An adaptive tuning device and method for an interference-resistant satellite communication antenna

Abstract

The application discloses an adaptive tuning device and method of an anti-interference satellite communication antenna, and relates to the technical field of satellite communication, wherein the device comprises a multi-mode reconfigurable surface antenna module, an environment sensing module, a prediction module, a cooperative tuning module and a closed-loop feedback correction module; the application constructs a multi-mode radiation system through the multi-mode reconfigurable surface antenna module and the environment sensing module, obtains an electromagnetic environment and a link state, the prediction module predicts interference and signal deviation based on multi-source data, the cooperative tuning module synchronously executes radiation mode switching and array amplitude and phase pre-tuning according to a prediction result, preset nulls are formed before interference arrives, and the closed-loop feedback correction module collects signal quality in real time and corrects a prediction model and a tuning strategy, so that a prediction type, closed-loop self-optimizing active anti-interference tuning mechanism is formed, interference is not arrived, and tuning is performed in advance, and the anti-interference capability and communication stability under a dynamic complex electromagnetic environment are significantly improved.

Description

Technical Field

[0001] This invention relates to the field of satellite communication technology, and in particular to an adaptive tuning device and method for an anti-jamming satellite communication antenna. Background Technology

[0002] The primary function of a satellite communication antenna is to convert the transmitter's radio frequency energy into directional electromagnetic waves and precisely project them onto an orbiting satellite, or to capture weak signals from space. Compared to conventional antennas, anti-jamming satellite communication antennas typically employ adaptive nulling antenna technology. This involves using array signal processing algorithms to automatically create a deep null in the direction of the interference source while maintaining the main beam aligned with the useful satellite. These antennas often combine spread spectrum and frequency hopping techniques to give the communication link a low probability of interception and strong anti-jamming capabilities.

[0003] Tuning a satellite communication antenna is essentially the process of impedance matching between the antenna and the feeder system, with the core objective of ensuring efficient transmission of radio frequency energy. In the receiver link, precise tuning can significantly improve the signal-to-noise ratio, ensuring the capture of weak satellite signals.

[0004] With the rapid growth of satellite communication services and the increasing complexity of the electromagnetic environment, satellite communication antennas face increasingly severe interference challenges. Existing anti-jamming satellite communication antenna tuning technologies have two major limitations: First, the tuning dimension is singular, mainly relying on spatial amplitude and phase adjustments, and multi-mode tuning only increases the number of modes, failing to achieve real-time linkage between mode selection and signal tuning; second, the tuning mechanism is lagging, generally following a passive logic of "interference arrival—detection—response," with significant response delays, making it difficult to cope with fast-moving interference sources and satellite orbital perturbations in highly dynamic scenarios.

[0005] For example, Chengdu Guoxing Communication Co., Ltd.'s "A Method to Improve the Convergence Speed ​​of Beidou Nulling Antennas" (CN120871182A), published in 2025, attempts to accelerate convergence by predicting interference direction through deep learning and combining it with the LMS algorithm for weight iteration. However, this scheme only accelerates the weight iteration process after interference arrives. The role of its deep learning module is limited to assisting in accelerating the convergence of existing nulling algorithms. The prediction results are not used to guide early tuning (such as pre-switching the antenna radiation mode or pre-configuring amplitude and phase weights). The system itself still lacks the ability to actively predict and compensate in advance, and there is still a significant response lag in interference transient scenarios. At the same time, the antenna radiation mode is fixed in this scheme, and tuning is only reflected at the weight adjustment level. It cannot adapt to different electromagnetic scenarios by switching the radiation mode. When the interference direction is close to the useful signal direction, the nulling process is prone to antenna pattern distortion and communication range loss.

[0006] Therefore, this invention proposes an adaptive tuning device and method for anti-interference satellite communication antennas to solve the problems existing in the prior art. Summary of the Invention

[0007] To address the aforementioned problems, the present invention aims to propose an adaptive tuning device and method for anti-interference satellite communication antennas, which combines a multi-mode reconfigurable surface antenna with integrated air-space-ground prediction to achieve active and intelligent anti-interference tuning.

[0008] To achieve the objectives of this invention, the invention is implemented through the following technical solution: an adaptive tuning device for an anti-interference satellite communication antenna, comprising a multi-mode reconfigurable surface antenna module, an environment sensing module, a prediction module, a cooperative tuning module, and a closed-loop feedback correction module, wherein: Multimode reconfigurable surface antenna module is used to provide a variety of switchable orthogonal radiation modes—that is, modes with different beam shapes and polarization characteristics achieved by changing the surface structure or excitation method, in order to adapt to different interference environments; The environmental perception module is used to collect electromagnetic environment data and satellite link status data in real time, and to build a dynamic model that represents the current electromagnetic environment. This model is a digital mapping of the real electromagnetic situation and is used to support subsequent predictions. The prediction module, connected to the environmental perception module, is used to fuse multi-source data and predict future interference situation and signal transmission offset, and output prediction information including tuning lead and optimal radiation mode combination. The lead provides time margin for subsequent tuning, so that the tuning action is completed before the interference arrives. The collaborative tuning module is connected to both the multimode reconfigurable surface antenna module and the prediction module. It has a built-in pattern recognition unit and a weight calculation unit. Based on the prediction information, the pattern recognition unit selects the optimal radiation mode combination to control the multimode reconfigurable surface antenna module to switch to the corresponding radiation mode. At the same time, the weight calculation unit calculates the amplitude and phase configuration weights of the antenna array and pre-tunes the excitation amplitude and phase of each channel to form a preset null in the predicted interference direction. This realizes the linkage between "mode switching" and "amplitude and phase pre-tuning", which is the core difference from traditional single tuning. The closed-loop feedback correction module is connected to the multi-mode reconfigurable surface antenna module, the prediction module, and the cooperative tuning module, respectively. It is used to extract the quality parameters of the tuned satellite signal, compare the quality parameters with the prediction information, and output correction information based on the comparison results to correct the prediction model parameters of the prediction module and the tuning strategy of the cooperative tuning module—forming a closed loop of "prediction-execution-evaluation-correction" to enable the system to have self-optimization capabilities.

[0009] Further improvements are made in that: the multimode reconfigurable surface antenna module adopts a single-fed low-profile metasurface structure—that is, beam control is achieved through a subwavelength element array, which is small in size and low in profile. The phase separation of degenerate modes of the same order is achieved through the arrangement design of anisotropic metamaterial elements—the anisotropy of the elements is used to break mode degeneracy, so that modes with the same frequency are separated in space, and a multimode radiation system containing at least 10 orthogonal broadband modes is constructed. The multimode radiation system covers half-mode, odd and even modes and higher-order modes—corresponding to different radiation patterns and polarization characteristics, respectively, and is suitable for different interference scenarios.

[0010] Further improvements are made in that the multimode reconfigurable surface antenna module can maintain stable circular polarization radiation characteristics within a 40% relative bandwidth—that is, within the operating frequency band, the axial ratio and main lobe width change little, ensuring signal transmission and reception quality.

[0011] Further improvements are made in the following ways: The pattern recognition unit uses a deep learning algorithm to analyze the signal-to-noise ratio, interference type, and satellite orbit perturbation parameters of the received signal to dynamically select the optimal combination of radiation modes—using a neural network to learn the mapping relationship between patterns and the environment from historical data, replacing fixed rules; The weight calculation unit uses an optimization algorithm combining recursive least squares and deep learning to dynamically adjust the amplitude and phase configuration and null depth—recursive least squares ensures the convergence speed, while deep learning provides initial value guidance to avoid local optima.

[0012] Further improvements include: the data acquisition frequency of the environmental perception module is no less than 100Hz, and the update frequency of the electromagnetic environment dynamic model is consistent with the data acquisition frequency—ensuring that the model can reflect millisecond-level electromagnetic changes and meet the requirements of high-dynamic scenarios.

[0013] Further improvements are made in the following aspects: The prediction module has a built-in AI prediction model and adopts an encoder-decoder architecture. The encoder extracts deep features from the input sequence, and the decoder generates the prediction sequence. The input layer receives at least the time series of satellite ephemeris data, the real-time sequence of atmospheric disturbance monitoring data, the trajectory prediction sequence of ground interference source location data, and the electromagnetic situation distribution map output by the electromagnetic environment dynamic model. The spatiotemporal correlation features of multi-source data are extracted through a multi-head attention mechanism. Multi-head attention can focus on the correlation of different time points and different data sources in parallel. The output layer generates the prediction of the interference source movement trajectory, the prediction of the signal angle of arrival offset, and the matching probability distribution of the optimal radiation mode combination. That is, it gives the future direction of the interference, the trend of signal angle change, and the applicability probability of each mode.

[0014] Further improvements are made in the following aspects: the signal quality parameters collected by the closed-loop feedback correction module include residual power detection results—reflecting the residual interference strength after interference suppression, receiver statistics—such as demodulation signal-to-noise ratio and bit error rate, and channel consistency status—amplitude and phase consistency fluctuation values ​​of each channel; the residual power detection results are used to reflect the residual interference level, the receiver statistics include demodulation signal-to-noise ratio and bit error rate, and the channel consistency status is used to monitor the amplitude and phase consistency fluctuations of each array element channel—these three factors evaluate the tuning effect from three dimensions: interference suppression, communication quality, and hardware status.

[0015] Further improvements are made in that: when the collaborative tuning module performs pre-tuning, it synchronously links radiation mode switching and signal weight tuning, wherein: for strong interference scenarios, it switches to a higher-order mode and forms nulls in the interference direction through the weight calculation unit—the higher-order mode has a narrow beam and strong directionality, and combined with nulls, it can suppress strong interference in a directional manner; for weak signal scenarios, it switches to a half-mode to improve signal acquisition capability—the half-mode has a wider beam and higher sensitivity, which is suitable for weak signal reception.

[0016] An adaptive tuning method for an adaptive tuning device of an anti-jamming satellite communication antenna includes the following steps: S1, System Initialization Configure the multimode reconfigurable surface antenna module to enter multimode standby mode—that is, a default mode that takes into account basic communication; start the data acquisition of the environmental perception module and build an initial electromagnetic environment dynamic model; initialize the AI ​​prediction model of the prediction module and the tuning parameters of the cooperative tuning module—set the initial amplitude and phase weights, null depth, etc. S2, Multi-source data acquisition and prediction The environmental perception module continuously collects and updates the electromagnetic environment dynamic model, and the prediction module integrates multi-source data and outputs the interference situation prediction and optimal radiation mode combination suggestion within a future time window—the prediction time window needs to balance lead time and accuracy. S3, Predictive Multimode Pretuning The cooperative tuning module controls the multimode reconfigurable surface antenna module to switch to the optimal radiation mode based on the prediction results, and at the same time pre-tunes the amplitude and phase configuration of the antenna array to form a preset null in the predicted interference direction. This step is completed before the interference actually arrives, which is the core manifestation of "active". S4. Performance Evaluation and Closed-Loop Optimization The closed-loop feedback correction module collects the signal quality parameters after tuning. If the quality parameters meet the preset requirements, the current tuning configuration is maintained. Otherwise, a correction command is output to correct the model parameters of the prediction module and the tuning strategy of the collaborative tuning module, and the pre-tuning step is repeated to achieve adaptive iterative optimization. S5, Dynamic Maintenance The system continuously updates the prediction model during operation. When environmental changes are detected, the prediction, pre-tuning, and feedback correction steps are re-executed to ensure that the system always tracks environmental changes and maintains optimal anti-interference performance.

[0017] Further improvements include: the future time window is 10ms to 100ms—balancing prediction lead and model response speed; the antenna beam pointing error after tuning is controlled within 0.1°—ensuring that the satellite alignment accuracy does not decrease significantly due to tuning; and the system tuning response speed is maintained at the millisecond level—that is, the total delay from environmental change to tuning completion is better than the traditional hundred-millisecond level.

[0018] The beneficial effects of this invention are as follows: This invention achieves joint optimization of radiation mode and amplitude and phase weights through the linkage mechanism of multi-mode reconfigurable surface antenna and cooperative tuning module, resulting in a more comprehensive tuning dimension and an improvement in interference suppression ratio of more than 15dB. At the same time, by using the air-space-ground multi-source data fusion prediction module to predict the interference situation in advance and perform pre-tuning, the response time is shortened from hundreds of milliseconds to milliseconds, significantly improving dynamic adaptability. In addition, the closed-loop feedback correction module enables the system to have self-optimization capability, significantly reducing the bit error rate when the environment changes abruptly. Furthermore, predictive pre-tuning avoids the frequent matrix inversion operations in traditional STAP, effectively reducing hardware resource consumption. Attached Figure Description

[0019] Figure 1 This is a schematic diagram of the frame structure of the adaptive tuning device for the anti-interference satellite communication antenna according to Embodiment 1 of the present invention; Figure 2 This is a schematic flowchart of the adaptive tuning method of the adaptive tuning device for the anti-interference satellite communication antenna according to Embodiment 2 of the present invention. Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] It should be noted that the technical means not described in detail in the following embodiments are all conventional means in the art, are not the key points of the invention, and will not be elaborated upon.

[0022] Example 1 See Figure 1This embodiment provides an adaptive tuning device for an anti-jamming satellite communication antenna. The device includes a multi-mode reconfigurable surface antenna module, an environment sensing module, a prediction module, a cooperative tuning module, and a closed-loop feedback correction module, wherein: 1. Multimode reconfigurable surface antenna module The multimode reconfigurable surface antenna module is the core hardware component of the device, employing a single-fed low-profile metamaterial surface structure. This module is implemented by arranging anisotropic metamaterial elements on a dielectric substrate. Each metamaterial element consists of a metal patch and metallized vias. The dielectric substrate uses Rogers RO4003C material with a thickness h = 1.524 mm and a dielectric constant εr = 3.55. The metal patch is square with a side length L = 8.2 mm and a center-to-center spacing p = 10 mm between adjacent elements. The metallized vias are located at the intersection of the patch diagonals, with a aperture d = 0.6 mm. Active phase separation of degenerate modes of the same order is achieved by adjusting the patch size, via position, and the on / off state of the loaded PIN diode. Specifically, through the on / off control of the PIN diode, the equivalent capacitance of the element is controlled at C... on =0.5pF (conduction) and C off The phase of the cell can be switched between 0.1 pF (cutoff) and 180° (on) to allow the cell phase to switch between two states. The coded state combinations of the 16×16 cell array can form different surface current distributions, thereby exciting the corresponding radiation modes.

[0023] In this embodiment, ten orthogonal broadband modes were constructed through optimized design, including two half-modes, four odd-even modes (two even modes and two odd modes), and four higher-order modes. The half-modes are achieved by controlling only half of the metasurface units in the excited state, characterized by a wide main lobe width (approximately 90°–120°) and low gain (approximately 3–5 dBi), suitable for weak signal acquisition scenarios. The odd-even modes are achieved through symmetrical settings of the array coding states; different combinations of odd-even modes can achieve differential feeding and beamforming. The higher-order modes are achieved by controlling the array to present a specific higher-order phase distribution, characterized by a narrow main lobe width (approximately 20°–30°), high gain (approximately 12–15 dBi), and strong beam directivity, suitable for scenarios with strong interference penetration. The modes are rapidly switched between each other through digital coding (i.e., the combination of PIN diode states of each unit), with a switching time of less than 1 μs. Actual measurements show that within 40% of the relative bandwidth (4.0–6.0 GHz, center frequency 5.0 GHz), the antenna module exhibits an axial ratio of less than 3 dB and a main lobe width fluctuation of less than 5%, meeting the requirements for wideband circular polarization operation. This module connects to the co-tuning module via an SPI bus, receives mode switching control signals, and dynamically switches the current radiation mode according to instructions.

[0024] 2. Environmental Perception Module The environmental perception module connects to spectrum monitoring sensors and a satellite orbit database. This module acquires ambient spectrum data in real time at a frequency of 100 Hz, performs time-frequency analysis using short-time Fourier transform (window length 256 points, overlap rate 75%), and extracts interference source characteristic parameters, including the bandwidth attributes (narrowband or wideband), transmission attributes (active or passive), center frequency, and power intensity of the interference signal. Simultaneously, it receives BeiDou / GPS satellite orbit monitoring data to obtain real-time perturbation parameters of the satellite orbits (including position deviation, velocity deviation, and orbital inclination changes). Based on the acquired multi-source data, this module constructs and dynamically maintains a dynamic electromagnetic environment model using a combination of finite element analysis and radial basis function neural networks. This model accurately replicates the transient distribution of the current electromagnetic environment in a 32×32×16 three-dimensional grid, with each grid point containing the electric field intensity E. x / E y / E z and magnetic field strength H x / H y / H z Six-dimensional physical quantities. This module communicates bidirectionally with the prediction module. On the one hand, it provides real-time acquired data via Ethernet interface using UDP protocol. On the other hand, it receives correction instructions from the closed-loop feedback correction module to update model parameters.

[0025] 3. Prediction Module The prediction module communicates bidirectionally with the environmental perception module and also accesses external data sources, including satellite ephemeris data (GPS broadcast ephemeris, updated at 1Hz), atmospheric disturbance monitoring data (from the National Meteorological Data Center, including temperature, humidity, air pressure and tropospheric delay parameters, updated at 5Hz), and ground interference source location and trajectory prediction data (from real-time reporting from the ground monitoring network and Kalman filter trajectory prediction, updated at 10Hz).

[0026] This module incorporates an AI prediction model and employs an encoder-decoder architecture. The encoder consists of four stacked Transformer encoding layers, each containing a multi-head self-attention mechanism (8 attention heads H) and a feedforward neural network (hidden layer dimension d). ff =512, activation function GELU), and residual connections and layer normalization. Input embedding dimension d of the encoding layer. model=256, maximum sequence length L=128. The decoder consists of four stacked Transformer decoding layers. Each layer, in addition to containing the same multi-head self-attention mechanism and feedforward network as the encoder, also adds an encoder-decoder cross-attention layer to extract contextual information related to the current decoding position from the encoder output. The model's input layer receives time series data from four categories through linear projection and sinusoidal position encoding: time series of satellite ephemeris data (time step Δt=10ms, sequence length T=64), real-time sequence of atmospheric disturbance monitoring data (T=32), trajectory prediction sequence of ground interference source location data (T=32), and electromagnetic situation distribution map output by the electromagnetic environment dynamic model (downsampled and flattened into a 256-dimensional feature vector). Through parallel computation of the multi-head attention mechanism, the model extracts the spatiotemporal correlation features of multi-source data from these four heterogeneous data categories. The model's loss function uses a weighted combination of smoothed L1 loss and cross-entropy loss, with a total loss L=L smoothL1 +0.5×L CE Batch size size =64, the optimizer used is AdamW, and the initial learning rate is set to 1×10. -4 It decreases by 0.9 every 20 epochs.

[0027] The output layer of this model generates three prediction results within a time window of 10–100 ms through linear mapping and Softmax normalization: interference source trajectory prediction (output in the form of two-dimensional coordinates (x,y) or three-dimensional coordinates (x,y,z), with confidence level), signal angle of arrival offset prediction (output in the form of azimuth angle Δθ and elevation angle Δφ), and the matching probability distribution of ten radiation modes as the optimal mode (a probability vector of length 10, with the sum of all components being 1). This module outputs the prediction results, including the tuning lead (usually the time point with the highest probability) and the optimal radiation mode combination (the 1–2 modes with the highest matching probability), to the co-tuning module via a high-speed serial interface (SPI, 10 Mbps).

[0028] 4. Coordinated Tuning Module The cooperative tuning module is connected to both the multimode reconfigurable surface antenna module and the prediction module. This module integrates a pattern recognition unit and a weight calculation unit, which exchange data in real time via a shared cache (1024×32bit). The pattern recognition unit transmits the recognition result (optimal pattern encoding) to the weight calculation unit in real time. The weight calculation unit incorporates the radiation pattern constraint corresponding to the pattern as prior information into the objective function when calculating the weights. The module uses an ARM Cortex-M7 core as the main control processor with a clock speed of 400MHz and 32MB of SDRAM for intermediate data caching.

[0029] The pattern recognition unit employs a deep learning algorithm that integrates convolutional neural networks and long short-term memory networks (CNN-LSTM hybrid network). It analyzes the signal-to-noise ratio of the received signal in real time (obtained through pilot symbol estimation, with an accuracy of ±0.5 dB), identifies interference types through time-frequency analysis (including four types: single-tone interference, narrowband interference, wideband sweep interference, and impulse interference), and extracts satellite orbital perturbation parameters from ephemeris (including position deviation Δr, velocity deviation Δv, and time deviation Δt), dynamically selecting the optimal combination of radiation modes. The CNN part contains three convolutional layers (kernel sizes of 3×3, 5×5, and 3×3, output channels of 32, 64, and 128 respectively, stride of 1, and padding of SAME) and two max-pooling layers (pooling window of 2×2, stride of 2) to extract local spatiotemporal features of the spectral data. The LSTM part contains two bidirectional LSTM layers (hidden layer dimensions of 128 and 256 respectively) to model long-term dependencies in the time series. The output layer outputs the probability distribution of ten modes through a fully connected layer and Softmax, and the mode with the highest probability is selected as the current optimal choice.

[0030] The weight calculation unit employs a hybrid optimization algorithm combining recursive least squares and deep reinforcement learning. The deep reinforcement learning utilizes a deep Q-network (DQN), with a state space containing the current interference direction, signal angle of arrival, and signal-to-noise ratio (SNR). The action space contains the coarse adjustment of the amplitude and phase weights (quantization step size Δw = 0.1), and the reward function R is defined as the change in output SNR ΔSINR (in dB). The DQN network structure consists of three fully connected layers (input dimension = state dimension, hidden layer dimensions are 128 and 64 respectively, and output dimension = action space size). The ε-greedy exploration strategy has an initial ε value of 0.9, decaying by 0.99 every 100 steps, and the target network is updated every 50 steps. The recursive least squares part uses a forgetting factor λ = 0.995 and a regularization parameter δ = 10. -4 The algorithm performs fine-grained iterative optimization based on the coarsely adjusted initial weights output by DQN. This hybrid algorithm provides global initial weight guidance through reinforcement learning, while the RLS algorithm converges quickly in the local range. It avoids the direct inversion of the covariance matrix and the sensitivity to local optima of the traditional RLS algorithm, thereby reducing the consumption of FPGA or DSP hardware resources (register usage is reduced by about 40%) while ensuring convergence accuracy.

[0031] During pre-tuning, this module synchronously links radiation mode switching and signal weight tuning: First, based on the optimal mode code output by the mode recognition unit, a mode switching command (8-bit control word) is sent to the multimode reconfigurable surface antenna module via the SPI bus. The antenna module completes the state reconfiguration of the surface PIN diode array within 1μs after receiving the command. At the same time, after receiving the same prediction result, the weight calculation unit calculates the amplitude and phase weights in a parallel pipeline manner (calculation time is about 150μs), and configures the calculated 32 sets of amplitude and phase weights (each set contains 16-bit precision data for both I and Q channels) into the amplitude and phase adjustment circuits of each channel through the JESD204B high-speed serial interface (line rate 3.125Gbps). For scenarios with strong interference (interference signal power exceeding the useful signal power by more than 10dB), the system automatically switches to a higher-order mode. This mode features a narrow beam (3dB beamwidth of approximately 25°) and high gain (approximately 13dBi). Through a weighting calculation unit, a null with a depth exceeding 30dB is formed in the predicted interference direction, effectively suppressing strong interference. For scenarios with weak signals (received signal power below -100dBm), the system automatically switches to a half-mode. This mode features a wide beam (3dB beamwidth of approximately 100°) and high receiving sensitivity (approximately -110dBm), prioritizing the acquisition capability of weak satellite signals.

[0032] 5. Closed-loop feedback correction module The closed-loop feedback correction module is connected to the multimode reconfigurable surface antenna module, the prediction module, and the cooperative tuning module, respectively. This module is implemented on a Xilinx Zynq UltraScale+ FPGA (model XCZU9EG), with approximately 65% ​​logic resource utilization and an operating frequency of 200MHz. The module extracts the tuned satellite signal from the output of the multimode reconfigurable surface antenna module via a coupler (coupling degree -20dB), and extracts three types of signal quality parameters in real time through an on-chip integrated signal analysis unit (including a power detector, demodulator, and vector network analysis functions). The residual power detection result is obtained using a broadband power detector (model ADL6010, detection range -60dBm to +10dBm, response time <10μs). The average of 100 samples continuously acquired by the FPGA's internal 12-bit ADC (sampling rate 1Msps) is used to quantify the residual interference level after interference suppression. The demodulated signal-to-noise ratio (SNR) and bit error rate (BER) are obtained through the receiver baseband processing unit: SNR estimation uses the pilot symbol-assisted minimum mean square error method, with a measurement range of -10dB to 30dB and an accuracy of ±0.5dB; BER is calculated by comparing the difference between the demodulated symbols and the locally reconstructed symbols, and statistics are performed once per 2048 bits per frame. Channel consistency status is obtained through the channel calibration circuit: a known calibration signal is injected during system idle time (approximately every 100ms), and the amplitude consistency (allowable deviation <±0.5dB) and phase consistency fluctuation (allowable deviation <±5°) between each RF transceiver channel are measured. Automatic calibration is triggered when the deviation exceeds the threshold.

[0033] This module compares the extracted real-time signal quality parameters with the prediction information output by the prediction module. The "interference source trajectory prediction" and "signal angle of arrival offset prediction" in the prediction information are used for point-by-point spatiotemporal consistency comparison with the measured residuals. The comparison results are processed by a fuzzy logic decision unit—this unit takes residual power, SNR, BER, and channel consistency deviation as inputs, and uses a Mamdani-type fuzzy inference system (3 input membership functions (low / medium / high), 9 fuzzy rules) to output a comprehensive health score between 0 and 1. When the health score is lower than a preset threshold (initially set to 0.7, which can be adaptively adjusted based on historical statistical results), it is determined that the tuning effect is unsatisfactory. At this time, the module generates correction instructions via the AXI-Lite bus using register mapping: sending a "model fine-tuning" instruction to the prediction module (with the most recent 10 input-output sample pairs, triggering online incremental learning, each update only corrects the weights of the last two network layers); and sending a "retuning" instruction to the co-tuning module (with a new tuning target—deep correction Δθ). target (or mode adjustment suggestions).

[0034] The above cycle forms a closed-loop self-optimization mechanism of "prediction → execution → evaluation → correction", which enables the device to maintain optimal anti-interference performance in a dynamically changing electromagnetic environment.

[0035] Example 2 See Figure 2 This embodiment provides an adaptive tuning method for an anti-interference satellite communication antenna. The method is based on the device in Embodiment 1 and specifically includes the following five steps.

[0036] S1, System Initialization After the device is powered on and started, the following sub-steps are executed: S11. The multimode reconfigurable surface antenna module enables the multimode standby mode, selecting half-mode and even-mode as the default operating mode combination (initially encoded as all 0s). At this time, the antenna has a medium gain (about 8dBi) and a wide beamwidth (about 60°), ensuring the initial communication establishment.

[0037] S12. The environmental perception module starts the data acquisition thread, which begins to collect spectrum data and satellite orbit data at a frequency of 100Hz. Based on the data of the initial 30 sampling cycles (about 300ms of historical data), an initial electromagnetic environment dynamic model is constructed. The model construction adopts a combination of finite element analysis and RBFNN. The initial model mesh resolution is 16×16×8 (slightly lower than the running accuracy to speed up the startup).

[0038] S13. The prediction module loads the pre-trained AI prediction model weight file (approximately 45MB) from local non-volatile memory, completes the initialization of all network parameters, performs layer-by-layer parameter verification during the loading process, and sets the model to running mode after the verification passes.

[0039] S14. The co-tuning module loads the preset initial tuning parameters: initial amplitude and phase weights w all =[w1,w2,...,wN] (N=32, initial weights for each channel are set to [1,0], i.e., amplitude normalization and phase zero, corresponding to the uniform array response in an untuned state), the mode matching decision threshold is set to 0.65 (i.e., when the Softmax output probability of the optimal mode is greater than 0.65, the mode is switched directly; otherwise, the current mode is maintained and the prediction is strengthened), and the initial null depth is set to -20dB. These parameters are written into the FPGA's configuration register.

[0040] S2, Multi-source data acquisition and prediction The environmental perception module continuously collects surrounding spectrum data, interference source characteristic parameters, signal propagation path information, and satellite orbit perturbation parameters at a frequency of 100Hz. Every 10ms, it pushes a data packet (containing a 256-point spectral power density vector, an 8-dimensional interference source characteristic vector, and a 6-dimensional satellite orbit parameter vector) to the prediction module. Simultaneously, the data push triggers incremental updates to the electromagnetic environment dynamic model. The updates employ a recursive least squares approach, with a new data weighting coefficient β=0.95 and a historical data attenuation factor of 0.05.

[0041] After receiving the data, the prediction module first performs time alignment and normalization on the four types of data (normalized to the [0,1] interval), and then performs inference calculations through the built-in encoder-decoder architecture AI prediction model. Inference is completed on an FPGA accelerator (fixed delay <3ms, power consumption <2W), and calculates and outputs the predicted trajectory of the interference source within the next 10-100ms time window (10 points are output in 10ms time steps, each point including the predicted location of the interference source (x,y,z) and confidence level c), and the predicted angle of arrival offset of the signal (Δθ and Δφ in the next 10ms). Based on the above outputs, the prediction module automatically generates the corresponding tuning advance parameters (taking the time point with the highest confidence in the prediction results as the advance, for example, if the confidence level reaches its peak at t+30ms, then the advance is set to 30ms) and a recommended list of optimal radiation mode combinations (arranged in descending order of matching probability, with the top ten modes retained as candidates).

[0042] S3, Predictive Multimode Pretuning The collaborative tuning module receives the prediction result output from step S2 and performs pre-tuning according to the following procedure: S31. The pattern recognition unit receives the optimal radiation mode combination recommendation list from the prediction results, combines it with the real-time characteristics of the currently received signal (signal-to-noise ratio, interference type identification results), and performs fusion inference through a CNN-LSTM network (inference time is about 50μs) to determine the final optimal radiation mode and generate an 8-bit mode switching control word.

[0043] S32. The cooperative tuning module sends a mode switching command to the multimode reconfigurable surface antenna module via the SPI bus. Upon receiving the command, the antenna module completes the state reconfiguration of the surface PIN diode array within 1μs, switching from the current operating mode to the selected optimal radiation mode. The mode switching process employs a "configure-then-disconnect" anti-interruption mechanism, i.e., first configuring the bias voltage of the new mode unit to the target state, waiting 2μs for stabilization, and then disconnecting the bias of the original mode to ensure that the antenna radiation power is not interrupted during the switching process.

[0044] S33. The weight calculation unit calculates the amplitude and phase configuration weights that match the optimal radiation mode combination based on the tuning lead parameter (e.g., 30ms). The calculation process first obtains coarse-tuned weights through a DQN network (inference time approximately 500μs), and then performs fine-tuning iterative optimization using the RLS algorithm based on the coarse-tuned values ​​(iterations K = 10-20, total time approximately 300μs). During the weight calculation process, utilizing the tuning lead time margin, the weight calculation can be performed in parallel with mode switching (in actual testing, mode switching takes 1μs, and weight calculation takes approximately 800μs, thus providing ample time to complete all configurations before interference arrives).

[0045] S34. After the weight calculation is completed, 32 sets of amplitude and phase weights (16-bit I-channel + 16-bit Q-channel per set) are configured into the amplitude and phase adjustment circuits of each channel via the JESD204B interface. The configuration completion time is approximately 50μs. Since the lead time (30ms) is much greater than the total time of all configuration actions (approximately 1ms), the pre-tuning can be completed before the physical interference signal reaches the receiver, forming a preset deep null in the direction of the predicted interference source, thus completing active pre-tuning.

[0046] S4. Performance Evaluation and Closed-Loop Optimization After the pre-tuning takes effect in step S3, the closed-loop feedback correction module captures the tuned satellite signal in real time and performs performance evaluation and closed-loop optimization according to the following process: S41. Signal Acquisition and Parameter Extraction: The module continuously acquires IQ data 64 times (1024 samples each time) through the ADC, and extracts signal quality parameters such as residual power detection results, demodulation signal-to-noise ratio, bit error rate, and channel consistency status. The extraction process uses a sliding window averaging method with a window length of 16 sampling periods to ensure parameter stability.

[0047] S42. Comprehensive Performance Evaluation: Based on a multi-dimensional performance evaluation index system, this system uses residual power detection results, receiver statistics (SNR and BER), and channel consistency status as core evaluation indicators. It employs a Mamdani-type fuzzy logic data fusion decision algorithm to comprehensively quantify and score the interference suppression effect, signal transmission quality, and channel stability of the current tuning configuration, outputting a health score H between 0 and 1. The scoring formula is: H = 0.5 × S res +0.3×S snr +0.2×S cons Where S res S snr S cons These are the normalized scores of residual power, SNR, and channel consistency (mapped to the 0-1 interval).

[0048] S43. Threshold Decision and Response: The comprehensive score H is compared with the preset communication quality threshold (initially set to 0.7). If H ≥ 0.7, the current tuning configuration is deemed to be effective. The currently effective tuning parameter set (including mode coding, amplitude and phase weight coefficients, and null depth parameters) is used as new baseline data and fed back to the prediction module to enhance the accuracy of its prediction model (by incrementally learning and updating network weights through a small number of samples).

[0049] S44. Iterative optimization: If H < 0.7, the tuning effect is determined to be unsatisfactory. A correction instruction is generated to trigger the prediction module to correct its model parameters (using the most recent 10 sets of data samples for model fine-tuning), and the collaborative tuning module is instructed to perform global iterative optimization of the tuning parameters (re-execute step S3). Then, the pre-tuning is re-executed until H ≥ 0.7 or the number of iterations exceeds the maximum limit (MaxIter = 5).

[0050] S5, Dynamic Maintenance After completing the initial tuning, the system continues to run, repeating steps S2 to S4. When the prediction module detects sudden changes in the satellite orbit (position deviation Δr exceeds 50m or velocity deviation Δv exceeds 0.5m / s), the appearance of a new interference source (a new high-power signal is detected in the spectrum and lasts for more than 10ms), or the disappearance of an existing interference source (power of the original interference band drops by more than 20dB) through real-time data monitoring, it immediately triggers a rapid recalculation of the prediction model (using the latest data as the starting point, only forward inference is performed, without backpropagation updates, and the inference delay is <3ms), and drives the cooperative tuning module to immediately start a new round of mode switching and weight retuning process.

[0051] In this embodiment, the tuning response speed of the dynamic maintenance process is maintained within 10ms (the total delay from environmental change detection to the completion of a new round of pre-tuning), and the error between the antenna beam main lobe pointing and the theoretical value after tuning is controlled within 0.1°. The above performance data comes from the following test scenario: the interference source moves relative to the satellite at a speed of 50° / s, the signal arrival angle change rate caused by satellite orbital perturbation is 2° / s, and the signal fading depth caused by atmospheric disturbance reaches a maximum of 8dB. Under the above scenario, the bit error rate of the system in this embodiment is maintained at 10... -5 The error rate is on the order of magnitude, while the untuned traditional system has a bit error rate of 10. -3 The tuning response speed is improved by 8 to 12 times compared to the traditional adaptive zeroing method based on LMS (response time of about 80 to 120 ms), and by 4 to 6 times compared to the traditional method based on RLS (response time of about 40 to 60 ms), which fully demonstrates the significant technical progress of the present invention.

[0052] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. An adaptive tuning device for an anti-interference satellite communication antenna, characterized in that, include: A multimode reconfigurable surface antenna module is used to provide multiple switchable orthogonal radiation modes; The environmental perception module is used to collect electromagnetic environment data and satellite link status data in real time, and to build a dynamic model representing the current electromagnetic environment. The prediction module is used to fuse multi-source data and predict future interference situations and signal transmission offsets, and outputs prediction information including tuning lead and optimal radiation mode combination. The collaborative tuning module has a built-in pattern recognition unit and a weight calculation unit. Based on the prediction information, the pattern recognition unit selects the optimal radiation mode combination to control the multimode reconfigurable surface antenna module to switch to the corresponding radiation mode. At the same time, the weight calculation unit calculates the amplitude and phase configuration weights of the antenna array and pre-tunes the excitation amplitude and phase of each channel to form a preset null in the predicted interference direction. The closed-loop feedback correction module is used to extract the quality parameters of the tuned satellite signal, compare the quality parameters with the prediction information, and output correction information based on the comparison results to correct the prediction model parameters of the prediction module and the tuning strategy of the cooperative tuning module.

2. The adaptive tuning device for an anti-interference satellite communication antenna according to claim 1, characterized in that: The multimode reconfigurable surface antenna module adopts a single-fed low-profile metamaterial surface structure. Through the arrangement design of anisotropic metamaterial units, it achieves active phase separation of degenerate modes of the same order and constructs a multimode radiation system containing at least 10 orthogonal broadband modes. The multimode radiation system covers half-mode, odd and even modes and higher-order modes.

3. The adaptive tuning device for an anti-interference satellite communication antenna according to claim 2, characterized in that: The multimode reconfigurable surface antenna module can maintain stable circular polarization radiation characteristics within a 40% relative bandwidth.

4. The adaptive tuning device for an anti-interference satellite communication antenna according to claim 1, characterized in that: The pattern recognition unit uses a deep learning algorithm to analyze the signal-to-noise ratio, interference type, and satellite orbit perturbation parameters of the received signal to dynamically select the optimal radiation mode combination; the weight calculation unit uses an optimization algorithm combining recursive least squares and deep learning to dynamically adjust the amplitude-phase configuration and null depth.

5. The adaptive tuning device for an anti-interference satellite communication antenna according to claim 1, characterized in that: The data acquisition frequency of the environmental sensing module is no less than 100Hz, and the update frequency of the electromagnetic environment dynamic model is consistent with the data acquisition frequency.

6. The adaptive tuning device for an anti-interference satellite communication antenna according to claim 1, characterized in that: The prediction module has a built-in AI prediction model and adopts an encoder-decoder architecture. The input layer receives at least the time series of satellite ephemeris data, the real-time sequence of atmospheric disturbance monitoring data, the trajectory prediction sequence of ground interference source location data, and the electromagnetic situation distribution map output by the electromagnetic environment dynamic model. The spatiotemporal correlation features of the multi-source data are extracted through a multi-head attention mechanism. The output layer generates the prediction of the interference source movement trajectory, the prediction of the signal angle of arrival offset, and the matching probability distribution of the optimal radiation mode combination.

7. The adaptive tuning device for an anti-interference satellite communication antenna according to claim 1, characterized in that: The signal quality parameters collected by the closed-loop feedback correction module include residual power detection results, receiver statistics, and channel consistency status. The residual power detection results are used to reflect the residual interference level, the receiver statistics include demodulation signal-to-noise ratio and bit error rate, and the channel consistency status is used to monitor the amplitude and phase consistency fluctuations of each array element channel.

8. The adaptive tuning device for an anti-interference satellite communication antenna according to claim 1, characterized in that: When the collaborative tuning module performs pre-tuning, it synchronously links radiation mode switching and signal weight tuning. Specifically, for strong interference scenarios, it switches to a higher-order mode and forms a null in the interference direction through the weight calculation unit. For weak signal scenarios, it switches to a half-mode to improve signal acquisition capability.

9. An adaptive tuning method applied to the adaptive tuning device of an anti-interference satellite communication antenna according to any one of claims 1-8, characterized in that, Includes the following steps: S1. Configure the multimode reconfigurable surface antenna module to enter the multimode standby mode, start the data acquisition of the environmental perception module and build an initial electromagnetic environment dynamic model, and initialize the AI ​​prediction model of the prediction module and the tuning parameters of the cooperative tuning module. S2. The environmental perception module continuously collects and updates the electromagnetic environment dynamic model, and the prediction module integrates multi-source data and outputs the interference situation prediction and optimal radiation mode combination suggestion within a future time window. S3. The cooperative tuning module controls the multimode reconfigurable surface antenna module to switch to the optimal radiation mode according to the prediction result, and at the same time pre-tunes the amplitude and phase configuration of the antenna array to form a preset null in the predicted interference direction. S4. The closed-loop feedback correction module collects the signal quality parameters after tuning. If the quality parameters meet the preset requirements, the current tuning configuration is maintained. Otherwise, a correction command is output to correct the model parameters of the prediction module and the tuning strategy of the collaborative tuning module, and the pre-tuning step is repeated. S5. The system continuously updates the prediction model during operation. When environmental changes are detected, the prediction, pre-tuning and feedback correction steps are re-executed.

10. The adaptive tuning method of the adaptive tuning device for an anti-interference satellite communication antenna according to claim 9, characterized in that: The future time window is 10ms to 100ms, the antenna beam pointing error after tuning is controlled within 0.1°, and the system tuning response speed is maintained at the millisecond level.

Images