate a phased array antenna monitoring direction finding report.

[0039] The present invention has the following beneficial effects: Using a bidirectional LSTM network combined with a Kalman-particle hybrid filter algorithm, it can accurately predict and calibrate phase drift caused by factors such as carrier vibration in real time. Using a federated learning mechanism for weighted aggregation and hierarchical optimization of preliminary positioning parameters, it achieves adaptive adjustment to the characteristics of data from different sources. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0041] Figure 1 Schematic diagram of the GPS interference phased array monitoring and direction-finding antenna system used in the 1575 frequency band.

[0042] Figure 2 Flowchart for generating phased array antenna monitoring direction finding report based on preliminary positioning parameters.

[0043] Figure 3 Flowchart of the process for obtaining a set of environmental parameters.

[0044] Figure 4 Flowchart of the process of obtaining the phase drift prediction value. DETAILED DESCRIPTION

[0045] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0046] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0047] Secondly, the term "one embodiment" or "embodiment" herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in various places throughout this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive of other embodiments.

[0048] Reference Figures 1 to 4 , is an embodiment of the present invention, which provides a GPS interference phased array monitoring and direction-finding antenna system for the 1575 frequency band, comprising the following steps:

[0049] The data integration module is used to integrate the data of satellite ephemeris, carrier attitude and airspace coverage through the weighted average method to generate an environmental parameter set.

[0050] The specific steps include:

[0051] The first step is to collect satellite ephemeris, carrier attitude, and airspace coverage data. Satellite ephemeris data is captured using high-precision GPS, which uses multi-frequency tracking to parse navigation messages, thereby collecting satellite ephemeris data in real time. Advanced anti-interference features are also included in the GPS to ensure the accuracy and purity of the satellite ephemeris data.

[0052] The carrier attitude data is collected using an inertial measurement unit (IMU). The IMU can accurately sense the dynamic parameters of the carrier's acceleration and angular velocity, and integrate and transform the dynamic parameters through built-in algorithms to obtain carrier attitude data, providing high-frequency and high-resolution data support for subsequent analysis.

[0053] Airspace coverage data is collected using distributed phased array antenna units. These units can provide all-around airspace coverage based on beam scanning control algorithms, capturing signal strength and phase information from different directions. Digital beamforming technology is used to integrate this data into airspace coverage data, providing a reliable data foundation for precise monitoring and positioning within the airspace.

[0054] The second step is to preprocess the collected satellite ephemeris, carrier attitude, and airspace coverage data. First, the satellite ephemeris data is processed: based on the multiple interpolation method, the satellite ephemeris data is multi-layered decomposition using the Daubechies8 wavelet basis to obtain the ephemeris time series. The ephemeris time series is then cleaned using the sliding window detection method. At the same time, the dynamic time warping (DTW) algorithm is used for spatiotemporal alignment to ensure the temporal continuity and spatial consistency of the satellite ephemeris data.

[0055] Secondly, the carrier posture data is processed: the Kalman filter is used to measure and update the carrier posture data to obtain posture parameters. The posture parameters are divided into data segments of five sampling points using the sliding average window method. Each data segment is weighted moving averaged using the Hanning window function to suppress noise. The carrier posture data is normalized using the Z-score normalization method, and outliers are removed using the box plot analysis method to ensure the purity and reliability of the carrier posture data.

[0056] Finally, the airspace coverage data is processed: Kriging interpolation is used to perform non-uniform point sampling on the airspace coverage data to obtain continuous airspace data; the local outlier factor (LOF) algorithm is used to perform density cluster analysis on the continuous airspace data to identify discrete points; and the Kriging residual analysis method is used to interpolate the discrete points to ensure the spatial continuity and signal strength consistency of the airspace coverage data.

[0057] The third step is to integrate the pre-processed satellite ephemeris, carrier attitude and airspace coverage data to obtain the environmental parameter set. In the specific operation, the entropy weight of the pre-processed satellite ephemeris, carrier attitude and airspace coverage data is calculated by weighted average method to obtain the fusion parameter set; the multi-modal correlation of the fusion parameter set is calculated by principal component analysis method to generate a weighted parameter matrix; the dynamic weight allocation algorithm is used to perform real-time weight optimization on the weighted parameter matrix to obtain dynamic weights;

[0058] Based on dynamic weights, the Kalman filter is used to further integrate satellite ephemeris, carrier attitude, and airspace coverage data to generate an environmental parameter set. Specifically, the quaternion Kalman filter (QKF) is used to solve the carrier attitude data, and the state vector is constructed through the vector splicing method. The least squares fitting method is used to extract orbital parameters from the satellite ephemeris data, and the phase double difference method is used to perform differential combination of the orbital parameters to construct the observation vector.

[0059] Based on the state vector and observation vector, the Kalman filter algorithm is used for prediction and update: in the prediction stage, the Newton-Euler equation is used to calculate the state transfer of the state vector to obtain the observation noise matrix. In the update stage, the weighted least squares method is used to calculate the inverse matrix of the observation noise matrix to obtain the weighted noise inverse matrix. The dynamic weights are then integrated into the weighted noise inverse matrix through the matrix Hadamard product method to generate the Kalman gain matrix. The Kalman gain matrix is ​​iterated through the Jacobi iteration method until the convergence threshold is reached, and the optimal environmental parameter set is output. The application of Kalman filtering can not only smooth the noise, but also make the generated environmental parameter set closer to the actual situation.

[0060] It should be noted that the convergence threshold is defined by a sliding window statistics method based on the 3σ criterion.

[0061] The prediction and calibration module is used to input the environmental parameter set into the bidirectional LSTM network to obtain the phase drift prediction value, and use the Kalman-particle hybrid filter algorithm to perform nonlinear calibration on the predicted value of the phase drift, and simultaneously use the closed-loop feedback mechanism to monitor the IQ signal.

[0062] The specific steps include:

[0063] The first step is to build a bidirectional LSTM network. Specifically, the LSTM is initialized in the PyTorch framework using the bidirectional parameter method and the bidirectional structure is enabled by the bidirectional=True parameter to obtain a two-way LSTM with forward and backward processing capabilities. Based on the two-way LSTM, the input layer, hidden layer, and fully connected layer are stacked using the Sequential container encapsulation method to build a complete bidirectional LSTM network. The input layer uses one-dimensional convolution to enhance local features. The hidden layer sets multiple bidirectional processing nodes and outputs bidirectional feature moments in a bidirectional splicing manner. The fully connected layer uses linear transformation and activation function to achieve feature dimensionality reduction.

[0064] Next, the constructed bidirectional LSTM network is trained. Specifically, the environmental parameter set is divided into a training set, a validation set, and a test set using stratified random sampling. Forward and backward propagation are performed on the training set using mini-batch gradient descent to obtain the error gradient. The Adam optimizer is used to adjust the momentum of the error gradient, while gradient clipping is used to prevent gradient explosion, generating updated LSTM network parameters. On the validation set, the early stopping mechanism is used to calculate the loss of the updated LSTM network parameters using the SmoothL1Loss function to obtain the smoothed L1 loss value. Overfitting is prevented by using the Dropout layer and L2 regularization. The training process is accelerated by GPU parallel computing. In the validation set, training is terminated when the smoothed L1 loss value stops decreasing and reaches the DCT threshold, and the trained bidirectional LSTM network is output synchronously.

[0065] It should be noted that the DCT threshold is defined based on the rate of change of the smoothed L1 loss value;

[0066] In the second step, the environmental parameter set is input into the bidirectional LSTM network through the time series reorganization method. The input layer performs local feature extraction and dimensionality increase to generate a three-dimensional time tensor. Specifically, the environmental parameter set is divided into multiple sliding windows through the time series reorganization method. The sliding windows are then reorganized using the sequence filling method to obtain a sequence of time series fragments. The sequence of time series fragments is then input into the input layer of the bidirectional LSTM network through the PyTorch DataLoader interface.

[0067] A one-dimensional convolutional layer is used in the input layer to extract local features of the temporal fragment sequence, and layer normalization is used to standardize the sequence to obtain standardized features. A linear projection layer is used to spatially map the standardized features, and a linear transformation is performed using the ReLU activation function to generate high-dimensional features. The high-dimensional features are restructured using the dimensionality reshaping method, and the dimensions are transposed using PyTorch to generate a three-dimensional time tensor.

[0068] The forward LSTM of the hidden layer performs temporal dependency analysis on the three-dimensional time tensor to obtain the hidden state sequence. Specifically, PyTorch's LSTMCell (Long Short-Term Memory Cell) is used to perform time step expansion calculations on the three-dimensional time tensor to obtain feature slices. A gating mechanism is used to perform linear transformation and bias superposition on the feature slices to obtain gated activation values. The gated activation values ​​are normalized using the sigmoid function, and the recursive update method is used for temporal fusion to capture long-term dependencies. The state output control method is used to extract the hidden states in the long-term dependencies, and the tensor splicing method is used to integrate them to output the hidden state sequence.

[0069] The backward LSTM of the hidden layer performs reverse temporal analysis on the three-dimensional time tensor to generate a reverse hidden state sequence. Specifically, the three-dimensional time tensor is expanded and calculated in reverse time order using LSTMCell to obtain reverse feature slices. The reverse gating mechanism is used to perform a linear transformation of the reverse feature slices in reverse time order to generate reverse gated activation values. The reverse gated activation values ​​are normalized using sigmoid and tanh functions, and reverse temporal analysis is performed using the reverse state recursive algorithm to obtain the reverse hidden state. The tensor splicing method is used to integrate them and synchronously output the reverse hidden state sequence.

[0070] The hidden state sequence and the reverse hidden state sequence are concatenated through depthwise separable convolution and normalized through layer normalization to generate a bidirectional feature matrix.

[0071] The fully connected layer uses the ReLU activation function to reduce the dimension of the bidirectional feature matrix to obtain the phase drift prediction value. Specifically, the bidirectional feature matrix is ​​flattened through the Flatten layer and converted into a two-dimensional matrix. Two layers of fully connected layers are used to reduce the dimension of the two-dimensional matrix: the first layer uses the ReLU activation function to reduce the dimension of the two-dimensional matrix to 64 dimensions, and the second layer uses 1×1 convolution to further compress the two-dimensional matrix to 1 dimension to obtain the original prediction value. The Sigmoid function is used to constrain the output of the original prediction value to the interval [0,1] to generate the final phase drift prediction value.

[0072] The third step is to use the Kalman-particle hybrid filter algorithm to perform nonlinear calibration on the phase drift prediction value. In the specific operation, the particle filter is used to decompose the error of the phase drift prediction value, obtain multiple initialized particles, and iteratively update the initialized particles through Monte Carlo sampling: in the prediction stage, the state transfer method is used to propagate the state of the initialized particles to obtain the predicted particle set. In the observation update stage, based on the 3σ criterion, the weighted likelihood estimation method is used to remove non-Gaussian noise from the predicted particle set, and the importance resampling method is used to normalize the weights of the denoised predicted particle set to generate the particle state distribution;

[0073] Next, the particle state distribution is weighted by the second-order moment operation to obtain the mean vector, which is then centered by vector subtraction to generate the particle deviation vector. The particle deviation vector is mean-normalized by matrix subtraction to generate the deviation matrix. The deviation matrix is ​​transposed to obtain the transpose of the deviation matrix. The deviation matrix is ​​multiplied by the transpose of the deviation matrix and matrix accumulation is performed using the weighted outer product summation algorithm to obtain the covariance matrix. Finally, a regularization term is added to the covariance matrix by the diagonal matrix superposition method to ensure the stability of the covariance matrix.

[0074] Then, the covariance matrix is ​​updated in time through Kalman filtering to obtain the initial state estimate. The Kalman gain is calculated on the initial state estimate through the innovation covariance method to obtain the optimal state estimate. Based on the optimal state estimate, the Gaussian noise in the covariance matrix is ​​removed through the Mahalanobis distance detection method, and the covariance is reset through the fixed ratio scaling method to generate the filter matrix.

[0075] The phase error parameters are obtained by performing least squares regression on the filter matrix through third-order polynomial fitting. The inverse function of the phase error parameters is derived through the matrix inversion algorithm to generate the phase compensation function. The phase compensation function is de-jittered through sliding window smoothing, and the curve is fitted using linear interpolation to generate a dynamic calibration curve.

[0076] The dynamic calibration curve is weighted by confidence using the Bayesian weight assignment method to obtain a weighted calibration curve; the weighted calibration curve is discretized and sampled using the linear interpolation method to generate a phase lookup table; based on the phase lookup table, the phase offset is calculated using the real-time table lookup method to obtain the offset value; based on the offset value, the phase drift prediction value is phase rotated using the phase compensation method to calibrate the phase drift prediction value;

[0077] In the fourth step, after the nonlinear calibration is completed, the IQ signal is monitored synchronously using the closed-loop feedback mechanism. In the specific operation, the phase drift prediction value after nonlinear calibration is subjected to error differential calculation through the closed-loop feedback mechanism to obtain the phase error gradient; the phase error gradient is adjusted proportionally, integrally and differentially through the PID controller, and the error weighted accumulation algorithm is used to generate a closed-loop compensation amount; according to the closed-loop compensation amount, real-time phase modulation is performed through the digitally controlled oscillator to perform error compensation on the phase drift prediction value; after the error compensation is completed, the FPGA timer is used to generate a timing trigger signal to drive the RF front-end equipment to collect the initial IQ signal of the phased array antenna, and the initial IQ signal is orthogonally demodulated by the digital down-conversion method to ensure the orthogonal accuracy of the IQ signal to obtain the final IQ signal.

[0078] The beamforming module is used to perform orthogonal modulation and RF up-conversion processing on the IQ signals, and generate Butler matrix beams using a hybrid topology beam space transformation algorithm.

[0079] The specific steps include:

[0080] The first step is to suppress the image frequency of the IQ signal using the Bessel interpolation method and perform orthogonal modulation using the QPSK modulation method to obtain a suppressed RF signal. Specifically, the final IQ signal is upsampled with high precision using a sixth-order Bessel interpolation filter to obtain a high-multiple upsampled signal. The high-multiple upsampled signal is then subjected to polyphase filtering calculations using the Bessel interpolation method to generate an image-suppressed signal. The sixth-order Bessel interpolation filter achieves efficient calculations through an optimized polyphase structure and uses 12-bit fixed-point arithmetic to ensure calculation accuracy. Next, the image-suppressed signal is spectrally shaped using a window function weighting method to suppress the image frequency components in the final IQ signal and generate a baseband signal without image interference.

[0081] After image suppression is completed, the QPSK modulation stage begins. In specific operations, based on the QPSK modulation method, the baseband signal stream without image interference is mapped to the orthogonal points according to the Gray code rule, and the amplitude is quantized by the linear quantization method to generate in-phase and orthogonal components.

[0082] Subsequently, the in-phase and quadrature components are bandwidth-limited using a cosine pulse shaping filter to effectively suppress out-of-band spectrum leakage. Complex mixing of the in-phase and quadrature components is then performed using a digital mixer to generate a modulated intermediate signal. Finally, digital quadrature up-conversion technology is used to quadrature modulate the modulated intermediate signal, while a carrier phase-locked loop is used for carrier synchronization to obtain a suppressed RF signal.

[0083] The application of QPSK modulation significantly improves spectrum utilization while ensuring baseband signal quality. It is particularly suitable for phased array antennas with strict requirements for image frequency suppression.

[0084] The second step is to use a polyphase filtering method to perform RF up-conversion processing on the suppressed RF signal to generate a multi-beam RF signal. Specifically, the polyphase filtering method is used to perform multi-phase decomposition processing on the suppressed RF signal. Each phase filter uses an equiripple FIR design method to band-limit the suppressed RF signal to generate a band-limited intermediate frequency signal. Subsequently, the band-limited intermediate frequency signal is digitally orthogonally up-converted using the CORDIC algorithm to generate an RF modulated signal.

[0085] Based on the RF modulated signal, the complex mixer uses direct digital frequency synthesis (DDS) to perform RF up-conversion processing on the suppressed RF signal, moving the suppressed RF signal to the target RF frequency. The beamforming (MVDR) method is then used to perform multi-channel amplitude and phase weighting on the moved suppressed RF signal to generate a multi-beam RF signal.

[0086] It should be noted that the target radio frequency is the standard operating frequency band definition of 5G base stations;

[0087] Digital pre-equalization (DPE) is used to perform nonlinear correction on multi-beam RF signals to ensure amplitude consistency across all channels. Pre-distortion compensation technology is also used to correct DAC nonlinearity in multi-beam RF signals, ensuring accuracy and efficiency.

[0088] The third step is to use a hybrid topology beam space transformation algorithm to perform spatial mapping on the multi-beam RF signal to obtain the Butler matrix beam. Specifically, in the digital baseband processor, the hybrid topology beam space transformation algorithm is used to perform fast Fourier transform (FFT) processing on the multi-beam RF signal to convert the multi-beam RF signal into a beam space signal. Subsequently, the digital phase shift weighting method is used to weight the beam space signal to obtain a phase calibration signal. The fixed phase gradient method is used to perform spatial mapping on the phase calibration signal to generate a primary beam signal.

[0089] In the amplitude equalizer, the automatic gain control (AGC) method is used to collect the actual power of each feeding channel and compare it with the reference power to obtain the power difference. The power difference is then subjected to a proportional-integral (PI) control operation to obtain the amplitude adjustment value of the feeding channel. The specific mathematical formula is as follows: ;

[0090] in, Indicates the Feed channels at time Amplitude adjustment value; represents the proportionality coefficient; Indicates the The power difference of the feeding channels at the current moment; represents the integral coefficient; Indicates the Feed channels at time The power difference; Indicates the sampling time interval; Indicates the feed channel index; Represents a time index; Indicates cumulative variables;

[0091] It should be noted that the reference power is based on the maximum radiated power Fallback point definition, the value range is ; The proportional coefficient is defined according to the actual response speed requirements and the value range is 0.05~0.2;

[0092] The coefficient is defined based on the actual steady-state error elimination requirements and has a value range of 0.001~0.01; the cumulative variable is defined based on the discrete time integral operation and has a value range of ;

[0093] According to the amplitude adjustment value of the feed channel, a digital attenuator is used to compensate the feed channel amplitude of the primary beam signal to obtain a balanced beam signal. Based on the balanced beam signal, a matrix operation algorithm is used to perform beam space transformation on the multi-beam RF signal to generate a Butler matrix beam. The Butler matrix beam not only reduces the complexity and cost of the phased array antenna, but also maintains stable beam pointing performance.

[0094] The anti-interference positioning module is used to trigger the phased array antenna to scan within the airspace coverage area through the Butler matrix beam. During the scanning process, space-time filtering is used to suppress GPS interference and obtain preliminary positioning parameters.

[0095] The specific steps include:

[0096] The first step is to use the Butler matrix beam to trigger the phased array antenna to scan the spatial coverage range. In the specific operation, the Butler matrix beam is first amplitude-weighted using the least squares error method to effectively suppress the sidelobe level and increase the mainlobe gain, thereby obtaining a weighted Butler matrix beam. The weighted Butler matrix beam is input into the directional coupler through the SPI digital control interface. The directional coupler distributes the weighted Butler matrix beam to each feeding channel of the phased array antenna through the impedance coaxial cable.

[0097] It should be noted that the sidelobe level refers to the maximum radiation level in the direction of the phased array antenna except the main lobe; the main lobe gain refers to the power gain of the phased array antenna in the direction of maximum radiation, which is a key indicator for measuring the directivity of the phased array antenna.

[0098] In the feeding channel, the beam pointing is optimized using the spatial phase gradient compensation algorithm: the least squares error (LSE) algorithm is used to calculate the channel phase of the weighted Butler matrix beam to obtain the phase difference between the feeding channels; the calculated phase difference between the feeding channels is matrixed using the phase quantization mapping algorithm to obtain a phase error distribution map; the phase error distribution map is subjected to error decomposition using the gradient projection method to obtain the phase compensation gradient field; based on the phase compensation gradient field, the optimal phase gradient is calculated using the matrix iteration method; based on the optimal phase gradient, the gradient iterative compensation method is used in the digitally controlled phase shifter to correct the phase gradient of each feeding channel, and the corrected beam signal is generated using the complex signal reconstruction algorithm; the signal components of the corrected beam signal are reconstructed using analog orthogonal modulation through a digital-to-analog converter, and the time domain waveform is synthesized using the inverse discrete Fourier transform (IDFT) to generate a scanning trigger signal;

[0099] Secondly, based on the scan trigger signal, the time-delayed scanning mechanism drives the phased array antenna to scan the spatial coverage range and acquire the array signal. In specific operations, the scan trigger signal is transmitted to the phased array antenna using a low-jitter LVDS differential transmission line. The T / R (transmitter / receiver) unit in the phased array antenna uses the time-delayed scanning mechanism to fine-tune the delay and generate the excitation signal.

[0100] The least mean square (LMS) algorithm is used to iterate the delay of the excitation signal to obtain the optimal delay. This optimal delay is then configured to the GaN RF front-end of each feeding channel via the SPI interface. The GaN RF front-end uses the fast Fourier transform (FFT) algorithm to decompose the optimal delay into beam directions and generate beam steering parameters. Based on the beam steering parameters, a digitally controlled phase shifter array (such as the HMC933LP6E) is used to drive the phased array antenna to scan within the spatial coverage area. During the scanning process, a high-speed ADC acquisition card is used to simultaneously acquire the array signal.

[0101] In the second step, during the scanning process, space-time filtering is used to suppress GPS interference and obtain preliminary positioning parameters. In the specific operation, first, space-time filtering performs multi-channel synchronous sampling on the array signal to obtain a discrete signal matrix; the discrete signal matrix is ​​divided into multiple windows using the sliding window method; within each window, statistical features are extracted using the FFT acceleration algorithm; based on the extracted statistical features, the space-time two-dimensional covariance matrix is ​​calculated using the Bartlett spectrum algorithm; the two-dimensional covariance matrix is ​​inverted in real time using the QR decomposition algorithm, and the weights are solved using the Wiener filter weighting algorithm to generate space-time weights; the space-time weights are synthesized in the time domain using multi-phase FIR filtering to output a time domain signal with strong interference suppression capability;

[0102] Secondly, the GPS interference in the time domain signal is suppressed by the space-time two-dimensional generalized sidelobe cancellation method, and the preliminary positioning parameters are generated using the RAIM integrity monitoring algorithm. In the specific operation, the space-time two-dimensional generalized sidelobe cancellation method is used to perform spatial filtering on the time domain signal to obtain a preliminary interference suppression signal. The GPS interference in the time domain signal is identified by the Doppler-code phase combined detection (DCPD) method. The GPS interference is removed from the preliminary interference suppression signal by the characteristic projection suppression (EPS) method to generate an optimized time domain signal.

[0103] The pseudorange features of the optimized time domain signal are extracted using the RAIM integrity monitoring algorithm, and the pseudorange features are converted into satellite navigation data using the least squares residual projection method. The positioning parameters of the satellite navigation data are solved using the weighted least squares method to generate preliminary positioning parameters.

[0104] The report generation module is used to perform weighted aggregation and hierarchical optimization on the preliminary positioning parameters using the federated learning mechanism to obtain the optimized positioning parameters, perform ephemeris consistency check and pseudorange residual analysis on the optimized positioning parameters, and generate a phased array antenna monitoring direction finding report.

[0105] The specific steps include:

[0106] The first step is to use a federated learning mechanism to suppress noise on the preliminary positioning parameters and perform weighted aggregation using the FedAvg algorithm to obtain multi-source fusion parameters. Specifically, the federated learning mechanism is used to perform local gradient calculation on the preliminary positioning parameters to obtain the positioning parameter update amount. The positioning parameter update amount is subjected to noise obfuscation processing using the differential privacy method and compressed using the gradient quantization compression method to reduce communication overhead and obtain compressed transmission parameters. The compressed transmission parameters are weightedly aggregated using the FedAvg (federated averaging algorithm) and the credibility weight of the compressed transmission parameters is calculated using the Mahalanobis distance metric to generate weighted fusion parameters. The weighted fusion parameters are then processed for spatiotemporal consistency using the Kalman filter method to generate multi-source fusion parameters.

[0107] In the second step, the multi-source fusion parameters are optimized hierarchically using the layer-by-layer alignment method to obtain the optimized positioning parameters. Specifically, the multi-source fusion parameters are projected hierarchically using the layer-by-layer alignment method, and dynamic noise injection is performed at each level using the differential privacy algorithm to ensure the privacy security and hierarchical consistency of the multi-source fusion parameters. At the same time, the weighted sliding average method is used to filter and smooth the multi-source fusion parameters after noise injection to generate the optimized positioning parameters.

[0108] The third step is to use the Neville interpolation method to perform ephemeris consistency check on the optimized positioning parameters and obtain orbit monitoring data. In the specific operation, the Neville interpolation method is used to interpolate and fit the optimized positioning parameters to obtain satellite trajectory data; and the residual sum of squares (RSS) method is used to calculate the deviation of the satellite trajectory data to obtain the orbit deviation; based on the real-time collected satellite ephemeris data, the satellite trajectory data is checked for ephemeris consistency using the chi-square test method. When the orbit deviation is lower than the deviation threshold, for example, the difference threshold is set to: radial error ≤ 5cm, tangential ≤ 10cm, then when the measured orbit deviation is 3cm radially and 8cm tangentially, the check is considered to have passed; and the sliding window mean filter method is used to smooth the satellite trajectory data after verification to generate orbit monitoring data;

[0109] It should be noted that the deviation threshold is defined based on the precise ephemeris accuracy standard published by the International Navigation Satellite Service (IGS);

[0110] Next, the double difference analysis method is used to perform pseudorange residual analysis on the optimized positioning parameters to obtain direction monitoring data. In the specific operation, the double difference analysis method is used to perform differential processing on the optimized positioning parameters to obtain double-difference pseudorange observations; the weighted least squares adjustment method is used to calculate and distribute the residuals of the double-difference pseudorange observations to generate a residual sequence; the moving window statistical test method is used to extract the deviation of the residual sequence to obtain the direction deviation vector; and the spatial vector decomposition method is used to perform three-dimensional direction solution on the direction deviation vector. The solution results are integrated using the weighted sliding average method to generate direction monitoring data;

[0111] The fourth step is to use a dynamic phase weighting algorithm to integrate the track monitoring data and the direction monitoring data to generate a phased array antenna monitoring direction finding report. In the specific operation, the phase difference compensation method is used to load the track monitoring data and the direction monitoring data into the same time and space for alignment.

[0112] Next, the track monitoring data and direction monitoring data are correlated using a dynamic phase weighting algorithm. The radial error component and tangential velocity vector from the track monitoring data are extracted using the least squares fitting method. The direction monitoring data is decomposed using the quaternion decomposition method, and the azimuth and elevation angle sequences are extracted using the sliding window method. A dynamic phase weighting algorithm is used to assign different fusion weights to the radial error component, tangential velocity vector, and azimuth and elevation angle sequences. Based on the fusion weights, the data are integrated using a weighted fusion engine, and a phased array antenna monitoring direction finding report is generated using statistical analysis tools (such as MATLAB).

[0113] Based on the phased array antenna monitoring direction-finding report, the Doppler-code phase joint analysis method is used to find and locate illegal signals that interfere with the GPS environment and obtain the characteristic parameters of the interference signal. In the specific operation, the phased array antenna monitoring direction-finding report is differentially processed using pseudo-range double difference, and the carrier phase smoothing pseudo-range method is used to remove noise to obtain a high-precision code phase observation sequence;

[0114] The Doppler-code phase joint analysis method is used to extract the time-frequency features of the code phase observation sequence, and the time-varying features are extracted by short-time Fourier transform (STFT). The time-frequency features and time-varying features are reduced in dimension by principal component analysis to obtain the time-frequency feature matrix.

[0115] The cyclostationary feature detection method is used to call the spectral correlation function (SCF) to perform cyclic frequency scanning on the time-frequency feature matrix to obtain cyclostationary feature parameters; the MUSIC algorithm is used to construct the spatial spectrum of the cyclostationary feature parameters to generate the spatial spectrum distribution map of the interference signal; the particle swarm optimization (PSO) algorithm is used to search for spectral peaks in the spatial spectrum distribution map of the interference signal to complete the search and location of illegal signals;

[0116] The fast Fourier transform (FFT) is used to extract the frequency domain features of the located illegal signal and generate the Doppler frequency shift. The phase-locked loop (PLL) is used to track the carrier of the located illegal signal and obtain the carrier phase. The sliding window energy detection method is used to divide the located illegal signal into windows, and the Welch power spectrum estimation is used to integrate the energy in each window to generate the signal strength state vector.

[0117] The entropy weight method is used to assign different weights to the Doppler frequency shift, carrier phase, and signal strength state vector. Based on the assigned weights, multimodal fusion is performed using the DS evidence theory (Dempster-Shafer Theory) to generate the interference signal dynamic feature matrix. The interference signal dynamic feature matrix is ​​subjected to space-time two-dimensional filtering using the space-time adaptive processing method to obtain the azimuth and elevation angles of the interference signal. The azimuth and elevation angles of the interference signal are parameterized using the generalized likelihood ratio test (GLRT) to generate the interference signal characteristic parameters of center frequency, bandwidth, azimuth, and signal type.

[0118] Finally, a dynamic clustering algorithm is used to classify the interference signal characteristic parameters and obtain the optimized phased array antenna monitoring direction finding report. In the specific operation, the dynamic clustering algorithm is used to perform density clustering on the interference signal characteristic parameters to generate initial clusters; the Mahalanobis distance metric is used to extract and iterate the center points of the initial clusters to obtain the optimized cluster center points;

[0119] The spectral clustering method is used to divide the cluster centers by Gaussian kernel similarity to generate a Laplace matrix. Based on the Laplace matrix, the cluster centers are divided into high-density core modes, transition edge modes, and isolated noise modes through the Lanczos iterative eigenvalue decomposition method. The FedAvg method is used to perform multi-node aggregation on the results of pattern classification to generate an optimized phased array antenna monitoring direction finding report.

[0120] In summary, the present invention utilizes a bidirectional LSTM network combined with a Kalman-particle hybrid filter algorithm to accurately predict and calibrate phase drift caused by factors such as carrier vibration in real time. It also utilizes a federated learning mechanism for weighted aggregation and hierarchical optimization of preliminary positioning parameters, enabling adaptive adjustment to the characteristics of data from different sources.

[0121] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. GPS jammer monitoring and direction-finding antenna system for the 1575 frequency band, characterized by: include, A data integration module is used to integrate satellite ephemeris, carrier attitude, and airspace coverage data using a weighted average method to generate an environmental parameter set. The airspace coverage data is collected by a distributed phased array antenna unit. The distributed phased array antenna unit can provide all-round airspace coverage based on a beam scanning control algorithm, capturing signal strength and phase information from different directions and integrating it into airspace coverage data using digital beamforming technology; The prediction and calibration module is used to input the environmental parameter set into the bidirectional LSTM network to obtain the phase drift prediction value, and use the Kalman-particle hybrid filter algorithm to perform nonlinear calibration on the phase drift prediction value. Synchronously, through a closed-loop feedback mechanism, dynamic error compensation is performed on the phase drift prediction value after nonlinear calibration, and the IQ signal is sampled using the digital down-conversion method; The beamforming module is used to perform orthogonal modulation and RF up-conversion on the IQ signals and generate Butler matrix beams using a hybrid topology beam space transformation algorithm; The anti-interference positioning module is used to scan the airspace coverage area through the Butler matrix beam triggering phased array antenna. During the scanning process, space-time filtering is used to suppress GPS interference and obtain preliminary positioning parameters; The report generation module is used to perform weighted aggregation and hierarchical optimization on the preliminary positioning parameters using the federated learning mechanism to obtain the optimized positioning parameters, perform ephemeris consistency check and pseudorange residual analysis on the optimized positioning parameters, and generate a phased array antenna monitoring direction finding report.

2. The GPS jamming phased array monitoring and direction-finding antenna system for the 1575 frequency band as claimed in claim 1, characterized in that: The specific operations for generating the environment parameter set are as follows: Satellite ephemeris data is cleaned and aligned in time and space using multiple interpolation methods; carrier attitude data is subjected to noise suppression and outlier removal using Kalman filtering and Z-score normalization; and spatial coverage data is interpolated at discrete points using Kriging interpolation. Dynamic weights are assigned to the pre-processed satellite ephemeris, carrier attitude and airspace coverage data through weighted averaging method, and fused through Kalman filtering to generate an environmental parameter set.

3. The GPS jamming phased array monitoring and direction-finding antenna system for the 1575 frequency band as claimed in claim 2, characterized in that: The specific operation of obtaining the phase drift prediction value is as follows: In the PyTorch framework, the input layer, hidden layer, and fully connected layer are integrated through the bidirectional parameter method to build a bidirectional LSTM network; The environmental parameter set is input into the bidirectional LSTM network through the time series reorganization method, and the input layer performs local feature extraction and dimension upgrading to generate a time three-dimensional tensor; The forward LSTM of the hidden layer performs time-series dependency analysis on the three-dimensional time tensor to obtain the hidden state sequence. The backward LSTM performs reverse time-series analysis on the three-dimensional time tensor to generate a reverse hidden state sequence. The hidden state sequence and the reverse hidden state sequence are concatenated to generate a bidirectional feature matrix. The fully connected layer uses the ReLU activation function to reduce the dimension of the bidirectional feature matrix and obtain the phase drift prediction value.

4. The GPS jamming phased array monitoring and direction-finding antenna system for the 1575 frequency band as claimed in claim 1, characterized in that: The Kalman-particle hybrid filter algorithm is used to perform nonlinear calibration on the phase drift prediction value. The specific operation is as follows: The particle filter processes the non-Gaussian noise in the phase drift prediction value through Monte Carlo sampling to obtain the particle state distribution; the particle state distribution is weighted by the second-order moment operation to generate the covariance matrix; Kalman filtering is used to remove Gaussian noise in the covariance matrix and generate a filter matrix. The inverse function of the filter matrix is ​​derived by third-order polynomial fitting to generate a dynamic calibration curve. Based on the dynamic calibration curve, the phase drift prediction value is calibrated using the Bayesian weight allocation method.

5. The GPS jamming phased array monitoring and direction-finding antenna system for the 1575 frequency band as claimed in claim 1, characterized in that: The IQ signal is subjected to orthogonal modulation and RF up-conversion processing, and a Butler matrix beam is generated using a hybrid topology beam space transformation algorithm. The specific operations are as follows: The IQ signal is image-suppressed using the Bessel interpolation method and quadrature modulated using the QPSK modulation method to obtain a suppressed RF signal. The suppressed RF signal is then up-converted using the polyphase filtering method to generate a multi-beam RF signal. A hybrid topology beam space transformation algorithm is used to perform spatial mapping on the multi-beam RF signal to obtain the Butler matrix beam.

6. The GPS jamming phased array monitoring and direction-finding antenna system for the 1575 frequency band as claimed in claim 1, characterized in that: The Butler matrix beam is used to trigger the phased array antenna to scan the airspace coverage range. The specific operations are as follows: The Butler matrix beam is input into the feed channel of the phased array antenna through a directional coupler. In the feed channel, the beam pointing is optimized using a spatial phase gradient compensation algorithm, and a scan trigger signal is obtained through a digital-to-analog converter. Based on the scanning trigger signal, the phased array antenna is driven to scan in the spatial coverage range through the time delay scanning mechanism to obtain the array signal.

7. The GPS jamming phased array monitoring and direction-finding antenna system for the 1575 frequency band as claimed in claim 6, characterized in that: The space-time filtering is used to suppress GPS interference and obtain preliminary positioning parameters. The specific operations are as follows: Space-time filtering is used to perform space-time two-dimensional weighting on the array signal to obtain the time domain signal. The GPS interference in the time domain signal is suppressed by the space-time two-dimensional generalized sidelobe cancellation method, and the preliminary positioning parameters are generated using the RAIM integrity monitoring algorithm.

8. The GPS jamming phased array monitoring and direction-finding antenna system for the 1575 frequency band as claimed in claim 1, characterized in that: The specific operations for obtaining the optimized positioning parameters are as follows: The federated learning mechanism is used to suppress noise of the preliminary positioning parameters, and the FedAvg algorithm is used for weighted aggregation to obtain multi-source fusion parameters. The hierarchical layer-by-layer alignment method is used to perform hierarchical optimization on the multi-source fusion parameters to obtain the optimized positioning parameters.

9. The GPS jamming phased array monitoring and direction-finding antenna system for the 1575 frequency band as claimed in claim 8, characterized in that: The specific operations for generating the phased array antenna monitoring direction finding report are as follows: The Neville interpolation method is used to perform ephemeris consistency check on the optimized positioning parameters to obtain orbit monitoring data. The double difference analysis method is used to perform pseudorange residual analysis on the optimized positioning parameters to obtain direction monitoring data. The track monitoring data and direction monitoring data are integrated using a dynamic phase weighting algorithm to generate a phased array antenna monitoring direction finding report.

Images