A real-time control system for air defense and disaster prevention early warning based on the third generation of Beidou communication

Abstract

The present application relates to the field of communication signal processing, in particular to a kind of air defense disaster early warning alarm real-time control system based on Beidou third-generation communication, system uses Daubechies wavelet family to carry out 5 layers discrete wavelet decomposition, and total satellite signal delay is decomposed into different scale coefficients. By accurately controlling the boundary scale, the system can effectively separate ionospheric error and tropospheric error. By inserting zero values between filter coefficients, information loss caused by downsampling operations is avoided. The system also adaptively selects the optimal wavelet basis function according to different atmospheric error characteristics. It maintains stable correction effect under extreme weather conditions and builds an efficient and practical air defense disaster real-time early warning control method.

Description

Technical Field

[0001] This invention relates to the fields of communication signal processing, atmospheric error correction and real-time early warning control, and specifically to a real-time control system for air defense and disaster prevention early warning based on BeiDou-3 communication. Background Technology

[0002] In modern information-based warfare environments and disaster early warning systems, precise positioning, navigation, and time synchronization services are fundamental to achieving accurate prevention and control. However, atmospheric conditions (including the ionosphere and troposphere) cause propagation delays, attenuation, and scattering of satellite navigation signals, severely impacting positioning accuracy and communication quality. When high-precision satellite navigation systems traverse the atmosphere, the signal propagation path and speed are affected by the atmospheric medium; in particular, changes in free electrons in the ionosphere and water vapor content in the troposphere cause significant propagation delays, leading to increased positioning errors. In military air defense and natural disaster early warning systems, such errors can result in failed target identification and tracking, or missed critical early warning opportunities, causing serious consequences.

[0003] Existing atmospheric error modeling methods are mostly based on static or semi-static models, which struggle to accurately describe the spatiotemporal variations of atmospheric parameters, especially under conditions of drastic weather changes, where model accuracy drops significantly. For example, while the International Reference Ionospheric Model (IRI) can describe the global distribution of the ionosphere, its temporal resolution is only on the hourly level, making it difficult to capture rapid changes in ionospheric disturbances. Although the Numerical Tropospheric Prediction Model (NWP) incorporates detailed physical processes, its high computational complexity makes it difficult to meet the needs of real-time correction.

[0004] Studies have shown that Multiple-Input Multiple-Output (MIMO) technology can significantly improve channel capacity and transmission reliability through spatial diversity and multiplexing; wavelet transform, due to its excellent time-frequency localization characteristics, has unique advantages in non-stationary signal analysis and denoising; the BeiDou-3 system provides high-precision positioning and short message communication services with global coverage, possessing inherent dual-use characteristics. MIMO technology, through multi-antenna transmission and reception, can create multiple parallel channels in the spatial domain, theoretically increasing channel capacity by N times (N being the number of antennas). Wavelet transform, through scaling and translation operations, can provide multi-resolution analysis capabilities, exhibiting excellent expressive power for both local features and global trends of signals. As my country's independently developed global satellite navigation system, the BeiDou-3 system provides multiple frequency signals (B1I / B3I / B1C / B2a / B2b), providing ample observational data for ionospheric error correction and precise positioning.

[0005] Therefore, by organically combining MIMO technology, wavelet transform theory and the BeiDou-3 communication system, the technical problems of insufficient accuracy, poor adaptability and weak real-time performance of traditional single signal processing methods in complex atmospheric environments are solved. By integrating MIMO antenna arrays, wavelet multi-resolution analysis and BeiDou-3 multi-frequency observation, accurate modeling and correction of atmospheric errors are achieved, providing high-precision navigation, positioning and communication support for air defense and disaster prevention applications around the clock. Summary of the Invention

[0006] This invention aims to solve the following technical problems:

[0007] How to effectively integrate the technological advantages of MIMO technology, wavelet transform, and BeiDou-3 communication system to achieve high-precision satellite navigation signal error correction in complex and ever-changing atmospheric environments, especially to maintain stable correction effects under extreme weather conditions (such as severe thunderstorms, typhoons, geomagnetic storms, etc.) in order to build an efficient and practical real-time early warning and control method for air defense and disaster prevention.

[0008] This invention proposes a real-time control system for air defense and disaster prevention early warning based on BeiDou-3 communication. The solution includes the following technical contents:

[0009] This invention employs a distributed MIMO antenna array structure to simultaneously receive BeiDou satellite signals through multiple antennas, thereby constructing a space diversity reception system.

[0010] ;

[0011] in, Represents the received signal matrix, with dimension 1. , For the number of receiving antennas, This represents the number of time samples. Represents the channel matrix, with dimensions of , Number of transmitting antennas; Represents the transmitted signal matrix, with dimension 1. ; This represents an additive white Gaussian noise matrix with dimension . .

[0012] For atmospheric propagation environments, channel matrix It can be further broken down into:

[0013] ;

[0014] in, It represents the channel characteristics in the spatial domain, describing the attenuation and phase shift of a signal during its propagation in space; It represents the time-domain channel characteristics and describes the dynamic changes of the channel over time; It represents the channel characteristics in the frequency domain and describes the differences in the propagation of signals at different frequencies; It represents the polarization characteristics of a signal and describes the changes in the polarization state of the signal during propagation.

[0015] Spatial domain channel characteristics This can be further expressed as:

[0016] ;

[0017] in, Represents from the transmitting antenna to receiving antenna Spatial propagation characteristics, Represents the signal wavelength. Represents from the transmitting antenna to receiving antenna The distance.

[0018] Time-domain channel characteristics This can be represented based on Jake's model as follows:

[0019] ;

[0020] in, represent and The time-related characteristics of time, Represents the zeroth-order Bessel function. This represents the maximum Doppler frequency shift.

[0021] Frequency domain channel characteristics It can be represented as:

[0022] ;

[0023] in, Representing frequency and Frequency correlation characteristics between them This represents the maximum delay spread.

[0024] Polarization channel characteristics It can be represented as:

[0025] ;

[0026] in, , , and These represent the coupling coefficients between horizontal-horizontal, horizontal-vertical, vertical-horizontal, and vertical-vertical polarization states, respectively.

[0027] This invention employs an encoding scheme based on Orthogonal Spatiotemporal Block Code (OSTBC), whose encoding matrix... satisfy:

[0028] ;

[0029] in, represent The conjugate transpose of the matrix. Represents a constant. represent An identity matrix of order 1.

[0030] For a 2-transmit, 2-receive MIMO system, the classic Alamouti coding is used:

[0031] ;

[0032] in, and Symbols representing two consecutive time slots transmitted. represent .

[0033] For a 4-transmit 4-receive MIMO system, extended OSTBC encoding is used:

[0034] ;

[0035] in, , , and Symbols representing four consecutive time slots transmitted.

[0036] To adapt to the characteristics of BeiDou satellite signals, this invention designs an adaptive modulation and coding scheme that dynamically adjusts the coding rate and modulation scheme based on Channel State Information (CSI). The modulation scheme is adaptively selected from BPSK, QPSK, 8PSK to 16QAM, and the coding rate is adaptively adjusted from 1 / 2 to 5 / 6. The selection method for the adaptive modulation and coding scheme is as follows:

[0037] ;

[0038] in, Indicates the use of modulation method and coding rate transmission rate Represents the signal-to-noise ratio The bit error rate.

[0039] This invention employs discrete wavelet transform to perform multi-scale analysis on the received signal in order to separate atmospheric errors into ionospheric errors and tropospheric errors.

[0040] ;

[0041] in, Representative signal wavelet coefficients, The scale parameter represents the degree of stretching or compression of the wavelet. The translation parameter represents the time position of the wavelet. This represents the complex conjugate of the wavelet basis functions.

[0042] A fast algorithm for discrete wavelet transform can be represented as a set of high-pass and low-pass filters:

[0043] ;

[0044] ;

[0045] in, Representing the Layer An approximation coefficient, Representing the Layer A detailed coefficient, and These represent the wavelet low-pass and high-pass filter coefficients, respectively, satisfying the orthogonality condition:

[0046] ;

[0047] ;

[0048] ;

[0049] in, Represents the Kronecker delta function, when The value is 1 if it is active, and 0 otherwise.

[0050] This invention employs the Daubechies wavelet family (db4-db20), the Symlet wavelet family (sym8-sym20), and the Coiflet wavelet family (coif3-coif5) as the main analytical tools, adaptively selecting the optimal wavelet basis based on different atmospheric error characteristics. The evaluation metrics for wavelet basis selection include:

[0051] ;

[0052] ;

[0053] ;

[0054] in, Represents the energy concentration of wavelet coefficients. Represents the sparsity of wavelet coefficients. The information entropy represents the wavelet coefficients. This represents the normalized energy distribution.

[0055] For BeiDou satellite signals, the ionospheric delay can be estimated using the following formula:

[0056] ;

[0057] in, Represents ionospheric delay (unit: seconds). Represents total electron content (unit: TECU). ), Represents signal frequency (unit: Hz). Represents satellite elevation angle Related mapping functions.

[0058] Mapping function The precise expression is:

[0059] ;

[0060] in, Represents the Earth's radius (approximately 6371 km). This represents the height of the ionosphere with the highest electron density (approximately 350 km).

[0061] For tropospheric delay, the Saastamoinen model is used:

[0062] ;

[0063] in, Represents tropospheric delay (unit: meters). Represents the zenith angle. Represents atmospheric pressure (unit: hPa). Represents temperature (unit: K). Represents water vapor pressure (unit: hPa).

[0064] Considering the high dependence on tropospheric delay, a height correction factor is introduced:

[0065] ;

[0066] in, Represents the tropospheric delay at sea level. Represents the high attenuation coefficient. Represents the altitude of the station (unit: km).

[0067] An atmospheric error separation algorithm based on wavelet transform is used to separate satellite signal delay. Decomposed into:

[0068] ;

[0069] in, Representing the Layer wavelet detail coefficients correspond to the high-frequency components of atmospheric disturbances. Representing the Layer wavelet approximation coefficients correspond to the low-frequency components of atmospheric disturbances. This represents the number of decomposition layers (generally 5-7 layers).

[0070] Based on the physical meaning of wavelet coefficients at different scales, an error separation model for the ionosphere and troposphere is established:

[0071] ;

[0072] ;

[0073] in, The boundary scale representing the error between the ionosphere and the troposphere (usually 2-3). This represents the proportion of ionospheric error in low-frequency components (typically 0.6-0.8, dynamically adjusted according to season and geographical location).

[0074] To improve the accuracy of wavelet decomposition, this invention employs Translation Invariant Wavelet Transform (TIWT) technology, eliminating the shift variability in traditional wavelet transforms and ensuring that the decomposition results are unaffected by the signal's temporal location. Inserting zero values ​​between wavelet filter coefficients avoids downsampling operations.

[0075] ;

[0076] ;

[0077] in, and Representing the Layer dilation filter coefficients:

[0078] ;

[0079] ;

[0080] To further improve the accuracy of atmospheric error separation, this invention proposes a fine-grained decomposition method based on wavelet packet transform (WPT), which subdivides the signal simultaneously in both high-frequency and low-frequency components, forming a complete binary structure. The WPT coefficient calculation formula is as follows:

[0081] ;

[0082] ;

[0083] in, Representing the Layer Wavelet packet coefficients of each node, Represents the node position parameter.

[0084] Based on the WPT coefficients, a more refined atmospheric error separation model is constructed:

[0085] ;

[0086] ;

[0087] in, and These represent the wavelet packet node sets corresponding to ionospheric and tropospheric errors, respectively, and are trained from historical data using machine learning methods.

[0088] This invention fully utilizes the multi-frequency signals provided by the BeiDou-3 satellite system, including B1I (1561.098MHz), B3I (1268.52MHz), B1C (1575.42MHz), B2a (1176.45MHz), and B2b (1207.14MHz), to construct a multi-frequency observation equation:

[0089] ;

[0090] ;

[0091] in, Representative receiver Receive satellite of Frequency pseudorange observations This represents the corresponding carrier phase observation value. Represents geometric distance. Represents the speed of light. and These represent the receiver clock bias and the satellite clock bias, respectively. represent The wavelength of the frequency point Represents the overall ambiguity throughout the week. and These represent pseudorange and carrier phase observation noise, respectively.

[0092] Geometric distance This can be further expressed as:

[0093] ;

[0094] in, Representative satellite coordinates Representative receiver The coordinates.

[0095] The ionosphere-free combination of the multi-frequency observation equations can be expressed as:

[0096] ;

[0097] ;

[0098] in, and Ionospherically-free combined observations, representing pseudorange and carrier phase respectively. and This represents two different frequency points.

[0099] The ionosphere-free combination eliminates the first-order ionospheric delay effect, but higher-order ionospheric effects still exist, and their influence can be expressed as follows:

[0100] ;

[0101] in, Represents the second-order ionospheric delay effect (unit: meters). Represents the intensity of the Earth's magnetic field (unit: Tesla). It represents the angle between the direction of signal propagation and the direction of the Earth's magnetic field.

[0102] To achieve high-precision positioning, Precise Point Positioning (PPP) technology is employed, utilizing precise ephemeris and clock bias products instead of broadcast ephemeris. The PPP observation equation is:

[0103] ;

[0104] ;

[0105] in, Represents precision satellite clock bias. Represents the wet delay mapping function, Represents zenith wet delay, This represents the carrier phase ambiguity term.

[0106] Considering that the BeiDou system includes satellites of different orbital types (GEO, IGSO, MEO), this invention proposes a differentiated weighting strategy:

[0107] ;

[0108] in, Representative satellite The weight, Represents the satellite's elevation angle. Representative satellite Carrier-to-noise ratio, This represents the reference carrier-to-noise ratio (typically 45 dB-Hz).

[0109] To fully utilize the short message communication function of the BeiDou-3 system, this invention designs an efficient differential correction data broadcasting scheme, encoding atmospheric error correction information into a compact binary message. The differential correction data message structure includes:

[0110] 1. Message header (4 bytes): contains message type, timestamp, and region identifier;

[0111] 2. Ionospheric grid correction (variable length): global TEC distribution is represented using spherical harmonic functions;

[0112] 3. Tropospheric grid correction (variable length): using parameters from the regional tropospheric delay model;

[0113] 4. Satellite orbit and clock corrections (variable length): Provides precise ephemeris correction parameters;

[0114] 5. Checksum (2 bytes): The CRC16 algorithm is used to ensure data integrity.

[0115] The ionospheric TEC distribution represented by the spherical harmonic function model is as follows:

[0116] ;

[0117] in, Representing latitude and longitude The total electron content at that location, Represents the normalized associated Legendre function, and Represents the spherical harmonic coefficient. This represents the maximum order (usually 8-12).

[0118] This invention constructs a three-layer early warning control system architecture:

[0119] (1) Data acquisition layer: It consists of MIMO antenna array, Beidou receiver and auxiliary sensors, and is responsible for receiving and preprocessing satellite signals at multiple frequencies;

[0120] (2) Data processing layer: It consists of wavelet transform module, atmospheric error separation module, multi-source data fusion module and signal enhancement module, and is responsible for realizing time-frequency analysis and error correction of the signal;

[0121] (3) Application Service Layer: It consists of a situation awareness module, a risk assessment module, a decision control module and an information release module, and is responsible for realizing air defense and disaster prevention early warning and control functions.

[0122] The functionality of the data acquisition layer can be described as a data acquisition function:

[0123] ;

[0124] in, represent The collection of data at any given time. Represents satellite signal observations. Represents the status of the communication link. Represents platform parameters (position, attitude, etc.).

[0125] The functionality of the data processing layer can be described as data processing functions:

[0126] ;

[0127] in, represent Real-time information products Represents historical data. This represents the parameters of the processing model.

[0128] The functionality of the application service layer can be described as a decision control function:

[0129] ;

[0130] in, represent Momentary action decisions Represents information products, Represents the risk assessment results. This represents the objective function.

[0131] The system's decision-making and control process can be described as a multi-objective optimization problem:

[0132] ;

[0133] The constraints are satisfied:

[0134] ;

[0135] ;

[0136] ;

[0137] in, Represents the objective function vector, containing One optimization objective, Represents a vector of decision variables. and These represent inequality and equality constraint functions, respectively. Represents the decision-making space.

[0138] Specific optimization objectives include:

[0139] ;

[0140] ;

[0141] ;

[0142] ;

[0143] in, , , and These represent the number of samples that are false positives, true negatives, false negatives, and true positives, respectively. Represents system response time. This represents an indicator of resource consumption.

[0144] This invention introduces an adaptive weight matrix into decision control. :

[0145] ;

[0146] in, Representing the The observation station for the first The influence weight of each observation station satisfies .

[0147] The weights are updated using reinforcement learning, dynamically adjusted based on historical data.

[0148] ;

[0149] in, Representing the The weight values ​​for the next iteration. This represents the learning rate (typically 0.01-0.1). Representing the The reward signal for the next iteration has a value range of [-1, 1].

[0150] Risk assessment uses a Bayesian network model to describe the conditional dependencies between different risk factors:

[0151] ;

[0152] in, Representing the One risk event Representing the One evidence variable, represent The set of parent nodes.

[0153] To achieve dynamic risk assessment, a dynamic Bayesian network (DBN) is introduced:

[0154] ;

[0155] in, and Represent and The risk status at all times, represent Observational evidence of the moment.

[0156] The system adopts a layered control architecture, including a strategic layer, a tactical layer, and an execution layer:

[0157] 1. Strategic Level: Responsible for long-term planning and resource allocation, with decision-making cycles ranging from hours to days.

[0158] 2. Tactical Level: Responsible for mid-term mission scheduling and coordination, with decision-making cycles at the minute level.

[0159] 3. Execution layer: Responsible for short-term real-time control and response, with a decision-making cycle in the second range.

[0160] To achieve a flexible response from the system, a five-level response mode is defined:

[0161] 1. Routine monitoring mode: daily surveillance and data collection

[0162] 2. Alert and monitoring mode: Increase the frequency of data collection and analysis.

[0163] 3. Early Warning Preparation Mode: Initiate preliminary early warning preparation and pre-configure resources.

[0164] 4. Early Warning Activation Mode: Issue a formal early warning and initiate an emergency response.

[0165] 5. Emergency Response Mode: Fully activate emergency response measures and execute rescue and disaster reduction operations.

[0166] To improve the system's ability to identify atmospheric anomaly patterns, we designed a hybrid model based on convolutional neural networks (CNNs) and long short-term memory networks (LSTMs) to identify anomalous patterns in the spatiotemporal distribution of atmospheric parameters. The model structure includes:

[0167] 1. Input layer: Receives multi-dimensional atmospheric parameter time-series data, with dimensions of... ,in Represents the number of time steps. and Represents spatial dimension, Represents the number of channels (parameter type);

[0168] 2. Spatiotemporal Convolution Module: Extracts spatiotemporal features through 3D convolution, with convolution kernel parameters as follows: Step size is ;

[0169] 3. Attention Mechanism: Introduce a self-attention mechanism to capture long-distance dependencies between different spatiotemporal locations;

[0170] 4. LSTM module: models temporal dependencies, with a hidden state dimension of 256;

[0171] 5. Fully connected output layer: Generates anomaly scores, which are mapped to... via the Sigmoid function. interval;

[0172] The model is trained using a contrastive learning method, with the loss function being:

[0173] ;

[0174] in, Representative input sample Feature representation, Represents positive samples (of the same category). Represents negative samples (different categories). Represents the cosine similarity function. This represents the temperature parameter (usually set to 0.07). This represents the set of negative samples.

[0175] To improve the model's performance on sparse outlier samples, a focal loss was used as a supplement:

[0176] ;

[0177] in, Represents the predicted probability. Represents category weight, This represents the focusing parameter (usually set to 2).

[0178] The final loss function is a weighted sum of the contrast loss and the focus loss:

[0179] ;

[0180] in, and The weighting coefficients are determined through cross-validation.

[0181] The model training data includes historical atmospheric parameter records and corresponding anomaly labels. Data augmentation techniques include random pruning, time window sliding, and Gaussian noise perturbation. To address the class imbalance problem, weighted sampling and synthetic minority class oversampling (SMOTE) were employed.

[0182] This invention has the following significant advantages over the prior art:

[0183] (1) The reliability and anti-interference capability of signal reception are improved by MIMO technology and space-time coding. The signal-to-noise ratio is improved by 3-5dB in complex electromagnetic environment and the bit error rate is reduced by an order of magnitude. Tests show that the system can still maintain a signal detection rate of more than 90% in the -15dB JSR environment, while the detection rate of traditional single antenna system is only about 50% in the -5dB JSR environment.

[0184] (2) Multi-scale wavelet transform was used to achieve accurate modeling and separation of atmospheric errors, and the accuracy of ionospheric delay correction was improved to over 90% and the accuracy of tropospheric delay correction was improved to over 85%. Under extreme atmospheric disturbance conditions (such as strong geomagnetic storms), the correction accuracy can still be maintained at over 75%, which is much higher than the 40%-60% correction rate of traditional models.

[0185] (3) By utilizing the multi-frequency collaborative observation of Beidou-3, high-precision positioning is achieved in all weather and all time, and the positioning accuracy reaches sub-meter level under complex atmospheric conditions; the system can achieve rapid integer ambiguity resolution within 60 seconds (FAST), and the initialization time is shortened by 40% compared with the traditional method. Attached Figure Description

[0186] Figure 1 The system architecture diagram provided for this invention. Detailed Implementation

[0187] The overall system architecture of this invention is as follows: Figure 1 As shown, the system mainly comprises four subsystems: a data acquisition subsystem, a signal processing subsystem, an atmospheric error correction subsystem, and an early warning and control subsystem. These subsystems exchange data and collaborate through standardized interfaces. The system adopts a modular design concept, allowing each functional module to be independently upgraded and replaced, thus improving the system's maintainability and scalability.

[0188] Data acquisition subsystem:

[0189] This subsystem implements multi-dimensional MIMO signal acquisition and space-time coding, employing a distributed MIMO antenna array structure. It consists of several receiving antenna elements. A typical configuration of the antenna array is as follows: or Depending on the application scenario, each antenna unit can simultaneously receive signals from five frequency points of BeiDou-3 satellites: B1I (1561.098MHz), B3I (1268.52MHz), B1C (1575.42MHz), B2a (1176.45MHz), and B2b (1207.14MHz).

[0190] The MIMO system is represented as:

[0191] ;

[0192] in, Represents the received signal matrix, with dimension 1. , For the number of receiving antennas, This represents the number of time samples. Represents the channel matrix, with dimensions of , Number of transmitting antennas; Represents the transmitted signal matrix, with dimension 1. ; This represents an additive white Gaussian noise matrix with dimension . .

[0193] For atmospheric propagation environments, channel matrix It can be further broken down into:

[0194] ;

[0195] in, It represents the channel characteristics in the spatial domain, describing the attenuation and phase shift of a signal during its propagation in space; It represents the time-domain channel characteristics and describes the dynamic changes of the channel over time; It represents the channel characteristics in the frequency domain and describes the differences in the propagation of signals at different frequencies; It represents the polarization characteristics of a signal and describes the changes in the polarization state of the signal during propagation.

[0196] Spatial domain channel characteristics This can be further expressed as:

[0197] ;

[0198] in, Represents from the transmitting antenna to receiving antenna Spatial propagation characteristics, Represents the signal wavelength. Represents from the transmitting antenna to receiving antenna The distance.

[0199] Time-domain channel characteristics This can be represented based on Jake's model as follows:

[0200] ;

[0201] in, represent and The time-related characteristics of time, Represents the zeroth-order Bessel function. This represents the maximum Doppler frequency shift.

[0202] Frequency domain channel characteristics It can be represented as:

[0203] ;

[0204] in, Representing frequency and Frequency correlation characteristics between them This represents the maximum delay spread.

[0205] Polarization channel characteristics It can be represented as:

[0206] ;

[0207] in, , , and These represent the coupling coefficients between horizontal-horizontal, horizontal-vertical, vertical-horizontal, and vertical-vertical polarization states, respectively.

[0208] The spatial layout of the antenna array adopts a non-uniform distribution, with an antenna spacing matrix. Defined as:

[0209] ;

[0210] in, Representing the Line number Distance between the antenna element and the center of the array (unit: meters).

[0211] Considering the wavelength differences at different frequencies, the antenna spacing design follows these principles:

[0212] ;

[0213] ;

[0214] in, and These represent the minimum and maximum antenna spacing, respectively. and These represent the minimum and maximum operating wavelengths, corresponding to the B2a and B3I frequencies, respectively.

[0215] This invention employs an encoding scheme based on Orthogonal Spatiotemporal Block Code (OSTBC), whose encoding matrix... satisfy:

[0216] ;

[0217] in, represent The conjugate transpose of the matrix. Represents a constant. represent An identity matrix of order 1.

[0218] For a 2-transmit, 2-receive MIMO system, the classic Alamouti coding is used:

[0219] ;

[0220] in, and Symbols representing two consecutive time slots transmitted. represent .

[0221] For a 4-transmit 4-receive MIMO system, extended OSTBC encoding is used:

[0222] ;

[0223] in, , , and Symbols representing four consecutive time slots transmitted.

[0224] To improve the directional performance of the array antenna, this invention designs an adaptive beamforming algorithm that dynamically adjusts the amplitude and phase characteristics of each antenna element. The beamforming weight vector calculation adopts the minimum variance distortionless response (MVDR) criterion.

[0225] ;

[0226] in, Represents the MVDR weight vector. The covariance matrix representing the received signal. Represents the target direction The guide vector, represent The conjugate transpose of .

[0227] To adapt to the characteristics of BeiDou satellite signals, this invention designs an adaptive modulation and coding scheme that dynamically adjusts the coding rate and modulation scheme based on Channel State Information (CSI). The modulation scheme is adaptively selected from BPSK, QPSK, 8PSK to 16QAM, and the coding rate is adaptively adjusted from 1 / 2 to 5 / 6. The selection criteria for the adaptive modulation and coding scheme are as follows:

[0228] ;

[0229] in, Indicates the use of modulation method and coding rate transmission rate Represents the signal-to-noise ratio The bit error rate.

[0230] The antenna array configuration parameters for different scenarios are shown in Table 1:

[0231] ;

[0232] in, The wavelength corresponding to the operating frequency (based on the B1I frequency point) varies in different scenarios, and the array size and antenna spacing are different to balance the requirements of system performance and physical size.

[0233] During signal acquisition, the signals received by each antenna element are amplified with low noise and converted from analog to digital. The sampling rate is 50MHz and the quantization accuracy is 16bit to ensure sufficient signal dynamic range. The front-end circuit integrates a temperature-compensated crystal oscillator (TCXO) and an automatic gain control (AGC) circuit to ensure stable operation within a temperature range of -40℃ to +85℃.

[0234] The data acquisition subsystem also integrates auxiliary sensors, including a high-precision barometer (accuracy ±0.1hPa), a temperature and humidity sensor (temperature accuracy ±0.1℃, humidity accuracy ±2%), and a three-axis gyroscope (accuracy better than 0.01° / s), to acquire local meteorological parameters and platform attitude information, providing input data for atmospheric error modeling.

[0235] In addition, the system is equipped with a high-stability atomic clock or rubidium clock as a time reference, with a frequency stability better than 5×10^-11 / day and a phase noise better than -130dBc / Hz@10Hz, effectively reducing the contribution of clock error to measurement error.

[0236] For applications in extreme environments, the system is also designed with environmental adaptation capabilities, including:

[0237] 1. Temperature Adaptive Function: Automatically adjusts the local oscillator frequency to maintain frequency stability when the temperature changes.

[0238] 2. Electromagnetic Adaptive: Automatically activates band-stop filtering when strong interference is detected to protect the receiving front end.

[0239] 3. Power Adaptive: Automatically adjusts transmission power based on communication distance and quality to save energy.

[0240] 4. Antenna Adaptive: Automatically selects the optimal antenna combination based on satellite visibility to optimize reception performance.

[0241] The overall technical specifications of the system are shown in Table 2:

[0242] ;

[0243] Signal processing subsystem:

[0244] This subsystem implements multi-scale wavelet transform atmospheric error modeling and separation, employs a software-defined radio (SDR) architecture, and collaboratively processes received signals through a field-programmable gate array (FPGA) and a general-purpose processor (GPU). The system utilizes a heterogeneous computing architecture, allocating tasks with different characteristics to the most suitable processing platform to improve overall computational efficiency.

[0245] The main hardware configuration of the signal processing platform includes: FPGA, GPU, dedicated DSP, and high-speed storage device.

[0246] The system signal processing flow is as follows:

[0247] (1) Multi-channel signal synchronization: The maximum likelihood estimation method is used to achieve time synchronization of signals received by multiple antennas, with a synchronization accuracy better than 5ns. The synchronization process is as follows:

[0248] ;

[0249] in, This represents the estimated time deviation. and Representing the first The and the first The signal received by each antenna.

[0250] In the specific implementation, the sliding correlation method is used to calculate the optimal delay:

[0251] ;

[0252] in, represent and The cross-correlation function, Represents the number of sampling points. This represents the maximum search latency.

[0253] To improve computational efficiency, the FFT method is used in the actual implementation to calculate the cross-correlation:

[0254] ;

[0255] in, and These represent the Fourier transform and the inverse Fourier transform, respectively. Represents complex conjugation.

[0256] (2) Channel estimation: Channel estimation is performed based on the pilot sequence, using the minimum mean square error (MMSE) criterion. The channel estimation formula is:

[0257] ;

[0258] in, Represents the estimated channel matrix. The cross-correlation matrix representing the received and transmitted signals. The autocorrelation matrix represents the transmitted signal.

[0259] To reduce computational complexity and improve robustness, the regularized MMSE method is used in the actual implementation:

[0260] ;

[0261] in, This represents the regularization parameter, which is typically set to the estimated value of the noise power. Represents the identity matrix.

[0262] (3) Space-time decoding: The maximum likelihood (ML) criterion is used for space-time code decoding. The decoding process can be represented as:

[0263] ;

[0264] in, This represents the estimated transmitted signal matrix. This represents the Frobenius norm.

[0265] For Alamouti encoding, there exists a closed-form solution:

[0266] ;

[0267] ;

[0268] in, and The symbol representing the estimated signal. and Represents the channel coefficient. and This represents the received signal.

[0269] (4) Wavelet decomposition: Multi-scale wavelet decomposition is performed on the decoded signal to realize the Discrete Wavelet Transform (DWT) described in the invention. The mathematical expression for wavelet decomposition is:

[0270] ;

[0271] in, Representative signal wavelet coefficients, The scale parameter represents the degree of stretching or compression of the wavelet. The translation parameter represents the time position of the wavelet. This represents the complex conjugate of the wavelet basis functions.

[0272] The fast algorithm for discrete wavelet transform is based on Mallat's multiresolution analysis, which uses a set of high-pass and low-pass filters:

[0273] ;

[0274] ;

[0275] in, Representing the Layer An approximation coefficient, Representing the Layer A detailed coefficient, and These represent the wavelet low-pass and high-pass filter coefficients, respectively, satisfying the orthogonality condition:

[0276] ;

[0277] ;

[0278] ;

[0279] in, Represents the Kronecker delta function, when The value is 1 if it is active, and 0 otherwise.

[0280] In implementation, a fast boosting scheme is adopted to improve computational efficiency:

[0281] ;

[0282] ;

[0283] ;

[0284] in, Represents the input signal. and These represent even-indexed and odd-indexed samples, respectively. and These represent the prediction and update operators, respectively. and These represent the approximation and detail coefficients, respectively.

[0285] This invention employs the Daubechies wavelet family (db4-db20), the Symlet wavelet family (sym8-sym20), and the Coiflet wavelet family (coif3-coif5) as the main analytical tools, adaptively selecting the optimal wavelet basis based on different atmospheric error characteristics. The evaluation metrics for wavelet basis selection include:

[0286] ;

[0287] ;

[0288] ;

[0289] in, Represents the energy concentration of wavelet coefficients. Represents the sparsity of wavelet coefficients. The information entropy represents the wavelet coefficients. This represents the normalized energy distribution.

[0290] To further improve decomposition efficiency, an adaptive wavelet packet decomposition strategy was designed for different signal characteristics:

[0291] ;

[0292] ;

[0293] in, Representative node energy, Representative node The Wavelet coefficients.

[0294] This invention proposes an atmospheric error separation algorithm based on wavelet transform to separate satellite signal delay. Decomposed into:

[0295] ;

[0296] in, Representing the Layer wavelet detail coefficients correspond to the high-frequency components of atmospheric disturbances. Representing the Layer wavelet approximation coefficients correspond to the low-frequency components of atmospheric disturbances. This represents the number of decomposition levels (5-7 levels).

[0297] Based on the physical meaning of wavelet coefficients at different scales, an error separation model for the ionosphere and troposphere is established:

[0298] ;

[0299] ;

[0300] in, The boundary scale representing the error between the ionosphere and the troposphere (2-3). This represents the proportion of ionospheric error in low-frequency components (0.6-0.8, dynamically adjusted according to season and geographical location).

[0301] To improve the accuracy of wavelet decomposition, this invention employs translation-invariant wavelet transform (TIWT) technology, which eliminates the displacement variability in traditional wavelet transform, ensuring that the decomposition results are unaffected by the signal's temporal location.

[0302] ;

[0303] ;

[0304] in, and Representing the Layer dilation filter coefficients:

[0305] ;

[0306] ;

[0307] To further improve the accuracy of atmospheric error separation, this invention proposes a fine-grained decomposition method based on wavelet packet transform (WPT), which subdivides the signal simultaneously in both high-frequency and low-frequency components, forming a complete binary tree structure. The WPT coefficients are:

[0308] ;

[0309] ;

[0310] in, Representing the Layer Wavelet packet coefficients of each node, Represents the node position parameter.

[0311] Based on the WPT coefficients, a more refined atmospheric error separation model is constructed:

[0312] ;

[0313] ;

[0314] in, and These represent the wavelet packet node sets corresponding to ionospheric and tropospheric errors, respectively, and are trained from historical data using machine learning methods.

[0315] The performance comparison of different wavelet basis functions is shown in Table 3.

[0316] Table 3:

[0317] ;

[0318] Depending on the specific application requirements and computational resources, a suitable wavelet basis function can be selected. Generally, the db4 wavelet basis function achieves a good balance between computational efficiency and correction accuracy.

[0319] The software architecture of the signal processing subsystem adopts a layered design, including:

[0320] 1. Hardware Abstraction Layer: Shields the underlying hardware differences and provides a unified interface;

[0321] 2. Data Management Layer: Responsible for data acquisition, storage, distribution, and synchronization;

[0322] 3. Core Algorithm Layer: Implements the core signal processing algorithm;

[0323] 4. Application Interface Layer: Provides standardized interfaces to upper-layer applications.

[0324] The system adopts a dataflow architecture and pipelined processing mode, with each processing unit working in parallel, which improves the system throughput. The overall performance indicators of the signal processing subsystem are shown in Table 4.

[0325] Table 4:

[0326] ;

[0327] Atmospheric error correction subsystem:

[0328] This subsystem is the core of this invention, responsible for realizing multi-frequency collaborative observation and error correction of the BeiDou-3 system. The processing flow is as follows:

[0329] (1) Ionospheric delay estimation: Using BeiDou-3 multi-frequency signals, a free combination of ionospheric signals is constructed to estimate the total electron content (TEC) distribution in the ionosphere. The relationship between ionospheric delay and TEC is as follows:

[0330] ;

[0331] in, The representative frequency is The delay (in meters) of the signal passing through the ionosphere. Represents the total electron content of the ionosphere (unit: TECU). ), Represents satellite elevation angle Related mapping functions:

[0332] ;

[0333] in, Represents the Earth's radius (approximately 6371 km). This represents the height of the ionosphere with the highest electron density (approximately 350 km).

[0334] The oblique TEC value can be estimated using dual-frequency observations:

[0335] ;

[0336] in, Represents the total content of oblique electrons. and Representing two frequencies, and This represents the corresponding pseudorange observation.

[0337] Considering the high accuracy of carrier phase observations, the actual TEC estimation uses phase smoothing code technology:

[0338] ;

[0339] in, Representing the Smooth pseudo-distance of each epoch, Represents the original pseudorange. This represents the carrier phase observation value.

[0340] This invention introduces a wavelet transform-based TEC estimation method, representing TEC as a two-dimensional wavelet expansion in the vertical and horizontal directions:

[0341] ;

[0342] in, Represents wavelet detail coefficients, Represents two-dimensional wavelet basis functions. Represents the wavelet approximation coefficients. Represents a two-dimensional scaling function. Represents the number of decomposition levels. and Representing the first Layer in and The number of wavelet coefficients in the direction.

[0343] To address the sparsity of observation data, a regularization constraint is introduced:

[0344] ;

[0345] in, This represents the regularization parameter, which is determined through cross-validation.

[0346] The ionospheric TEC distribution represented by the spherical harmonic function model is as follows:

[0347] ;

[0348] in, Representing latitude and longitude The total electron content at that location, Represents the normalized associated Legendre function, and Represents the spherical harmonic coefficient. This represents the maximum order (usually 8-12).

[0349] (2) Tropospheric delay estimation: The tropospheric delay includes two parts: dry delay and wet delay. The total delay can be expressed as:

[0350] ;

[0351] For the dry delay portion, an improved Saastamoinen model is used:

[0352] ;

[0353] in, Represents dry delay (unit: meters). Represents atmospheric pressure at the Earth's surface (unit: hPa). Represents the latitude of the observation point. Represents the altitude of the observation point (unit: km). Represents the satellite's elevation angle (unit: radians).

[0354] For the wet delay component, a neural network-based prediction model is used:

[0355] ;

[0356] in, Represents wet delay (unit: meters). This represents a neural network model, with input parameters including temperature. (Unit: °C), Relative Humidity (Percentage), air pressure (Unit: hPa), Satellite elevation angle (Unit: radians) and dates of the year (1-366).

[0357] (3) Multi-frequency collaborative observation of BeiDou-3: This invention fully utilizes the multi-frequency signals provided by the BeiDou-3 satellite system, including B1I (1561.098MHz), B3I (1268.52MHz), B1C (1575.42MHz), B2a (1176.45MHz), and B2b (1207.14MHz), to construct a multi-frequency observation equation:

[0358] ;

[0359] ;

[0360] in, Representative receiver Receive satellite of Frequency pseudorange observations This represents the corresponding carrier phase observation value. Represents geometric distance. Represents the speed of light. and These represent the receiver clock bias and the satellite clock bias, respectively. represent The wavelength of the frequency point Represents the overall ambiguity throughout the week. and These represent pseudorange and carrier phase observation noise, respectively.

[0361] Geometric distance This can be further expressed as:

[0362] ;

[0363] in, Representative satellite coordinates Representative receiver The coordinates.

[0364] The ionosphere-free combination of the multi-frequency observation equations can be expressed as:

[0365] ;

[0366] ;

[0367] in, and Ionospherically-free combined observations, representing pseudorange and carrier phase respectively. and This represents two different frequency points.

[0368] The ionosphere-free combination eliminates the first-order ionospheric delay effect, but higher-order ionospheric effects still exist, and their influence can be expressed as follows:

[0369] ;

[0370] in, Represents the second-order ionospheric delay effect (unit: meters). Represents the intensity of the Earth's magnetic field (unit: Tesla). It represents the angle between the direction of signal propagation and the direction of the Earth's magnetic field.

[0371] To achieve high-precision positioning, this invention employs Precise Point Positioning (PPP) technology, utilizing precise ephemeris and clock bias products instead of broadcast ephemeris. The PPP observation equation is:

[0372] ;

[0373] ;

[0374] in, Represents precision satellite clock bias. Represents the wet delay mapping function, Represents zenith wet delay, This represents the carrier phase ambiguity term.

[0375] Considering that the BeiDou system includes satellites of different orbital types (GEO, IGSO, MEO), this invention proposes a differentiated weighting strategy:

[0376] ;

[0377] in, Representative satellite The weight, Represents the satellite's elevation angle. Representative satellite Carrier-to-noise ratio, This represents the reference carrier-to-noise ratio (typically 45 dB-Hz).

[0378] (4) Error Correction Fusion: The ionospheric and tropospheric delay estimation results are fused and corrected with the multipath signal received by MIMO. The fusion process uses the weighted least squares method:

[0379] ;

[0380] in, This represents the corrected delay estimate. Represents the design matrix. Represents the weight matrix. This represents the initial delay estimate.

[0381] weight matrix Dynamically adjusted based on satellite elevation angle, signal strength, and receiver noise level:

[0382] ;

[0383] in, , Representing the The elevation angle of the satellite, Represents the carrier-to-noise ratio. This represents the standard deviation of the observed noise.

[0384] To improve the calibration accuracy, this invention further introduces a residual calibration model to capture systematic errors and model defects:

[0385] ;

[0386] in, Represents residual delay. This represents the spatiotemporal bicubic spline interpolation function.

[0387] The final calibration model is:

[0388] ;

[0389] Under extreme atmospheric conditions, the system automatically switches to robust mode, using anomaly-resistant M-estimation instead of the conventional least squares method:

[0390] ;

[0391] in, Represents residuals, Representing the robust loss function, either the Huber function or the Tukey function can be used:

[0392] ;

[0393] ;

[0394] in, The threshold parameter is typically set to 1.345 times (Huber function) or 4.685 times (Tukey function) the standard deviation of the residuals.

[0395] The performance evaluation results of the atmospheric error correction system are shown in Table 5.

[0396] Table 5:

[0397] ;

[0398] Early warning and control subsystem:

[0399] This subsystem implements a real-time early warning and control system architecture, constructing a three-layer early warning and control system architecture:

[0400] (1) Data acquisition layer: It consists of MIMO antenna array, Beidou receiver and auxiliary sensors, and is responsible for receiving and preprocessing satellite signals at multiple frequencies;

[0401] (2) Data processing layer: It consists of wavelet transform module, atmospheric error separation module, multi-source data fusion module and signal enhancement module, and is responsible for realizing time-frequency analysis and error correction of the signal;

[0402] (3) Application Service Layer: It consists of a situation awareness module, a risk assessment module, a decision control module and an information release module, and is responsible for realizing air defense and disaster prevention early warning and control functions.

[0403] The functionality of the data acquisition layer can be described as a data acquisition function:

[0404] ;

[0405] in, represent The collection of data at any given time. Represents satellite signal observations. Represents the status of the communication link. Represents platform parameters (position, attitude, etc.).

[0406] The functionality of the data processing layer can be described as data processing functions:

[0407] ;

[0408] in, represent Real-time information products Represents historical data. This represents the parameters of the processing model.

[0409] The functionality of the application service layer can be described as a decision control function:

[0410] ;

[0411] in, represent Momentary action decisions Represents information products, Represents the risk assessment results. This represents the objective function.

[0412] The system's decision-making and control process can be described as a multi-objective optimization problem:

[0413] ;

[0414] The constraints are satisfied:

[0415] ;

[0416] ;

[0417] ;

[0418] in, Represents the objective function vector, containing One optimization objective, Represents a vector of decision variables. and These represent inequality and equality constraint functions, respectively. Represents the decision-making space.

[0419] Specific optimization objectives include:

[0420] ;

[0421] ;

[0422] ;

[0423] ;

[0424] in, , , and These represent the number of samples that are false positives, true negatives, false negatives, and true positives, respectively. Represents system response time. This represents an indicator of resource consumption.

[0425] This invention introduces an adaptive weight matrix into decision control. :

[0426] ;

[0427] in, Representing the The observation station for the first The influence weight of each observation station satisfies .

[0428] The weights are updated using reinforcement learning, dynamically adjusted based on historical data.

[0429] ;

[0430] in, Representing the The weight values ​​for the next iteration. This represents the learning rate (typically 0.01-0.1). Representing the The reward signal for the next iteration has a value range of [-1, 1].

[0431] Risk assessment uses a Bayesian network model to describe the conditional dependencies between different risk factors:

[0432] ;

[0433] in, Representing the One risk event Representing the One evidence variable, represent The set of parent nodes.

[0434] To achieve dynamic risk assessment, a dynamic Bayesian network (DBN) is introduced:

[0435] ;

[0436] in, and Represent and The risk status at all times, represent Observational evidence of the moment.

[0437] The main functions of the early warning and control subsystem include:

[0438] (1) Space Threat Assessment: Based on high-precision positioning information and combined with target motion characteristics, potential threats are assessed in real time. Threat assessment indicators include:

[0439] ;

[0440] in, This represents the collision risk index, with values ​​ranging from [0, 1]. Represents the minimum approach distance. This represents the distance uncertainty.

[0441] ;

[0442] in, Represents the speed risk index, Represents the target speed. Represents the speed threshold. Represents the maximum reference speed. This represents the Heaviside step function.

[0443] ;

[0444] in, Represents the acceleration risk index, Represents the target acceleration. Represents the acceleration threshold. This represents the maximum reference acceleration.

[0445] The comprehensive risk index is calculated through weighted fusion:

[0446] ;

[0447] in, , , and Each represents the weight of the respective risk factor, satisfying the following conditions: , The risk of abnormal trajectory is calculated by comparing it with the historical trajectory database.

[0448] The risk level classification criteria are shown in Table 6.

[0449] Table 6:

[0450] ;

[0451] (2) Meteorological Disaster Assessment: Based on the atmospheric anomaly features extracted by wavelet transform and combined with historical data, a meteorological disaster assessment model is constructed. The atmospheric anomaly feature vector is defined as:

[0452] ;

[0453] in, Represents the ionospheric TEC anomaly index, Represents the TEC gradient. Represents the tropospheric water vapor anomaly index. Represents the water vapor gradient. Represents wind speed. This represents the probability of rainfall.

[0454] The formula for calculating the ionospheric TEC anomaly index is:

[0455] ;

[0456] in, Represents the observed TEC value, Represents the TEC value of the reference model. This represents the standard deviation of TEC.

[0457] The formula for calculating the TEC gradient is:

[0458] ;

[0459] The formula for calculating the tropospheric water vapor anomaly index is:

[0460] ;

[0461] in, Represents the observable precipitation. The representative reference model shows the amount of precipitation. This represents the standard deviation of precipitable water.

[0462] Meteorological disaster risk levels are determined using a fuzzy inference system:

[0463] ;

[0464] in, Represents the risk level (range 0-1). Representing the The weight of each rule, Representing the Membership function of a rule.

[0465] The correspondence between meteorological disaster types and characteristic parameters is shown in Table 7.

[0466] Table 7:

[0467] ;

[0468] (3) Warning Information Issuance: Based on the threat assessment results, warning information is issued through the BeiDou short message communication function. The warning information encoding adopts an efficient variable-length encoding scheme, which improves the encoding efficiency by more than 30%. The structure of the warning message is shown in Table 8.

[0469] Table 8:

[0470] ;

[0471] This invention optimizes BeiDou short message communication, including:

[0472] Low-density parity-check codes are used to improve coding efficiency, with a code rate of 0.8 and a minimum Hamming distance of 12.

[0473] Adaptive modulation technology is used to dynamically adjust the modulation scheme based on channel quality. The mapping relationship between channel quality and modulation scheme satisfies the formula:

[0474] ;

[0475] in, This represents the optimal modulation scheme. Represents bit rate, Represents the bit error rate;

[0476] Introducing a time-slot random multiple access mechanism reduces channel collisions;

[0477] Implement a message hierarchical transmission strategy, allocating different transmission priorities and retransmission counts based on the warning level.

[0478] The sending of warning messages uses a priority queue management system, with higher-priority messages being sent first:

[0479] ;

[0480] in, Represents the message content, Represents priority (1-10, the larger the value, the higher the priority). This indicates the time the message was generated.

[0481] The queue sorting rules are as follows:

[0482] ;

[0483] Message transmission uses a reliable transport protocol, including acknowledgment and retransmission mechanisms:

[0484] ;

[0485] in, This represents the maximum number of retransmissions. This represents the highest priority value (usually 10).

[0486] (4) Control Command Execution: The system automatically generates corresponding control commands for different types of threats, including air defense system activation, personnel evacuation, and resource allocation. Control command generation is based on a pre-set response strategy library and real-time status assessment results. Control command execution employs a hierarchical structure:

[0487] a. Strategy Layer: Determine the overall response strategy and resource allocation.

[0488] ;

[0489] in, Represents the response strategy. Represents the risk assessment results. Represents the current system state. It represents the historical record of decision-making.

[0490] b. Tactical Level: Determine specific execution plans and task allocation.

[0491] ;

[0492] in, Representing tactical plans, Represents the response strategy. Represents available resources. It represents the state of the environment.

[0493] c. Execution layer: Executes specific control actions and makes real-time adjustments.

[0494] ;

[0495] in, Represents control action, Representing tactical plans, This represents the feedback status. This represents real-time intervention.

[0496] The system supports comprehensive early warning functions for multiple disasters, including meteorological disasters, geological disasters, marine disasters, space disasters, and man-made disasters; it uses a Bayesian network model to integrate multi-source disaster information and construct a disaster chain reaction prediction model.

[0497] ;

[0498] in, Representing the Types of disasters represent The set of parent nodes, Represents environmental conditions; the system supports multi-hazard integrated risk assessment, integrating risk indices from different hazard types:

[0499] ;

[0500] in, Represents a comprehensive risk index. Representing the Risk index of various disasters Represents the weighting factor, satisfying .

[0501] To improve the system's ability to identify atmospheric anomaly patterns, we also designed a hybrid model based on convolutional neural networks (CNN) and long short-term memory networks (LSTM) to identify anomaly patterns in the spatiotemporal distribution of atmospheric parameters.

[0502] The model is trained using a contrastive learning method, with the loss function being:

[0503] ;

[0504] in, Representative input sample Feature representation, Represents positive samples (of the same category). Represents negative samples (different categories). Represents the cosine similarity function. This represents the temperature parameter (usually set to 0.07). This represents the set of negative samples.

[0505] To improve the model's performance on sparse outlier samples, a focal loss was used as a supplement:

[0506] ;

[0507] in, Represents the predicted probability. Represents category weight, This represents the focusing parameter (usually set to 2).

[0508] The final loss function is a weighted sum of the contrast loss and the focus loss:

[0509] ;

[0510] in, and The weighting coefficients are determined through cross-validation.

[0511] The model training data includes historical atmospheric parameter records and corresponding anomaly labels. Data augmentation techniques include random pruning, time window sliding, and Gaussian noise perturbation.

[0512] The software architecture of this system adopts a microservice architecture, which mainly includes the following service modules:

[0513] (1) Signal Acquisition Service: Responsible for the acquisition and preprocessing of MIMO antenna array signals; implements adaptive gain control, frequency synchronization and preliminary filtering, and supports a maximum sampling rate of 50MHz and 16-bit quantization accuracy. Adopts a streaming processing architecture with a minimum processing latency of less than 1ms.

[0514] (2) Wavelet analysis service: realizes multi-scale wavelet decomposition and reconstruction of signals; supports 18 wavelet basis functions, with a maximum decomposition level of 12 levels, adopts multi-threaded parallel processing to improve efficiency, and the processing delay is less than 5ms.

[0515] (3) Atmospheric error modeling service: Construct ionospheric and tropospheric delay models; integrate multiple atmospheric models (including IRI, NeQuick, GPT2, etc.), support real-time parameter inversion and historical data training, and the model update frequency can reach up to 1Hz.

[0516] (4) Positioning service: Calculates high-precision positioning results based on the corrected observations; supports three modes: single-point positioning, differential positioning and precise single-point positioning (PPP), with a position update rate of up to 100Hz, and supports integer ambiguity fast fixation (FAST) technology.

[0517] (5) Early warning assessment service: Analyzes potential threats and generates risk assessment reports; integrates multiple early warning models and prediction algorithms, supports multi-source data fusion and historical data comparison, and the assessment update frequency is 1-10Hz.

[0518] (6) Control and scheduling service: Generate control commands and supervise the execution process; adopt a hierarchical control architecture, support three control modes: manual, semi-automatic and fully automatic, and monitor and provide feedback on command execution status in real time.

[0519] (7) Data storage service: manages observation data and system status information; adopts a hybrid storage strategy, with hot data stored in an in-memory database (Redis), warm data stored in a NoSQL database (MongoDB), and cold data stored in a distributed file system (HDFS).

[0520] (8) User interface service: Provides visual interfaces for Web and mobile terminals; developed based on React and Flutter frameworks, supports responsive layout and multi-terminal adaptation, with an interface refresh rate of no less than 30Hz, and supports 2D / 3D visualization and data drill-down analysis.

[0521] The software stack configuration is shown in Table 9.

[0522] Table 9:

[0523] ;

[0524] The system performance of this invention was tested in different application scenarios, and the main performance indicators are shown in Table 10.

[0525] Table 10:

[0526] ;

[0527] Among them, standard conditions refer to clear skies with no significant atmospheric disturbances; slight disturbances refer to light rain or slight ionospheric fluctuations; moderate disturbances refer to thunderstorms or moderate ionospheric storms; and severe disturbances refer to typhoons, heavy rain, or severe ionospheric storms.

[0528] The positioning accuracy was monitored over a long period of time, and the results are shown in Table 11.

[0529] Table 11:

[0530] ;

[0531] The system outperforms existing technologies under various conditions, especially under strong disturbance conditions, where the positioning accuracy can still be maintained within 1.2 meters and the system response time does not exceed 100 milliseconds, fully meeting the requirements for real-time early warning and control for air defense and disaster prevention.

[0532] Comparative test results show that, compared with traditional single-model correction methods, especially under extreme atmospheric conditions, the performance advantage is more significant, as shown in Table 12.

[0533] Table 12:

[0534] ;

[0535] This invention solves the technical problems of insufficient accuracy and weak real-time performance of traditional atmospheric error correction methods in complex environments, and provides reliable technical support for air defense and disaster prevention early warning control systems. It has strong application prospects and social and economic value.

Claims

1. A real-time control system for air defense and disaster prevention early warning based on BeiDou-3 communication, characterized in that, Includes the following subsystems: The data acquisition subsystem includes a distributed MIMO antenna array and auxiliary sensors. The distributed MIMO antenna array is used to simultaneously receive multi-frequency signals from BeiDou-3 satellites B1I / B3I / B1C / B2a / B2b. The signal processing subsystem is used to realize multi-channel signal synchronization, channel estimation, space-time decoding, and wavelet decomposition, and to delay satellite signals. Decomposed into: ;in, Representing the Layer wavelet detail coefficients correspond to the high-frequency components of atmospheric disturbances. Representing the Layer wavelet approximation coefficients correspond to the low-frequency components of atmospheric disturbances. Represents the number of decomposition levels; The atmospheric error correction subsystem is used to realize multi-frequency collaborative observation and error correction of BeiDou-3. The processing flow includes ionospheric delay estimation, tropospheric delay estimation, multi-frequency collaborative observation of BeiDou-3, and error correction fusion. Error correction fusion fuses the ionospheric and tropospheric delay estimation results with the multipath signals received by MIMO for correction. The early warning and control subsystem is used to realize space threat assessment, meteorological disaster assessment, early warning information dissemination and control command execution based on the corrected high-precision positioning information and combined with the Beidou short message communication function.

2. The system according to claim 1, characterized in that, The spatial layout of the MIMO antenna array adopts a non-uniform distribution, with an antenna spacing matrix. Defined as: ； in, This represents the distance between the antenna element in the Mth row and Nth column and the center of the array; an adaptive beamforming algorithm is used, and the beamforming weight vector calculation adopts the minimum variance distortionless response criterion.

3. The system according to claim 1, characterized in that, The wavelet decomposition employs a fast algorithm based on discrete wavelet transform, expressed as follows: ； ； in, Representing the Layer An approximation coefficient, Representing the Layer A detailed coefficient, and These represent the wavelet low-pass and high-pass filter coefficients, respectively. The system adaptively selects the optimal wavelet basis from the Daubechies, Symlet, and Coiflet wavelet families based on different atmospheric error characteristics, and eliminates the displacement variability in the traditional wavelet transform through translation-invariant wavelet transform technology.

4. The system according to claim 1, characterized in that, The multi-frequency observation equations for BeiDou-3 multi-frequency point collaborative observation are as follows: ； ； in, Representative receiver Receive satellite of Frequency pseudorange observations This represents the corresponding carrier phase observation value. Represents geometric distance. Represents the speed of light. and These represent the receiver clock bias and the satellite clock bias, respectively. represent The wavelength of the frequency point Represents integer ambiguity; the system employs precise point positioning technology and a differentiated weighting strategy to process satellite data of different orbit types. and These represent pseudorange and carrier phase observation noise, respectively.

5. The system according to claim 1, characterized in that, The ionospheric delay estimation uses the following formula: ； in, The representative frequency is The delay in signal transmission through the ionosphere Represents the total electron content of the ionosphere. Represents satellite elevation angle The relevant mapping function; the system introduces a TEC estimation method based on wavelet transform, representing TEC as a two-dimensional wavelet expansion: ； in, Represents wavelet detail coefficients, Represents two-dimensional wavelet basis functions. Represents the wavelet approximation coefficients. Represents a two-dimensional scaling function. Represents the number of decomposition levels. and Representing the first Layer in and The number of wavelet coefficients in the direction.

6. The system according to claim 1, characterized in that, The tropospheric delay estimation includes two parts: dry delay and wet delay. ； The dry delay part uses the improved Saastamoinen model: ； in, Represents dry delay, Represents atmospheric pressure at the Earth's surface. Represents the latitude of the observation point. Represents the height of the observation point. Represents satellite elevation angle; the wet delay component uses a neural network-based prediction model. ； in, This represents a neural network model, with input parameters including temperature. relative humidity air pressure Satellite elevation angle and the dates of the year .

7. The system according to claim 1, characterized in that, The early warning and control subsystem is constructed with a three-layer architecture: Data acquisition layer: consists of MIMO antenna array, BeiDou receiver and auxiliary sensors; The data processing layer consists of a wavelet transform module, an atmospheric error separation module, a multi-source data fusion module, and a signal enhancement module. Application service layer: consists of situation awareness module, risk assessment module, decision control module and information dissemination module.

Images