Radar dynamic anti-jamming method and system based on jammer positioning

Abstract

The application relates to the field of information technology and discloses a radar dynamic anti-interference method and system based on interference source positioning. The method comprises the following steps: acquiring current signal data and a historical signal sequence, and determining an initial deviation value; extracting a historical deviation sequence according to the initial deviation value to obtain a deviation change trend vector; calculating a trend slope and a fluctuation amplitude, and determining an adjustment trigger signal; extracting a characteristic interaction influence according to the adjustment trigger signal to determine a weight coefficient update increment; iteratively optimizing in combination with a convergence speed parameter to obtain an optimized convergence speed value; adjusting a deviation compensation model parameter according to the convergence speed value, and outputting a deviation compensation result when residual error is lower than a preset threshold value; and performing positioning calculation according to the deviation compensation result to determine a corrected positioning coordinate. The application dynamically optimizes the model through technologies such as a sliding window and gradient descent, solves the problem of positioning deviation caused by complex environmental signal interference, and improves positioning accuracy and response speed.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of information technology, in particular to a radar dynamic anti-interference method and system based on interference source positioning. BACKGROUND

[0002] As a core perception device in navigation, national defense, transportation and other fields, the positioning accuracy of radar directly determines the system performance. However, in complex environments, multipath effect, obstacle shielding, non-line-of-sight propagation and dynamic interference source switching can easily cause unpredictable fluctuations in the strength, frequency and modulation mode of the received signal, leading to positioning deviation and seriously affecting the reliable operation of radar in complex scenarios. Therefore, an efficient dynamic anti-interference technology is needed to solve this core problem.

[0003] Existing radar anti-interference technologies have significant defects when dealing with complex dynamic interference. Firstly, most technologies rely on a single static bias compensation model, which cannot adapt to the spatiotemporal dynamic changes of signal characteristics. For example, in urban canyons, high-rise building reflections can cause signal mutations, or in indoor scenarios, wall attenuation can cause changes in bias patterns. Static models have difficulty in adjusting compensation strategies in real time, resulting in low positioning accuracy. Secondly, the feature interaction processing capability is insufficient. The interaction between signal strength, frequency and modulation mode has a significant impact on bias, but existing technologies lack accurate analysis of such feature interactions, and the weight coefficient update is lagging, which cannot timely reflect the dynamic contribution of each feature to the bias, exacerbating the compensation residual error. Thirdly, the balance between model optimization and real-time performance is difficult. Although some technologies use sliding windows or iterative optimization algorithms, the window size is fixed and cannot match the signal change frequency. Moreover, when dealing with high-dimensional features, algorithms such as gradient descent can easily cause slow convergence or high computational complexity, resulting in response delays when the signal mutates, making it difficult to meet the dual requirements of real-time performance and low residual error. In addition, existing technologies do not establish a cooperative mechanism between interference source positioning and anti-interference, and only compensate for bias without associating with dynamic interference sources, resulting in insufficient anti-interference robustness.

[0004] To address the above-mentioned defects, the present application dynamically adapts to interference changes through the method of "signal data collection-bias trend analysis-triggering dynamic adjustment-weight iterative optimization-model parameter adaptation-positioning coordinate correction", improving the radar anti-interference capability and positioning accuracy. SUMMARY

[0005] The present application relates to the field of information technology and discloses a radar dynamic anti-interference method and system based on interference source positioning. By dynamically optimizing the model using sliding windows and gradient descent techniques, the positioning deviation caused by signal interference in complex environments is addressed, and the positioning accuracy and response speed are improved.

[0006] In a first aspect, the present application provides a radar dynamic anti-interference method based on interference source positioning, which comprises:

[0007] Step S101, obtaining current signal data and historical signal sequence, and extracting signal strength, frequency and modulation mode characteristic parameters, and calculating an initial deviation value according to the signal strength, frequency and modulation mode characteristic parameters;

[0008] Step S102, extracting a historical deviation sequence from the initial deviation value based on a dynamic sliding window, and performing time series analysis on the historical deviation sequence to generate a deviation change trend vector;

[0009] Step S103, calculating a trend slope and a fluctuation amplitude using the deviation change trend vector, and determining an adjustment trigger signal according to a relationship between the trend slope and the fluctuation amplitude and a preset threshold size;

[0010] Step S104, extracting a related feature interaction from the deviation change trend vector according to the adjustment trigger signal, determining a correlation weight between features, and calculating a weight coefficient update increment according to the correlation weight;

[0011] Step S105, updating the convergence speed parameter according to the weight coefficient update increment, and optimizing the updated convergence speed parameter through a gradient descent iterative optimization algorithm to obtain an optimized convergence speed value;

[0012] Step S106, adjusting a deviation compensation model parameter according to the optimized convergence speed value, calculating a compensation residual of the adjusted deviation compensation model, and outputting a deviation compensation result if the compensation residual is lower than a preset residual threshold;

[0013] Step S107, adjusting an input parameter weight of a positioning algorithm according to the deviation compensation result, and performing weighted calculation on the adjusted input parameter through a weighted positioning algorithm to determine a corrected positioning coordinate.

[0014] Optionally, the step S101 comprises:

[0015] Obtaining current interference signal data and historical signal sequence comprises: collecting current interference signal data in real time through a radar signal receiving device, and calling historical signal sequence within a preset time period matched with a current radar working mode from a pre-established signal database;

[0016] Calculating an initial deviation value according to characteristic parameters of the current interference signal and the historical signal sequence comprises: respectively calculating strength deviation components, frequency deviation components and modulation mode deviation components of the interference signal and the historical signal;

[0017] Determining a weight coefficient of the deviation component according to a variance of the historical signal sequence, calculating an initial deviation value through weighted summation, and storing the initial deviation value in association with a calculation time stamp, a radar working mode and an interference scene label.

[0018] Optionally, the step S102 comprises:

[0019] The historical deviation sequence is extracted from the pre-stored deviation database by using a sliding window technique, and a window size of the sliding window technique is dynamically adjusted according to a signal change frequency;

[0020] The historical deviation sequence is subjected to time series analysis to generate a deviation change trend vector;

[0021] The deviation change trend vector is subjected to normalization processing to obtain a standardized deviation change trend vector.

[0022] Optionally, the step S103 comprises:

[0023] A time domain linear regression slope component in the deviation change trend vector is extracted, and the time domain linear regression slope component is determined as a trend slope in a time window;

[0024] A fluctuation amplitude is calculated by using a standard deviation algorithm on time domain and frequency domain feature components of the deviation change trend vector;

[0025] If the trend slope exceeds a preset slope threshold value and the fluctuation amplitude satisfies an abnormality determination condition, a mutation mode is marked and an adjustment trigger signal is generated.

[0026] Optionally, the step S104 comprises:

[0027] An interference scene label, a timestamp, a trend slope and a fluctuation amplitude are analyzed and extracted according to the adjustment trigger signal;

[0028] A scene data set in the historical trend database is matched according to the interference scene label, and feature interaction data is extracted from the scene data set according to the timestamp, the trend slope and the fluctuation amplitude;

[0029] A degree of correlation between features in the feature interaction data set is analyzed by using a Pearson correlation coefficient matrix, and correlation weights of three groups of feature interactions of signal strength-frequency, signal strength-modulation mode and frequency-modulation mode are determined;

[0030] A basic weight coefficient of a current deviation compensation model is obtained, and a weight coefficient update increment is calculated according to the basic weight coefficient, the correlation weights and the dynamic adjustment coefficient.

[0031] Optionally, the step S105 comprises:

[0032] A current convergence speed parameter is obtained from a parameter buffer area of a radar deviation compensation model, and the parameter buffer area contains a convergence speed parameter, an update timestamp and a corresponding interference scene label;

[0033] The update amount of the convergence speed parameter is calculated by weighting and summing the update increment of the weight coefficient with the intensity influence coefficient, frequency influence coefficient and modulation influence coefficient respectively. Then, the update amount of the convergence speed parameter is superimposed with the current convergence speed parameter to obtain the updated convergence speed parameter. The intensity influence coefficient, frequency influence coefficient and modulation influence coefficient are determined by statistical analysis of historical data.

[0034] The updated convergence speed parameters are iteratively optimized using the gradient descent method.

[0035] Optionally, step S106 includes:

[0036] Retrieve the initial parameters of the current deviation compensation model, including the deviation compensation coefficient and the compensation offset;

[0037] Based on the optimized convergence speed value, the deviation compensation coefficient and the compensation offset are adjusted so that the adjusted parameters are adapted to the convergence speed of the deviation compensation model.

[0038] The characteristic parameters of the current signal are input into the adjusted deviation compensation model. After the moving average filtering process, the real-time compensation value is output. Then, the ideal compensation value that matches the current interference scene and signal characteristics is retrieved. The absolute value error between the real-time compensation value and the ideal compensation value is the compensation residual.

[0039] If the compensation residual is lower than the residual threshold, the output includes the real-time compensation value after filtering, the new compensation coefficient, the new compensation offset, the timestamp, and the scene label. If the compensation residual is higher than the residual threshold, return to step S105 to iterate and optimize the convergence speed parameters again.

[0040] Optionally, step S107 includes:

[0041] The deviation compensation result is decomposed into values ​​corresponding to the original radar positioning coordinates. x Directional components and y Directional components;

[0042] According to the above x Directional components and the y The input parameters of the directional component adjustment positioning algorithm include the distance estimation value corresponding to the signal strength, the phase offset value corresponding to the frequency, and the positioning error correction value corresponding to the modulation method.

[0043] The weighted positioning method is adopted, which obtains the corrected coordinate components by multiplying each adjusted input parameter by its corresponding adjusted weight and then summing the product results.

[0044] Secondly, this application provides a radar dynamic anti-jamming system based on interference source localization, the system comprising:

[0045] The signal acquisition module is configured to acquire current signal data and a historical signal sequence, calculate and determine an initial deviation value, and ensure the real-time and accuracy of data acquisition.

[0046] The deviation analysis module is configured to extract a historical deviation sequence according to the initial deviation value, generate a deviation change trend vector through processing, and intuitively reflect the change characteristics of the deviation over time.

[0047] The trigger judgment module is configured to calculate a trend slope and a fluctuation amplitude using the deviation change trend vector, determine and generate an adjustment trigger signal if the slope exceeds a preset threshold, and start a subsequent compensation adjustment process.

[0048] The weight optimization module is configured to extract feature interaction according to the adjustment trigger signal, analyze feature correlation to determine a correlation weight, and calculate a weight coefficient update increment to provide a basis for model parameter updating.

[0049] The convergence optimization module is configured to acquire a current convergence speed parameter, update the parameter in combination with the weight coefficient update increment, optimize through an iterative optimization algorithm to obtain an optimized convergence speed value, and improve the response speed of the model.

[0050] The compensation output module is configured to adjust deviation compensation model parameters according to the optimized convergence speed value, calculate a compensation residual, and output a deviation compensation result if the residual is lower than a preset threshold, which is used for subsequent positioning correction.

[0051] The positioning correction module is configured to integrate the deviation compensation result into a positioning calculation process, adjust positioning algorithm input parameters, calculate and determine a corrected positioning coordinate through a positioning algorithm, and realize accurate positioning.

[0052] In a third aspect, the present application provides a radar dynamic anti-interference device based on interference source positioning, which comprises a memory and a processor, the memory stores a computer program capable of running on the processor, and the processor implements the radar dynamic anti-interference method based on interference source positioning when executing the computer program.

[0053] The present application provides a radar dynamic anti-interference method and system based on interference source positioning, which is suitable for anti-interference positioning scenarios of radar in complex environments, and can solve the problems of poor static model adaptability, insufficient feature interaction processing, unbalanced model optimization and real-time performance, and weak anti-interference robustness in the prior art. Compared with the prior art, the beneficial effects of the technical solution of the present application are at least as follows:

[0054] First, by obtaining the initial deviation value of the current signal data and the historical signal sequence, the historical deviation sequence is dynamically extracted and the deviation trend vector is generated by combining the sliding window technology, which breaks through the limitation of the traditional static deviation compensation model, can adaptively adjust the analysis dimension according to the signal change frequency, adapt to the spatiotemporal dynamic change of signal characteristics in complex environment, and avoid the problem of sudden drop of positioning accuracy when signal mutation or deviation mode changes.

[0055] Second, based on the adjustment trigger signal, the feature interaction among signal strength, frequency and modulation mode is extracted, the weight coefficient update increment is determined through feature correlation analysis, the deficiency of the prior art in feature interaction processing capability is made up, the weight coefficient can reflect the dynamic contribution of each feature to the deviation in real time, the compensation residual caused by weight update lag is reduced, and the precision of deviation compensation is improved.

[0056] Third, the optimized convergence speed value is obtained by combining the weight coefficient update increment and the convergence speed parameter, using gradient descent iterative optimization algorithm, which not only ensures the scientificity of model parameter adjustment, but also balances the calculation complexity and convergence efficiency through dynamic optimization, avoids the problem of slow convergence or response delay of traditional algorithm in processing high-dimensional features, and meets the dual needs of radar positioning real-time and low residual.

[0057] Fourth, the deviation compensation result is integrated into the positioning calculation process, the corrected positioning coordinates are determined through weighted positioning algorithm, the cooperative mechanism of interference source positioning and anti-interference is established, not simply compensating the deviation, but correlating the dynamic characteristics of interference source to optimize the positioning process, enhancing the anti-interference robustness of radar in complex scenes such as multi-interference source switching and multipath effect, and ensuring the stable and reliable operation of radar system. BRIEF DESCRIPTION OF DRAWINGS

[0058] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0059] Figure 1 Flow chart of a radar dynamic anti-interference method based on interference source positioning of the present application;

[0060] Figure 2 Process chart of a radar dynamic anti-interference processing process based on interference source positioning of the present application;

[0061] Figure 3 System structure schematic diagram of a radar dynamic anti-interference system based on interference source positioning of the present application;

[0062] Figure 4A radar dynamic anti-interference device structure diagram based on interference source positioning. DETAILED DESCRIPTION

[0063] The embodiment of the application provides a radar dynamic anti-interference method and system based on interference source positioning. The terms "first", "second", "third", "fourth" and the like (if any) in the specification and claims of the application and the above drawings are used to distinguish similar objects, and do not have to be used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged under appropriate circumstances, so that the embodiments described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "include" or "have" and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, system, product or device including a series of steps or units does not have to be limited to those steps or units clearly listed, but can include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0064] The application relates to the field of information technology and discloses a radar dynamic anti-interference method and system based on interference source positioning. The method comprises the following steps: acquiring current signal data and a historical signal sequence, determining an initial deviation value; extracting a historical deviation sequence according to the initial deviation value to obtain a deviation change trend vector; calculating a trend slope and a fluctuation amplitude, and determining an adjustment trigger signal if the trend slope exceeds a preset threshold; extracting a feature interaction according to the adjustment trigger signal to determine a weight coefficient update increment; iteratively optimizing in combination with a convergence speed parameter to obtain an optimized convergence speed value; adjusting a deviation compensation model parameter according to the convergence speed value, and outputting a deviation compensation result when residual error is lower than a preset threshold; and performing positioning calculation according to the deviation compensation result to determine a corrected positioning coordinate. The application dynamically optimizes a model through technologies such as a sliding window and gradient descent, solves the problem of positioning deviation caused by complex environmental signal interference, and improves positioning accuracy and response speed.

[0065] For the convenience of understanding, the specific process of the embodiment of the application is described below. Please refer to Figure 1 One embodiment of the radar dynamic anti-interference method based on interference source positioning in the embodiment of the application comprises the following steps:

[0066] In step S101, current signal data and a historical signal sequence are acquired, and signal strength, frequency and modulation mode characteristic parameters are extracted, and an initial deviation value is calculated according to the signal strength, frequency and modulation mode characteristic parameters.

[0067] In a specific embodiment, the process of performing step S101 can specifically comprise the following steps:

[0068] Obtaining current interference signal data and historical signal sequence, including: collecting current interference signal data in real time through a radar signal receiving device, and calling historical signal sequence within a preset time period matched with a current radar working mode from a pre-established signal database;

[0069] Calculating an initial deviation value according to a characteristic parameter of the current interference signal and the historical signal sequence, including: respectively calculating an intensity deviation component, a frequency deviation component and a modulation mode deviation component of the interference signal and the historical signal;

[0070] Determining a weight coefficient of the deviation component according to a variance of the historical signal sequence, calculating an initial deviation value through weighted summation, and storing the initial deviation value in association with a calculation time stamp, a radar working mode and an interference scene label.

[0071] Specifically, a signal receiving device (such as a high-sensitivity antenna module, adapted to a radar working frequency band such as X-band or S-band) carried by a radar system is started to sample interference signals in a current environment in real time, a sampling frequency needs to match a radar working frequency (such as 100MHz sampling rate commonly used for X-band radar), and current interference signal data collected contains time domain waveforms (such as continuous time sequence voltage amplitude data) and spectral characteristics (such as power distribution of signals at different frequency points), which can be directly used for subsequent feature extraction. Meanwhile, historical signal sequence matched with a current radar working mode (such as search mode or tracking mode) is called from a pre-established signal database, the signal database stores signal records under different interference scenes (such as electromagnetic interference, multipath reflection interference, dynamic interference source switching scene), each record contains a time stamp, a signal type label (such as “urban canyon multipath interference” and “airport electromagnetic interference”) and corresponding signal intensity, frequency and modulation mode parameters, and when calling, a time stamp filtering mechanism is used to select historical signal sequence within the last 30 minutes to construct a comparison benchmark, ensuring the timeliness of historical data.

[0072] Secondly, for signal strength, the received signal strength indication is calculated through the power detection module built-in the signal receiving device, which is quantified as dBm unit (the value range is usually -40dBm to -100dBm, such as the strength indication of the current interference signal is -68dBm); for frequency, the fast Fourier transform algorithm is applied to the time domain waveform of the current interference signal data, the time domain signal is converted into frequency domain signal, the main frequency component in the frequency domain spectrum is identified and extracted through the peak detection algorithm, the digital filtering technology is used in the extraction process to eliminate out-of-band noise, and the resolution of the frequency parameter is improved through the cubic spline interpolation algorithm; for modulation mode, the change law of signal amplitude and frequency with time is analyzed through the demodulation circuit of the radar system, and the modulation mode (such as ASK, FSK, QPSK or FM, AM) adopted by the interference signal is identified combined with the maximum likelihood estimation algorithm. Through the three types of feature parameters extracted, the problem of distortion of deviation evaluation caused by incomplete feature extraction and insufficient precision is solved.

[0073] Further, a multi-dimensional difference quantification method is used to calculate the deviation components of the three types of feature parameters: the signal strength deviation component is the algebraic difference between the current signal strength value and the average signal strength in the historical signal sequence under the same interference scenario, such as the current strength indication is -68dBm, and the average strength indication of the "urban multipath interference" scenario in the historical signal sequence is -72dBm, then the signal strength deviation component is 4dBm; the frequency deviation component is calculated by the Euclidean distance algorithm, that is, the difference between the current extracted main frequency and the frequency mean value of the historical signal sequence, such as the current main frequency is 10.03GHz, and the frequency mean value of the historical signal sequence is 10GHz, then the frequency deviation component is 0.03GHz; the modulation mode deviation component is assigned a qualitative value according to the pattern matching degree, if the current modulation mode is consistent with the modulation mode of the historical signal sequence, then the component is 0, if not, a preset qualitative deviation value is assigned. Then, the weight coefficients of each deviation component are determined according to the statistical variance of the historical signal sequence, if the variance of the signal strength in the historical data is large, then the weight of the signal strength deviation component is set to 0.5, the weight of the frequency deviation component is set to 0.3, and the weight of the modulation mode deviation component is set to 0.2, the initial deviation value is calculated by the weighted sum formula (initial deviation value = signal strength deviation component × strength weight + frequency deviation component × frequency weight + modulation mode deviation component × modulation weight).

[0074] Finally, the calculated initial bias value is stored in the local cache of the radar system or the associated bias database, and the storage format needs to include the calculation timestamp (such as "2028-08-21 15:48:55"), the current radar working mode (such as "tracking mode") and the interference scene label (such as "urban multipath interference"). During storage, a ring buffer is used to manage data, and the newly generated initial bias value automatically covers the expired data in the database that exceeds the preset storage time, maintaining storage efficiency. At the same time, index optimization technology (such as a composite index based on the interference scene label and the timestamp) is used in the database query link to accelerate the retrieval of the historical bias sequence by the sliding window technology in the subsequent steps, solving the problem of low efficiency in subsequent processing caused by scattered and disordered data storage.

[0075] Step S102, extracting a historical bias sequence from the initial bias value based on a dynamic sliding window, performing time series analysis on the historical bias sequence to generate a bias change trend vector.

[0076] In a specific embodiment, the process of performing step S102 can specifically include the following steps:

[0077] Applying a sliding window technique to extract a historical bias sequence from a pre-stored bias database, the window size of the sliding window technique being dynamically adjusted according to the signal change frequency;

[0078] Performing time series analysis on the historical bias sequence to generate a bias change trend vector;

[0079] Normalizing the bias change trend vector to obtain a standardized bias change trend vector.

[0080] Specifically, when extracting the historical bias sequence of the recent time period from the pre-stored bias database, the sliding window technique is used, and the current time corresponding to the initial bias value is taken as the reference, a time interval is set forward, and the sliding window is started. The initial size of the sliding window is set to 10 seconds of data points, and the number of data points in the window is dynamically adjusted according to the radar signal sampling rate. If the radar working sampling rate is 100MHz (i.e. 100x10 6If there are 1000 signal data points, then a 10-second window contains 1000 data points. Window boundary processing uses a mirror extension method to avoid data truncation; the extension length is set to 5% of the window size (e.g., a 10-second window extends by 0.5 seconds of data points) to prevent incomplete sequences due to missing boundary data. The signal frequency change rate is calculated. If the change rate exceeds 5% / second (e.g., a 5% change in 1 second when the signal frequency suddenly changes from 10GHz to 10.5GHz), the window is reduced to 5 seconds of data points; if the frequency is stable (change rate less than 0.5% / second), the window is expanded to 20 seconds of data points. The adjustment step size is limited by the system's minimum response time of 1 millisecond to ensure that window adjustment does not lag behind signal changes. Simultaneously, data processing employs a double buffering mechanism: the front buffer stores the original deviation sequence retrieved from the deviation database in real time, while the back buffer temporarily stores sequence segments to be processed, avoiding data read / write conflicts. In addition, the dynamic window adjustment algorithm also monitors the change of signal spectrum entropy. If the entropy value changes abruptly and exceeds the preset threshold, it immediately triggers window reconfiguration. The window size is negatively correlated with the rate of change of entropy (the higher the rate of change of entropy, the smaller the window). This can solve the problems of fixed windows being unable to match the frequency of signal changes, data redundancy during high-frequency interference, and insufficient information during low-frequency stability. It avoids the distortion of historical deviation sequence features caused by traditional fixed windows and ensures that the extracted sequence can accurately reflect the transient and steady-state characteristics of the deviation.

[0081] Secondly, the multi-scale feature extraction of the historical deviation sequence extracted by the sliding window is carried out. Firstly, the instantaneous change amount of adjacent deviation values is calculated by a first-order difference algorithm (for example, the first-order difference result of the sequence [2.009, 2.015, 2.021, 2.018, 2.025] is [0.006, 0.006, -0.003, 0.007]); then the acceleration of change is calculated by a second-order difference algorithm (for example, the second-order difference based on the first-order difference result is [0, -0.009, 0.01]); subsequently, a linear regression algorithm is used to fit the original historical deviation sequence, so as to obtain a regression coefficient reflecting the time evolution trend of the deviation (for example, the slope is 0.004 / s, and the intercept is 2.005), and the coefficient constitutes a time domain basic component of the deviation change trend vector. Meanwhile, the Fourier transform is applied to the historical deviation sequence to detect the periodic characteristics in the sequence (for example, it is identified that the period of the interference signal is 0.5 seconds), and the amplitude of the main frequency component (for example, 0.005) is normalized to be used as a frequency domain feature of the trend vector. The dimension of the deviation change trend vector is fixed as 6, and the first three dimensions correspond to the time domain features (respectively, the sequence mean, the linear regression slope, and the second-order difference curvature, such as [2.015, 0.004, 0.01]), and the last three dimensions store the frequency domain features (respectively, the normalized amplitudes of the first three main frequency components, such as [0.8, 0.15, 0.05]). In the feature extraction stage, a parallel computing architecture is adopted, and independent operation units are allocated to the time domain analysis (difference, linear regression) and the frequency domain analysis (Fourier transform) to improve the processing efficiency; the fixed-point number operation optimization is adopted in the vector generation algorithm, the time domain feature calculation accuracy is maintained at the level of 0.001, and the frequency domain feature resolution reaches 0.1 Hz. By fusing the time domain and frequency domain features, the problem of lag in the response of the pure time domain analysis to the periodic interference is solved, so that the deviation change trend vector can not only reflect the instantaneous change speed and direction of the deviation, but also capture the frequency characteristics of the periodic interference.

[0082] Finally, the normalized deviation change trend vector is obtained by normalizing the deviation change trend vector, and the maximum and minimum value method is adopted for the normalization. The value range of each dimension feature is first calculated from the training data set of the radar system (containing deviation trend vector samples under different working modes and different interference scenarios): the time domain mean is 1.8-2.2, the linear regression slope is -0.01-0.01 / s, the second-order difference curvature is -0.02-0.02, and the normalized amplitude of the first three main frequency components in the frequency domain is 0-1. The normalized value of each dimension is calculated according to the formula "standardized component=(original component-minimum value of this dimension) / (maximum value of this dimension-minimum value of this dimension)", for example, the deviation change trend vector [2.015, 0.004, 0.01, 0.8, 0.15, 0.05] is normalized to , the linear regression slope is normalized to , and the second-order difference curvature is normalized to After standardization of the three frequency domain components, the original values ​​are retained, resulting in a standardized deviation trend vector [0.5375, 0.7, 0.75, 0.8, 0.15, 0.05]. The normalization module has a built-in automatic parameter calibration function with a calibration cycle of 1 hour. Each calibration extracts historical trend vector data from the deviation database for the most recent 24 hours, recalculates the value range of each dimension, and updates the normalization parameters to ensure parameter timeliness. Simultaneously, the normalization process eliminates the dimensional differences in deviation features under different radar operating modes (such as search mode and tracking mode), making the standardized deviation trend vector comparable across scenarios and solving the problem of inconsistent feature scales under different modes, leading to weight imbalances in subsequent threshold comparisons.

[0083] Step S103: Calculate the trend slope and fluctuation amplitude using the deviation change trend vector, and then determine the adjustment trigger signal based on the relationship between the trend slope and the fluctuation amplitude and the preset threshold.

[0084] In one specific embodiment, the process of executing step S103 may specifically include the following steps:

[0085] Extract the time-domain linear regression slope component from the deviation change trend vector and determine it as the trend slope within the time window;

[0086] The fluctuation amplitude is calculated using the standard deviation algorithm for the time-domain and frequency-domain characteristic components of the deviation change trend vector;

[0087] If the trend slope exceeds a preset slope threshold and the fluctuation amplitude meets the abnormality determination conditions, it is marked as a mutation mode and an adjustment trigger signal is generated.

[0088] Specifically, when calculating the trend slope within the time window, it is necessary to extract the time-domain linear regression slope component from the deviation change trend vector. This component originates from the least squares fitting result of the historical deviation sequence. For example, the extracted historical deviation sequence is [2.009, 2.015, 2.021, 2.018, 2.025] (corresponding to a 5-second sliding window, with timestamps of [timestamps not specified]). t 1 to t 5) Obtain the fitted straight line equation through linear regression fitting. y =0.004 t +2.005 ( t The unit is seconds. yIf the slope of the straight line is 0.004 / s, which is the trend slope, the value directly reflects the speed of the deviation value changing over time in the time window. If the slope is positive, the deviation shows an upward trend, and if the slope is negative, the deviation shows a downward trend. The greater the absolute value of the slope, the more dramatic the change in the deviation. The calculation of the trend slope here strictly corresponds to the definition of the slope value of the deviation change trend vector, and the time window is consistent with the sliding window of step S102 (such as 5 seconds), ensuring the continuity of the data time dimension and avoiding distortion of the trend judgment due to window mismatch. This processing solves the problem of the lack of quantification of the speed of deviation change in the prior art, which cannot accurately identify signal mutations.

[0089] Further, when calculating the fluctuation amplitude of the deviation change trend vector [0.5375, 0.7, 0.75, 0.8, 0.15, 0.05], the standard deviation is 0.294. The standard deviation value is the fluctuation amplitude, which reflects the stability of the deviation change. The smaller the fluctuation amplitude, the more stable the deviation change, and vice versa. The correspondence between the difference sequence and the fluctuation amplitude ensures that the fluctuation amplitude can accurately quantify the dispersion degree of the deviation change, solving the problem of the lack of effective evaluation of the deviation fluctuation in the prior art, which easily misjudges noise as a signal mutation.

[0090] Finally, when judging whether the trend slope exceeds the preset threshold and generating an adjustment trigger signal, the preset threshold needs to be determined based on historical interference scene data statistics of the radar system. Different interference scenes correspond to different thresholds. For example, in the urban multipath interference scene, the deviation mutation is frequent due to the reflection of buildings on the signal, and the preset threshold is set to 0.005 / s. In the airport electromagnetic interference scene, the signal is relatively stable, and the preset threshold is set to 0.008 / s. By comparing the calculated trend slope with the preset threshold corresponding to the scene, if the trend slope exceeds the threshold, it is marked that the current interference signal is in the mutation mode, and an adjustment trigger signal is generated. The signal is a digital level signal, which uses a high level (logic 1) to indicate triggering and a low level (logic 0) to indicate non-triggering. The signal needs to carry a timestamp, a current interference scene label (such as "urban multipath interference"), and trend slope and fluctuation amplitude data, which are transmitted to the next step through the internal bus of the radar system to start the subsequent deviation compensation adjustment process. If the trend slope does not exceed the threshold, no mutation mode is marked, and no adjustment trigger signal is generated, and the system maintains the current deviation compensation model parameters. This threshold judgment mechanism solves the problem that the static model cannot adjust the compensation strategy in real time when the signal is mutated. By quickly generating an adjustment trigger signal, it ensures that the system can start the compensation process in time to avoid a sharp drop in positioning accuracy. At the same time, the fluctuation amplitude can be used as an auxiliary judgment basis. If the trend slope slightly exceeds the threshold but the fluctuation amplitude is very small, it can be determined that it is accidental noise, and the adjustment is not triggered, further improving the judgment accuracy.

[0091] Step S104, extracting relevant feature interaction from the deviation change trend vector according to the adjustment trigger signal, determining the correlation weight between features, and calculating the weight coefficient update increment according to the correlation weight.

[0092] In a specific embodiment, the process of step S104 can specifically include the following steps:

[0093] According to the adjustment trigger signal, analyze and extract the interference scene label, timestamp, trend slope and fluctuation amplitude;

[0094] According to the interference scene label, match the scene data set in the historical trend database, and extract feature interaction data from the scene data set according to the timestamp, trend slope and fluctuation amplitude;

[0095] Using Pearson correlation coefficient matrix to analyze the correlation degree between each feature in the feature interaction data set, and determine the correlation weight of signal intensity-frequency, signal intensity-modulation mode and frequency-modulation mode three groups of feature interaction;

[0096] Obtain the basic weight coefficient of the current deviation compensation model, and calculate the weight coefficient update increment according to the basic weight coefficient, the correlation weight and the dynamic adjustment coefficient.

[0097] Specifically, when extracting relevant feature interaction from the historical trend database according to the adjustment trigger signal, the adjustment trigger signal carries key retrieval information, including the calculation timestamp (such as "2025-08-21 15:48:55"), the interference scene label (such as "airport electromagnetic interference"), the trend slope (0.004 / s) and the fluctuation amplitude (such as 0.0060), which are used as composite conditions for database retrieval. The historical trend database stores feature data under various interference scenes in the past operation of the radar, and each record contains real-time sampling values of signal intensity, frequency and modulation mode, as well as interaction records between the three types of features. In the extraction process, the interference scene label is matched with the same scene data set in the database, the records within 1 hour closest to the current time are selected according to the timestamp, and the feature interaction data of similar interference change trend is further selected by combining the trend slope and the fluctuation amplitude range, for example, from the 1-hour data of the "airport electromagnetic interference" scene, 20 groups of signal intensity-frequency-modulation mode interaction records are extracted, which meet the slope and amplitude range, to ensure that the extracted feature interaction is highly consistent with the current interference state.

[0098] Further, in analyzing the feature interaction to determine the correlation weight between features, the extracted 20 groups of feature interaction data are input, and the feature correlation analysis is carried out among the signal strength, frequency and modulation mode. The present application calculates the linear correlation degree between features by using Pearson correlation coefficient. For example, when calculating the correlation coefficient between signal strength and frequency, the mean value of signal strength and the mean value of frequency in 20 groups of data are calculated first, and the correlation coefficient is calculated by formula (wherein is the correlation coefficient, x is the signal strength, is the signal strength average, y is the frequency, is the frequency average). When the correlation coefficient , it indicates that the signal strength and the frequency are strongly positively correlated, and the frequency has a greater impact on the strength; when calculating the correlation coefficient between signal strength and modulation mode, the modulation mode is quantized first, and the correlation coefficient , it indicates that the signal strength and the modulation mode are weakly correlated, and the modulation mode switching has a smaller impact on the strength; when calculating the correlation coefficient between frequency and modulation mode , it indicates that the frequency and the modulation mode are moderately correlated, and part of the modulation mode switching is accompanied by a slight frequency fluctuation. According to the absolute value of the correlation coefficient, the correlation weight is allocated, and the greater the absolute value of the correlation coefficient, the more significant the influence of the corresponding feature interaction on the deviation. Therefore, the correlation weight of the signal strength-frequency interaction is set to 0.55, the signal strength-modulation mode interaction is set to 0.2, and the frequency-modulation mode interaction is set to 0.25. This process solves the problem that the existing technology cannot quantify the contribution degree of feature interaction to the deviation and the subjective problem of weight allocation.

[0099] The weight coefficient update increment according to the correlation weight calculation is specifically: the basic weight coefficients of the current radar bias compensation model are called (such as the signal strength basic weight 0.4, the frequency weight 0.3, and the modulation mode weight 0.3), and the weight coefficient update increment is calculated by using the weighted average method, and the specific formula is: weight coefficient update increment = (signal strength - frequency interaction weight x signal strength basic weight + signal strength - modulation mode interaction weight x modulation mode basic weight + frequency - modulation mode interaction weight x frequency basic weight) x dynamic adjustment coefficient. Among them, the dynamic adjustment coefficient is determined according to the fluctuation amplitude, the larger the fluctuation, the larger the adjustment coefficient, so as to ensure that the increment can match the degree of interference change. Assuming that the adjustment coefficient is set to 0.12 when the fluctuation amplitude is 0.294, the weight coefficient update increment is (0.55x0.4+0.2x0.3+0.25x0.3)x0.12=0.0426. Further distribute the update increment to each feature weight: signal strength weight increment = 0.55x0.0426≈0.0234, frequency weight increment = 0.25x0.0426≈0.0107, modulation mode weight increment = 0.2x0.0426≈0.0085, and finally obtain the weight coefficient update increment vector [0.0234, 0.0107, 0.0085]. The calculation result is stored in the system weight cache area, and is associated with the timestamp of the adjustment trigger signal and the interference scene label, so as to solve the problem that the weight coefficient update lags behind and cannot reflect the dynamic contribution of feature interaction to the bias in real time.

[0100] Step S105, updating the convergence speed parameter according to the weight coefficient update increment, and optimizing the updated convergence speed parameter by a gradient descent iterative optimization algorithm to obtain an optimized convergence speed value.

[0101] In a specific embodiment, the process of performing step S105 can specifically include the following steps:

[0102] Obtain the current convergence speed parameter from the parameter cache area of the radar bias compensation model, and the parameter cache area contains the convergence speed parameter, the update timestamp and the corresponding interference scene label;

[0103] Weighted sum the weight coefficient update increment with the intensity influence coefficient, the frequency influence coefficient and the modulation influence coefficient to calculate the update amount of the convergence speed parameter, and then superimpose the update amount of the convergence speed parameter with the current convergence speed parameter to obtain the updated convergence speed parameter, wherein the intensity influence coefficient, the frequency influence coefficient and the modulation influence coefficient are determined by historical data statistics;

[0104] Iteratively optimize the updated convergence speed parameter by using the gradient descent method.

[0105] Specifically, when obtaining the current convergence speed parameter, the parameter is called from the parameter cache area of the radar bias compensation model. The parameter cache area stores real-time parameters according to radar working modes (such as tracking mode and search mode). Each record contains a convergence speed parameter, an update timestamp, and a corresponding interference scene label (such as "urban multipath interference" and "airport electromagnetic interference"). The convergence speed parameter reflects the response speed of the bias compensation model, and the value range is 0.1-0.8. The larger the value, the faster the model responds to changes in bias. For example, in the tracking mode, the initial value of the current convergence speed parameter is set to 0.3 because the interference changes caused by the need to quickly adapt to target movement. In the search mode, the initial value of the parameter is set to 0.2 because the signal coverage is wide and the interference is relatively stable. When calling, the current radar working mode label is used to accurately match the parameters, and the parameter update timestamp is verified to ensure the timeliness of the parameters.

[0106] When updating the convergence speed parameter according to the weight coefficient update increment, the weight coefficient update increment (such as vector [0.0234, 0.0107, 0.0085], corresponding to the weight increment of signal strength, frequency, and modulation mode, respectively) generated in step S104 is used as input. The parameter update amount is calculated using the weighted sum formula: parameter update amount = (signal strength weight increment x strength influence coefficient + frequency weight increment x frequency influence coefficient + modulation mode weight increment x modulation influence coefficient). The influence coefficient is determined according to historical data statistics to determine that signal strength has the most significant influence on the response speed of the model, and the strength influence coefficient is set to 0.5. Frequency is second, and the frequency influence coefficient is set to 0.3. Modulation mode has less influence, and the modulation influence coefficient is set to 0.2. The weight coefficient update increment is calculated as follows: parameter update amount = (0.0234 x 0.5 + 0.0107 x 0.3 + 0.0085 x 0.2) = 0.01661. Then, the update amount is superimposed with the current convergence speed parameter to obtain the updated convergence speed parameter: 0.3 + 0.01661 = 0.31661 in tracking mode, and 0.2 + 0.01661 = 0.21661 in search mode. During the update process, the maximum value of the parameter should be limited to not more than 0.8, and the minimum value should be limited to not less than 0.1, to avoid parameter abnormalities causing model response to be too fast or too slow.

[0107] The updated convergence speed parameter is optimized by gradient descent method. The learning rate is set to 0.01 when initializing the gradient descent parameter, the upper limit of the iteration number is set to 100 times, and the convergence threshold is set to 0.001. The iteration process takes the square sum of the deviation compensation residual as the loss function, and the loss function value is obtained by real-time calculation of the difference between the current model output and the target value. The gradient of the loss function with respect to the convergence speed parameter is calculated in each iteration, and the gradient value is obtained by numerical differentiation method, and the differential step is set to 0.001. The parameter update is carried out in the opposite direction of the gradient, and the update step is determined by the product of the learning rate and the gradient value. The iteration termination condition is that the loss function value is lower than 0.001 or the maximum iteration number is reached, and the convergence speed parameter value obtained at this time is the optimized convergence speed value. Gradient descent iteration optimization solves the problems of slow convergence, high computational complexity and delayed response when signal suddenly changes in traditional iteration algorithm, and ensures that the optimized convergence speed value can quickly adapt to the current interference change.

[0108] In step S106, the deviation compensation model parameters are adjusted according to the optimized convergence speed value, the compensation residual of the adjusted deviation compensation model is calculated, and if the compensation residual is lower than the preset residual threshold, the deviation compensation result is output.

[0109] In an embodiment, the process of step S106 can specifically include the following steps:

[0110] The initial parameters of the current deviation compensation model are called, including the deviation compensation coefficient and the compensation offset;

[0111] The deviation compensation coefficient and the compensation offset are adjusted according to the optimized convergence speed value, so that the adjusted parameters are adapted to the convergence speed of the deviation compensation model;

[0112] The characteristic parameters of the current signal are input into the adjusted deviation compensation model, and the real-time compensation value is output after the sliding average filtering process. The ideal compensation value matching the current interference scene and signal characteristics is called, the absolute value error between the real-time compensation value and the ideal compensation value is the compensation residual;

[0113] If the compensation residual is lower than the residual threshold, the deviation compensation result containing the filtered real-time compensation value, the new compensation coefficient, the new compensation offset, the time stamp and the scene label is output, and if the compensation residual is higher than the residual threshold, the convergence speed parameter is iteratively optimized again in step S105.

[0114] Specifically, when adjusting the deviation compensation model parameters according to the optimized convergence speed value, the current deviation compensation coefficient and compensation offset are first called from the parameter storage unit of the deviation compensation model - the deviation compensation coefficient is used to quantify the compensation weight of the signal characteristics on the deviation (such as the deviation compensation amount corresponding to a change of 1 dBm in signal strength), and the compensation offset is used to correct the system deviation caused by fixed interference, and the initial values of the two need to match the radar working mode, for example, in tracking mode, the initial deviation compensation coefficient is 1.2 (high response requirement), and the compensation offset is 0.05 (small fixed deviation); in search mode, the initial coefficient is 1.0 (smooth requirement), and the offset is 0.03. The optimized convergence speed value (such as 0.31661 obtained in step S105 in tracking mode) is used as the basis for adjustment, and the correlation logic between the parameters and the convergence speed value is established: the deviation compensation coefficient is adjusted in positive correlation, and the formula is "new compensation coefficient = original compensation coefficient x (1 + optimized convergence speed value x 0.1)", and the data is substituted to obtain a new compensation coefficient = 1.2 x (1 + 0.31661 x 0.1) = 1.238, which ensures that the faster the convergence speed, the larger the coefficient to enhance the compensation sensitivity; the compensation offset is adjusted in increments, and the formula is "new compensation offset = original compensation offset + optimized convergence speed value x 0.02", and the data is substituted to obtain a new compensation offset = 0.05 + 0.31661 x 0.02 = 0.0563, which makes the offset adapt to the dynamic changes of interference with the convergence speed.

[0115] Further, when calculating the compensation residual of the adjusted deviation compensation model, the characteristic parameters (signal strength -68 dBm, frequency 10.03 GHz, modulation mode FM) of the current interference signal are input into the adjusted deviation compensation model (containing the new compensation coefficient 1.238 and the new compensation offset 0.0563), and the compensation value is output through model operation - the model operation logic is "compensation value = (signal strength deviation component x compensation coefficient) + (frequency deviation component x compensation coefficient x 0.8) + compensation offset" (the frequency component weight 0.8 is derived from historical characteristic correlation analysis), and if the signal strength deviation component is 4 dBm and the frequency deviation component is 0.03 GHz, then the compensation value = (4 x 1.238) + (0.03 x 1.238 x 0.8) + 0.0563 = 5.038. Then the compensation target value is determined, which is called from the recorded historical trend database, and the ideal compensation value (such as 5.036) matching the current interference scene (such as "urban multipath interference") and signal characteristics is selected, and the compensation residual is calculated according to "compensation residual = |adjusted model output compensation value - compensation target value|", and the data is substituted to obtain a compensation residual = |5.038 - 5.036| = 0.002. During the calculation process, digital filtering technology (such as sliding average filtering, window 3 data points) is used to eliminate the output fluctuations caused by interference signal noise, to ensure the accuracy of residual calculation, and to solve the problem that the residual calculation is disturbed by noise and cannot truly reflect the compensation effect.

[0116] Finally, if the compensation residual is lower than the preset threshold, the preset threshold needs to be set in combination with the radar working mode and the interference scene when outputting the deviation compensation result: in the tracking mode, the threshold is set to 0.005 because high-precision positioning is required; in the search mode, the threshold is set to 0.01 because the signal coverage range is wide. Comparing the calculated compensation residual 0.002 with the tracking mode threshold 0.005, it is determined that the residual is lower than the threshold, and then the deviation compensation result is generated, which includes the compensation value 5.038 of the adjusted model output, the new deviation compensation coefficient 1.238, the new compensation offset 0.0563, the calculation timestamp (such as "2025-08-21 15:48:55"), and the interference scene label "urban multipath interference", and is stored in the system compensation result cache area and simultaneously transmitted to the next process through the internal data bus. If the compensation residual is higher than the threshold, return to step S105 to reiterate optimization until the residual meets the standard. This threshold judgment and result output mechanism solves the problem of accumulated positioning deviation caused by using unverified compensation results, and ensures that the output deviation compensation result can provide reliable support for subsequent positioning correction.

[0117] Step S107, adjusting the input parameter weight of the positioning algorithm according to the deviation compensation result, and determining the corrected positioning coordinates by weighted calculation on the adjusted input parameters through the weighted positioning algorithm.

[0118] In a specific embodiment, the process of performing step S107 can specifically include the following steps:

[0119] The deviation compensation result is decomposed into a x direction component and a y direction component corresponding to the original positioning coordinates of the radar.

[0120] The input parameters of the positioning algorithm are adjusted according to the x direction component and the y direction component, and the input parameters of the positioning algorithm include the distance estimation value corresponding to the signal strength, the phase offset value corresponding to the frequency, and the positioning error correction value corresponding to the modulation mode.

[0121] The weighted positioning method is adopted, the adjusted input parameters are multiplied by the corresponding adjusted weights respectively, and then the sum of the product results is obtained to get the corrected coordinate components.

[0122] Specifically, when the deviation compensation result is integrated into the positioning calculation process, the deviation compensation result needs to include the adjusted compensation value, the deviation compensation coefficient, the compensation offset, and the associated interference scene label (such as urban multipath interference and airport electromagnetic interference) and the calculation timestamp. The positioning calculation process originally includes an original coordinate generation step, and at this time, a coordinate correction step based on the deviation compensation result is inserted to decompose the compensation value into components corresponding to the original positioning coordinates (such as the originalx coordinate correspondence x direction compensation component, original y coordinate correspondence y direction compensation component), ensuring that the compensation data and the coordinate components are one-to-one, such as the original x coordinate is 50 meters, if the compensation value is 0.0965 meters, then directly associate the original coordinate component for subsequent correction, to avoid correction errors caused by data misplacement. x direction compensation component is 0.0965 meters, then directly associate the original coordinate component for subsequent correction, to avoid correction errors caused by data misplacement.

[0123] Further, when adjusting the input parameters of the positioning algorithm according to the deviation compensation result, the input parameters of the positioning algorithm include the distance estimation value corresponding to the signal strength, the phase offset value corresponding to the frequency, and the positioning error correction value corresponding to the modulation mode. The initial weights of these parameters need to be adjusted in combination with the weight increment in the deviation compensation result. For example, if the signal strength weight increment in the deviation compensation result is 0.0234, the frequency weight increment is 0.0107, and the modulation mode weight increment is 0.0085, and the initial weights of the signal strength, frequency, and modulation mode in the positioning algorithm are 0.5, 0.3, and 0.2 respectively, then the adjusted weights are 0.5+0.0234=0.5234, 0.3+0.0107=0.3107, and 0.2+0.0085=0.2085 respectively. In this adjustment process, the weight change of each input parameter is directly associated with the corresponding weight increment in the deviation compensation result, ensuring that the input parameter weight can reflect the dynamic contribution of each signal feature to positioning under current interference in real time, solving the compensation residual problem caused by insufficient feature interaction processing capability and lagging weight coefficient update in the prior art.

[0124] Finally, when calculating the corrected positioning coordinates by the weighted positioning method, the adjusted input parameters (such as the signal strength distance estimation value, frequency phase offset value, and modulation mode error correction value after adjusting the weights) are multiplied by the corresponding adjusted weights respectively, and then the sum of the products is obtained to get the corrected coordinate components. Taking x direction coordinate calculation as an example, if the adjusted signal strength distance estimation value is 30 meters, multiplied by the adjusted weight 0.5234, the result is 15.702; the x component corresponding to the frequency phase offset is 15 meters, multiplied by the adjusted weight 0.3107, the result is 4.0065; the x modulation mode corresponding to the direction error correction value is 5 meters, multiplied by the adjusted weight 0.2085, the result is 1.0425; the sum of the three is 15.702+4.0065+4.0065=20.751 meters, which is the x corrected coordinate in the direction. yThe direction coordinate calculation logic is the same. If the estimated signal strength distance after adjustment is 20 meters, multiplying it by 0.5234 gives 10.468; the frequency phase offset corresponds to... y The component is 10 meters, multiplied by 0.3107, resulting in 3.107; the modulation method corresponds to... y The direction error correction value is 3 meters, multiplied by 0.2085, we get 0.6255; summing, we get 10.468 + 3.107 + 0.6255 = 14.2005 meters, which is... y The coordinates after direction correction. In this weighted calculation process, the input parameters, adjusted weights, and coordinate components form a clear data correspondence. The algorithm prioritizes signal features that have a significant impact under the current interference through weight allocation, solving the problem that static models cannot adapt to the dynamic changes of signals in complex environments and the sharp drop in positioning accuracy. At the same time, it associates interference scene labels to achieve the synergy between interference source localization and anti-interference, enhancing the anti-interference robustness under multipath effects and dynamic interference source switching scenarios.

[0125] The radar dynamic anti-jamming method based on interference source localization in the embodiments of this application has been described above. The radar dynamic anti-jamming processing procedure 200 based on interference source localization in the embodiments of this application will be described below. Please refer to... Figure 2 The processing procedure 200 revolves around a closed-loop operation of radar dynamic anti-interference based on interference source localization. It consists of seven processes: signal acquisition 201, deviation analysis 202, slope and threshold judgment 203, weight optimization 204, convergence optimization 205, compensation output 206, and positioning correction output 207. These processes work together to achieve signal processing and positioning correction functions. The specific process is as follows:

[0126] In the signal acquisition process 201, the signal receiving device samples the current interference signal in real time according to the radar operating frequency. At the same time, it retrieves historical sequences from the database storing historical signals of the same operating mode for the past 30 minutes by dynamically matching the signal frequency. After sampling, it extracts features such as signal strength, frequency (extracted by fast Fourier transform + peak detection), and modulation mode (identified by demodulation circuit + maximum likelihood estimation). The deviation is calculated by multi-dimensional difference quantification (the strength deviation is the difference between the current and historical mean, the frequency deviation is Euclidean distance, and the modulation mode deviation is qualitatively assigned and weighted according to the historical variance).

[0127] The bias analysis process 202 enables an initial 10-second (including 1000 data points) sliding window, according to the signal frequency change rate (5 seconds for more than 5% / second, 20 seconds for less than 0.5% / second, and a double buffering mechanism to ensure read-write conflict-free), to carry out frequency domain analysis (Fourier transform to extract frequency domain components) and time domain analysis (first-order / two-order difference algorithm to extract time domain components), to obtain regression slope, main amplitude, etc., and to generate a 6-dimensional (3 time domain + 3 frequency domain) trend vector through maximum and minimum value normalization (based on training data set statistical range, 1 hour calibration parameters).

[0128] The slope and threshold judgment process 203 first extracts the time domain linear regression slope based on the trend vector, calculates the vector standard deviation for bias analysis, judges whether the slope is higher than the preset scene threshold (0.005 / second for urban multipath interference, 0.008 / second for airport electromagnetic interference), and generates a high-level trigger signal with timestamp and scene label if it is higher than the threshold. If it is not higher than the threshold, it further judges whether the fluctuation amplitude is higher than the threshold, and generates an adjustment trigger signal if it is higher than the threshold, and re-executes the bias analysis process if it is not higher than the threshold.

[0129] The weight optimization process 204 is driven by the adjustment trigger signal, extracts 20 groups of similar interaction data from the historical database according to the timestamp, scene label, slope / amplitude of the trigger signal; calculates the correlation degree between each pair of features (such as intensity-frequency, R∈(0.6,1) when the weight is set to 0.55) through Pearson correlation coefficient and normalizes to get the correlation weight; calculates the weight increment as “basic weight + (interaction weight x basic weight) x dynamic adjustment coefficient (according to fluctuation amplitude, such as 0.12 when amplitude is 0.294)” and outputs to the convergence optimization process.

[0130] The convergence optimization process 205 receives the weight update increment, retrieves the current convergence speed parameter (tracking mode 0.3, search mode 0.2, verify timestamp) from the parameter cache area; according to “parameter update amount = intensity increment x 0.5 + frequency increment x 0.3 + modulation increment x 0.2”, the current parameter is added (limited to 0.1-0.8); iteratively optimized by gradient descent method (learning rate 0.01, iteration 100 times, convergence threshold 0.001, residual sum of squares as loss function), and outputs the optimized speed value if the convergence condition is met, otherwise continue iteration.

[0131] The compensation output process 206 inputs the optimization convergence speed, adjusts the compensation model parameters (such as the tracking mode compensation coefficient 1.2 -> 1.238) according to "compensation coefficient = original coefficient x (1 + convergence speed x 0.1), offset = original offset + convergence speed x 0.02"; calculates the residual of the model output compensation value and the historical library ideal value (such as 5.038-5.036 = 0.002, which is denoised by sliding average filtering); judges whether the residual is lower than the scene threshold (0.005 for the tracking mode and 0.01 for the search mode), and outputs the compensation result with parameters, timestamps and scene labels if yes, otherwise returns to the convergence optimization process for re-adjustment.

[0132] The positioning correction output process 207 inputs the compensation result, and decomposes it into x / y The coordinate component is inserted into the positioning flow correction step; the positioning algorithm input parameter weight is adjusted according to the weight increment (such as the intensity weight 0.5 -> 0.5234); the correction coordinates are output after fusion coordinates are calculated by the weighted positioning method (the sum of the adjusted parameters x the corresponding weights, such as x direction 15.702 + 4.0065 + 1.0425 = 20.751 meters) calculation, completing the whole process processing from signal acquisition to positioning accuracy correction.

[0133] The radar dynamic anti-jamming method based on interference source positioning in the embodiments of the application is described above, and the radar dynamic anti-jamming system based on interference source positioning in the embodiments of the application is described below. Please refer to Figure 3 , one embodiment of the radar dynamic anti-jamming system based on interference source positioning in the embodiments of the application includes:

[0134] The signal acquisition module 301 is configured to acquire current signal data and historical signal sequences, calculate and determine an initial deviation value, and guarantee the real-time and accuracy of data acquisition.

[0135] The deviation analysis module 302 extracts a historical deviation sequence according to the initial deviation value, generates a deviation change trend vector after processing, and intuitively reflects the change characteristics of the deviation over time.

[0136] The trigger judgment module 303 calculates the trend slope and fluctuation amplitude by using the deviation change trend vector, determines and generates an adjustment trigger signal immediately if the slope exceeds a preset threshold, and starts the subsequent compensation adjustment process.

[0137] The weight optimization module 304 extracts feature interaction according to the adjustment trigger signal, analyzes feature correlation to determine a correlation weight, and calculates a weight coefficient update increment, thereby providing a basis for model parameter update.

[0138] The convergence optimization module 305 acquires a current convergence speed parameter, updates the parameter by combining a weight coefficient update increment, optimizes by an iterative optimization algorithm to obtain an optimized convergence speed value, and improves the model response speed.

[0139] The compensation output module 306 adjusts the deviation compensation model parameter according to the optimized convergence speed value, calculates a compensation residual, and outputs a deviation compensation result if the residual is lower than a preset threshold, which is used for subsequent positioning correction.

[0140] The positioning correction module 307 integrates the deviation compensation result into a positioning calculation process, adjusts the positioning algorithm input parameter, calculates and determines a corrected positioning coordinate by the positioning algorithm, and realizes accurate positioning.

[0141] The above Figure 3 The radar dynamic anti-interference system based on interference source positioning in the embodiment of the application is described in detail from the perspective of a modular functional entity, and the radar dynamic anti-interference device based on interference source positioning in the embodiment of the application is described in detail from the perspective of hardware processing.

[0142] Referring to Figure 4 In the embodiment of the application, a radar dynamic anti-interference device 400 based on interference source positioning is also provided, which can be a server, and its internal structure can be as shown in Figure 4 The radar dynamic anti-interference device 400 based on interference source positioning includes a processor 402, a memory 403, a display screen 404, an input device 405, a network interface 406 and a database 407 connected through a system bus 401. The processor 402 of the computer design is used to provide computing and control capabilities. The memory 403 of the radar dynamic anti-interference device based on interference source positioning includes a non-volatile storage medium 4031 and an internal memory 4032. The non-volatile storage medium 4031 stores an operating system and a computer program. The internal memory 4032 provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database 407 of the radar dynamic anti-interference device based on interference source positioning is used to store the corresponding data in the embodiment. The network interface 406 of the radar dynamic anti-interference device based on interference source positioning is used to communicate with external terminals through network connection. The computer program executed by the processor can implement the above method.

[0143] Those skilled in the art can understand, Figure 4The structures shown are merely block diagrams of some structures related to the present invention and do not constitute a limitation on the radar dynamic anti-jamming device based on interference source localization applied thereto. Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0144] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A radar dynamic anti-jamming method based on interference source localization, characterized in that, The method includes: Step S101: Obtain current signal data and historical signal sequences, and extract signal strength, frequency and modulation mode characteristic parameters, and calculate the initial deviation value based on the signal strength, frequency and modulation mode characteristic parameters; Step S102: Extract historical deviation sequences from the initial deviation values ​​based on a dynamic sliding window, and perform time series analysis on the historical deviation sequences to generate a deviation change trend vector; Step S103: Calculate the trend slope and fluctuation amplitude using the deviation change trend vector, and then determine the adjustment trigger signal based on the relationship between the trend slope and the fluctuation amplitude and the preset threshold. Step S104: Extract relevant feature interaction effects from the deviation change trend vector according to the adjustment trigger signal, determine the correlation weight between each feature, and calculate the weight coefficient update increment according to the correlation weight. Step S105: Update the incremental convergence speed parameter according to the weight coefficient, and optimize the updated convergence speed parameter by gradient descent iterative optimization algorithm to obtain the optimized convergence speed value. Step S106: Adjust the parameters of the deviation compensation model according to the optimized convergence speed value, calculate the compensation residual of the adjusted deviation compensation model, and output the deviation compensation result if the compensation residual is lower than the preset residual threshold. Step S107: Adjust the weights of the input parameters of the positioning algorithm according to the deviation compensation result, and perform weighted calculation on the adjusted input parameters through a weighted positioning algorithm to determine the corrected positioning coordinates.

2. The radar dynamic anti-jamming method based on interference source localization according to claim 1, characterized in that, Step S101 includes: Acquiring current interference signal data and historical signal sequences includes: real-time acquisition of current interference signal data through radar signal receiving equipment, and retrieval of historical signal sequences within a preset time period that match the current radar operating mode from a pre-established signal database; The initial deviation value is calculated based on the characteristic parameters of the current interference signal and the historical signal sequence, including: calculating the intensity deviation component, frequency deviation component and modulation mode deviation component of the interference signal and the historical signal respectively; The weighting coefficients of the deviation components are determined based on the variance of the historical signal sequence. The initial deviation value is obtained by weighted summation and then stored in association with the calculation timestamp, radar operating mode, and interference scene label.

3. The radar dynamic anti-jamming method based on interference source localization according to claim 2, characterized in that, Step S102 includes: By applying the sliding window technique, historical deviation sequences are extracted from a pre-stored deviation database. The window size of the sliding window technique is dynamically adjusted according to the frequency of signal changes. Time series analysis is performed on the historical deviation sequence to generate a deviation change trend vector; The deviation change trend vector is normalized to obtain the standardized deviation change trend vector.

4. The radar dynamic anti-jamming method based on interference source localization according to claim 1, characterized in that, Step S103 includes: Extract the time-domain linear regression slope component from the deviation change trend vector and determine it as the trend slope within the time window; The fluctuation amplitude is calculated using the standard deviation algorithm for the time-domain and frequency-domain characteristic components of the deviation change trend vector; If the trend slope exceeds a preset slope threshold and the fluctuation amplitude meets the abnormality determination conditions, it is marked as a mutation mode and an adjustment trigger signal is generated.

5. The radar dynamic anti-jamming method based on interference source localization according to claim 3, characterized in that, Step S104 includes: Based on the adjustment trigger signal analysis, the interference scene label, timestamp, trend slope, and fluctuation amplitude are extracted. Based on the interference scene label, match the scene dataset in the historical trend database, and extract feature interaction data from the scene dataset based on the timestamp, the trend slope, and the fluctuation amplitude; The correlation between features in the feature interaction dataset was analyzed using the Pearson correlation coefficient matrix to determine the correlation weights of the three sets of feature interactions: signal strength-frequency, signal strength-modulation mode, and frequency-modulation mode. Obtain the basic weight coefficients of the current deviation compensation model, and calculate the weight coefficient update increment based on the basic weight coefficients, the correlation weights, and the dynamic adjustment coefficients.

6. The radar dynamic anti-jamming method based on interference source localization according to claim 1, characterized in that, Step S105 includes: The current convergence speed parameters are obtained from the parameter buffer of the radar bias compensation model. The parameter buffer includes the convergence speed parameters, update timestamps, and corresponding interference scene labels. The update amount of the convergence speed parameter is calculated by weighting and summing the update increment of the weight coefficient with the intensity influence coefficient, frequency influence coefficient and modulation influence coefficient respectively. Then, the update amount of the convergence speed parameter is superimposed with the current convergence speed parameter to obtain the updated convergence speed parameter. The intensity influence coefficient, frequency influence coefficient and modulation influence coefficient are determined by statistical analysis of historical data. The updated convergence speed parameters are iteratively optimized using the gradient descent method.

7. The radar dynamic anti-jamming method based on interference source localization according to claim 1, characterized in that, Step S106 includes: Retrieve the initial parameters of the current deviation compensation model, including the deviation compensation coefficient and the compensation offset; Based on the optimized convergence speed value, the deviation compensation coefficient and the compensation offset are adjusted so that the adjusted parameters are adapted to the convergence speed of the deviation compensation model. The characteristic parameters of the current signal are input into the adjusted deviation compensation model. After the moving average filtering process, the real-time compensation value is output. Then, the ideal compensation value that matches the current interference scene and signal characteristics is retrieved. The absolute value error between the real-time compensation value and the ideal compensation value is the compensation residual. If the compensation residual is lower than the residual threshold, the output includes the real-time compensation value after filtering, the new compensation coefficient, the new compensation offset, the timestamp, and the scene label. If the compensation residual is higher than the residual threshold, return to step S105 to iterate and optimize the convergence speed parameters again.

8. The radar dynamic anti-jamming method based on interference source localization according to claim 7, characterized in that, Step S107 includes: The deviation compensation result is decomposed into values ​​corresponding to the original radar positioning coordinates. x Directional components and y Directional components; According to the above x Directional components and the y The input parameters of the directional component adjustment positioning algorithm include the distance estimation value corresponding to the signal strength, the phase offset value corresponding to the frequency, and the positioning error correction value corresponding to the modulation method. The weighted positioning method is adopted, which obtains the corrected coordinate components by multiplying each adjusted input parameter by its corresponding adjusted weight and then summing the product results.

9. A radar dynamic anti-jamming system based on interference source localization, characterized in that, For implementing the radar dynamic anti-jamming method based on interference source localization as described in any one of claims 1-8, the radar dynamic anti-jamming system based on interference source localization comprises: The signal acquisition module is used to acquire current signal data and historical signal sequences, calculate and determine the initial deviation value, and ensure the real-time performance and accuracy of data acquisition. The deviation analysis module extracts historical deviation sequences based on initial deviation values, processes them to generate a deviation change trend vector, and intuitively reflects the characteristics of deviation changes over time. The trigger judgment module uses the deviation change trend vector to calculate the trend slope and fluctuation amplitude. If the slope exceeds the preset threshold, it immediately determines and generates an adjustment trigger signal to start the subsequent compensation and adjustment process. The weight optimization module extracts the interaction effects of features based on the adjustment trigger signal, analyzes the correlation of features to determine the correlation weight, and calculates the weight coefficient update increment to provide a basis for updating model parameters. The convergence optimization module obtains the current convergence speed parameter, updates the parameter incrementally in conjunction with the weight coefficient, and optimizes it through an iterative optimization algorithm to obtain the optimized convergence speed value, thereby improving the model response speed. The compensation output module adjusts the parameters of the deviation compensation model according to the optimized convergence speed value, calculates the compensation residual, and outputs the deviation compensation result if the residual is lower than the preset threshold for subsequent positioning correction. The positioning correction module integrates the deviation compensation results into the positioning calculation process, adjusts the input parameters of the positioning algorithm, calculates and determines the corrected positioning coordinates through the positioning algorithm, and achieves accurate positioning.

10. A radar dynamic anti-jamming device based on interference source localization, characterized in that, It includes a memory and a processor, the memory storing a computer program that can run on the processor, and the processor executing the computer program to implement the radar dynamic anti-jamming method based on interference source localization as described in any one of claims 1 to 8.

Images